Alexandra Y. Aikhenvald, R. M. W. Dixon (2014) The Grammar of Knowledge - A Cross-Linguistic Typology-Oxford University Press

The Grammar of Knowledge
E X PL O R AT IO N S I N L I N G U I S T IC T Y P OL O G Y
general editors: Alexandra Y. Aikhenvald and R. M. W. Dixon
The Cairns Institute, James Cook University
This series focuses on aspects of language that are of current theoretical interest and
for which there has not previously or recently been any full-scale cross-linguistic
study. Its books are for typologists, fieldworkers, and theory developers, and
designed for use in advanced seminars and courses.
published
1 Adjective Classes
edited by
R. M. W. Dixon and Alexandra Y. Aikhenvald
2 Serial Verb Constructions
edited by
Alexandra Y. Aikhenvald and R. M. W. Dixon
3 Complementation
edited by
4 Grammars in Contact
edited by
5 The Semantics of Clause Linking
edited by
6 Possession and Ownership
edited by
7 The Grammar of Knowledge
edited by
published in association with the series

Areal Diffusion and Genetic Inheritance
Problems in Comparative Linguistics
edited by
The Grammar
of Knowledge
A Cross-Linguistic Typology
Edited by
A L E X A N DR A Y. A I K H E N VA L D and
R . M . W. DI XON
Language and Culture Research Centre

James Cook University
1
3
Great Clarendon Street, Oxford, ox2 6dp,
United Kingdom
Oxford University Press is a department of the University of Oxford.
It furthers the University’s objective of excellence in research, scholarship,
and education by publishing worldwide. Oxford is a registered trade mark of
Oxford University Press in the UK and in certain other countries
© Editorial matter and organization Alexandra Y. Aikhenvald and R. M. W. Dixon 2014
© The chapters their several authors 2014
The moral rights of the authors have been asserted
First Edition published in 2014
Impression: 1
All rights reserved. No part of this publication may be reproduced, stored in
a retrieval system, or transmitted, in any form or by any means, without the
prior permission in writing of Oxford University Press, or as expressly permitted
by law, by licence or under terms agreed with the appropriate reprographics
rights organization. Enquiries concerning reproduction outside the scope of the
above should be sent to the Rights Department, Oxford University Press, at the
address above
You must not circulate this work in any other form
and you must impose this same condition on any acquirer
Published in the United States of America by Oxford University Press
198 Madison Avenue, New York, NY 10016, United States of America
British Library Cataloguing in Publication Data
Data available
Library of Congress Control Number: 2013944500
ISBN 978–0–19–870131–6
As printed and bound by
CPI Group (UK) Ltd, Croydon, CRO 4YY
Links to third party websites are provided by Oxford in good faith and
for information only. Oxford disclaims any responsibility for the materials
contained in any third party website referenced in this work.
Contents
Preface x
Notes on the contributors xii
Abbreviations xvi
1 The grammar of knowledge: a cross-linguistic view of evidentials

and the expression of information source 1
Alexandra Y. Aikhenvald
1 Evidentiality and information source 2
2 Evidentiality: a bird’s-eye view 3
3 Information source through other means 19
4 Choosing, and using, an evidential 29
5 How to talk about knowledge 37
6 About this volume 41
Appendix. On terminological clarity 44
References 45
2 The grammar of knowledge in Hinuq 52
Diana Forker
1
Introduction 52
2
General properties of the evidential system and formal marking 52
3
Lexical expression of information source 61
4
Reported speech constructions, the quotative enclitic, and
evidentiality 63
5 Cultural conventions 65
6 Summary 67
References 67
3 The expression of knowledge in Tatar 69
Teija Greed
1 Introduction 69
2 Tatars: cultural context and society 69
3 General characteristics of Tatar 70
4 Grammatical expression of information source in Tatar 71
5 Evidential functions of particles 75
6 Evidentials in use 81
7 Lexical expression of knowledge and attitudes towards it 83
8 Historical development and pathways of evidentials 86
vi Contents
9 Conclusion 87
References 87
4 The grammar of knowledge in Saaroa 89
Chia-jung Pan
1 Preliminaries 89
2 Organization of the evidential system 92
3 Expression of reported evidentials 93
4 Semantics of reported evidentiality 97
5 The reported evidential and person 98
6 Origins of the reported evidential 100
7 Reported speech as an evidentiality strategy 101
8 Evidential-like meanings through special markers 102
9 Verbs of perception as lexical means to express knowledge 104
10 Evidentiality, communication, and cultural knowledge 105
11 Summary 106
References 106
5 The grammar of knowledge in Kurtöp: evidentiality, mirativity,
and expectation of knowledge 108
Gwendolyn Hyslop
1 Introduction 108
2 Background 109
3 Perfective 113
4 Imperfective aspect 117
5 Future tense 119
6 Copulas 120
7 Particles 123
8 Putting it together 127
References 130
6 The expression of knowledge in Ersu 132
Sihong Zhang
1 The language 132
2 The data 133
3 An overview 133
4 Evidentials and their semantics 134
5 Evidentiality and other grammatical categories 138
6 Co-occurrence of evidentials 142
7 Information source conveyed through other means 142
8 Summary 145
References 146
Contents vii
7 Evidentiality in Kalmyk 148

Elena Skribnik and Olga Seesing
1
Kalmyk 148
2
Organization of the evidential system 152
3
Direct evidential 152
4
Indirectivity 155
5
Inferred evidential 157
6
Assumed evidential 159
7
Prospective evidential 160
8
Reported evidential 161
9
Common knowledge 163
10
Evidentials in discourse 164
11
Evidentiality strategies: complementation 165
12
Evidentiality strategy: demonstratives and interpretation
of evidence 167
13 Mirativity 168
14 Summary 169
Text corpus 170
References 170
8 The non-visible marker in Dyirbal 171

R. M. W. Dixon
1 Introduction 171
2 Grammatical background 172
3 Noun markers 174
4 Verb markers 180
5 Bound forms 182
6 Interrogative noun and verb markers 182
7 Meanings of ŋa- markers 183
8 Conclusion 188
Appendix 188
References 189
9 The grammar of knowledge in Maaka (Western Chadic, Nigeria) 190

Anne Storch and Jules Jacques Coly
1 Introduction 190
2 Evidentiality, information source, and truth 195
3 Transmission of knowledge 202
4 Lexical expression of perception and cognition 204
5 Conclusions 206
References 207
viii Contents
10 Expression of information source meanings in Ashéninka Perené

(Arawak) 209
Elena Mihas
1 Introduction 209
2 Information source extensions of modal operators and of the verb amampy
‘be in the shade’ 211
3 Other means of expressing evidential-like meanings 220
4 Conclusions 225
References 226
11 Nominalization, knowledge, and information source in Aguaruna
(Jivaroan) 227
Simon E. Overall
1 Background information and typological profile 227
2 Cultural practices 228
3 Expressing and talking about knowledge 229
4 Nominalization in Aguaruna 232
5 Why nominalize a verb if you’re going to use it to head a finite clause? 239
6 The areal context 242
References 242
12 The grammar of knowledge in Tima 245
Gerrit J. Dimmendaal
1
Preliminary observations 245
2
The temporal dimension: construction-type effects 247
3
The spatial dimension: the viewer-centred perspective 250
4
The referential dimension: logophoricity marking as an evidential hedging
strategy 253
5 Sharing experiential knowledge: the expressive use of ideophones 256
6 Some preliminary conclusions 257
References 258
13 Saying, seeing, and knowing among the Karawari of Papua New Guinea 260
Borut Telban
1 Speaking 262
2 The creation of knowledge 268
3 Aykapɨkan ‘remember, learn, reflect, know’ 271
4 Hearing and seeing 273
5 Mirativity 274
Contents ix
6 Conclusion 275
References 277
Index of authors 279

Index of languages, language families, and linguistic areas 282
Index of subjects 285
Preface
Every language has a way of talking about knowledge, and expressing information
source. Some languages have a grammatical system of evidentials; others employ addi-
tional means to express information source and the ways in which speakers know what
they are talking about. The marking, and the conceptualization of knowledge, vary
across languages and cultures. This volume aims at investigating the varied facets of
evidentiality, information source, and associated notions.
The volume starts with a typological introduction outlining the marking, and the
meaning, of evidentials and other ways of marking information source, together with
cultural and social aspects of the conceptualization of knowledge in a range of speech
communities. It is followed by revised versions of twelve of the fourteen presentations
from the International Workshop ‘The grammar of knowledge’, held at the Language
and Culture Research Centre, James Cook University, 16–21 July 2012. An earlier ver-
sion of Chapter 1 had been circulated to the contributors, with a list of points to be
addressed, so as to ensure that their detailed studies of individual languages were cast
in terms of a common set of typological parameters. (This is the seventh monograph
in the series Explorations in Linguistic Typology, devoted to volumes from the Inter-
national Workshops organized by the co-editors.)
The week of the workshop was intellectually stimulating and exciting, full of good
discussions and cross-fertilization of ideas. Each author has undertaken intensive
fieldwork, in addition to experience of working on linguistic typology, historical com-
parative issues, and problems of areal diffusion. The analysis is cast in terms of basic
linguistic theory—the cumulative typological functional framework in terms of which
almost all descriptive grammars are cast—and avoids formalisms (which provide rein-
terpretations rather than explanations, and come and go with such frequency that any
statement made in terms of them is likely soon to become inaccessible).
It is our hope that this volume will provide a consolidated conceptual and analytic
framework. We aim at covering the major parameters of variation in the expression of evi-
dentiality, information source, and knowledge in general across languages of the world.
We are grateful to all the participants in the Workshop and colleagues who took part
in the discussion and provided feedback on presentations at various stages, particu-
larly, Michael Wood, Yongxian Luo, Grant Aiton, Hannah Sarvasy, Mikko Salminen,
Dineke Schokkin, Esther Stützle-Csaja, and Lidia Suarez. We owe a special debt of
gratitude to Brigitta Flick and to Elena Rhind, for helping us organize the Workshop in
a most efficient manner. Brigitta Flick’s support and editorial assistance was, as always,
invaluable. A very big ‘thank you’ goes to Amanda Parsonage, for her assistance and
cheerful support during the preparation of the final manuscript.
Preface xi
The Workshop was made possible through the Australian Research Council
iscovery Project ‘The grammar of knowledge: a cross-linguistic view of evidential-
D
ity and epistemological expressions’. We gratefully acknowledge financial assistance
from the Faculty of Arts and Social Sciences and from the Cairns Institute at James
Cook University.
As with all previous volumes emanating from our International Workshop (also
published in the series Explorations in Linguistic Typology), we owe a considerable debt
to John Davey, our editor at Oxford University Press. His support, and encouragement,
make our books feel welcome.
Notes on the contributors
Alexandra Y. Aikhenvald is Distinguished Professor, Australian Laureate Fellow and Direc-
tor of the Language and Culture Research Centre at James Cook University. She is a major
authority on languages of the Arawak family, from northern Amazonia, and has written gram-
mars of Bare (1995) and Warekena (1998), plus A Grammar of Tariana, from northwest Amazonia
(Cambridge University Press, 2003), in addition to essays on various typological and areal fea-
tures of South American languages. Her other major publications, with OUP, include Classifiers:
a typology of noun categorization devices (2000), Language contact in Amazonia (2002), Eviden-
tiality (2004), The Manambu language from East Sepik, Papua New Guinea, (2008), Imperatives
and commands (2010), Languages of the Amazon (2012), and The art of grammar (forthcoming).
Address: The Cairns Institute, James Cook University, PO Box 6811, Cairns, NQld 4780, Aus-
tralia; e-mail: Alexandra.Aikhenvald@jcu.edu.au
Jules Jacques Coly received his PhD from the University of Cologne, where he is currently occupy-
ing a four-year postdoctoral position. His principal research has been on Jóola and Wolof (Senegal)
and Maaka (Nigeria) on which he has published several articles. His book Morphosyntaxe du Jóola
Kuwaataay was published in 2012 (Munich, Lincom Europa). Address: Institut für Afrikanistik, Uni-
versität zu Köln, Meister-Ekkehard Str-7, D-50923 Cologne, Germany; e-mail: colyjules@hotmail.com
Gerrit J. Dimmendaal is Professor of African Studies at the University of Cologne, Germany.

He specializes in Nilo-Saharan and neighbouring language families belonging to Afroasiatic
and Niger-Congo groupings. He is the author of The Turkana language (Dordrecht: Foris, 1983).
His recent publications include an edited volume, Coding participant marking: Construction
types in twelve African languages (2009), a monograph, Historical linguistics and the comparative
study of African languages (2011), and a collection of papers The leopard’s spots (forthcoming,
Leiden: Brill). Address: Institut für Afrikanistik, Universität zu Köln, Meister-Ekkehard Str -7,
D-50923 Cologne, Germany; e-mail: gerrit.dimmendaal@uni-koeln.de
R. M. W. Dixon is Adjunct Professor and Deputy Director of the Language and Culture Research
Centre at James Cook University. He has published grammars of a number of Australian lan-
guages (including Dyirbal and Yidiñ), in addition to A grammar of Boumaa Fijian (Univer-
sity of Chicago Press, 1988), The Jarawara language of Southern Amazonia (Oxford University
Press, 2004, paperback 2011) and A semantic approach to English grammar (Oxford University
Press, 2005). His works on typological theory include Where have all the adjectives gone? and
other essays in semantics and syntax (Mouton, 1982) and Ergativity (Cambridge University Press,
1994). The rise and fall of languages (Cambridge University Press, 1997) expounded a punctuated
equilibrium model for language development; this is the basis for his detailed case study Austral-
ian languages: Their nature and development (Cambridge University Press, 2002). He is also the
author of the three-volume work Basic linguistic theory (Oxford University Press, 2010–12) and
of an academic autobiography I am a linguist (Brill, 2011). Address: The Cairns Institute, James
Cook University, PO Box 6811, Cairns, NQld 4780, Australia; e-mail: robert.dixon@jcu.edu.au
Notes on the contributors xiii
Diana Forker teaches in the Department of General Linguistics at the University of Bamberg.
She completed her PhD at the Max Planck Institute for Evolutionary Anthropology. Her main
interests are languages of the Caucasus, typology, and morphosyntax. She is currently working
on the documentation of the Nakh-Daghestanian language Sanzhi Dargwa and on a typologi-
cal investigation of referential devices in natural discourse. Among her recent and forthcoming
publications are A grammar of Hinuq (Mouton grammar library, 2013) and several articles on
different aspects of Hinuq and other Nakh-Daghestanian languages. She was awarded a pres-
tigious Otto-Hahn Medal for the excellence of her research and a Feodor Lynen Postdoctoral
Fellowship. Address: Universität Bamberg, Lehrstuhl für Allgemeine Sprachwissenschaft, Obere
Karolinenstr. 8 96049 Bamberg, Germany; e-mail: diana.forker@uni-bamberg.de
Teija Greed obtained her first Master’s degree at the University of Tampere, Finland, in Sla-
vonic Philology. She then studied linguistics at SIL International, before moving to Russia to
study the Tatar language at Kazan State University, after which she assisted in the translation of
the New Testament into Tatar (published 2001). She has a particular interest in the interaction
of source and receptor texts in translation. Her Master’s thesis is on Tatar evidentiality. She is
currently a linguistics consultant with SIL International, and is involved in a PhD programme at
the General Linguistics Department at the University of Helsinki, studying evidentiality in the
non-Slavonic languages of Russia. Address: SIL International, Hämeenlinnantie 20 A 3, 15800
Lahti, Finland; e-mail: teija_greed@sil.org
Gwendolyn Hyslop received her PhD in Linguistics from the University of Oregon in 2011.
She is currently a Research Fellow in Linguistics at the College of Asia and the Pacific at the
Australian National University. She has worked on several Tibeto-Burman languages and is a
specialist on the East Bodish languages of Bhutan and Arunachal Pradesh. Publications include
articles on tonogenesis, ergativity, historical linguistics, and a forthcoming grammar of Kurtöp
with Brill. She is also co-director of the Bhutan Oral Literature and Language Documentation
Projects. She has been awarded a prestigious Visiting Fellowship of the Cairns Institute for 2013.
Address: ANU College of Asia and the Pacific, Australian National University, Canberra ACT
2601, Australia; e-mail: Gwendolyn.Hyslop@anu.edu.au
Elena Mihas has been doing research into Ashéninka/Asháninka varieties of Kampan
(Arawak) languages of Peru since 2008. She earned her PhD in Linguistics from of the Uni-
versity of Wisconsin-Milwaukee in 2010. Her doctoral dissertation deals with the essentials of
Ashéninka Perené grammar. Currently, she holds a position of Postdoctoral Research Asso-
ciate at the Language and Culture Research Centre, James Cook University and is a Visit-
ing Scholar at the Center for Latin American and Caribbean Studies at UW-Milwaukee. Her
most recent publications include ‘Ideophones in Alto Perené’ (Studies in Language) and ‘Sub-
ordination strategies in Ashéninka Perené (Arawak) from Central-Eastern Peru’ (Rivista di
Linguistica/Italian Journal of Linguistics). Address: The Cairns Institute, James Cook University,
PO Box 6811, Cairns, NQld 4780, Australia; e-mail: Elena.Mihas@jcu.edu.au
Simon E. Overall received his PhD in 2008 from RCLT, La Trobe University, with a thesis on the
grammar of Aguaruna. He has taught linguistics at La Trobe University and Otago University, and
has published on aspects of the syntax and phonology of Aguaruna. His current research focus is
on the diachrony of nominalizations and their involvement in discourse and switch-reference, as
xiv Notes on the contributors
well as the linguistic situation in the eastern foothills of the Andes. He holds a position of Post-
doctoral Research Fellow at the Language and Culture Research Centre, James Cook University
and is working on Candoshi, an isolate of Peru. Address: The Cairns Institute, James Cook Uni-
versity, PO Box 6811, Cairns, NQld 4780, Australia; e-mail: Simon.Overall@gmail.com; Simon.
Overall@jcu.edu.au
Chia-jung Pan is Assistant Professor at the School of Literature, Nankai University, P. R. China.
His PhD thesis A Grammar of Lha ’ alua, an Austronesian language of Taiwan, was completed
at the Language and Culture Research Centre, Cairns Institute, James Cook University in 2012.
In 2013, he was a Postdoctoral Research Fellow at the Institute of Linguistics, Academia Sinica.
Currently, he is continuing his research into the Saaroa language and investigating a number of
grammatical topics on Saaroa, Tsou, and Kanakanavu languages. Address: School of Literature,
Fansun Building, Nankai University, No. 94 Weijin Road, Tianjin 300071, P. R. China; e-mail:
chiajung.pan216@gmail.com
Olga Seesing studied linguistics at the Kalmyk State University (Russian Federation), and
completed her PhD on temporal clauses in Kalmyk at the Ludwig Maximilian University of
Munich. As a Postdoctoral Research Fellow at the Ludwig Maximilian University of Munich
she worked on evidentiality and mirativity in Kalmyk (supervisor: Elena Skribnik), and is cur-
rently working on temporal clause construction in Kalmyk. Address: Institute of Finno-Ugric
and Uralic Studies, Ludwig Maximilian University of Munich, Ludwigstr. 31, D-80539 Munich,
Germany; e-mail: olga.seesing@web.de
Elena Skribnik is Professor and Director of the Institute of Finno-Ugric and Uralic Studies at
the Ludwig Maximilian University of Munich. Her main areas of research are syntax, especially
grammatical categories and grammaticalization processes, and language contact of the languages
of Siberia. She has carried out fieldwork on a number of Altaic and Uralic languages of Siberia
(1977–2008) and published studies on these languages. She has participated in educational pro-
grammes for representatives of indigenous peoples at the Universities of Novosibirsk and Khanty-
Mansiysk (Russian Federation), and is co-author of the first Mansi teaching manual intended for
students of Mansi national schools and pedagogical institutions with insufficient knowledge of
their heritage language. She led the ESF EuroCORES/EuroBABEL project ‘Ob-Ugric languages:
conceptual structures, lexicon, constructions, categories’, and is currently working on a reference
grammar of Mansi, and on a digital constructicon of temporal clauses in Mongol, Buryat, and
Kalmyk. Address: Institute of Finno-Ugric and Uralic Studies, Ludwig Maximilian University of
Munich, Ludwigstr. 31, D-80539 Munich, Germany; e-mail: skribnik@lmu.de
Anne Storch is Professor of African Languages and Linguistics at the University of Cologne.
Her principal research has been on the various languages of Nigeria (including Jukun and
Maaka), on the Atlantic language region, on Western Nilotic, and on African speech registers.
Her publications include Das Hone und seine Stellung im Zentral-Jukunoid (Cologne 1999), The
noun morphology of Western Nilotic (Cologne 2005), Secret manipulations (New York 2011), and
several edited volumes. Her book Repertoires and choices in African languages, co-authored
by Friederike Lüpke, was published in 2013, by Mouton de Gruyter. She is currently finishing
a grammar of Luwo (Sudan). Address: Institut für Afrikanistik, Universität zu Köln, Meister-
Ekkehard Str-7, D-50923 Cologne, Germany; e-mail: anne.storch@uni-koeln.de
Notes on the contributors xv
Borut Telban is Research Advisor and Associate Professor at the Scientific Research Centre of
the Slovenian Academy of Sciences and Arts. His anthropological research among the Karawari-
speaking people of the East Sepik Province, Papua New Guinea, has spanned more than twenty
years including ten months of ethnographic fieldwork in Ambonwari village between February
and December 2011. In his published articles and chapters he has explored in detail Ambonwari
cosmology, kinship and social organization, ritual, death, poetics, and socio-cultural change
pertaining to the impact of the Catholic charismatic movement. He is the author of Dancing
through time: A Sepik cosmology (Clarendon Press, Oxford, 1998) and is currently working on an
encyclopaedic and ethnographic dictionary of Karawari language. Address: Institute for Anthro-
pological and Spatial Studies, Scientific Research Centre of the Slovenian Academy of Sciences
and Arts, Novi trg 2, 1000 Ljubljana, Slovenia; e-mail: Borut.Telban@zrc-sazu.si
Sihong Zhang is Associate Professor and Vice-dean of the School of International Education
and Exchange, Anhui University of Chinese Medicine. He has just completed his PhD at
the Language and Culture Centre within the Cairns Institute of James Cook University. He
is working on Ersu, an underdocumented language spoken in Sichuan Province, China. His
research area is in anthropological linguistics (Tibeto-Burman languages), applied linguis-
tics (ESL), and corpus linguistics (pedagogy). Address: Anhui University of Chinese Medi-
cine, 103 Mishan Road, Hefei, Anhui 230038, China; e-mail: Sihong.Zhang@my.jcu.edu.au;
zhang-sihong@qq.com
Abbreviations
1 1st person
2 2nd person
3 3rd person
III third nominal group
IV fourth nominal group
V fifth nominal group
VII seventh nominal group
I–V gender markers
A transitive subject
ABL ablative
ABS absolutive
ABST absent at the present moment
ACC accusative
ACHI achievement
ADJ adjective
ADV adverb
AFF affect (Chapter 10)
AFF affirmative (Chapter 7)
AGR agreement
ALL allative
ALOC animate location
ALSO also (no abbreviation)
ANAPH anaphoric pronoun
ANT anterior
AP antipassive
APPLIC applicative
ASP aspect
ASSERT assertive
ASSUM assumed
AUG augmentative
AV actor voice
Abbreviations xvii
AUX auxiliary
BEN benefactive
BR bound root
CATEG categorical future (tense)
CAUS causative
CEXP counter-expectation
CL classifier
CMPLZ complementizer
CNTF counterfactual
COMIT comitative
COMPL completive
COMPL.CL complement clause
CONC concessive
COND conditional
CONJ conjunction
CONT continuous
CONTA location with contact
CONVB converb
COP copula
CORE core case
COS change of state
CTM co-temporal
DAT dative
DBT dubiative
DEC declarative
DEF definite
DEM demonstrative
DER derivational affix
DESID desiderative
DET determiner
DIM diminutive
DIR directional
DM discourse marker
DS different subject
du, DU dual
DUB dubitative
xviii Abbreviations
-DY past tense suffix in Tatar

e eyewitness
E extension to core (Chapter 11)
E edible (Chapter 8)
EGO position of the speaker (Chapter 12)
EGO Egophoric (Chapter 5)
EP epenthetic
EMPH emphasis
EQ equative
EQUAT equative
ERG ergative
EVID evidential
EVID.MOD modified situation
EVID.NPOT anti-potential
exc exclusive
EXCLAM exclamative
EXIS existential
EXIST existential
EXP experiential
EXPE experiential
FEM, F, f feminine
FOC focus
FP far past
FUT future
-GAn resultative past tense suffix in Tatar
GEN genitive
HAB habitual
HESIT hesitation pro-form
HON honorific
HPL human plural
HT high transitivity
ICP intransitive copy pronoun
IDEO ideophone
IMMED immediate
IMP imperative
Abbreviations xix
IMPER imperfect
IMPERS impersonal
IMPERV imperfective
IN location ‘in’
inc inclusive
INCH inchoative
incl inclusive
INCOM incompletive
INDEF indefinite
INDEP independent
INDIC indicative
INDIR indirectivity
INESS inessive
INFER inferred
INFIN infinitive
INST instrumental
INT intentional
INTER interrogative
INTERJ interjection
INTR intransitive
IRR irrealis
JOINT joint perception
JUS jussive
KIN kinship
LAT lative
LINK linker
LOC locative
LOG logophoric
LV locative voice
MASC, M, m masculine
MC Mandarin Chinese
min, MIN minimal
MIR mirative
MIRAT mirative
MOD modal
xx Abbreviations
NARR narrative
NCL noun class
NEG negation
NEUT, N neuter
NF non-final
nf non-feminine
NFIRSTH non-firsthand evidential
NHPL non-human plural
NIGHT nightime
NMASC non-masculine
NOM nominative
NOMZ nominalizer
NONVIS non-visual
NP noun phrase
NPAST non-past
NPOSSD non-possessed
nsg non-singular
NWIT non-witnessed evidential
O transitive object
OBJ object
OBL oblique
ONOM onomatopoeia
OTR first person ‘other’
OPT optative
ORD ordinal number
p person
PART particle
PARTIC participle
PASS passive
PAST past
PERV perfective
PF pause filler
pl, PL plural
PN pronoun
POSS possessive
Abbreviations xxi
POT, pot potential

PR polite request
PRED predicative (word)
PREF prefix
PREP preposition
PRES present
PRES.NONVIS present non-visual
PREST present speaker
PRN proper noun
PROG progressive
PROHIB prohibitive
PROS prospective
PURP purposive
PV patient voice
Q question particle
QP question particle
QUOT quotative
RC relative clause
REAL realis
REAS reason
REC recent
REC.P.NONVIS recent past non-visual
RECIP reciprocal
REDUP reduplicated
REF referential
REFL reflexive
REGR regressive
REL relative
REM remote
REP reported
REPET repetitive
RES resultative
REV reversive
S intransitive subject
s.th. something
xxii Abbreviations
Sa ‘active’ S, marked like A

SEQ sequential
sg, SG singular
SIM simultaneous
So ‘stative’ S, marked like O
SLF first person ‘self ’ (narrator)
SPEC speculative
SPECUL speculative
SPR location ‘on’
STAT stative
SU subject
SU>OBJ subject of marked clause is object of controlling clause
SUB subordinator
SUBORD subordinate
SUPRESS superessive
SVC serial verb construction
TAG tag particle
TAM tense-aspect-mood
TEL telic
TEMP temporal
TERM terminative
TOP topic
TOP.NON.A/S topical non-subject
TR transitive
UNCERT uncertain(ty)
UNWIT unwitnessed
VBLZ verbalizer
VEN ventive
VIS visible (Chapter 8)
VIS visual
VN verbal noun
VOC vocative
WIT witnessed
1
The grammar of knowledge:

a cross-linguistic view of evidentials
and the expression of information
source
A L E X A N D R A Y. A I K H E N VA L D
Every language has a way of speaking about how one knows what one says, and what
one thinks about what one knows. In any language, there are ways of phrasing infer-
ences, assumptions, probabilities, and possibilities, and expressing disbelief. These
epistemological meanings and their cultural correlates are the subject matter of the
present volume.
In a number of the world’s languages, every sentence must specify the information
source on which it is based—whether the speaker saw the event, or heard it, or inferred
it based on visual evidence or on common sense, or learnt it from another person. As
Frans Boas (1938: 133) put it, ‘while for us definiteness, number, and time are obliga-
tory aspects, we find in another language location near the speaker or somewhere
else, source of information—whether seen, heard, or inferred — as obligatory aspects’.
‘Evidentiality’ is grammaticalized marking of information source. This is a bona
fide grammatical category, on a par with tense, aspect, mood, modality, directional-
ity, obviation, negation, and person. Just as ‘person’ can be fused with ‘gender’ and
‘number’, evidentiality may be fused with tense or aspect or mood. Its expression,
and meanings, may correlate with sentence types: evidentials in questions may have
overtones different from evidentials in statements. Exclamatory sentences may have
no evidentials at all. Evidentials in commands are very limited in their meanings.
In §1, we briefly revisit the relationship between evidentiality and information source.
§2 presents a potted summary of evidentials and their meanings across the world. In §3,
we turn to the means other than grammatical evidentials which can cover information
source, and attitude to information. Evidentials may have non-evidential extensions.
2 Alexandra Y. Aikhenvald
Their use tends to reflect cultural norms and conventions. And their meanings change
as new techniques of acquiring information become available. These are the topics of
§4. Ways of talking about knowledge in languages with, and without, evidentials are
addressed in §5. In the last section (§6) we offer a brief outline of this volume.
1 Evidentiality and information source

It is not uncommon for a linguistic term to have a counterpart in the real world. The
idea of ‘time’ in the real world translates into ‘tense’ when expressed in a language.
‘Time’ is what our watch shows and what often passes too quickly; ‘tense’ is a grammat-
icalized set of forms we have to use in a particular language. Not every time distinction
acquires grammatical expression in the language: the possibilities for time are unlim-
ited, and for tense they are rather limited. Similarly, an ‘imperative’ is a category in the
language, while a command is a parameter in the real world. Sex distinctions between
males and females is often translated into the category of ‘gender’. Along similar lines,
‘evidentiality’ is a linguistic category whose real-life counterpart is information source.
Information source can be expressed in a variety of ways. These may include lexi-
cal means, including verbs of perception (‘see’, ‘hear’, ‘smell’) and cognition (‘know’,
‘understand’, and so on). Modal verbs, particles, parentheticals of various sorts, and
even facial expressions, can be used to express inference, assumption, and attitude
to information—whether the event is considered probable, possible, or downright
unlikely. In any language, there is a way of reporting what someone has said. All lan-
guages use quotations, and many have direct and indirect speech reports. Their use
may interrelate with attitude to the information quoted or cited. For example, a ver-
batim quote in Arizona Tewa implies that the speaker does not vouch for the informa-
tion quoted (we return to this in §3.2.4). To sound neutral a speaker would prefer an
indirect speech report.1
Any means of expressing information source may correlate with attitudes to infor-
mation, and communicative strategies—which information is considered more valu-
able, and how it is expressed in culturally appropriate ways in each language. There
are, however, significant differences between evidentials and non-primarily evidential
means ‘co-opted’ to cover some information sources.
In a nutshell: grammatical evidential systems are closed and restricted, with lim-
ited choices available. The scope of grammatical evidentials is usually the clause, or
the sentence. Only very occasionally can a noun phrase have its own evidentiality
specification, different from that of a verb; we return to this in §2.2. In contrast, other
means of expressing information source offer open-ended options in terms of their
semantics, and can be more flexible in their scope. Expressions related to information
1
Kroskrity (1993: 146); Aikhenvald (2004: 139); Aikhenvald (2011c: 322) for typological features of speech
reports.
1 Cross-linguistic view 3
source are heterogeneous and versatile. They include closed classes of particles and
modal verbs, and a potentially open-ended array of verbs of opinion and belief. The
term ‘lexical evidentiality’ is misleading in that it obscures these differences (we briefly
turn to this in the Appendix).2
In languages with evidentials, these are never the only means of expressing informa-
tion source. Verbs, adjectives, adverbials, and speech reports may provide additional
detail, to do with attitude to knowledge—the sum of what is known and the informa-
tion this is based on.
Our main concern within this volume is a cross-linguistic investigation of expres-
sion of knowledge through evidentials as a major grammatical means to express the
information source, and through other means. We also focus on their correlations with
types of knowledge including traditional knowledge, information acquired through
more modern means, cultural conventions, and speech practices. The ‘grammar’ of
knowledge subsumes the principles of expressing how one knows things, and what this
knowledge is based on. Every language will have ways of talking about these issues,
but these will vary. There may be constraints on how specific, or how vague, one is
expected to be. The practices of talking about what one knows, and how one knows it,
may turn out to be shared in languages with and without evidentials.
Evidentials often come from grammaticalized verbs of perception, from modal
markers, and from verbs of speech. A typology of lexical and other expressions of
information source will help us trace the origins of evidential systems.
Terminological clarity is essential in any branch of science, and linguistics is no
exception. The Appendix lists a few common misconceptions about evidentiality and
‘evidential’ meanings.
2 Evidentiality: a bird’s-eye view3

Evidentiality is a grammatical marking of how we know something—whether we saw
it happen, or heard it, or smelt it, or inferred what was happening based on logical
assumption, or on a result we can see, or whether we were just told about it. In perhaps
a quarter of the world’s languages, marking a selection of information sources is a
2
Further details on evidentiality as a grammatical category, its meanings and developments are sum-
marized in Aikhenvald (2004, 2006); grammaticalization of evidentials is discussed in Aikhenvald (2011b);
evidentials and other means of expression of information source are contrasted in Aikhenvald (2008). An
up-to-date bibliography on evidentials is in Aikhenvald (2011a). There are useful papers in Aikhenvald and
Dixon (2003), Johanson and Utas (2000) and some in Chafe and Nichols (1986). Earlier approaches to evi-
dentiality which are strongly recommended, include Boas (1938), Jakobson (1957); and especially Jacobsen
(1986). On the opposite side of the coin, a warning should be noted that Willett (1988), de Haan (2005), and
van der Auwera and Plungian (1998) and many of the papers in Diewald and Smirnova (2010) are of decided-
ly mixed quality (see the review in Aikhenvald 2012b). The Appendix addresses some issues of terminology.
3
The generalizations here and in all my work on evidentials are based on the analysis of grammars of c.
600 languages (since the publication of Aikhenvald 2004, I have had access to further grammars). I avoid
limiting myself to any artificially constructed ‘samples’ of languages, since these are likely to engender
skewed results.
obligatory.4 We now summarize a few points relevant for an understanding of eviden-

tial systems (§2.1). A brief overview of geographical distribution of evidentials is in
§2.2. Information source expressed within noun phrases is the topic of §2.3.
2.1 Evidentiality systems: a snapshot

As a category in its own right, evidentiality is a relatively recent ‘arrival’ on the lin-
guistic scene—in contrast to other categories such as person, gender, number, and
tense which have been household concepts in linguistics for many centuries (see Rob-
ins 1967). This may well be the reason why the proper limits of evidentiality are still
debated by some.
The idea of obligatory marking of information source encoded in grammar goes
back to Boas, and his sketch of Kwakiutl (Boas 1911: 443, 496). ‘The source, or nature,
of human knowledge (known by actual experience, by hearsay, by inference)’ is listed
by Sapir (1921: 108–9) alongside other grammatical concepts, such as person, modal-
ity, number, and tense. Since Boas’ work, the notion of grammatical evidentiality has
made its way into many grammars of North American languages.
In languages with obligatory evidentiality, a closed set of information sources has
to be marked in every clause—otherwise the clause is ungrammatical, or the speaker
incompetent, or even not quite right in his mind (Weber 1986: 142). Evidentiality is a
category in its own right. Evidentials may occur together with exponents of modality,
tense, or mood.5 Evidentiality can be intertwined with other categories—including
tense, aspect, epistemic modalities, expectation, and ‘sharing’ of knowledge.
2.1.1 What meanings are expressed in evidentiality systems Languages with grammatical
evidentials divide into a number of types depending on how many information sources
are assigned a distinct grammatical marking. Some languages mark just information
reported by someone else.
Nhêengatú, a Tupí-Guaraní lingua franca of north-west Amazonia, has a reported
evidential marker paá (Floyd 2005). Suppose you saw Aldevan go fishing. After that,
Aldevan’s aunt Marcilha arrives at the house and asks where he has gone. You then
reply, u-sú u-piniatika (3sg-go 3sg-fish) ‘He went fishing’. Then a friend comes to visit
4
The term ‘evidential’ as a label for the grammatical category of information source was first introduced
by Jakobson in 1957; and became established by the mid-1960s (see Jacobsen 1986: 4–7; Aikhenvald 2004:
10–17). Lazard (1957) was among the first French linguists to have discussed evidential meanings (‘inféren-
ciel’), based on the material from Tajik, an Iranian language.
5
Statements that evidentiality is a type of verbal modality can be found in Palmer (1986), van der Auwera
and Plungian (1998), and Willett (1988) are not borne out by the facts of languages, and are mistaken. See
the arguments in de Haan (1999), Lazard (1999, 2001), and DeLancey (2001), and the general summary in
Aikhenvald (2004: 3–10). Some scholars whose experience is limited to a handful of familiar European
languages tend to assume that evidentials are a kind of modal largely because of their absence in most major
European languages, thus trying to explain an unusual category in terms of some other, more conventional,
notion.
and asks Marcilha where Aldevan has gone. She replies, using a reported evidential—
she did not see the man go:
(1) u-sú u-piniatika paá

3sg-go 3sg-fish rep
He went fishing (they say/I was told)
Systems with just one, reported, evidential are widespread across the world. Saaroa, an
Austronesian language from Taiwan, also has just one reported evidential (Chapter 4).
Other languages distinguish firsthand and non-firsthand information sources. A
typical conversation in Jarawara, an Arawá language from Brazil, is as follows. One
speaker asks the other:
(2) jomee tiwa na-tafi-no awa?

dog(masc) 2sgO caus-wake-immed.past.noneyewitness.masc seem.masc
Did the dog wake you up?
He uses a non-firsthand evidential in his question: he didn’t himself see or hear the
dog; perhaps he was just told about it. The other speaker—who had indeed been
woken by the dog and thus saw it or heard it or both—answers using the firsthand
evidential:
(3) owa na-tafi-are-ka

1sgO caus-wake-immed.past.eyewitness.masc-dec.masc
It did wake me up (I saw it or heard it)
Evidentials in Jarawara are distinguished in past tense only. This is the case in many
languages with evidentials (including Hinuq: Chapter 2 and Tatar: Chapter 3): the
source of information is easier to gather for what has already occurred. In Tatar, the
marking of non-firsthand information source is also associated with the resultative.
A further type of small evidential system involves having a marker for informa-
tion acquired through a non-firsthand source, and leaving any other information
unmarked, or ‘source-neutral’. This is frequent in Caucasian, Turkic, some Finno-
Ugric languages, and some languages of the Andes (Johanson and Utas 2000; Aikhen-
vald 2012a: ch. 9). Within the present volume, this kind of system is described for
Hinuq, a Nakh-Daghestanian language (Chapter 2) and Tatar, a Turkic language
(Chapter 3).
The ‘non-firsthand’ term has an array of meanings covering reported or hearsay, and
logical deduction or inference. The evidential marker -rke- in Mapudungun, an isolate
spoken in the Andean areas of Chile and west central Argentina, is a prime example
(Smeets 2007: 246–7, 110). In (4), it refers to reported information:
(4) Kuyfi miyaw-ürke-y mawida mew

long.ago walk-nonfirsthand-indicative forest through
Long ago s/he wandered through the forest (it is said)
In (5), the same -rke- describes what one has inferred:
(5) weðweð-pe-rke-la-y
crazy-proximity-nonfirsthand-indic-3person
He must be crazy (that one, he travelled through all that rain)
The ‘information-source-neutral’ form may refer to a variety of information sources.

It is the most likely one to be used in translations and elicitation, as a ‘default’ choice
(see §2.2.2 of Chapter 2, on Hinuq). The information-source neutral form in Tatar
often acquires a firsthand reading (see Figure 4 in §6.3 of Chapter 3, and an alternative
analysis of Hinuq mentioned in Chapter 2).
What could be the reason for this? In many languages, the least formally marked
verb in a language with evidentiality would acquire a visual, or a firsthand reading. In
Ersu, if a clause contains a verb without an evidential, this is interpreted as based on
‘direct’ information source (§4.1 of Chapter 6, and see Aikhenvald 2004: 72–3). Thus,
no overt marking for information source may imply that the source will be understood
as visual or firsthand by ‘default’. Having firsthand information formally unmarked
is a strong tendency, but by no means a universal rule. A formally unmarked verb in
Kurtöp encodes uncertain future (§5.2 of Chapter 5); the expression of information
source in this language depends on the choices made in the aspect system, and is
intertwined with a number of further verbal categories.
Firsthand experience can be contrasted with what one has inferred, and with what
one assumes. If a speaker of Matses, a Panoan language, has experienced something
directly—that is, seen it, heard it, or smelt it—they would use an ‘experiential’ eviden-
tial. To answer a question ‘How many wives do you have?’, a Matses would say:
(6) daëd ik-o-�h

two be-recent.past.experiential-3p
There are (lit. were) two
According to Fleck (2007), this can be understood as something like ‘last time I
checked, they were two’. Evidentials are there only in the past: again, this makes sense
because information source is clearer for events which have happened.
If a speaker sees a dead man, and there is no natural cause for death in sight, they
would use an inferential evidential:
(7) nënëchokid-n ak-ak

shaman-erg kill-rec.past.inferred
(I infer that) a shaman (must have) killed him
And if the speaker has not seen the corpse yet, and assumes that the shaman may have
killed the man, the ‘conjecture’ evidential is the right choice:
(8) nënëchokid-n ak-a�h

shaman-erg kill-rec.past.conjecture
(I guess that) a shaman (may have) killed him
Neither the inferential nor the conjecture evidentials imply any uncertainty. If the
speaker thinks that the shaman might have killed the man, but they are not sure, they
will use a counterfactual suffix -en on the subject:
(9) nënëchokid-n-bi-en ak-chito-ak

shaman-erg-emph-counterfactual kill-uncert-rec.past.inferred
A shaman (might have) killed him
Ersu, a Tibeto-Burman language, marks three information sources (§4 of Chapter 6).
If information is acquired directly, that is, through seeing, hearing, feeling, or smelling,
the verb is formally unmarked. There is a special marker pà for inferred and assumed
information, and a reported evidential used if the speaker’s statement is based on
something someone else had told them. A quotative marker is in the process of being
grammaticalized.
What one saw can be contrasted with what one learnt through hearing and smelling,
and through various kinds of inference. Tariana, an Arawak language from north-west
Amazonia, and its many East Tucanoan neighbours, have five options. If I see José play
football, I will say ‘José is playing-naka’, using the visual evidential. If I heard the noise
of the play (but didn’t see it), I will say ‘José is playing-mahka’, using the non-visual. If
all I see is that José’s football boots are gone and so is the ball, I will say ‘José is playing-
nihka’, using the inferential. If it is Sunday and José is not home, the thing to say is ‘José
is playing-sika’ since my statement is based on the assumption and general knowledge
that José usually plays football on Sundays. And if the information was reported to me
by someone else, I will say ‘José is playing-pidaka’, using the reported marker. Omitting
an evidential will produce ungrammatical and unnatural sentences.
Recent studies in grammatical evidential systems have revealed the existence of fur-
ther terms. Yongning Na (Mosuo), a Tibeto-Burman language (Lidz 2007), has a
direct or visual, an inferential, a reported, and a quotative evidential, and a further
term which covers general knowledge. This marker is illustrated in (10) (numbers
represent tones):
(10) nɑ13 bu33 nɑ13 bu33 ʐi33 tu33-kuɔ 33 di31 tɑ13

Na poss Na poss family-loc all comparative
tɔ31 tsɑ33=dʐɔ33 ʐi33 mi33 thi33 li33 ni31
important/busy=common.knowledge.evid hearth room this cl cop
In Na families, more important than anything, as everyone knows, is the hearth room
Kalmyk (Chapter 7) has a special evidential for ‘common knowledge’, and so does
Mamaindê, a Nambiquara language. Mamaindê has two further evidentials, for sec-
ondhand and for thirdhand information.
No spoken language has a special evidential to cover smell, taste, or feeling: this
complex of meanings is typically covered by a non-visual, a non-firsthand, or experi-
ential evidential. However, Catalan sign language is reported to have a special eviden-
tial marking smell (Sherman Wilcox, p.c.).
Amazonian languages may have further terms. In the Southern Nambiquara dialect
complex, there is an obligatory marking on the verb for, among others (Lowe 1999):
• whether a statement is eyewitness—that is, implying that the speaker had seen
the action they are reporting;
• whether a statement is inferred or assumed, whereby ‘the speaker’s claim is based
either on seeing an associated simultaneous action and making an interpreta-
tion therefrom, or on seeing a set of circumstances which must have resulted
from a previous action and making an inference; different suffixes mark these
two options’;
• whether it is reported, that is if ‘the speaker is simply passing on information
they have heard from another speaker’; or
• whether there is ‘internal support’—if ‘the speaker reports their “gut feeling”
that which they assert must be so’.6
The meaning of ‘gut feeling’ or ‘internal support’ can be expressed through means
other than an evidential. Tariana has a lexical verb with a similar meaning (see exam-
ple (23)), and Ashéninka Perené has a bound marker -amampy ‘have suspicions, mis-
givings’ which may have developed out of a verb (see §2.4 of Chapter 10).
6
See Eberhard (2009) on Mamaindê, and Lowe (1999: 275–6) on Southern Nambiquara.
2.1.2 Summing up: semantic parameters in grammatical evidentiality The semantic

parameters employed in languages with grammatical evidentiality cover physical
senses, several types of inference and of report. The recurrent terms are:
I. Visual covers evidence acquired through seeing.

II. Sensory covers evidence acquired through hearing, and is typically extended
to smell and taste, and sometimes also touch.
III. Inference is based on visible or tangible evidence or result.
IV. Assumption is based on evidence other than visible results: this may include
logical reasoning, assumption, or simply general knowledge.
V. Reported, for reported information with no reference to who it was report-
ed by.
VI. Quotative, for reported information with an overt reference to the quoted
source.
These semantic parameters group together in various ways, depending on the sys-
tem’s internal organization. The most straightforward grouping is found in three-
term systems—where sensory parameters (I and II), inference (III and IV), and
reported (V and VI) are grouped together, as in Quechua, Shilluk, and Bora (Aikhen-
vald 2004: 145–6; 159–66). Numerous languages of Eurasia group parameters (II–VI)
under a catch-all non-firsthand evidential, for example Hinuq and Tatar (Chapters 2
and 3 of this volume), and also Abkhaz and Yukaghir. This kind of system is uncom-
mon in Amazonia (although it has been described for Mapudungun, in the Andean
region).
Alternatively, an evidentiality system may allow one to specify—or not—the exact
information source (in line with Aikhenvald 2003a: 3; Johanson 2003). Kalmyk, a Mon-
golic language (Chapter 7), distinguishes direct and indirect evidentials. The ‘direct’
term combines reference to sensory parameters (I and II). The indirect term covers
the rest. The speaker may choose to be more specific as to ‘indirect’ evidentiality—
there is then the choice of inferred, assumed, prospective, reported, and common
knowledge.
We now turn to further features of evidentiality systems, highlighting those
described within this volume.
2.1.3 Evidentials, and other categories Just like most other grammatical catego-
ries, evidentials interrelate with mood. The maximum number of evidential speci-
fications tends to be distinguished in declarative main clauses. The most frequent
evidential in commands is reported (meaning ‘do what someone else told you to!’).
Evidentials in questions may reflect the information source of the answerer (as in
Tsafiki, Quechua, Tariana, and Tucano: Aikhenvald 2004: 245–6) or the questioner’s
assumptions concerning the information source of the addressee; this has different
consequences for their use. Typically, only reported evidentiality can be expressed in
commands (see Aikhenvald 2010: 138–40, for details). An evidential can be

questioned, as in Wanka Quechua.7
Evidentiality can interact with negation. In Mỹky, an isolate from Brazilian Amazo-
nia (Monserrat and Dixon 2003), no evidentials at all can be distinguished if the clause
is negative. In contrast, negative existential copulas in Kurtöp (§6.1 of Chapter 5) have
an evidential-cum-epistemic distinction not found in their affirmative counterparts:
indirect evidence or inference with or without ‘doubt’. An evidential may be within the
scope of negation, as in Akha, a Tibeto-Burman language.
The maximum number of evidential specifications is found in past tenses. In Hinuq,
Tatar, Jarawara, and Matses evidentiality is only distinguished in the past tense. The
choices made in the tense system thus determine the choices made in the system of
evidentials (in line with Aikhenvald and Dixon 2011: 190).
The choice of evidentiality in Kurtöp depends on the choice made in the aspect
system. This adds a further dependency to the list in Aikhenvald and Dixon (2011: 91)
(there, we predicted the existence of such a dependency, but did not have any examples
to illustrate it). Evidential distinctions made in perfective aspect in Kurtöp cover per-
sonal knowledge versus lack thereof, and shared versus non-shared knowledge. These
evidential meanings form one system with expectation of knowledge, and overlap
with the epistemic notion of ‘certainty’ (see Figure 2, Chapter 5). Only ‘expectation of
knowledge’ is distinguished within the imperfective aspect.
Future and various modalities—conditional, dubitative and so on—may allow fewer
evidential specifications than the indicative. In many languages—including Matses
(Fleck 2007)—information source is ‘irrelevant’ for statements about the future. Pro-
jection of information source and thus marking of evidentiality in future may have
special epistemic overtones, as in Wanka Quechua (Floyd 1999: 75, see Aikhenvald
2004: 261–2), and Ersu (Sihong Zhang, p.c.). In Kurtöp, only epistemic meanings of
‘certainty’ versus ‘uncertainty’ are relevant for the future (Figure 4, Chapter 5).
Prospective evidential in Kalmyk (Table 2 and §7 of Chapter 7) expresses predic-
tions based on information inferred by the speaker at different times. This takes us to
the next section.
2.1.4 Time reference of an event, and time reference of an evidential The time of
verbal report about something happening may be the same as that of the happening
itself. Or it may be made post factum. In Saaroa (§3.5 of Chapter 4), Tariana, and
a number of other languages there can be a time gap between the event and the
speaker’s information on it (see Aikhenvald 2004: 99–102; 2012a: 259–61).
The time of a report, and the time of the event can be marked separately, with-
in one word. In Tariana, if the speaker had just been told that information would
happen in the future, a future marker will be used with the recent past reported
evidential:
7
Contrary to flawed assertions by Willett (1988), a source to be avoided.
(11) du-dia-karu-pida-ka
3sgnf-return-purposive-reported-rec.p
She will return reportedly (the speaker has been told recently)
If the speaker has been told a long time ago, they will use the remote past reported
evidential:
(12) du-dia-karu-pida-na
3sgnf-return-purposive-reported-rem.p
She will return reportedly (the speaker was told a long time ago)
Matses (Fleck 2007) allows combination of different evidentials, each with its own
time reference. If the hunter saw tracks of a white-lipped peccary a long time ago, and
the tracks were fresh, he will use distant past experiential. This can be accompanied by
the recent past inferred evidential, since the inference relates to the fact that the pec-
caries had been here recently with respect to the time when the hunter had seen them:
(13) �hëktenamë kuen-ak-onda-�h

white.lipped.peccary pass.by-rec.past.inferred-distant.past.experiential-3
White-lipped peccaries evidently passed by (here) (fresh tracks were discovered a
long time ago)
The source of inference may be specified by two evidentials in one clause. In Yongn-
ing Na (Mosuo), a Tibeto-Burman language, the quotative evidential may co-occur
with inferred evidential (Lidz 2007: 67), meaning that the act of speech (and thus the
quotation) was inferred:
(14) “ɕi33 gi13 ze33” pi33 phæ33 di33

rain change.of.state.marker quotative inferred
It is inferred (that) s/he says, “It’s raining”
In Ersu (§6 of Chapter 6) the inferred evidential can occur together with either report-
ed or quotative evidential. Similarly to Yongning Na, one information source specifies
the other.
Making an inference or an assumption implies that information was first obtained
and then interpreted. It may have been obtained before the speech act, or simultane-
ously with it. This creates an option for a language to make additional distinctions
within inferred evidentiality (see Aikhenvald 2004: 95–6, on how this can be reflected
in co-occurrence of two evidentials in one clause; similar instances are mentioned
in San Roque and Loughnane 2012: 118). Kalmyk distinguishes two inferred eviden-
tials, depending on whether the information was obtained prior to the inference or
simultaneously with it. The assumed evidential does not warrant such distinctions. A
prediction can be made based on an inference (this is called ‘prospective’ evidential by
Skribnik and Seesing, Chapter 7). Kalmyk distinguishes three forms: one for a predic-
tion based on information obtained in the past, one based on prediction on the basis
of information obtained simultaneously with it, and one based on the information one
expects to obtain (Table 2, Chapter 7).
The possibility of ‘double tense’ with evidentials, and the co-occurrence of eviden-
tials in one clause is one of the many features that make evidentials special, compared
to any other verbal category.
Inference and assumption may not be as reliable as what one sees with one’s own
eyes. We now turn to epistemic meanings of some evidentials, and further categories
involved.
2.1.5 Evidentials, epistemic meanings, shared experience, and expectation of
knowledge Epistemic modality and evidentiality are different categories. A modal
and an evidential marker can occur in one verb (see, for example, Chirikba 2003,
on Abkhaz). Some evidentials may have epistemic extensions, to do with probability
and the speaker’s evaluation of the trustworthiness of information.8 Visual evidential
in Quechua can refer to information the speaker vouches for. Not so in Tariana
or Tucano. The direct and the indirect evidentials in Kalmyk have no epistemic
overtones. Meanings of ‘uncertainty’ and ‘certainty’ are a feature of just some
inferred and assumed evidentials (Table 2, Chapter 7).
Reported evidential in Estonian has an overtone of doubt: saying ‘he is-
reported.evidential a doctor’ would mean that I doubt his qualifications or abilities.
In English, ‘they say’ may imply that the speaker does not really believe what is being
reported. Similarly, the reported evidential in Saaroa (§4.1 of Chapter 4) may be used if
the information is not reliable. This is akin to how the ubiquitous diz que has overtones of
doubt in many varieties of South American Spanish.9 In contrast, in Quechua, Shipibo-
Konibo, and Tariana, the reported evidential does not imply any of that. These languages
have a plethora of other categories, which express doubt, belief, disbelief, and so on.
As Valenzuela (2003: 57) remarks for Shipibo-Konibo, the selection of reported evi-
dential over the direct evidential ‘does not indicate uncertainty or a lesser degree of
8
Readers should be warned against gratuitously dividing languages into those where evidentials have
epistemic extensions, and those where they do not (as did Plungian 2001). As shown in Chapter 5 of Aikhen-
vald (2004), in the same language one evidential may have an epistemic extension, and another one may not.
9
Summary in Aikhenvald (2004), also Kany (1944: 171); Travis (2006); Olbertz (2005, 2008), Babel
(2009).
reliability but simply reported information’. The reported markers in Hinuq (Chapter 2),
Tatar (Chapter 3), and Maaka (§2.2.1 of Chapter 9) have no overtones of disbelief or doubt.
Neither does the reported clitic =ri in Kurtöp (§7.2 of Chapter 5). Note that this clitic is
widely used in story-telling. I suspect that if a reported evidential is used as the mark of a
narrative genre, hardly any epistemic meanings would be attached to it.
The complementizer kònò in Maaka introduces a speech report: it is an evidential-
ity strategy rather than an evidential. Unlike the dedicated reported evidential nà, the
complementizer always has epistemic overtones of doubt. And, unlike the reported
evidential whose scope is the clause, the scope of the complementizer may vary, from
a clause to an NP (§2.2.1 of Chapter 9).
Whether or not a reported evidential implies doubt can depend on the position
of the evidential within a clause. The reported evidential nana in Tsou (Tsouic, For-
mosan, Austronesian) indicates that information was acquired through hearsay or a
speech report if the marker appears before the verb of speech (Yang 2000: 72–3), as in
(15). The speech report is in square brackets.
(15) nehucma o-si nana [eainca to amo-su

yesterday aux-3sg reported say oblique father-2sg
maine’e hohucma]
go.home tomorrow
I heard from other people yesterday that your father said ‘(you) go home
tomorrow’
If the marker nana occurs within the reported clause, the implication is that the speak-
er is not certain of the information in the speech report:
(16) o-si eUsvUta a'o [nehucma tena cu la

aux-3sg tell aux-3sg? yesterday fut perf hab
nana bumemealU]
reported work.hard
Yesterday she told me that she would work hard from then on (but I am not sure
about ‘work hard’)
The grammatical marking of information source in the Kurtöp verb is intertwined

with epistemic meanings (certainty of knowledge), the expectation of knowledge and
‘knowledge sharing’—whether only the speaker is privy to the information, or it is also
shared by speech-act participants (§3 and Figure 2 of Chapter 5). The information can
be expected or unexpected for the speaker and the addressee: this is the core of the
category of expectation of knowledge, or ‘mirativity’ (also see Aikhenvald 2012c, for
an overview of mirativity and the meanings subsumed under this term). The meanings
of expectation of knowledge are linked with the meanings of ‘certainty’ in existential
and equational copulas—another area of Kurtöp grammar where direct and indirect
information source are also relevant.
The conventionalized attitude to hearsay as a source of information determines
whether or not a reported evidential, or a speech report in general, has epistemic
extensions (see also §5.2 of Chapter 11). We return to this in §5.
2.1.6 Scattered coding of information source Evidential meanings can be expressed

within the same language without relating to just one category or type of expression
(see Aikhenvald 2004: 80–2). Different evidential meanings in Jarawara can be
realized in three different slots within the predicate (Dixon 2003: 185). In Hinuq
(Chapter 2), neutral versus non-witnessed evidentiality is expressed in synthetic past
tense. Quotative and reported particles each form a separate system. This ‘scattered’
expression of evidentiality is also found in Tatar. In this language, reported,
quotative, and assumed particles constitute a ‘layer’ of evidentiality marking which is
separate from—and independent of—the neutral/non-firsthand distinction marked
within the verbal past tense (Chapter 3). Along similar lines, in Maaka (Chapter 9)
some meanings to do with information source are expressed on the verb, and others
within a noun phrase.
Meanings to do with information source in Kurtöp are expressed through verbal
suffixes, and equational and existential copulas; there are also two additional forms,
one marking reported information and the other one quotes.
In many instances, the reported evidential stands apart from the other evidential
meanings, in terms of its grammatical status and properties (also see Aikhenvald 2004:
82–7). Different ways of expressing information source, and different information sourc-
es in different parts of the grammar can be analysed as different evidentiality subsystems.
Other categories may also behave in a similar manner. Number, and gender, are
often expressed differently on verbs, nouns, and pronouns. They may also have dif-
ferent semantic distinctions. In some languages of northern Amazonia the choice of
a classifier depends on the modifier type, or the type of construction (e.g. Aikhenvald
2000: 68 for a discussion of nominal and pronominal genders; Aikhenvald 2007 on
different subsystems of classifiers in different contexts).
Information source can be expressed through other, essentially non-evidential cat-
egories. This creates an opening for further ‘scattering’ of the way one can talk about
the source of one knowledge in a given language (also see §4 of Chapter 10).
2.2 The geography of evidentials

Not every linguistic area or language family is of equal relevance for our study of
evidentiality. An overwhelming number of languages with evidentials are spoken in
Amazonia and the adjacent areas of the Andes. They boast the richest array of evi-
dentials in the world, comparable only to North American Indian languages and lan-
guages of the Tibeto-Burman domain.
Evidentials are relatively poorly represented in familiar European languages (see

Squartini 2007; Pusch 2008), in the Australian area, and in many languages from
the large Austronesian family (other than Formosan languages which tend to have a
reported evidential). Numerous Papuan languages spoken in the Highlands of New
Guinea appear to have evidential systems. Some genetic groups, such as Semitic, do
not have them at all. Hardly any evidential systems have been described for the lan-
guages of Meso-America (except for Uto-Aztecan languages which tend to have a
reported evidential).
Evidentials are a prominent feature in many Turkic, Iranian, and Uralic languages
(see Comrie 2000, and Johanson and Utas 2000), and in most north-east Caucasian
languages. Basque, an isolate, has a reported evidential (Alcazar 2010, forthcoming).
Only a handful of evidential systems have been described for African languages. The
few descriptions include two Nilotic languages, Shilluk (Miller and Gilley 2007) and
Luo (Storch 2013), !Xun, a Central Khoisan language (König 2013), Fur, a Nilo-Saharan
language from the Sudan (Waag 2010), Laal, an isolate spoken in the Moyen-Chari
prefecture in Chad (Boyeldieu 1982: 125–6), Sissala, a Gur (Voltaic) language from
Burkina Faso (Blass 1989), and Lega, a Bantu language (Botne 2003).
Evidentials are easily diffused in language contact. They spread together with the
diffusion of speech practices and speech etiquette, from one neighbour to the next.
They are a salient feature of the Vaupés River Basin linguistic area (Aikhenvald 2012a:
chs 2 and 9), and of the Balkans as a linguistic area (Aikhenvald 2004: 288–-98). The
presence of evidentials in Turkic and Iranian languages spoken next to each other
may also be accounted for by areal diffusion. That is, we typically find evidentials in
contiguous areas.
Amazonia is a high spot for evidentiality. Many Amazonian groups share the cul-
tural convention of being ‘precise’. Yet, not every Amazonian language has a gram-
maticalized evidentiality system. Ashéninka Perené (Chapter 10) and Aguaruna
(Chapter 11) compensate for this ‘gap’ with a plethora of evidentiality strategies and
other ways of specifying how one knows things.
Once a language becomes obsolescent, evidentials tend to be lost. The last speaker
of Baré, an Arawak language from north-west Amazonia, with whom I chanced to
work, did not use evidentials. Speakers who were around twenty years before that did
employ a reported evidential. The rarity of reports on evidentials in the Australian
languages may be due to the advanced stage of their obsolescence (see Dixon 2002,
on the state of affairs there). The reported evidential continues to be widely used in
Saaroa, a moribund Austronesian language of Taiwan (Chapter 4); however, in many
instances the exact meanings and conditions of its use are hard to pinpoint.
Quality and reliability of descriptions is another matter. Many sketch grammars
and grammatical descriptions cast in prescriptive frameworks do not have a place
for evidentiality. A brief sketch of Shilluk, by Westermann (1911), does not mention
any evidential distinctions. Neither does Migliazza’s (1972) study of Yanomami: his
concern was to fit the language into a transformationalist framework, rather than to
see what distinctions were expressed. Some grammarians are more insightful than
others. Carlin’s (2004) study of Trio discusses evidentiality; Meira’s (1999) grammar
of the same language does not mention it. I suspect that the ‘lack’ of evidentiality in
Africa is due to an oversight, and not to its absence.
Only through detailed investigation of languages based on intensive immersion
fieldwork can we ever expand our general knowledge about the potential of human
languages to mark information source in their grammars. It is as yet premature to try
and map the geographical distribution of evidentials across the world: such an attempt
would reflect how little we know about the expression of information source in Africa,
New Guinea, and the Austronesian domain, and not what the facts are.
2.3 The scope of evidentials: information source of a noun phrase

Evidentials typically have sentential, or clausal, scope. Dependent clauses usually can-
not have an evidential value different from that of a main clause. Then, the scope of
an evidential covers the whole sentence: this is the case in Hinuq (§2.4 of Chapter 2),
Saaroa (§3.4 of Chapter 4), Ersu (§4 of Chapter 6) and numerous other languages. In
Tatar, a language with ‘scattered’ expression of information source, verbal evidentials
have the whole sentence in their scope, while the quotative particle dip extends to the
whole speech report, and the scope of the reported particle di ‘can range from a word
or phrase . . . to a whole text/discourse’ (§4 of Chapter 3).
The scope of verbal evidentials and information-source-marking copulas in Kurtöp
is the clause. In contrast, the clitic which marks reported speech (§7.2 of Chapter 5)
may have scope over an entire stretch of discourse. But it can also be a noun phrase, if
it attaches to just one word, indicating a direct quote. In contrast, epistemic markers
tend to have a clause or a sentence as their scope. So do most evidentiality strategies.
Having a noun phrase within the scope of a grammatical evidential appears to be a
rarity. In Jarawara (Dixon 2004, and p.c.) several information sources can be marked
in one sentence, one on the verb and one on a noun phrase. A speaker was talking
about what had happened to him and his companions, using far past tense (referring
to what had happened more than two years ago): they had seen a place which had been
reported to be another group’s old village:
(17) [[mee tabori botee]-mete-moneha]NP:OotaaA awa-hamaro ama-ke

3nsg home:f old -fpnf-repf 1nsg.exc see-fpef extent-decf
We were seeing in the far past what was reported to be their old camp from far past
The speaker used the remote past (to reflect that it was some time ago) and a firsthand
evidential (to reflect that he had been there and had seen everything himself). And
he used the non-firsthand version of far past tense plus the reported evidential suffix
with the name of the location — ‘reportedly’ known to have been another group’s old
village. This is why the ‘old village’ is marked with the reported evidential.
If I were to attempt to translate this word for word, I would come up with a tortured
and clumsy sentence, such as one reads in English newspapers now and again—The
reported killer was allegedly seen to be captured by the police. But unlike English, the
Jarawara sentence is natural and compact. The same set of tense-cum-evidentiality
markers is to express information source at a clause, and at a noun phrase level.
In a number of languages, information source is marked only at the NP level. We call
it ‘non-propositional’ marking of information source. These appear to always include
a term with visual, or firsthand reference.
Dyirbal, an Australian language (Dixon 1972: 44–57, 2010: 244, and Chapter 8), has
a three-term system of noun markers which combine reference to visibility and spatial
distance of the noun:
bala- ‘referent is visible and there (that is, not near speaker)’;
yala- ‘referent is visible and here (near speaker)’; and
ŋala- ‘referent is not visible’.
There is an additional series of verbal markers which accompany nouns in a periph-
eral locational case (§4 of Chapter 8), with the same meanings. The scope of all these
markers is a noun phrase.
These distinctions are reminiscent of a cross-linguistically rather common eviden-
tial system, with a basic opposition between ‘firsthand’ and ‘non-firsthand’ informa-
tion source (A1 in Aikhenvald 2004; comparable perhaps to Tatar: Chapter 3). The
‘non-visible’ marker covers something that is not seen but heard, or only known from
its noise; something previously visible but now just audible; something neither visible
nor audible; or something remembered from the past and not currently visible (§7 of
Chapter 8).
Somewhat more complex systems of non-propositional evidentiality whose
scope is just the noun phrase have been described for Mataco-Mataguayan lan-
guages of Argentina and Paraguay. The markers combine reference to information
source and to the distance of the nouns’ referent. Chorote distinguishes the follow-
ing markers: visually perceived; distant (or dead/consumed); not visible now but
visible before; invisible or unknown (used in myth) (Carol 2011). The information-
source-related markers in Maká, from the same family, cover the meanings of:
close (can be reached by hand); close (cannot be reached by hand); far and visible;
far and non-visible; absent, seen before; absent, never seen before (Gerzenstein
1994: 166).
Perceptual meanings are encoded within the case system in Tsou, a Formosan lan-
guage (Pan 2010, based on Tung 1964). The ‘nominative’ and the ‘oblique’ case markers
combine information on how distant the object is from the speaker and the addressee,
and whether the object was seen by both speaker and hearer, or by the speaker, or not
seen at all (but is nearby and can be heard, or is known to both). In (18), the speaker
cannot see the child, but its cries can be heard:
(18) m-o mongsi co oko

affix-realis cry(actor.voice) nominative:not.seen.heard child
The child is crying
Neither of these languages have any grammatical marking for the information source
of a clause.
Nominal markers in Maaka indicate the information source of a topicalized noun
phrase (§2.1 of Chapter 9)—that is, they correlate with the discourse-pragmatic sta-
tus of a noun. They encode visually acquired information, assumed information and
information acquired through ‘joint perception’ by the speaker and the addressee. This
meaning is reminiscent of the category of ‘shared knowledge’ described for Kurtöp in
Chapter 5. The ‘joint perception’ noun marker in Maaka also has epistemic overtones
of veracity and ‘truth’. There is also a clausal evidential—the reported marker nà (§2.2.1
of Chapter 9). Its meanings are different from those of the NP-level non-propositional
evidentiality.
The Southern Nambiquara dialect complex has a remarkably complex set of nomi-
nal tense markers fused with information source; nouns are also specified for whether
they are definite, or not, and represent given or new information (raised numbers
stand for tones). Here are some examples, for wa3 lin3-su3-a2 (manioc-cl:bone.like-
def) ‘the manioc root’:
(19) wa3lin3-su3-ai2na2
manioc-classifier:bone.like-definite.current
This manioc root which we both see before us now
(20) wa3lin3-su3-ait3tã2
manioc-classifier:bone.like-observational.middle.past.new
The manioc root that I saw some time past at some distant place (but you
didn’t)
(21) wa3lin3-su3-nũ)1tã2
manioc-classifier:bone.like-inferential.definite.unmarked
The manioc root that must have been at some time past, as inferred by me
(but not by you)
The verbal categories of tense, aspect, evidentiality, and given information are differ-
ent, in form and in meanings (Lowe 1999).
Southern Nambiquara and Maaka are the only languages we know of with different
systems of evidentiality expressed on a clausal, and on an NP level. This is reminiscent
of how tense can be expressed independently within an NP, and within a clause (see
Nordlinger and Sadler 2004). In most cases discussed here, NP-level realization of
evidentiality is intertwined with distance in space. Establishing the existence of NP-
level evidentiality is a major insight within this volume.
There may be more examples of NP-level evidentiality. Santali (Munda: Neukom
2001: 42–4) has a special series of demonstrative pronouns referring to what is seen,
or to what is heard. Both distinguish six degrees of distance combined with emphasis.
The semantic extensions of these demonstratives are parallel to those in evidential-
ity systems: the visual demonstrative can refer to ‘what is evident’, while the auditive
one may also refer to smell, taste, and feeling (Neukom 2001: 42). Note that a two-
term audible versus inaudible demonstrative system has not been recorded in any
language.
Perceptual meanings are often encoded in the system of demonstratives (Aikhen-
vald 2004: 130–1). There, reference to spatial distance can be combined with vis-
ibility or lack of it. The obligatory ‘visible/invisible’ distinction in demonstratives
in Kwakiutl, a Wakashan language, combines with three degrees of spatial distance,
yielding a six-term system (Boas 1911: 41): ‘visible, near me’, ‘visible, near thee’, ‘vis-
ible, near him’, ‘invisible, near me’, ‘invisible, near thee’, ‘invisible, near him’ (see
Dixon 2010, for further examples). The choice of a locational marker in Tima (§3 of
Chapter 12) correlates with the presence of the speaker as a witness of the event or an
object. The category of ventive, roughly translatable as ‘move to where the speaker
is’ in Tima, and in a number of neighbouring languages, also relates to the speaker
being witness to the event, and to potential visibility. In each of these cases however
the information-source related meanings of demonstratives can be understood as a
corollary of their deictic functions: pointing at something is linked to whether you
can see it or not.
3 Information source through other means

3.1 Evidential strategies in grammar
Meanings to do with how people know things may be expressed without develop-
ing a dedicated form whose primary meaning is information source. Non-evidential
categories frequently acquire evidential extensions. This is what is known as ‘evi-
dential strategies’. A conditional mood, a perfect aspect, or a passive can develop an
evidential-like meaning as a ‘side effect’ (also see the discussion in Lazard 1999).
Conditionals and other non-declarative moods may acquire overtones of informa-
tion the speaker cannot vouch for. One of the best-known examples is the condi-
tional in French (known as ‘conditionnel d’information incertaine’) used to relate facts
obtained from another source for which the speaker does not take any responsibility.10
The modal marker mixa in Ersu (§7.1.2) has a range of meanings similar to that of the
assumed evidential.
Further markers may be related to attitude to knowledge. Cavineña, a Tacana lan-
guage with a reported evidential, has a special marker =tukwe ‘contrary to evidence’.
Tucano, with five evidential specifications, also has baa to mark ‘obvious evidence’.
Perfect aspect can be extended to express non-firsthand evidential meanings in
Georgian. This development is shared with many nearby Turkic, Iranian, and north-
east Caucasian languages. Nominalizations and participles often develop connota-
tions similar to non-firsthand evidentials. For example, participles in Lithuanian have
inferential and hearsay meanings. Marking of assertion may correlate with speaker’s
attitudes to information and—indirectly—to its sources. Gascony Occitan has a num-
ber of particles which mark speaker’s assertion intertwined with certainty and ‘general
knowledge’.11 In Ersu (§7.1.1 of Chapter 6), the meanings of the ‘experiential’ aspect
partly overlap with those of the ‘direct’ evidential.
Or the choice of a complementizer or a type of complement clause may serve to
express meanings related to how one knows a particular fact. In English, different
complement clauses distinguish an auditory and a hearsay meaning of the verb hear:
saying I heard John cross the street implies that I did hear John stamping his feet,
while I heard that John crossed the street implies a verbal report of the result. That is, a
that- clause with perception verbs can refer only to indirect knowledge (see a concise
analysis of complement clauses with verbs of perception in English in the context
of complementation in general, by Dixon 2005: 270–1).12 Similar principles apply in
Kalmyk. A participial complement clause of the verb ‘hear, listen’ implies actual hear-
ing, and a clause with the complementizer marks information obtained through hear-
say (§11 of Chapter 7). In Acholi, a Western Nilotic language, a perception verb without
a complementizer implies direct perception (Hieda 2012).
In Aguaruna, a deverbal nominalization has developed nuances of non-firsthand
evidentiality (§§4.2 and 5.2 of Chapter 11). This is reminiscent of Mansi, Nenets, and
Purépecha, where nominalizations have developed similar overtones (references in
Aikhenvald 2004: 117–20).
Evidentiality strategies often develop a range of meanings characteristic of report-
ed and non-firsthand evidentials: they combine reference to inference and to verbal
report. And they are not averse to having epistemic extensions to do with probability,
10
Dendale (1993) and Dendale and Van Bogaert (2007); see Squartini (2007) on how the conditional in
Italian can cover reported information.
11
Giacalone Ramat and Topadze (2008); Hewitt (1995: 259, 93) on Georgian; Wiemer (2008), Grone-
meyer (1997) and Timberlake (1982) on Lithuanian; Comrie (1976: 110), Aikhenvald (2004: 289–96) on
perfect aspect; Pusch (2008) on Occitan.
12
Also see Kirsner and Thompson (1976) on a difference between ‘direct perception of a situation’ and
‘deducing a situation’ in their analysis of complements of sensory verbs in English.
inference based on results or assumption hearsay
-----------------------------------------------------------------------------------------------------
modalities, perfects, resultatives, passives, nominalizations
_________________________________________________________________________________________
hearsay inference based on results or assumption
-------------------------------------------------------------------------
------------------------------------
reported speech, particles derived from 'say', de-subordinated speech complements including nominalizations
Scheme 1 The semantic range of evidentiality strategies.
and also expressing speakers’ attitudes to the veracity of what is being said. Mean-
ings of evidentiality strategies expressed through modalities, perfects, resultatives,
passives, and nominalizations range from inference to hearsay. In contrast, reported
speech, particles derived from ‘say’ and desubordinated speech complements may
develop inferential meanings out of a primary meaning of ‘hearsay’. These pathways
are shown in Scheme 1.
No language has been found to have a special evidentiality strategy for each
of the evidential meanings which can be expressed (I–VI in §2.1). Many of the
features outlined for grammatical evidentials in §2.1 are not characteristic of evi-
dential strategies. For example, no evidential strategy can have scope over a noun
phrase.
Evidentiality strategies in Ashéninka Perené include a number of modalities. Dubi-
tative modality has inferential and assumptive meanings, and affect modality express-
es speculation (§§2.1–3 of Chapter 10). The counterfactual conditional marker has
overtones of speakers’ reliance on their general knowledge and the ensuing expecta-
tion that something similar will occur. The bound ‘intuitive suspicion marker’ is used
in reports about previous experience with speakers relying on their gut feelings as a
basis for assumptions concerning future predictions (§§2.3–4 of Chapter 10). These
meanings are comparable to ‘prospective meanings’ in the Kalmyk evidential system
(Chapter 7), and also to the ‘gut feeling’ evidential in Nambiquara (Lowe 1999). How-
ever, the meanings of evidentiality strategies go beyond what is typically expressed in
closed systems of grammatical evidentiality.
Over time, an evidential overtone of a non-evidential category may conventionalize
as its major meaning. In other words, evidential strategies may develop into grammati-
cal evidentials. For instance, a future tense can give rise to a dedicated non-firsthand
evidential, as happened in Abkhaz (Chirikba 2003: 262–4).
The exact line between an evidentiality strategy on the way towards becoming a
grammaticalized evidential and a fully grammaticalized evidential may be hard to
draw. Nominalizations in Aguaruna are better considered evidentiality strategies

(with a range of non-firsthand meanings) and not fully-fledged evidentials, because
they retain the full range of nominal functions and nominal morphology. Expressing
information source is ‘clearly not their primary function’ (§4.3.2 of Chapter 11). And
they do not always have to refer to non-firsthand information: their evidential over-
tones are optional.
3.2 Information source in its further guises

Every language can express doubt, inference, and assumption. The means vary—from
open classes of verbs, adverbs, and adjectives (§3.2.1), and parentheticals (§3.2.2), to
more restricted subsets of modal verbs and particles (§3.2.3). Various kinds of speech
report constructions are another, almost universal, device (§3.2.4) for talking about
what one has learnt from someone else.
3.2.1 Open lexical classes: verbs, adverbs, and adjectives Verbs of cognition tend
to be linked to information source, as in French penser ‘think’, trouver ‘think,
judge’, avoir l’impression ‘have the impression’ and so on.13 English has an array of
reporting verbs and opinion verbs like think, suppose, find, claim, state, or allege,
in addition to verbs to do with seeming or appearing. Each of these is semantically
versatile and they display subtle grammatical differences (see Dixon 2005:
200–6). One can say It looks like rain, or This idea sounds good, or I hear you are
getting married—each of these ways of saying things in English can be replicated
in German, Dutch, French, Italian, Spanish, and Swedish. These are the ways in
which familiar European languages allow us to express some of the meanings
which must be expressed grammatically in languages like Quechua, Tariana,
Matses, Western Apache, and Shipibo-Konibo (where they form an obligatory
closed system).
This does not imply that ‘exotic’ languages are bereft of verbs of opinion, ‘thinking’,
‘claiming’, and so on. They are not—on the contrary, Tariana (Aikhenvald 2003b) has
an array of verbs to do with mental states. And one can use them to complement the
restricted number of five choices imposed by obligatory evidentials. By using the verb
-hmeta in (22), I specify the fact that the use of the assumed evidential is based on my
intuitive suspicion:
(22) nu-hmeta-ka du-ñami-sita-sika

I-feel.intuitively/think-subordinator she-die-already-assumed.evidential.recent.past
She (assumed) has already died, as I intuitively feel (my gut feeling tells me that she is
dead)
13
See Dendale and Van Bogaert (2007), Pietrandrea (2008), and Giacalone Ramat and Topadze (2008).
I can also use the verb -awada- ‘reason, think’. In (23), I stress that the assumption—
encoded in the evidential—is based on logical reasoning:
(23) nu-awada-ka du-ñami-sita-sika

I-think.by.reasoning-subordinator she-die-already-assumed.evidential.recent.past
She (assumed) has already died, as I reason (that she is dead is a logical conclusion
based on my reasoning)
Tariana has no evidentials to describe intuition and reasoning. The lexical means of
marking information source are much more versatile than the grammatical options.
The interaction between these two is what makes Tariana discourse fascinating. There
are many more options in the details one may want to express through lexical means
than through grammar.
In every language, one can talk about perception—‘see’, ‘hear’, and ‘smell’. The mean-
ings of individual items in individual languages vary: Warekena, an Arawak language
from north-west Amazonia, has just one verb -eda covering these three sources of
perception. In Yukaghir, a Paleo-Siberian isolate, the verb of auditory perception can
refer to vision.14 Ashéninka Perené has one verb kim covering all non-visual sensory
perception (hearing, smelling, touching, and tasting). Its cognate in Tariana, -hima,
has the same range of meanings, in addition to ‘understand’ (-hmeta in (22) is a causa-
tive form of -hima). Most chapters within this volume briefly address lexical verbs of
perception.
Lexical verbs can express further meanings. These may include cultural scripts for
talking about objects and events removed from immediate perception (that is, wheth-
er one ‘sees’ a mental image, or whether there is a separate way of referring to ‘clear
appearance in a vision’: see §3.2 of Chapter 10, for Ashéninka Perené). Dyirbal does
not have a general verb ‘know’: this would be too vague, and go against the cultural
requirement to be ‘precise’ (Chapter 8). We return to this in §5.
Most languages of the world have a way of badging one piece of information as ‘true’
and another as ‘unreliable’. This can be done through adverbial expressions. They may
express possibility, probability, doubt, and can also extend to refer to inference, assump-
tion, validity of information, and attitude to it—that is, they may be used to refer to
information source. English adverbs reportedly, supposedly, and allegedly are a case in
point. One can opt to use an adjective to express a similar meaning: one hears reference
to an alleged drug-dealer, or a supposedly false statement. The choices are many.
Prepositional constructions may express opinion, belief, inference, and so on, for
example Italian secondo me ‘according to me’, or Portuguese ao meu ver (lit. to my
seeing) ‘in my opinion’. Manambu, a Ndu language from the Sepik area, with no
14
Aikhenvald (1998) on Warekena, Maslova (ms) on Yukaghir.
evidentials, employs wuna mawulam ‘in my thinking, lit. in my inside’. Ashénin-

ka Perené (Chapter 10), Karawari (Chapter 13), Aguaruna (Chapter 11), and Maaka
(Chapter 9) have a plethora of expressions to do with belief, disbelief, doubt, and
veracity.
Further means and word classes correlate with the ways in which one ‘knows’ what
is being talked about, and what one thinks about this. Gestures—such as eye-gaze
or hand movements—can be used to indicate doubt or speculation. Ideophones and
onomatopoeia reflect sensory perception, and may correlate with sensory information
sources (see §5 of Chapter 12, on Tima). The plethora of such means goes well beyond
the scope of this volume.
3.2.2 Parentheticals A parenthetical is roughly defined as ‘a word, phrase, or
sentence which interrupts a sentence and which bears no syntactic relation to that
sentence at the point of interruption’ (Trask 1991: 199). It expresses more than ‘source
of evidence’: it is a way of referring to one’s opinion, judgement, belief, inference,
assumption, doubt, attitude, and more.15 European languages tend to have a plethora
of parentheticals, such as English I think, I suppose, Spanish parece, Italian sembra;
and French dit-on and paraît-il.
Parentheticals in English are an open class. A parenthetical construction ‘can paral-
lel any kind of sentence which includes a THAT complement clause coming after the
verb’ (Dixon 2005: 234). A parenthetical may consist of a subject and verb, and also
an object, and provides a comment on the clause. It may express opinion (I think) or
information source (I am told or I suspect). Adverbs and adjectives may be used in
parentheticals. Parentheticals may occupy the same syntactic position as sentential
adverbs, with a similar semantic effect:
(24) a. She will, regrettably, have to sell her car

b. She will, I regret, have to sell her car
A verb or an adjective used in parentheticals may not have a corresponding adverb.

Or the adverb may have a different meaning. In The King will, it is correct, enter by the
front door, the parenthetical with an adjective indicates that this is a correct statement
of what the King will do. But in The King will, correctly, enter by the front door, with the
corresponding parenthetical adverb, the King will act in a correct manner.
The meanings of parentheticals—just like lexical verbs and adverbs—are broad-
er than those of grammatical evidentials, or of epistemic markers. This is what one
expects of an open class. Parentheticals help qualify the utterance; they also have spe-
cial intonation properties and may require a pause. Parentheticals are not uncommon
15
Urmson (1952) is a classic study of parentheticals. Dixon (2005: 233–8) provides a typological frame-
work and an in-depth study of parentheticals in English, in terms of their form and their function. (Dehé
and Kavalova (2007) and Thompson and Mulac (1991) address a number of facts).
across the world (a further discussion of their grammatical properties in European

languages is in Kaltenböck, Heine, and Kuteva 2011). The ubiquitous phrase nu-a-ka
nhua (1sg-say-subordinator I) ‘I am saying’ in Tariana can be analysed as a parentheti-
cal, functionally similar to ‘I think’ or ‘I am saying’ in Modern English (Aikhenvald
2003b: 583–4). Parentheticals in Hinuq (§3 of Chapter 2) express opinions and assump-
tions. Parentheticals in Ersu (§7.4 of Chapter 6) may mark, inter alia, reported infor-
mation. Ashéninka Perené has an array of parentheticals with reportative meanings
(including a self-report marker) (§3.2 of Chapter 10).
The range of meanings of the parenthetical ‘sense, intuit’ in Ashéninka Perené is rem-
iniscent of the notion of ‘internal feeling’ (Evans and Wilkins 2000: 554; Aikhenvald
and Storch 2013). The range of meanings of parentheticals as exponents of information
source go well beyond the limits of recurrent semantic parameters for evidentials as
a closed system. Parentheticals help express subtleties of culturally specific ways of
perception and cognition. Many chapters within this volume illustrate this.
3.2.3 Modal verbs and particles So-called ‘modal verbs’ frequently combine reference
to information source with whatever other meaning they may have. In agreement with
Dixon’s (2005) classification, modal verbs express secondary concepts, ‘those providing
semantic modification of some other verb with which they are in a syntactic or
morphological construction’ (2005: 96). In many languages they are a closed subclass.
Secondary verbs of the same semantic group as seem (2005: 203–5), and verbs of
obligation and permission often extend to cover probability, inference, and assumption.
In Dixon’s (2005: 204) words,
seem is used when the Arbiter is not fully certain whether the adjectival description is appropri-
ate, or whether the statement of the complement clause in a construction like It seems that Mary
found the body or Mary seems to have found the body is correct—perhaps when there is not quite
enough evidence. Appear has the same syntactic possibilities and a very similar meaning, but
may imply ‘can be observed by me’ in contrast to seem ‘can be inferred by me’.
A link with information source is obvious—yet information source is an overtone of

seem, rather than its only meaning. Modal verbs in Hinuq (§3 of Chapter 2) express
attitude to knowledge and epistemic meanings, with overtones of inference. Tatar has
a number of modal verbs and auxiliary constructions with meanings to do with infer-
ence and assumption (§7 of Chapter 3).
A plethora of particles referring to verbal report, or inference, or both may form a
largish but closed class. Lithuanian has over twenty-five particles referring to verbal
report or inference (Wiemer 2008). None of them is obligatory. Many come from
depleted reanalysed verbs of perception, as does girdì, literally ‘you hear’, used to
mark reported information; or speech, as does tariamai, a present passive participle
of the verb ‘say, pronounce’. This is another non-obligatory, and yet non-lexical, way
of expressing information source.
3.2.4 Speech report constructions Every language has a way of reporting what
someone else has said. This can be cast as a direct, or often an indirect speech report.16
Multiclausal speech report constructions can be viewed as lexical ‘paraphrases’ of
meanings grammaticalized in closed evidential systems. And, in many languages,
speech reports acquire epistemic overtones. Saying ‘He says he is a doctor’ may be
meant to cast doubt over his qualifications. In some languages, speech reports are
used to transmit something one does not really believe.17 These connotations are far
from universal—they have not been attested in Hinuq, Tatar, or Saaroa.
What someone else had said can be represented in a number of ways—via quota-
tion, direct, indirect, or semi-direct speech report. It will not be appropriate to go into
the details of how to differentiate a quote from a direct speech report; and what are the
properties of indirect and semi-direct speech; see a summary in Aikhenvald (2011c).
Dimmendaal (Chapter 12) discusses some features of indirect and semi-direct speech
in Tima.
Speech report techniques may have their own epistemic connotations. Here is an
example. If a speaker of Arizona Tewa (Kroskrity 1993: 146) chooses to specify who
said what, a direct quotation is used. The construction includes the reported evidential
accompanied by a complementizer:
(25) 'o-he: gi-ba na-t�

1sgstative.perf-sick that-reported 3sgstative.prefix-say
‘I’m sick’, he said
An alternative would be to use indirect speech, where the third person prefix
replaces the first person. The complementizer remains, and the reported evidential
is removed:
(26) na-he: gi na-t�

3sgstative.prefix-sick that 3sgstative.prefix-say
He said that he is sick
The two alternatives are not fully synonymous. The difference between (25) and (26)
lies in the speaker’s attitude to the veracity of the information. Example (25) means ‘“I
am sick”, he is quoted as saying’ and implies that the speaker does not vouch for the
information reported. That is, for the native speakers of Tewa, the direct quotation
‘lacks the reliability of facticity of its indirect counterpart’ (Kroskrity 1993: 146). In
contrast, the indirect speech in (26) does not contain any overtones of doubt. Similar
16
See Aikhenvald (2011c) for a summary, and further references.
17
See, for instance, Dimmendaal (2001), on reported speech as a ‘hedging’ device.
effects of direct speech quotations have been reported for Gahuku, Usan, and Tauya,
all from Papua New Guinea.18 Tauya also has a reported evidential, and Usan and
Gahuku have no evidentials. In these three languages direct quotes indicate a false
presupposition on the part of the speaker.
Direct quotes in Ashéninka Perené serve to ‘mitigate responsibility for the quoted
statement, at the same time meeting ‘the expectation of being accurate and precise’
(§3.2 of Chapter 10).
Direct quotes in the Ambonwari variety of Karawari have epistemic overtones. In
Telban’s words (§1.4 of Chapter 13), ‘the Ambonwari are inclined to put words into
other people’s mouths in this way to (re)create their intentions’; ‘it is therefore thought
that direct quotation is more an expression of opinion or presupposition (anxious,
desired, assumed, false, or exaggerated) of the reported than of the original speaker
or the person involved in the event’. And so, ‘direct quotations can contain overtones
of doubt and lack reliability’. Throughout my fieldwork with the Manambu, another
group in the Sepik area of Papua New Guinea, I observed the same principle at work.
A self-quotation in Ambonwari may have epistemic overtones: a listener is likely to
doubt the veracity of a self-quote. A self-quotation can be a means of ‘lifting speaker’s
importance’, or to justify why the speaker had done a particular thing. This is reminis-
cent of how speech reports in general can have overtones of causation and intention
(Aikhenvald 2011c: 319 and references there; Overall 2008, with special relevance to
multifunctional speech reports in Aguaruna).
In other languages, a self-quotation is a way of stressing the veracity of what one
is talking about (see Michael 2008, on speech practices in Nanti, a Campa Arawak
language). In Ashéninka Perené, a combination of a reportative parenthetical ‘they
say’ with the assertive marker emphasizes the credibility of what is being talked about
(§3.2 of Chapter 10). Or it may be employed to make sure the author of the report is
specified. In Kalapalo, ‘the emotions and motives of characters . . . are realized through
their quoted speech’ (Basso 1995: 295).
Reported evidentials and reported speech (including quotations) do essentially the
same job: they indicate that the information was acquired from someone else. It is
no wonder, therefore, that they can acquire similar semantic extensions. A marker
of speech report, or a generic reported parenthetical (as in Ashéninka Perené, §3.2
of Chapter 10), allows the speaker to leave the author of the speech report vague.
Other techniques for expressing reported speech allow the source to be stated (see, for
instance, §7.2 of Chapter 4, on Saaroa).19
It thus comes as no surprise that a speech report construction is a frequent source
for developing reported evidentials. One such grammaticalization path involves rea-
nalysis of a biclausal quotation or reportative construction whereby the matrix clause
18
Deibler (1971: 105) on Gahuku; Reesink (1986: 259) on Usan, and MacDonald (1990) on Tauya.
19
Further comparison is in Aikhenvald (2004: 135–40).
with the verb ‘say’ and a complement clause of this verb become a single clause via the
loss or reinterpretation of the subordinator (Aikhenvald 2004: 273–4; 281–3). This is
what we see in a marker of reported speech, dizque, in Colombian Spanish. Example
(27) illustrates this (Travis 2006):
(27) y eso, dizque es peligroso ¿no?

and this reported is dangerous no
And this, it is said to be dangerous, isn’t it?
Similar scenarios—whereby speech report constructions have given rise to reported

and also to quotative markers—have been described for Hinuq (Chapter 2), Tatar
(Chapter 3), Saaroa (Chapter 4), and Ersu (Chapter 6).
3.3 Information source: a summary

Meanings associated with information source can be expressed as extensions of non-
evidential grammatical categories and also through members of open classes. For the
latter, the range of meanings is wider than that of grammatical evidentials. Closed
classes of particles and modal verbs tend to share their meanings with evidential
strategies.
The choice of a grammatical evidential often depends on the mood or tense of the
clause (see §2). The choice of a parenthetical or an adverb depends on what the speaker
wants to say. A parenthetical, an adverb, or a modal verb can have an NP or a whole
clause in its scope. For grammatical evidentials, these options are restricted. None of
the means listed in §§3.1–2 forms a paradigm of any sort. In contrast, grammatical
evidentials do.
What may justify putting the various verbs, adverbs, and parentheticals discussed
in §§3.2.1–4 together with modal verbs and particles is the fact that they all vaguely
relate to the ways in which one knows things. All these devices for marking infor-
mation source combine reference to inference, assumption, and often speech reports
with increasing ‘subjectification’ — a ‘historical pragmatic-semantic process whereby
meanings become increasingly based in the speaker’s subjective belief state, or attitude
toward what is said’ (Traugott 1996: 185). This is what sets them apart from closed evi-
dential systems—whose primary meaning has nothing to do with subjectification—
and makes them similar to prototypical modalities.
One question concerns the means a language may use to express knowledge. Anoth-
er, and a trickier one, is how and when it is appropriate to talk about knowledge and
the ways in which it is acquired. How people discuss knowledge and express attitudes
to its reliability and trustworthiness may depend on their status in society, on their age
category, and even on their sex. Women and men may differ in their preferred hedg-
ing strategies (see, for instance, Bradac, Mulac, and Thompson 1995). Question tags
seeking confirmation are widely—but perhaps incorrectly—believed to be features of

women’s speech (Dubois and Crouch 1975). All of this may depend on accepted con-
ventions within a society, on types of knowledge and on types of acceptable informa-
tion source. In a small Amazonian society one does not use a hedging strategy when
talking about a shamanic vision—this is simply not done.
4 Choosing, and using, an evidential

An evidential may have an additional meaning, going beyond information source.
This may have to do with certainty or lack thereof, probability, and unexpected infor-
mation and surprise. Choosing and using an appropriate evidential shapes communi-
cation, and may itself be shaped by established and by emerging cultural conventions.
Different experiences may require different choices from the evidential system.
4.1 Beyond information source

4.1.1 Certainty, control, and first person Evidentials may have non-evidential
overtones. One is likely to be certain about what one has seen with one’s own eyes.
The visual or a direct experience evidential tends to have overtones of commitment
to the truth of utterance, control over the information, and certainty. It may cover
information acquired through seeing, and also generally known and observable
facts. Every Peruvian knows that there are monkeys in the rainforest. This generally
known fact is expressed using a visual evidential. Example (28) comes from Cuzco
Quechua:20
(28) Yunka-pi-n k'usillu-kuna-qa ka-n

rainforest-loc-direct.evidential monkey-pl-topic be-3person
In the rainforest, there are monkeys
East Tucanoan languages and Tariana have two sensory evidentials—one for visual,
and one for non-visual information. You cannot ‘see’ how you feel—so it is appropri-
ate to use a non-visual evidential when talking about yourself this way, in Tucano
(Ramirez 1997: 133, 135):
(29) yɨ'ɨ-re upî-ka pũri-sa'

I-top.non.a/s tooth-cl:round hurt-pres.nonvis.nonthird.p
My tooth hurts
20
Eberhard (2009: 464–5) provides similar examples for visual evidential in Mamaindê, a Nambiquara
language.
When you talk about how someone else feels, you judge by what you see yourself: you
cannot get into their skin and feel what they feel. So, a visual evidential is then appro-
priate. If I see that Pedro looks very sick I will say, in Tucano:
(30) Péduru do'âtigi' wee-mí

Pedro sick be/do-pres.vis.third.person.masculine
Pedro is sick (I see it)
A non-visual evidential may refer to something I cannot quite see, and am not quite
sure about. A Mamaindê man has just taken a second wife, but is not quite certain if
he has done the right thing, and so he uses the non-visual evidential in talking about
this (Eberhard 2009: 466).
When used with a first person subject, the non-visual, non-firsthand evidentials
and reported evidentials in systems of various types may acquire additional meanings
to do with lack of intention, control, awareness and volition on the part of the speaker.
Visual evidential has an overtone of certainty—I am sure of what I see. But if I am
talking about myself, I can use the non-visual evidential if whatever happened was out
of my control. Suppose I broke a plate by accident—it slipped out of my hands. I will
then say, in Tucano:21
(31) bapá bope-ásɨ

plate break-rec.p.nonvis.nonthird.p
I broke a plate by accident
This is what the literature on evidentials calls the ‘first person effect’: when I talk about
myself, evidentials have somewhat different overtones. If I was drunk or unconscious,
and do not really remember what I did, I can even use a reported evidential to talk
about myself: ‘I spent the night drinking-reported’ takes away all the responsibility
from my being drunk all night. In Hinuq (§2.2.1 of Chapter 2), if the unwitnessed
evidential is used with a first person subject, this implies the speaker’s lack of control
over what happened to them, or simply their lack of memory. The reported evidential
may occur with a first person subject, with similar meanings—‘lack of control or an
unintended, unconscious participation’ (§2.5 of Chapter 2). The reported evidential in
Saaroa with the first person has a similar semantic effect (§5.2 of Chapter 4).
Verbs covering internal states may require obligatory evidential choice depending
on person: for instance, one may use the non-visual evidential to refer to one’s own
state, and the visual or inferred one to refer to a state experienced by someone else
(Aikhenvald 2004: 224–5). As a result, evidentials may acquire the implicit value of
21
Ramirez (1997, vol. 1: 133).
person markers (this is similar to the distinction between ‘Self ’ and ‘Other’ in eviden-
tial use in Pastaza Quichua: see Nuckolls 2008).
Evidentials may interrelate with person in other ways. In Mỹky, an isolate from
Brazil, visual and non-visual evidential distinctions are obligatory for second and
third person, and are not made at all in the first person.
4.1.2 Expectation of knowledge, and mirativity Suppose I see something which I
did not expect. I can then use a non-firsthand or a non-visual evidential. A speaker
of Jarawara saw a dead sloth—he was surprised that the sloth was dead, and used the
non-firsthand evidential despite the fact that he actually saw it:22
(32) jo abohi home-hino

sloth(masc) be.dead+compl.cl lie-immediate.past.nonfirsthand.masc
A dead sloth lay (there) (non-firsthand: as a marker of surprise)
This meaning is known as ‘mirative’. Many languages employ non-visual evidentials,

with this meaning. In Mapudungun, this is one of the uses of the non-firsthand evi-
dential -rke (see (4)–(5) above). Unwitnessed past forms in Hinuq (§2.2.1 of Chapter 2)
have mirative overtones in the context of first-person subject, especially ‘when there is
something mysterious or unexplainable in the event’. That is, an overtone of ‘surprise’
is linked to the lack of previous knowledge of the speaker. Along similar lines, mira-
tive overtones of the non-firsthand past in Tatar (§4.1.2)—found just with the first-
person subject—can be linked to the lack of ‘involvement’ by the speaker. Overtones
of surprise of the speaker in Saaroa (§4.2 and §5.2 of Chapter 4) may also occur if one
is talking about oneself using the reported evidential.
The inferred evidential expresses ‘surprise’ in Mamaindê. When David Eberhard
returned to the Mamaindê village after a long absence, speakers were surprised that he
could still speak the language, and commented on this, using the inferred evidential.
They also use ‘inferred’ in humorous songs about nature which may take surprising
twists: for instance, that baby toucans have (inferred) very soft beaks, unlike the young
of many other bird species. The indirective marker in Kalmyk can have mirative mean-
ings (§13 of Chapter 7). The particle ikän in Tatar usually expresses assumption (§5.2.1
of Chapter 3). In conjunction with the indefinite future tense, it conveys overtones
of surprise (example (25), Chapter 3). Just like the inferred evidential in Mamaindê,
this particle can be used in joking contexts. How ‘mirative’ extensions—or a special
mirative marking if there is one—can be manipulated in jokes and other genres is a
further fascinating topic.
22
This example comes from Dixon (2003: 171). ‘Surprise’ can be expressed through other means. A highly
unusual system of ‘mirative’ pronouns in Hone was described by Storch (1999).
Mirative meanings cover ‘expectation of knowledge’ rather than information source

(also see Aikhenvald 2012c). It is thus distinct from evidentiality (see Aikhenvald
2012c, for a cross-linguistic typology and further references). Kurtöp (Chapter 5) is
a prime example of a language where information source, expectation of knowledge,
and knowledge sharing are linked together (and yet distinguishable). Saaroa has an
enclitic with a gamut of mirative meanings (§8 of Chapter 4). Ashéninka Perené has
what can be described as a ‘mirative’ strategy: the affect modality in content questions
imparts overtones of the speaker’s surprise at some unexpected information (§2.2 of
Chapter 10). ‘Surprise’ in Karawari is marked through incorporation of the interjec-
tion kambay- ‘alas!’ between two verbs, one of which, or both, tend to be perception
verbs ‘hear’ or ‘see’ (§5 of Chapter 13). Mirative meanings in Hinuq are expressed with
a special verb (Chapter 3). Kalmyk has a mirative particle whose meanings cover sur-
prise, and also newly acquired and unexpected information (§13 of Chapter 7).
4.2 How to choose the correct evidential

4.2.1 Evidentials: an enviable feature? Having to always express information
source in one’s language is often viewed as an enviable feature. Speakers of languages
without evidentials wish they had been compelled to always be so ‘precise’. In
Palmer’s (1996: 200) words, ‘what a lot of breath and ink this might save us in
English if we had evidential suffixes that we could use in the courtroom. Using the
Wintun suffix, we might say, for example, “The defendant shoplift-be [be is a visual
evidential] the compact disc”, thereby eliminating the need to ask the inevitable
question: “Did you actually see her take it?”’ And, as Boas (1942: 182) put it, ‘we
could read our newspapers with much greater satisfaction if our language would
compel them to say whether their reports are based on self-experience, inference, or
hearsay!’
Evidentiality is ingrained in speech habits and conventions—breach of which may
result in losing face and reputation. And the adoption of new means of acquiring
information, such as television or internet, results in extending the meanings of evi-
dential categories.
If a language has obligatory evidentials, leaving them out produces a grammati-
cally awkward ‘incomplete’ sentence. Those who have evidentials in their languages
complain that languages without evidentials—Portuguese and Spanish included—are
somehow deficient and inadequate. Hence the perception of ‘white people’—those
outsiders who do not have information markers in their speech—as ‘liars’. Tariana and
Tucano speakers grumble that Portuguese sounds like a ‘shortcut’. Hardman (1986: 133)
reports how difficult it is for Jaqi (Aymara) speakers to imagine how one can speak a
language which does not mark the information source. Finally she and her colleagues
had to ‘adjust their English’ and always specify how they knew things, so as not to
upset their Jaqi-speaking friends. Speakers are often conscious of how evidentials are
to be used.
Among the Mamaindê, a typical way to refer to a ‘good, trustworthy person’ is to

call them ‘one who speaks well’. Someone who is ‘untrustworthy or of a questionable
moral reputation is labelled as one who does not speak well’ (Eberhard 2009: 468).
The correct use of evidentials is the ‘token’ of a good speaker—and henceforth, of a
good person.
The same principle applies in Huallaga Quechua (Weber 1986: 142), and among
the Tariana and the East Tucanoan peoples in the Vaupés River Basin area. The late
José Manuel, a Tariana elder, was sneered at and said (behind his back) to be ‘useless’
because he was not using the correct remote past reported evidential. A major token
of ‘correct’ Tariana is the ability to use evidentials in the right way.23
4.2.2 Cultural conventions in evidential use: dreams, spirits, shamans, and
authority Fixed evidential choices may always describe certain types of experience.
We call them cultural conventions. Consider dreams. In Jarawara and Wanka
Quechua, dreams are ‘seen’; they are part of ‘everyday experienced reality’.24 In
Hinuq and in Tatar, dreams are ‘seen’, and are recounted as if they were personal
experience (§5 of Chapter 2). The Ashéninka Perené use the parenthetical ‘appear
clearly in a vision’ and the verb ‘see’ to talk about dreams (§3.2 of Chapter 10).
Dreams are ‘seen’ in Karawari, Aguaruna, and also Manambu and Kwoma (from
the same area as Manambu). In Kalmyk dreams are ‘seen’, but recounted using the
indirect information source marker (example (18) in §4, Chapter 7).
But dreams experienced by ordinary humans are not part of reality in Shipibo-
Konibo—and so they are recounted using the reported evidential =ronki. However, if
a shaman has a dream or a vision induced by the hallucinogenous ayahuasca he will
retell this experience using direct evidential. In Tariana and Tucano, a simple mortal
will talk about their dream using the non-visual evidential. Only a powerful shaman
will use the visual evidential when recounting what they saw in a dream. That is, evi-
dentials can be linked to person’s status, access to knowledge and power, in societies
which we are accustomed to consider egalitarian at heart.
In their traditional life, speakers of Kagwahiv (a Tupí-Guaraní language, from the
Upper Madeira River basin) used to rely on dreams a lot. Dreams were used to forecast
the presence of game, to plan the day’s hunt, and to foresee illness and death. In times
of wars, dreams were relied upon to predict the victor. Relating a dream and discussing
what it may possibly mean used to be an important part of Kagwahiv interactions. And
every sentence in a dream contains an evidential, ra'ú—a marker of non-firsthand
information. This may appear odd: as Kracke (2010: 69) puts it, ‘in our way of thinking
about dreams, it would seem that dreams are par excellence events witnessed by the
person telling them’. The Kagwahiv ‘dream-marker’ ra'ú is cognate to words meaning
23
See details in Aikhenvald (2002: 213–20; 2004: 336–7).
24
See Floyd (1999: 65), on Wanka Quechua, Dixon (2004: 203), on Jarawara, and Aikhenvald (2004:
345–6).
‘ghost’, ‘augury’, and relating to ‘falsehood’ in general. Dreams can provide informa-
tion about the future; but they are regarded as essentially deceptive and unreliable: a
dream ‘is a message, a message from an unknown source. Hence it cannot be coded
as personal experience’ (2010: 73). Or, in the words of Western psychoanalyst Lacan
(1988: 135), ‘someone other than ourselves talks in our dreams’. What we have to learn
from a psychoanalyst, speakers of languages with evidentials know through evidenti-
ality conventions. Along similar lines, dreams in Tsou (Yang 2000: 82) are recounted
using the reported evidential.
Talking about spirits in Dyirbal requires the non-visible noun marker (§7 of Chap-
ter 8). A spirit may appear in disguise—as a rainbow, or a woman. One can see the dis-
guise, or the manifestation, but not the ‘true’ self of the spirit. The non-visible marker is
a way of talking about a not-quite-real reality. Along similar lines, my Tariana teachers
often talk about their encounters with evil spirits in the jungle. This experience always
involves the non-visual evidential.
Knowing which evidential to use, and when, provides an important way of impos-
ing one’s authority. A warrior-ancestor of the Carib-speaking Kalapalo is a strong
character, who wishes to make a point in his speech. This attitude is reflected in the
evidentials he uses. In Basso’s (1990: 137–40) words, the most assertive and imposing
part is marked with distant past firsthand evidential, and ‘the tone is something like,
“I bear witness”’.
But if you are neither a respectable authority nor a shaman, and the community feels
you have no reason to over-use an evidential, you may be in trouble. Weber (1986: 142)
describes a speaker who was using the direct evidential -mi too much. To many, this
sounded ‘incautious with respect to the information’ conveyed; the man was judged to
be ‘not a member of a Quechua speaking community which values his stature’.
Breaching conventions of evidential use results in possible social exclusion.
There is an obvious connection here with knowledge and its expression, as part of
societal norm, and knowledge as a social phenomenon: see §5.
4.2.3 Evidentials, new technologies, and change What happens if speakers of a
language acquire access to new ways of knowing things? As Boas put it, ‘when
changes of culture demand new ways of expression, languages are sufficiently
pliable to follow new needs’ (1942: 183). New practices—reading, television, radio,
telephone, and internet—help us understand just how pliable the systems are. A
Shipibo-Konibo speaker will now employ reported =ronki to talk about what they
read in a book. And for this a speaker of Tariana or Tucano will use an assumed
evidential, typically used for information acquired by interpretation, reasoning, or
common sense.
If a Shipibo-Konibo watches something on television, this implies ‘experiencing the
event oneself, since one actually “sees” what is happening’—and so they would use
the direct evidential =ra. Tariana and Tucano speakers would use a visual evidential.
The Ashéninka Perené use the verb ‘see’ to talk about television shows (§3.2 of Chap-
ter 10). But if a Shipibo-Konibo hears something on the radio, or hears a TV report
without seeing the picture, they will use the reported =ronki. A similar system has
been described for Yongning Na, a Tibeto-Burman language (Lidz 2007). A Tariana
or a Tucano would use a non-visual evidential.
In contrast, Magar (Grunow-Hårsta 2007), from the Tibeto-Burman family, employs
the inferred evidential to recount what one saw on television. This is consistent with
how this evidential is employed in narratives: it is a way of describing a picture book.
The reported evidential is only used to recount what one has heard.
Before Tariana speakers acquired regular access to phones, they used a non-visual
evidential for the occasional reports of phone conversations. Ten years on, a phone
is part of their lives, and a conversation on the phone is being treated as the same as
a face-to-face talk. A visual evidential is now preferred in this context. However, one
speaker who does not have a phone at home and uses it only occasionally keeps using
a non-visual evidential. And for the few speakers who now are in the habit of regularly
chatting over the internet, this is also like face-to-face: a visual evidential is preferred.
When a speaker of Hinuq or of Tatar retells something they have seen on TV or
heard on the radio, they use unwitnessed forms—since they were not there to see the
event for themselves, and are relying on someone else’s account (§2.3 of Chapter 2 and
§6.1 of Chapter 3). An evidentiality-neutral form can also be used. For talking about
live broadcasts, only neutral past forms are appropriate. In fictional stories, unwit-
nessed and evidentiality-neutral forms can be employed, with different stylistic effects:
using the unwitnessed form makes the story sounds like a traditional tale, while using
an evidentiality-neutral form sounds as if they actually witnessed the event. A phone
conversation in Hinuq is treated just like a face-to-face talk (§5 of Chapter 2).
The attitude to knowledge, and the means of acquiring it, may change over time.
Telban (Chapter 13) has been working with the Karawari people for more than twenty-
five years. Back in the old days, what people knew was reliable and steady—based on
‘the internal knowledge of the past’. Nowadays, with new technologies coming in, the
pace of life quickening and communications improving, speculations are pervasive.
And people constantly complain about ‘unreliability of information’ and untrustwor-
thiness of the sources. The frequency of assumptions and ‘wishful thinking’ in actual
discourse has dramatically increased (§2 of Chapter 13). As a consequence, the value
of different kinds of ‘knowledge’ has shifted.
4.2.4 Evidentials and genres Types of stories may always go together with just
one evidential. We call these tokens of a genre. In the overwhelming majority of
languages, ancestral stories and legends are told using reported evidential. Traditional
tales in Jarawara are told using non-firsthand evidential, which in 90 per cent of the
cases is followed by the reported suffix. A story about what happened to the speaker
can be firsthand, as in Jarawara, experiential, as in Matses, or visual, as in Tariana,
Tucano, and other languages with many evidential options. Fairy tales and legends in
Hinuq are cast in the unwitnessed evidential (§2.2.1 of Chapter 2). The non-firsthand
evidential is used in legends, folk-tales, and historical accounts in Tatar (§6.2 and
Table 1 in Chapter 3 offer an overview of evidentials as tokens of genre in this
language). In Ersu, traditional legends are told using the reported evidential (§5.1 of
Chapter 6). In Kalmyk, a story to which the speaker was a witness is told using the
direct evidential. The indirective marker is preferred in traditional narratives and folk-
tales. Reported evidential is used in the modern newspaper style. New genres and new
means of communication require new speech practices. In Kalmyk the prospective
evidential -x bolv is used in newspaper language. It marks reported information about
planned future events obtained in personal interviews with their planners (point (c) in
§8 of Chapter 7).
Evidentials, and evidentiality strategies, can serve as grammatical means for dif-
ferentiating genres of stories. A historical account in Aguaruna is typically told using
nominalizations as an evidentiality strategy with a non-firsthand meaning. A tradi-
tional myth employs this same strategy in addition to the narrative modality marker
(§5.2 of Chapter 11). In Tariana, an autobiographical story will be told using the visual
evidential. A traditional tale would be cast in reported, and a historical account about
one’s ancestors—which is based on visible traces of their movements—will involve
inferred evidentials. A story about shamans’ activities is often cast in non-visual (the
explanation given to me was ‘because it all happens in the shaman’s mind’).
Evidentials may be used several times in a clause to make the narrative more vivid:
that is, for stylistic effect. This is the case in Hinuq and in Tatar (Chapters 2 and 3). In
Saaroa, the reported evidential can be repeated if a constituent on which it occurs is
contrastive (§3.3 of Chapter 4).
4.2.5 Truth, lies, and evidentials The ‘truth value’ of an evidential may be different
from that of the verb in its clause. It is simply not the case that those who speak a
language with evidentials never lie. Evidentials can be manipulated to tell a lie. As
Eberhard (2009: 468) puts it in his grammar of Mamaindê (based on living with the
people for eighteen years):
I do not see any basis for the supposition that they have a stronger than a normal concern for
truth. The evidential system, in fact, can be taken advantage of and exploited quite ingeniously
for the express purpose of lying, not only about the content, but also one’s degree of involvement
in a given situation.
One can give correct information source and wrong information, as in saying ‘He is dead-
reported’, when you were told that he is alive, or correct information and wrong informa-
tion source, as in saying ‘He is alive-visual’, when in fact you were told that he is alive, and
did not see this. Having evidentials is not about needing to be ‘truthful’. In a similar vein,
one does not need to speak a language with grammatical tense to be punctual.
Evidentials are a powerful device for manipulating knowledge. Using them in an

appropriate and accepted fashion is a means of safeguarding your reputation. Speak-
ing properly, and using the right evidentials, are a means of saving face. They are often
the basis for a value judgement: a good and proper person is the one ‘who speaks well’
(Eberhard 2009: 468). This takes us to our next section.
5 How to talk about knowledge

Epistemological devices—that is, evidentials and other means of expressing informa-
tion source—reflect the means of acquiring knowledge, and attitudes to it. ‘Knowl-
edge’ is a ‘social phenomenon, an aspect of social relations between people’ (Hill and
Irvine 1992: 17). Proper linguistic expression of knowledge promotes mutual under-
standing as a basis for empathy. But expressing your knowledge in an inappropriate
way may result in social exclusion and rejection.
5.1 On being precise

Being precise in one’s information source goes together with cultural conventions
which appear to be particularly strong in languages with evidentials. Such conventions
may include:
(i) whether one should be as specific as possible when speaking, or whether a high
degree of vagueness is a normal social expectation, and
(ii) how much information is to be shared—whether one should tell people what
they want to know, or whether ‘new information’ is regarded as prized goods,
only to be disseminated for some appropriate return.
‘Information sharing’ is an obligatory verbal category in Kurtöp (Chapter 5). A similar

principle is reflected in the marker of ‘joint perception’ in Maaka (Chapter 9).
In many linguistic communities with evidentiality, being as specific as possible
about what one has to say is obligatory. Those who do not obey the cultural conven-
tions of evidential usage are not to be trusted. Quechua cultural postulates summa-
rized by Weber (1986: 138) point in the same direction. These are:
1. (Only) one’s own experience is reliable.

2. Avoid unnecessary risk, as by assuming responsibility for information of
which one is not absolutely certain.
3. Don’t be gullible. (Witness the many Quechua folk-tales in which the villain is
foiled because of his gullibility.)
4. Assume responsibility only if it is safe to do so. (The successful assumption of
responsibility builds stature in the community.)
That is, one should provide the information required, and be specific about it. In a
similar vein, McLendon (2003: 113) reports:
Eastern Pomo speakers from whom I have learned Eastern Pomo since 1959 remembered that
when they were children their grandparents constantly reminded them to be careful how they
spoke. They were told to be especially careful to speak well to, and about, other people, because
if they didn’t the person spoken about, or to, might be offended and try to ‘poison’ them, that
is, use ritual or other means to bring them misfortune, illness, or even death. Evidentials which
distinguish non-visual sensory experience, inference, memory, and knowledge seem a useful
means of speaking with care, asserting only what one has evidence for, and making one’s evi-
dence clear.
In the context of Amazonian societies, the requirement to be precise in one’s informa-

tion sources may be related to the common belief that there is an explicit cause—most
often, sorcery—for everything that happens. So as not to be blamed for something
that in fact they had no responsibility for, a speaker is careful always to be as explicit
as possible about what they have done (see Aikhenvald 2004). This relates to the desir-
ability of stating the evidence for everything that is said, visually obtained information
being the most valuable. The speaker is also careful not to impute their assumption and
their information source onto another person. Such imputation could be potentially
dangerous: if the speaker is perceived as having access to how other people know
things, they may well be regarded as a sorcerer. In a society where sorcery is the most
dangerous crime of all, to be accused of it is hardly desirable. Different conventions in
stating information source may create conflicts, miscommunication and social exclu-
sion. Evidentials provide grammatical backing for Grice’s (1989) Maxim of Manner,
helping avoid ‘obscurity of expression’ and ‘ambiguity’.
Being informative has different implications in different cultures. For the everyday
exchange of information in Malagasy in remote rural communities ‘the basic norm
concerning free exchange of information simply does not apply’. New information
is a ‘premium’ to be imparted piece by piece. As Keenan and Ochs (1979: 149) put it,
‘possession of new information is possession of a scarce good allowing the posses-
sor to command the attention of others’. But to a Westerner, the Malagasy ways of
information exchange sound vague and uninformative. Does this feature of Malagasy
discourse correlate with the absence of evidentials in the language? Does it correlate
with any specific strategies of talking about what one knows, and the structure of the
semantic field of cognition? One wonders.
It is however not the case that once you have evidentials in your language you
have to be precise. In some languages with a small evidential system there may be a
non-firsthand term which lacks the ‘precision’ of information source (examples from
Mapudungun are in (4) and (5)). There is no indication that many languages with
just a reported evidential—such as Basque or Estonian—have a requirement to be
‘precise’.
The correlations between the requirement to be precise in one’s information source
imposed by the grammar, and cultural conventions, are very tempting, but highly ten-
tative. Moreover, the same requirements and conventions appear to hold in languages
with no grammatical evidentiality. Neither Tuvaluan (Besnier 1992) nor Weyewa

(Kuipers 1992) have grammatical evidentials. And yet what was said above applies to
speakers of these languages, too. Being precise and explicit is important for speakers of
Aguaruna (Chapter 11) and Ashéninka Perené (Chapter 10). However, these languages
do not have grammatical evidentials.
Australian Aboriginal communities value explicitness. One should be as specific
as possible in identification and in description. Only a few Australian languages have
grammatical marking for information source. In others, this is achieved by lexical
means. In Dyirbal, for instance, there is no verb ‘to know’; it would simply be too
vague. When R. M. W. Dixon enquired how to say ‘I know where the money is hidden’,
he was told that details had to be provided. One could say ‘I saw where the money is
hidden’ or ‘My father told me where the money is hidden’. Evidentials in Dyirbal are
limited to a noun phrase (see §2.3 and Chapter 8).
Being specific as to one’s information source appears to correlate with the size
of a community. In a small community everyone keeps an eye on everyone else,
and the more precise one is in indicating how information was acquired, the less
the danger of gossip, accusation, and so on. No wonder that most languages with
highly complex evidential systems are spoken by small communities. On the other
hand, why is it that some languages spoken in small closed communities have only
a reported evidential? Fortescue (2003: 301) is also convincing when he speculates
that
presumably life in very small, scattered Arctic communities, where everyone is likely to know
of everyone else’s doings and where rumours spread easily, is such as to make being vague
[emphasis mine] about one’s source of information . . . a generally sensible strategy.
Speech styles and genres may also correlate with degree of precision, and how attitudes
to knowledge and ‘truth’ can be cast. An open debate—or a longhouse address—in a
traditional society may reveal power relationships reflected in talking about ‘truth’
and how it is known (see, for example, Lindstrom 1992, on Tanna, an Austronesian-
speaking group in Vanuatu). The expression of how one knows things may be different
in everyday language, and in a secret language, including ‘avoidance’ styles. The lan-
guage used in spirit possession and by mediums may also differ from the ‘normal’ style
in how knowledge is talked about (see, for instance, Storch 2011, on secret languages
and special ‘spirit’ languages in Africa).
The requirement for precision is a feature of the ordinary Dyirbal. In traditional
times, there used to be another register employed when speaking in the presence of
in-laws (this is known as ‘avoidance style’). In Dixon’s words (§8 of Chapter 8), ‘while
a high value is placed on precision in normal speech, it is considered appropriate to
be deliberately vague in an avoidance situation. There you have it—different levels of
specificity and generality, each in its proper place’.
5.2 The value of knowledge

A requirement to be precise may correlate with the value placed by the community
on a particular type of knowledge (which may be restricted). If something was seen,
heard, inferred, and also assumed, the ‘visual’ source is likely to be the first option in a
‘hierarchy’ of evidential choices. Visual perception is the most ‘valuable’ source.
In many languages with evidentials, it is a feature which speakers are prepared to
discuss. An evidential can be ‘rephrased’ with a verb of perception which roughly
corresponds to its meaning. The late Américo Brito was the only person among the
Tariana to have witnessed the Offering Ritual. He told a story about it using visual
evidentials. Some of the audience looked at him incredulously. Reacting to this, he
added ‘I saw this-visual-remote-past’. Metalinguistic perception of evidentials opens
a gateway to our understanding of representation of knowledge, and non-evidential
ways of expressing epistemological nuances (see also Aikhenvald 2008).
Being a ‘good’ speaker—proficient and ‘correct’ in one’s evidential choice—is equat-
ed to a being a reliable citizen in numerous Amazonian societies (see Eberhard 2009:
468, on Mamaindê, and Aikhenvald 2013, on the Vaupés area and beyond).
Speakers of mainstream European languages tend to think of ‘hearsay’ as unreli-
able. But for predominantly oral cultures, valuable knowledge used to be embodied
in, and transmitted through, traditional stories and speech reports. As Overall (§5.2
of Chapter 11) puts it, ‘the canon of oral literature . . . is a means by which the com-
munity can maintain a body of knowledge that is more than one person can handle
alone’. Within an essentially oral tradition of transmitting knowledge, ‘marking a
narrative as hearsay’ has the opposite effect of what you expect in English: ‘it imparts
the legitimacy of precedent to the narrative being related’. This provides a reason why
in many essentially oral cultures reported information has no epistemic overtones
of doubt.
The value of types of knowledge changes over time. With the spread of new technol-
ogies and new and faster means of communication, the Karawari started relying more
on assumptions and speculations than on the traditional knowledge of the past (§1.4 of
Chapter 13). And as oral cultures throughout Amazonia acquire literacy systems (see
Aikhenvald 2012a: 60, 381), the value of the ‘written word’ increases dramatically. The
impact on evidentiality systems is yet to come.
Epistemological expressions (which subsume evidentials) tell us something about
the speech community. They may be used to show power, authority, and agency. An
omniscient shaman ‘sees’ everything. That his special knowledge can be cast in visual
evidential highlights his power and authority. As Duranti (1990) puts it, ‘language does
not simply reflect the world, it also shapes it, fashions it’.
This is directly related to language ideologies and theories of knowledge, including
the nature of theories of mind, the role of intention in linguistic communication and
social interaction, and the importance of empathy. The requirement to mark informa-
tion source in Western Apache (a language with evidentiality: de Reuse 2003: 96) may
go together with ‘Athabaskan attitudes about the autonomy of the person . . . , resulting
in a reluctance to speak for another person, or to impute feelings to another person’.
The use of evidentials in Pastaza Quichua (Nuckolls 2008) reflects the distinction
between ‘Self ’ and ‘Other’. It appears that the linguistic expression might help us solve
the puzzle of the ‘opacity’ of other’s minds, in an endeavour to understand ‘other’s’
motivation and to cooperate with each other (in the spirit of Robbins and Rumsey
2008, Rumsey 2008 and Duranti 2008).
It is undoubtedly the case that some categories are particularly open to diffusion
and contact-induced change. Evidentials are a case in point. It appears that the atti-
tudes to precision in communicating one’s information and information source are
as well.
5.3 Why evidentials?

When we look for extralinguistic explanations for linguistic categories, we should
avoid the danger of being circular. Do Tucano or Quechua have an elaborate system
of evidentials because of a cultural requirement to be precise about one’s information
source lest one is accused of sorcery? Or is the explanation the other way round?
At present, all that can be suggested is that some communities in some areas—for
instance, in the Amazonian area, and those in the adjoining Andean region—in some
way share a common set of beliefs, mental attitudes, and behavioural conventions, as
well as discourse genres; and that these are compatible with the independent devel-
opment of evidential systems with their requirement to be as precise and as specific
as possible about information source. This could help explain why evidentiality has
independently evolved in at least six (possibly, more) places in Amazonia, and also
why it is so susceptible to being diffused in their language contact.
Knowledge correlates with power and control. Storch and Coly (§5 of Chapter 9)
hypothesize that a tendency towards exclusive control of knowledge among the Maaka
may be responsible for the development of its complex system of evidentials and epis-
temics. The requirement to be precise, and the importance of expressing oneself well,
appear to be a major motivation for having evidentials in one’s language. As Eberhard
(2009: 469) puts it,
The avoidance of being wrong is intrinsically related to the avoidance of losing face. The entire
Mamaindê evidentiality system, then, may have the larger social function of providing the
speaker with a way to avoid losing face within a society where one’s words are connected to
one’s character.
6 About this volume

We aim at a cross-linguistic overview of the gamut of epistemological devices across a
selection of languages in terms of parameters and issues outlined in this chapter. The
volume contains in-depth discussion of twelve languages, from a variety of families
and areas. None of them have been previously described with regard to grammatical
and other expressions of knowledge, and its sociolinguistic status. Some languages
included here have evidentiality systems, others do not. Each chapter systematically
addresses grammatical and other deviecs involved in the expression of knowledge
and information source, and their cultural and sociolinguistic features, and practices.
Eight of the twelve languages described here have grammatical evidentiality. Lan-
guages with small systems of grammatical evidentials are discussed first. Chapter 2, by
Diana Forker, addresses ‘The grammar of knowledge in Hinuq’, a Nakh-Daghestanian
(north-east Caucasian) language with a non-firsthand (or non-witnessed) evidential
contrasted to evidentiality-neutral terms. Chapter 3, by Teija Greed, focuses on evi-
dentials and epistemic expressions in Tatar (‘The expression of knowledge in Tatar’).
Evidentiality in Tatar, a Turkic language, is of the same type as that of Hinuq. Both lan-
guages have a plethora of other grammatical and non-grammatical means of express-
ing knowledge, including speech reports and quotes. Saaroa, an obsolescent Formosan
language, is discussed by Chia-jung Pan in Chapter 4 (‘The grammar of knowledge in
Saaroa’). This language has just one reported evidential. Pan addresses the intricacies
of its use, and other means of referring to how one knows things.
We then turn to languages with more elaborate systems of grammatical evidential-
ity. Chapter 5, by Gwendolyn Hyslop, ‘The grammar of knowledge in Kurtöp: eviden-
tiality, mirativity, and expectation of knowledge’, focuses on a highly complex system
of intertwined information source, attitude to knowledge, and whether or not it is
unexpected to the speaker or the audience. This is followed by Sihong Zhang’s discus-
sion of ‘The expression of knowledge in Ersu’, another Tibeto-Burman language, with
a more straightforward three-term system of evidentials, and a plethora of evidential-
ity strategies and other means of expressing source of information and its reliability.
Kalmyk, a Mongolic language from Central Russia, is discussed, by Elena Skribnik
and Olga Seesing, in Chapter 7 (‘Evidentiality in Kalmyk’). This discussion reveals a
highly complex system of grammatical marking of information source with a basic
distinction between direct and indirect evidentials, with an additional option to spec-
ify inference, assumption, and reported information in various tenses and aspects
(including the future). The time of inference does not have to be the same as the time
of the actual event; this is also reflected in the evidential system of Kalmyk.
Information source may be encoded just within a noun phrase. Dyirbal, an Aus-
tralian language, analysed by R. M. W. Dixon in Chapter 8 (‘The non-visible marker
in Dyirbal’) is a prime example of this. In Chapter 9, ‘The grammar of knowledge
in Maaka (Western Chadic, Nigeria)’, Anne Storch and Jules Jacques Coly address a
complicated system of expressing information source within a noun phrase, and also
within a clause. This is in addition to further evidentiality strategies with their own
epistemic overtones.
How do languages without grammatical evidentiality express knowledge? In Chap-
ter 10, ‘Expression of information source meanings in Ashéninka Perené (Arawak)’,
Elena Mihas offers a detailed discussion of evidentiality strategies, parentheticals, and

other knowledge-related expressions in the language. In his ‘Nominalization, knowl-
edge, and information source in Aguaruna’ (Chapter 11), Simon E. Overall discusses
various ways of expressing knowledge in this language from the small Jivaroan family
in the Andean foothills. These include a narrative modality, and nominalizations. In
Aguaruna- and in Ashéninka Perené-speaking communities, being precise in stating
one’s information source is a cultural requirement. These languages make do with
numerous devices, none of which are fully grammaticalized. Tima, from the small
Katla-Rashad group in the Nuba mountains of the Sudan, does not have grammatical-
ized evidentials. In Chapter 12, ‘The grammar of knowledge in Tima’, Gerrit J. Dim-
mendaal looks at the ways in which knowledge, its sources and speakers’ attitudes
to it can be expressed through various constructions, including ventive, logophoric
markers, speech reports, and ideophones.
Chapter 13, ‘Saying, seeing, and knowing among the Karawari of Papua New Guin-
ea’, is somewhat different from the rest. Correlations between linguistic expression of
knowledge and the conceptualization of its value in a changing society are an impor-
tant issue to address. The Sepik region of Papua New Guinea is renowned for its focus
on the value of knowledge in its varied guises. This value easily translates into mon-
etary terms: as shown by Harrison (1990), words, spells, and other pieces of knowledge
can be bought and sold. Karawari is a highly synthetic language from Lower Sepik
family. Similarly to other languages of the region, there are no grammatical eviden-
tials. And yet, the ways of talking about knowledge, its sources and reliability, are
highly elaborate. Based on more than twenty-five years of work among the Karawari,
Borut Telban, an eminent anthropologist, explores the nature and the expression of
knowledge, and concomitant changes in recent years.
Evidentials as closed grammatical systems are different from information source
marked in other ways (just like time, a real-life concept, is different from tense, real-
ized in grammar). Meanings related to information source may be expressed through
open classes of verbs (of perception, opinion, speech, and others), adverbs, and
parentheticals.
These tend to be richer in their semantic range than closed systems of grammatical
evidentials. Information source may be expressed via a closed subclass of modal verbs,
or via particles (often grammaticalized from verbs). They are much closer to gram-
matical evidentials in their nature, and their meanings.
As Heine (1997: 14–15) put it:
the way people in Siberia or the Kalahari Desert experience the world around them can imme-
diately be held responsible for the way they shape their grammars. Although conceptualisation
strategies are perhaps the main driving force for linguistic categorisation, conceptualisation is
not the only force that can be held responsible for why grammar is structured the way it is. . . .
Another, equally important, force is communication.
This volume spans languages with and without evidentiality systems, focusing on
knowledge across the borders of grammar and lexicon. How do communicative prac-
tices shape the expression of knowledge through grammatical and other means? This
is what the volume is about.
Appendix. On terminological clarity

The term ‘evidential’ primarily relates to information source as a closed grammatical
system. The term ‘information source’ relates to the corresponding conceptual cat-
egory. This is akin to the distinction between the category of ‘tense’, as grammaticalized
location in time, and the concept of ‘time’. This was addressed in §1 of the main chapter.
The view of evidentiality which goes back to Boas’ work, and is followed here, is that
it is a grammatical category in its own right with information source as its primary
meaning. Talking about ‘lexical evidentiality’ is unhelpful. It would be similar to refer-
ring to words like today and tomorrow in English as ‘lexical tense’.
Evidentiality is a verbal grammatical category in its own right. It does not bear
any straightforward relationship to truth, the validity of a statement, or the speak-
er’s responsibility. Evidentials may have extensions to do with certainty, uncertainty,
probability, doubt, and commitment or lack thereof. But the presence of such exten-
sions does not make evidentials into ‘modals’, a subcategory of epistemic or any other
modality, nor of irrealis. This can be compared to gender systems: in many languages
feminine gender is associated with diminution, or endearment (see numerous exam-
ples in Aikhenvald 2000), and masculine gender with augmentative. This however
does not mean that gender is a type of diminutive or augmentative category.
Evidentiality does not offer ‘justification’ for a statement, nor ‘evidence’ (as one expects
in a court). Neither does a tense on a verb offer ‘justification’, or evidence for some-
thing being done in a particular time frame. Gender marking and agreement in Indo-
European languages is not a means of ‘justification’ for the existence of men and women.
We can now summarize a number of misconceptions concerning evidentials.
Some misconceptions concerning evidentials
1. Evidential marking provides justification for a statement: wrong.

2. An evidential reflects attitude to evidence: wrong.
3. Evidentiality is a type of modality, mood, or aspect: wrong.
4. Evidentiality is universal, because every language has a way of expressing how
one knows things: wrong.
5. If a language has verbs meaning ‘see’, ‘hear’, and ‘smell’, it has evidentiality:
wrong.
6. If a language has a way of saying ‘probably’, it has evidentiality: wrong.
7. Evidentiality is a gradient category: wrong.
8. Evidentiality is the same as evidence: wrong, just as grammatical gender is

not the same as biological sex.
9. Speakers of languages with evidentials have to always tell the truth: wrong.
10. Languages with evidentials divide into those where evidentials have epistemic
extensions and those where they do not: wrong.
References
Aikhenvald, Alexandra Y. 1998. ‘Warekena’, pp. 215–439 of Handbook of Amazonian languages,
Volume 4, edited by Desmond C. Derbyshire and Geoffrey K. Pullum. Berlin: Mouton de
Gruyter.
———. 2000. Classifiers: A typology of noun categorization devices. Oxford: Oxford University
Press.
———. 2002. Dicionário Tariana-Português e Português-Tariana. Boletim do Museu Paraense
Emílio Goeldi, 17: 1.
———. 2003a. ‘Evidentiality in typological perspective’, pp. 1–31 of Aikhenvald and Dixon 2003.
———. 2003b. A grammar of Tariana, from northwest Amazonia. Cambridge: Cambridge Uni-
versity Press.
———. 2004. Evidentiality. Oxford: Oxford University Press.
———. 2006. ‘Evidentiality in grammar’, pp. 320–5 of Encyclopedia of language and linguistics,
2nd edition, Volume 4, edited by Keith Brown. Oxford: Elsevier.
———. 2007. ‘Classifiers in multiple environments: Baniwa of Içana/Kurripako: a North Arawak
perspective’, International Journal of American Linguistics 27: 475–500.
———. 2008. ‘Information source and evidentiality: what can we conclude?’, Rivista di Lin-
guistica 19: 207–27, Mario Squartini (ed.). Special issue on evidentiality between lexicon and
grammar.
———. 2010. Imperatives and commands. Oxford: Oxford University Press.
———. 2011a. ‘Evidentiality’. Oxford Bibliography Online. New York: Oxford University Press.
———. 2011b. ‘The grammaticalization of evidentiality’, pp. 602–10 of The Oxford Handbook of
Grammaticalization, edited by Bernd Heine and Heiko Narrog. Oxford: Oxford University
Press.
———. 2011c. ‘Speech reports’, pp. 290–326 of Language at large, by Alexandra Y. Aikhenvald
and R. M. W. Dixon. Leiden: Brill.
———. 2011d. ‘The wonders of the Chacos: setting the scene’, Revista Indiana 28: 171–81.
———. 2012a. Languages of the Amazon. Oxford: Oxford University Press.
———. 2012b. ‘Review of Diewald and Smirnova (2010)’, Studies in Language 36(2): 431–9.
———. 2012c. ‘The essence of mirativity’, Linguistic Typology 16: 435–85.
———. 2013. ‘The value of language and the language of value: A view from Amazonia’, HAU
3(2): 55–73.
———and Dixon, R. M. W. (eds). 2003. Studies in evidentiality. Amsterdam: John Benjamins.
———and ———. 2011. Language at large. Essays on syntax and semantics. Leiden: Brill.
———and Storch, Anne. 2013. ‘Linguistic expression of perception and cognition: A typological
glimpse’, pp. 1–46 of Linguistic expression of perception and cognition, edited by Alexandra
Aikhenvald and Anne Storch. Leiden: Brill.
Alcazar, Asier. 2010. ‘Information source in Spanish and in Basque: A parallel corpus study’,
pp. 131–56 of Diewald and Smirnova (2010).
———. forthcoming. ‘On the scope of evidentials and the lexico-grammatical divide’.
Babel, Anna. 2009. ‘Dizque, evidentiality and stance in Valley Spanish’, Language in Society 38:
487–511.
Basso, Ellen B. 1990. ‘The last cannibal’, pp. 133–53 of Native Latin American cultures through their
discourse, edited by Ellen B. Basso. Bloomington: Indiana University.
———. 1995. The last cannibal. Austin: University of Texas Press.
Besnier, Nico. 1992. ‘Reported speech and affect on Nukulaelae atoll’, pp. 161–81 of Hill and
Irvine 1992.
Blass, Regina. 1989. ‘Grammaticalization of interpretive use: The case of rein Sissala’, Lingua
79: 299–326.
Boas, Frans. 1911. ‘Kwakiutl’, pp. 423–557 of Handbook of American Indian languages,
Part 1, edited by Franz Boas. Smithsonian Institution, Bureau of American Ethnology
Bulletin 40.
———. 1938. ‘Language’, pp. 124–45 of General anthropology, edited by Frans Boas. Boston, New
York: D. C. Heath and Company.
———. 1942. ‘Language and culture’, pp. 178–84 of Studies in the history of culture: The disciplines
of the humanities. Menasha: The George Banta Publishing Co.
Botne, Robert. 2003. ‘Lega (Beya dialect) (D25)’, pp. 422–49 of The Bantu languages, edited by
Derek Nurse and Gérard Philippson. London: Routledge.
Boyeldieu, P. 1982. Deux études laal (Moyen-Chari, Tschad). Berlin: Verlag von Dietrich Reimer.
Bradac, James J., Mulac, Anthony, and Thompson, Sandra A. 1995. ‘Men’s and women’s use
of intensifiers and hedges in problem-solving interaction: molar and molecular analysis’,
Research on Language and Social Interaction 28: 93–116.
Carlin, Eithne B. 2004. A grammar of Trio, a Cariban language of Suriname. Frankfurt am Main:
Peter Lang.
Carol, Javier J. 2011. ‘Determinantes demonstrativos en chorote (mataguayo). Interrelación con
la modalidad, la temporalidad y la evidencialidad’, Indiana 28: 227–54.
Chafe, Wallace L. and Nichols, Johanna (eds). 1986. Evidentiality: The linguistic coding of episte-
mology. Norwood, NJ: Ablex.
Chirikba, Viacheslav. 2003. ‘Evidential category and evidential strategy in Abkhaz’, pp. 243–72
of Aikhenvald and Dixon 2003.
Comrie, Bernard. 1976. Aspect. Cambridge: Cambridge University Press.
———. 1985. Tense. Cambridge: Cambridge University Press.
———. 2000. ‘Evidentials: Semantics and history’, pp. 1–12 of Johanson and Utas 2000.
Cornillie, Bert. 2008. ‘The continuum between lexical and grammatical evidentiality: A func-
tional analysis of Spanish parecer’, Rivista di Linguistica 19: 109–28.
de Haan, Ferdinand. 1999. ‘Evidentiality and epistemic modality: Setting boundaries’, Southwest
Journal of Linguistics 18: 83–102.
———. 2005. ‘Semantic distinctions of evidentiality’, pp. 318–21 of World atlas of language struc-
tures, edited by Martin Haspelmath, Matthew S. Dryer, David Gil, and Bernard Comrie.
Oxford: Oxford University Press.
Dehé, N. and Kavalova, Y. 2007. Parentheticals. Amsterdam: John Benjamins.
Deibler, E. W. 1971. ‘Uses of the verb ‘to say’ in Gahuku’, Kivung 4: 101–10.
DeLancey, Scott. 2001. ‘The mirative and evidentiality’, Journal of Pragmatics 33: 369–82.
Dendale, Patrick. 1993. ‘Le conditionnel de l’information incertaine: marqueur modal ou mar-
queur évidentiel?’ pp. 165–76 of Proceedings of the XXe Congrès International de Linguistique
et Philologie Romanes, Tome I, Section I. La phrase, edited by Gerold Hilty. Tübingen: Francke.
———and Van Bogaert, Julie. 2007. ‘A semantic description of French lexical evidential markers
and the classification of evidentials’, Rivista di Linguistica 19: 65–90.
Diewald, Gabriele, and Smirnova, Elena. 2010. Linguistic realization of evidentiality in European
languages. Berlin: De Gruyter Mouton.
Dimmendaal, Gerrit J. 2001. ‘Logophoric marking and represented speech in African languages
as evidential hedging strategies’, Australian Journal of Linguistics 21: 131–57.
Dixon, R. M. W. 1972. The Dyirbal language of North Queensland. Cambridge: Cambridge Uni-
versity Press.
———. 1994. Ergativity. Cambridge: Cambridge University Press.
———. 2002. Australian Languages: Their nature and development. Cambridge: Cambridge Uni-
versity Press.
———. 2003. ‘Evidentiality in Jarawara’, pp. 165–87 of Aikhenvald and Dixon 2003.
———. 2004. The Jarawara language of southern Amazonia. Oxford: Oxford University Press.
———. 2005. A semantic approach to English grammar. Oxford: Oxford University Press.
———. 2010. Basic linguistic theory. Volume 2. Grammatical topics. Oxford: Oxford University
Press.
———and Aikhenvald, Alexandra Y. (eds). 1999. The Amazonian languages. Cambridge: Cam-
bridge University Press.
Dubois, Betty and Crouch, Isabel. 1975. ‘The question of tag questions in women’s speech: They
really don’t use more of them’, Language in Society 4: 289–94.
Duranti, Alessandro. 1990. ‘Politics and grammar: Agency in Samoan political discourse’, Amer-
ican Ethnologist 17: 646–66.
———. 2008. ‘Further reflections on reading other minds,’ Anthropological Quarterly 81: 483–94.
———and Goodwin, Charles. 1992. Rethinking context. Language as an interactive prenomenon.
Cambridge: Cambridge University Press.
Eberhard, David. 2009. A grammar of Mamaindê: A northern Nambikwara language and its
cultural context. Amsterdam: LOT.
Evans, Nicholas and Wilkins, David. 2000. ‘In the mind’s ear: The semantic extensions of per-
ception verbs in Australian languages’, Language 76: 546–92.
Fleck, David. 2007. ‘Evidentiality and double tense in Matses,’ Language 83: 589–614.
Floyd, Rick. 1999. The structure of evidential categories in Wanka Quechua. Summer Institute of
Linguistics and University of Texas, Arlington.
Floyd, Simeon. 2005. ‘The poetics of evidentiality in South American storytelling’, Santa Bar-
bara Papers in Linguistics 16: 28–41.
Fortescue, Michael. 2003. ‘Evidentiality in West Greenlandic: A case of scattered coding’, pp.
291–306 of Aikhenvald and Dixon 2003.
Gerzenstein, Anna. 1994. Lengua maká: Estudio descriptivo. Archivo de Lenduas Indoameri-
canas. Buenos Aires: Universidad de Buenos Aires, Facultad de Filosofía y Letras, Instituto
de Lingüística.
Giacalone Ramat, Anne and Manana Topadze. 2008. ‘The coding of evidentiality: A compara-
tive look at Georgian and Italian’, Rivista di Linguistica 19: 109–28.
Grice, P. 1989. Studies in the Way of Words. Cambridge, MA: Harvard University Press.
Gronemeyer C. 1997. ‘Evidentiality in Lithuanian’, Working Papers 46: 93–112, Lund University,
Department of Linguistics.
Grunow-Hårsta, Karen. 2007. ‘Evidentiality and mirativity in Magar’, Linguistics of the Tibeto-
Burman Area 30(2): 151–94.
Hardman, Martha J. 1986. ‘Data-source marking in the Jaqi languages’, pp. 113–36 of Chafe and
Nichols 1986.
Harrison, Simon J. 1990. Stealing people’s names. History and politics in a Sepik River cosmology.
Heine, Bernd. 1997. Cognitive foundations of grammar. Oxford: Oxford University Press.
Hewitt, B. George. 1995. Georgian grammar. Amsterdam: John Benjamins.
Hieda, Osamu. 2012. ‘Complementation in Kuman’, pp. 93–124 of Descriptive studies of Nilotic
morphosyntax, edited by Osamu Hieda. Studies in Nilotic linguistics, vol. 2. Tokyo: Research
Institute for Language and Culture of Asia and Africa.
Hill, Jane H. and Irvine, Judith T. (eds). 1992. Responsibility and evidence in oral discourse. Cam-
bridge: Cambridge University Press.
Jacobsen, William H. Jr. 1986. ‘The heterogeneity of evidentials in Makah’, pp. 3–28 of Chafe
and Nichols 1986.
Jakobson, Roman O. 1957. ‘Shifters, verbal categories, and the Russian verb,’ Cambridge, MA:
Harvard University.
Johanson, Lars. 2003. ‘Evidentiality in Turkic’, pp. 273–91 of Aikhenvald and Dixon 2003.
———and Utas, B. (eds). 2000. Evidentials. Turkic, Iranian and neighbouring languages. Berlin:
Mouton de Gruyter.
Kaltenböck, Gunther, Heine, Bernd, and Kuteva, Tania. 2011. ‘On thetical grammar’, Studies in
Language 35: 852–97.
Kany, Charles. 1944. ‘Impersonal dizque and its variants in American Spanish’, Hispanic Review
12: 168–77.
Keenan, Edward L. and Ochs, Elinor. 1979. ‘Becoming a competent speaker of Malagasy’,
pp. 113–58 of Languages and their speakers, edited by Timothy Shopen. Cambridge, MA:
Winthrop.
Kirsner R. S. and Thompson, Sandra A. 1976. ‘The role of pragmatic inference in semantics: A
study of sensory verb complements in English’, Glossa 10: 200–40.
König, Christa. 2013. ‘Source of information and unexpected information in !Xun—evidential,
mirative and counterexpectation marker’, pp. 69–94 of Linguistic expression of perception and
cognition, edited by Alexandra Aikhenvald and Anne Storch. Leiden: Brill.
Kracke, Waud H. 2010. ‘Dream as deceit, dream as truth: The grammar of telling dreams’,
Anthropological Linguistics 51: 61–77.
Kroskrity, Paul V. 1993. Language, history, and identity. Ethnolinguistic studies of the Arizona
Tewa. Tucson and London: The University of Arizona Press.
Kuipers, Joel C. 1992. ‘Obligations to the word: Ritual speech, performance, and reponsibility
among the Weyewa’, pp. 88–104 of Hill and Irvine 1992.
Lacan, Jacques. 1988. The ego in Freud’s theory and in the technique of psychoanalysis, 1954–1955.
New York: W. W. Norton.
Lazard, Gilbert 1957. ‘Caractères distinctifs de la langue tadjik’, Bulletin de la Société de Linguis-
tique de Paris 52: 117–86.
———. 1999. ‘Mirativity, evidentiality, mediativity, or other?’, Linguistic Typology 3: 91–110.
———. 2001. ‘On the grammaticalization of evidentiality,’ Journal of Pragmatics 33: 359–67.
Lidz, Liberty A. 2007. ‘Evidentiality in Yongning Na (Mosuo)’, Linguistics of the Tibeto-Burman
Area 30: 45–88.
Lindstrom, Lamont. 1992. ‘Context contests: Debatable truth statements on Tanna (Vanuatu)’,
pp. 77–99 of Rethinking context. Language as an interactive phenomenon, edited by Alessan-
dro Duranti and Charles Goodwin. Cambridge: Cambridge University Press.
Lowe, Ivan. 1999. ‘Nambiquara’, pp. 269–92 of Dixon and Aikhenvald 1999.
MacDonald, Lorna. 1990. ‘Evidentiality in Tauya’, Language and Linguistics in Melanesia 21:
31–46.
McLendon, Sally. 2003. ‘Evidentials in Eastern Pomo with a comparative survey of the category
in other Pomoan languages’, pp. 101–30 of Aikhenvald and Dixon 2003.
Maslova, Elena. Ms. ‘A universal constraint on sensory lexicon, or when hear can mean “see”.’
Meira, Sérgio. 1999. ‘A Grammar of Tiriyó’. PhD dissertation, Rice University.
Michael, Lev D. 2008. ‘Nanti Evidential Practice: Language, Knowledge and Social Action in an
Amazonian Society’. PhD dissertation, University of Texas at Austin.
Migliazza, Ernest C. 1972. ‘Yanomama Grammar and Intelligibility’. PhD thesis, Indiana
University.
Miller, Cynthia L. and Gilley, Leona G. 2007. ‘Evidentiality and mirativity in Shilluk’, pp. 191–206
of Proceedings of the 8th International Nilo-Saharan Linguistics Colloquium, edited by Mech-
thild Reh and Doris L. Payne. Cologne: Rüdiger Köppe.
Monserrat, Ruth and Dixon, R. M. W. 2003. ‘Evidentiality in Mỹky’, pp. 237–42 of Aikhenvald
and Dixon 2003.
Neukom, L. 2001. Santali. Munich: Lincom Europa.
Nordlinger, Rachel and Sadler, Louise. 2004. ‘Nominal tense in cross-linguistic perspective’,
Language 80: 776–806.
Nuckolls, Janis B. 2008. ‘Deictic selves and others in Pastaza Quichua evidential usage’, Anthro-
pological Linguistics 50: 67–89.
Olbertz, Hella. 2005. ‘Dizque en el español andino ecuatoriano: conservador e innovador’, pp.
77–94 of Encuentros y conflictos. Bilingüismo y contacto de lenguas en el mundo andino, edited
by Hella Olbertz and Pieter Muysken. Madrid: Iberoamericana.
———. 2008. ‘Dizque in Mexican Spanish: The subjectification of reportative meaning’, Rivista
di Linguistica 19: 151–72.
Overall, Simon. 2008. ‘A Grammar of Aguaruna’. PhD thesis, La Trobe University.
Palmer, F. R. 1986. Mood and modality. Cambridge: Cambridge University Press.
Palmer, Gary. 1996. Towards a theory of cultural linguistics. Austin: University of Texas Press.
Pan, Chia-jung. 2010. ‘Linguistic expression of perception and cognition in Lha’alua and Tsou:
A preliminary study’. Paper presented at a Workshop ‘Linguistic expression of perception
and cognition’, LCRG, CI, JCU.
Pietrandrea, Paola. 2008. ‘The grammatical nature of some epistemic-evidential adverbs in spo-
ken Italian’, Rivista di Linguistica 19: 39–64.
Plungian, Vladimir A. 2001. ‘The place of evidentiality within the universal grammatical space’,
Journal of Pragmatics 33: 349–58.
Pusch, Claus D. 2008. ‘Is there evidence for evidentiality in Gascony Occitan?’, Rivista di Lin-
guistica 19: 91–108.
Ramirez, Henri. 1997. A Fala Tukano dos Yepâ-masa. Tomo I. Gramática. Manaus: Inspetoria
Salesiana.
Reesink, G. P. 1986. Structures and their functions in Usan, a Papuan language of New Guinea.
Amsterdam: John Benjamins.
de Reuse, Willem J. 2003. ‘Evidentiality in Western Apache’, pp. 79–100 of Aikhenvald and
Dixon 2003.
Robbins, Joel and Rumsey, Alan. 2008. ‘Cultural and linguistic anthropology and the opacity of
other minds’, Anthropological Quarterly 81: 407–20.
Robins, R. H. 1967. A short history of linguistics. London: Longmans.
Rumsey, Alan. 2008. ‘Confession, anger and cross-articulation in Papua New Guinea’, Anthro-
pological Quarterly 81: 455–72.
San Roque, Lila and Loughnane, Robyn. 2012. ‘The New Guinea Highlands evidentiality area’,
Linguistic Typology 16: 111–67.
Sapir, Edward. 1921. Language. New York: Harcourt, Brace and World.
Smeets, Ineke. 2007. Mapuche. Berlin: Mouton de Gruyter.
Squartini, Mario. 2007. ‘Lexical vs grammatical evidentiality in French and Italian’, Linguistics
46: 917–47.
Storch, Anne. 1999. Das Hone und seine Stellung im Zentral-Jukunoid. Cologne: Rüdiger Köppe.
———. 2011. Secret manipulations. The language of marginality in Africa. New York: Oxford
University Press.
———. 2013. ‘Knowing, smelling and telling tales in Luwo’, pp. 47–68 of Linguistic expression of
perception and cognition, edited by Alexandra Aikhenvald and Anne Storch. Leiden: Brill.
Thompson, Sandra A. and Mulac, Anthony. 1991. ‘A quantitative perspective on the grammatici-
zation of epistemic parentheticals in English’, pp. 313–29 of Approaches to grammaticalization,
Volume II, edited by Elizabeth C. Traugott and Bernd Heine. Amsterdam: John Benjamins.
Timberlake, Alan. 1982. ‘The impersonal passive in Lithuanian’, Proceedings of the Eighth Annual
Meeting of the Berkeley Linguistics Society, 508–23.
Trask, R. L. 1991. A dictionary of grammatical terms in linguistics. London: Routledge.
Traugott, Elizabeth C. 1996. ‘Subjectification and the development of epistemic meaning: The
case of promise and threaten’, pp. 185–210 of Modality in Germanic languages: Historical
and comparative perspectives, edited by Toril Swan and Olaf J. Westvik. Berlin: Mouton de
Gruyter.
Travis, Catherine. 2006. ‘Dizque: A Colombian evidentiality strategy’, Linguistics 44: 1269–97.
Tung, Tung-ho. 1964. A descriptive study of the Tsou language, Formosan. Taipei: Institute of
History and Philology, Academia Sinica.
Urmson, J. O. 1952. ‘Parenthetical verbs’, Mind 61: 480–96.
Valenzuela, Pilar. 2003. ‘Evidentiality in Shipibo-Konibo, with a comparative overview of the
category in Panoan’, pp. 33–62 of Aikhenvald and Dixon 2003.
van der Auwera, Johan and Plungian, Vladimir A. 1998. ‘On modality’s semantic map’, Linguistic
Typology 2: 79–124.
Waag, C. 2010. The Fur verb and its context. Cologne: Rüdiger Köppe.
Weber, David J. 1986. ‘Information perspective, profile, and patterns in Quechua’, pp. 137–55 of
Chafe and Nichols 1986.
Westermann, D. 1911. A short grammar of the Shilluk language. Berlin: Reimer Verlag.
Wheatley, J. 1973. ‘Knowledge, authority and individualism among the Cura (Bacairi)’, Anthro-
pological Linguistics 15: 337–44.
Wiemer, Björn. 2008. ‘Lexical markers of evidentiality in Lithuanian’, Rivista di Linguistica 19:
173–208.
Willett, Thomas. 1988. ‘A cross-linguistic survey of the grammaticization of evidentiality’, Stud-
ies in Language 12: 51–97.
Yang, Gloria Fan-pei. 2000. ‘The semantics-pragmatics of the hearsay evidential nana in Tsou’,
National Taiwan University Working Papers in Linguistics 3: 69–86.
2
The grammar of knowledge in Hinuq
DI A NA FOR K ER
1 Introduction
Hinuq belongs to the Tsezic sub-branch of the Nakh-Daghestanian language fam-
ily. Most of its approximately 600 speakers live in the village of Hinuq in Daghestan
(Russia), in the Caucasian mountains. Hinuq is an oral language used by Hinuq peo-
ple within their own community as the basic means of everyday communication. All
speakers over six years old speak Hinuq and Russian to various degrees. Most people
also have a fair command of other neighbouring Nakh-Daghestanian languages such
as Tsez, Bezhta, and Avar. Avar used to be a lingua franca in the region where Hinuq
speakers live and is still taught at school as ‘mother tongue’. However, during the last
50 years it has lost its important role and status, which has been gained by Russian.
Hinuq has a rich case system, with ergative alignment. The case system includes six
grammatical and thirty-five spatial cases. Gender plays an important role in the grammar
of Hinuq. Nouns belong to five genders. Various parts of speech (many verbs, some adjec-
tives, very few adverbs, and all demonstrative pronouns) show agreement in gender and
number. The verb is marked for gender/number agreement, tense, mood, aspect, and evi-
dentiality, making up a paradigm with more than twenty simple and periphrastic forms.
2 General properties of the evidential system and formal marking

2.1 Verbal evidentiality
Hinuq has a grammaticalized evidentiality system that comprises inflectional suffixes
on verbs and two enclitics (reportative and quotative). In this section, I treat only
the verbal evidentiality system and the reportative enclitic. The quotative enclitic is
discussed in Section (§4).
The verbal evidentiality system is fused with the tense system and restricted to all
verb forms with past time reference. The evidentiality markers cannot be omitted and
they can appear only once per clause. The fused tense and evidentiality suffixes occur
in the same slot as other tense, modality, converbal, and participial suffixes, with which
marking for evidentiality is mutually exclusive. Within the framework developed by
Aikhenvald (2003, 2004), I analyse Hinuq as having an A2 system (see Khalilova (2011)
for a different opinion). That is, it distinguishes unwitnessed (‘non-firsthand’) from
neutral (‘everything else’). All neutral forms have the suffix -s, all non-firsthand forms
the suffix -n. The unwitnessed forms are both formally and functionally more marked
than the neutral past forms (see Forker 2012 for an analysis of the Hinuq evidential
systems in terms of markedness). In synthetic verb forms the suffixes are added to the
lexical verb, in analytic verb forms they appear on the auxiliary zoq’we- ‘be’.
Hinuq has two non-finite verb forms that are cognate with the unwitnessed and the
neutral past forms respectively. The narrative converb, which is heavily used in clause
chaining to advance the plot of a narration and in the formation of the pluperfect, is
formed with a suffix that is formally identical to the unwitnessed past (-n). However,
functionally the two verbal forms are easily distinguishable because the narrative con-
verb lacks an absolute time reference and an evidential value. Diachronically, the suf-
fix has probably first been used for the expression of perfective meaning in adverbial
clauses. This use has been extended to independent clauses by adding the auxiliary
‘be’, which led to periphrastic verb forms formally and functionally resembling per-
fects and pluperfects in other languages. These verbal forms have been employed as
evidentiality strategies, and following the loss of the auxiliary, they became grammati-
calized evidentials. This path of development is very typical for languages of that area
(cf. Friedman 2000 for Balkan languages; Johanson 2000, 2003 for Turkic languages;
Tatevosov 2001, 2007 for some Nakh-Daghestanian languages; Dahl 1985: 152–3; Bybee
et al. 1994: 95–7).
The neutral past suffix is a cognate of the resultative participle suffix -s, but again
both forms can easily be distinguished on semantic grounds. The resultative partici-
ple expresses resultative aspect, but lacks any absolute time reference. The diachronic
scenario is probably similar to that of the narrative converb and the unwitnessed past
suffixes. Creissels (2009) has proposed a similar development for participles in the
Nakh-Daghestanian language Akhvakh.
2.2 Semantics of verbal evidentials

The unwitnessed forms indicate that the described situation or event was not wit-
nessed by the speaker (§2.2.1). The neutral past forms only imply the situation or event
itself, but do not express that it was directly witnessed by the speaker (§2.2.2). Non-
firsthand evidentiality cannot be expressed in anything other than the past tenses, nor
can it be expressed in non-indicative moods and non-finite verb forms employed in
subordinate clauses, although it is possible in interrogative clauses (§2.4).
There is a weak interaction between verbal semantics and evidentiality, basically
restricted to the verb -eti- ‘want’ and occasionally also found with -iči- ‘be, stand,
sit’. The simple unwitnessed past of -eti- lacks the normal temporal and evidential
54 Diana Forker
meaning otherwise found with this category because it has present time reference and
no evidential value.
2.2.1 Unwitnessed evidentials Unwitnessed past forms are used in traditional
narratives such as fairy tales and legends (example (1)). Speakers are highly aware
of this usage of unwitnessed past forms, which is already found among children.
Any use of neutral past tense forms in these contexts is judged as blatantly wrong.
Unwitnessed forms are also employed for anecdotes, stories about the past, and for
all other circumstances when the speaker reports a situation in which s/he was not
present. The full semantic range of unwitnessed evidentials covers:
• situations not seen or otherwise perceived by the speaker;
• inferred information based on perceived results;
• lack of participation and control on part of the speaker.
If the speaker was not present at the situation, s/he might have obtained the relevant
information through the words of others or via reading or other kinds of media.
(1) haɬu-y gotto zoq’e-n mecxer

this:obl-erg pour:imperv be-unwit money
It (i.e. the donkey) did not pour money.
Inferred information is based on present perceivable results of past events. The results
can be visually observed by the speaker, or obtained by means of other senses such
as taste or smell. For instance, sentence (2) was uttered by a speaker who d iscovered
that his gun had been stolen without having seen the burglary. Similarly, in (3) the
speaker reports that when the Hinuq people returned back to Daghestan they found
all their houses destroyed and concluded from the remains that the houses were
burnt down.
(2) pistolet r-iɣ-no

gun(v) v-take-unwit
(My) gun was stolen.
(3) eli hadi-š gučibiq-o�’o, r-ek’we-r-no

1pl here-abl resettle:hpl-sim nhpl-burn-caus-unwit
When we resettled from here, (all houses) were burnt down.
Usually, unwitnessed past forms are not used with the first person in assertions. If the
first person is combined with unwitnessed past forms, then the speaker expresses his/
her lack of control of the situation, for instance, if s/he was drunk or otherwise lack-
ing awareness of his or her actions. In example (4) the speaker describes an event that
happened in his childhood, about 50 years ago. He does not remember the event very
well, during which he almost lost consciousness when he was lying on the ground after
falling from a high tree.
(4) hibayi-š Ø-i�’i-n c’ox-oru, de q’idi ħuħel=no

there-abl i-fall-convb fall-partic 1sg down breath=and
šit’-no �exwe-n de
close-convb remain-unwit 1sg
When I fell down from there, I remained without breathing.
The use of unwitnessed past forms in this context creates mirative overtones, especially
when there is something mysterious or unexplainable in the event, or when the speaker
assumes that a higher entity like God or the devil was involved (5).1 Thus, the Hinuq
unwitnessed past behaves in a very similar way to non-firsthand evidentials of many
other languages of the world in acquiring mirative overtones (cf. Greed, Chapter 3, on
Tatar; Skribnik and Seesing, Chapter 7, on Kalmyk).
(5) hibayɬi-š=no de Allahli c’unzi Ø-u:-n

there-abl=and 1sg Allah.erg save i-do-unwit
And from there Allah also saved me.
2.2.2 Neutral past forms Neutral past forms conventionally imply that the speaker
was an eye-witness of the situation, or perceived it with the appropriate senses, but
under specific circumstances the implication can be cancelled. They are mainly
used in dialogues in natural conversation and in autobiographical narrations (6). In
traditional narratives neutral past forms appear only in reported speech.
(6) uq’ira qura seda e�a, q’ono e�a

four:obl twenty:obl one:obl ord two ord
�eba-ɬ Hinuq r-aɣi-š hor�’oqeb iškola

year:obl-conta Hinuq v-open-past middle school:v
In 1981, 1982 an intermediate school was opened in Hinuq.
Occasionally, neutral past tenses occur when speakers report events that they either
did not attend at all or in which they participated, but not as conscious agents. Such
events represent well-known facts mostly belonging to the personal knowledge sphere
of the speaker, as for instance their own birthday or the forced resettlement of the
1
This use of the unwitnessed past resembles the use of non-visible markers in the description of actions
carried out by spirits in Australian languages (cf. Dixon on Dyirbal, Chapter 8, this volume).
56 Diana Forker
Hinuq people to Chechnya in 1944 and back to Daghestan in 1957. If in a sentence such
as (7) a speaker uses an unwitnessed past form, this means that s/he thought that s/he
was born on another day or in another place and then afterwards just discovered that
s/he in fact was born in 1943 in the village of Hinuq.
(7) de Ø-u:-s Hinuq a�-a ʡazalno ʡač’ino

1sg i-do-past Hinuq village-in 1000 9
bišonno q’ono quno ɬono e�a �eba-ɬ

100 2 20 3 ord year:obl-conta
I (male) was born in the village of Hinuq in 1943.
Note that previous analyses (cf. Khalilova 2011) have considered these verbal forms
as expressing firsthand evidentials and they have been termed ‘witnessed past’. How-
ever, the neutral past forms are used when speakers report events that they did not
consciously experience themselves (7). They are also occasionally employed when
speakers tell anecdotes from the lives of other people with whom they stand in a close
personal relationship (e.g. relatives). These anecdotes describe events not witnessed by
the speaker. For instance, example (8) is from a narration about a famous man in the
village. The speaker did not attend the event he narrated, but he knew the protagonist
himself and therefore uses the neutral past form.
(8) saqu k’al-mo-zo beceː Ø-iš-no gaː-n,

once fasting-obl-gen month:in i-eat-convb drink-convb
hago godek’an-i-r Ø-aq’e -s

3sg:m godekan-in-lat i-come-past
Once during the month of Ramadan he came drunk to the godekan.
Furthermore, neutral past forms are also used for conveying certain types of
encyclopaedic knowledge of which speaker does not have firsthand information
(§5). Finally, neutral past forms are used as default forms in elicitation, that is,
speakers always give neutral past forms when they formulate sentences with past
time reference except for those cases where the context is explicitly described as
unwitnessed.
Thus, the neutral past conveys information belonging to the personal knowledge
sphere of the speaker. In the majority of cases the speaker was in fact a witness of the
events in question, but firsthand evidentiality is just an implicature that arises when
neutral past forms are used. It is not part of the meaning of these verbal forms and can
therefore, under the appropriate circumstances, be cancelled.
2.3 Ways of information acquisition

Neutral past forms are normally employed when the speaker was present at the
reported event, unwitnessed evidentials otherwise. Therefore, unwitnessed eviden-
tials also include information obtained by reading a book, a newspaper, or, nowadays,
texts on the Internet. For instance, after reading a newspaper, one speaker uttered
sentence (9).
(9) huɬ Maskwu-�’o y-aɣi-n memoriyalnaya

yesterday Moscow-spr vi-open-unwit memorial
q’il šaʡir rasul ħamzatow-ez

board:iv poet Rasul Gamzatov-dat
Yesterday in Moscow a memorial board for the poet Rasul Gamzatov
was inaugurated.
The situation is different when you get a personal letter from somebody who writes
to you about his activities. If you then recount the activities of the letter-writer you
use neutral past forms. The reason is probably that you have direct information
from the person you are talking about, not mediated through a third party (e.g.
some unknown journalist who wrote an article for a newspaper). Furthermore,
information obtained through personal letters normally implies emotional prox-
imity between the letter-writer and the recipient, that is, the information belongs
to the personal knowledge sphere of the speaker (cf. the discussion in §2.2.2).
When TV and radio news are narrated, speakers usually use unwitnessed past
forms, since it is not the events themselves being broadcast but only a retrospective
account presented by a third person (10). However, this does not seem to involve a
strict rule, because occasionally neutral past forms are possible. Thus, in (11) a speaker
retells a part of the TV news, which consisted not only of the telecaster reading the
news but also of a short film showing the situation in Magadan.
(10) q’ono quno ɬono rek’we ɬuqɬezi Ø-iq-no,

2 20 3 man:i wound i-happen-unwit
uq’ino Ø-uhe-n, oc’eno ɬeno balnicaː-r kezi
4 i-die-unwit 10 4 hospital:in-lat meet
Ø-iq-no
i-happen-unwit
43 men were wounded. Four (men) died, 14 were brought to the hospital.
58 Diana Forker
(11) Magadan-�’o iše y-aq’e-n, hune-be=n

Magadan-spr snow:iv iv-come-convb way-pl=and
r-oc’-no, tok=no gosme ahlu ƛexwe-s

nhpl-cut-convb electricity=and without people remain-past
It snowed in Magadan, the streets were closed, and the people remained with-
out electricity.
It is different if the information comes from real-time broadcasts such as, for
instance, soccer games. In this case it is not important whether the speaker listened
to the real-time broadcast on the radio or whether s/he watched TV, in both cases
neutral past forms are used to retell the event (12). Similarly, when speakers are
presented with audio or video recordings of events and are asked to narrate what
happened, they employ neutral past forms even if the recordings have been made
some time ago (13).
(12) Anži-y žiqu b-eg e�-iš futbol, amma

Anzhi-erg today iii-well play-past soccer:iii but
axir�’odo Dinamo-y haw b-egwe-r-iš

in.the.end Dynamo-erg it iii-lose-caus-past
Anzhi played soccer well, but in the end Dynamo beat them.
(13) ʡoloq-be b-aq’e-n, zikru=n b-uː-n,

youth-pl hpl-come-convb dhikr:iii=and iii-do-convb
b-iš-an=no b-iš-no b-i�’-iš

hpl-eat-redup=and hpl-eat-convb hpl-go-past
The youth came, sang the dhikr, ate and went away.
The expression of fictive events such as the description of films is more open to
variation than the narration of TV, radio, or newspaper news (i.e. real events). It
is common to use verb forms with present-time reference (‘historical present’),
which is also found in other narratives such as fairy tales or anecdotes. If forms
with past time reference are employed, unwitnessed past forms as well as neutral
past forms are admissible. Speakers who use unwitnessed past forms normally
start the narration with a formulaic expression such as ‘Once upon a time there
was . . . ’, found at the beginning of fairy tales and legends. In general, narrations
based on fictive video films show considerable inter-speaker and intra-speaker
variation (cf. Forker 2011).
2.4 Scope properties of verbal evidentials

Although the unwitnessed past suffixes appear only on the predicate of the main
clause, they have the whole sentence in their scope. That is, if the verb in the main
clause is marked with an unwitnessed past form then predicates in subordinate clauses
also have the evidentiality value ‘unwitnessed’, although they do not bear any marking
(see example (3)).
Evidentials do not fall within the scope of negation, that is, what is negated is the
situation or event itself, not the source of the information. Interrogative clauses have
the same distinction between neutral and unwitnessed verb forms. These clauses are
often marked by an interrogative enclitic that marks the focus of the question. In ques-
tions with neutral past tense forms, the interrogative enclitic obligatorily occurs on the
verb in the main clause thereby replacing the suffix -s. Optionally, a second enclitic
may be added to a focused constituent. In questions containing an unwitnessed past
form, the enclitic is optionally added to the unwitnessed past suffix or to the focused
constituent.
The use of the evidentials in questions is based on the speaker’s assumptions of
the knowledge of the addressee. If the speaker assumes that the addressee knows the
answer from first-hand knowledge then s/he uses a neutral past form (14). In contrast,
if s/he assumes that the addressee was not present at the event in question, an unwit-
nessed past tense form occurs (15). Consequently, in questions with second persons,
unwitnessed past forms are not used.
(14) bu�e b-aši=ye? b-aši-š

house:iii iii-find=inter iii-find-past
Did (you) find the house? (I) did.
(15) se deru r-iq-no, hago uži Ø-aši-n=e?

what how v-happen-unwit that boy:i i-find-unwit=inter
What happened, did they find the boy?
2.5 The reportative enclitic

The reportative enclitic =(e)ƛ predominantly occurs in traditional narrations. Al-
though its form as well as its function are somewhat similar to the quotative enclitic,
both morphemes are easily distinguishable (for the quotative enclitic and its relation
to the expression of evidentiality in Hinuq see §4). The reportative enclitic can be com-
bined with all TAM forms, including unwitnessed past forms, non-finite verb forms
and questions, but not with non-indicative verb forms such as the imperative or the
optative. Its origin is not completely transparent, though it may originate from the verb
eƛi- ‘say’. Usually the reportative enclitic appears on the finite verb, but in principle it
can occur on any other constituent or on the predicate of a subordinate clause. It can
60 Diana Forker
never appear twice in the same clause or sentence. The use of the reportative enclitic
marks the sentence as based on hearsay or report, but normally leaves the origin (i.e.
the author) of the information unexpressed. Thus, it is similar to Russian particle mol.
Frequently, the reportative enclitic occurs only in the first introductory sentence of a
narrative (16) and then has scope over the whole narrative.2 But some Hinuq speakers
also use it in every sentence of the story and combine it with neutral past forms (17)
(see also Section 5 for an explanation of this phenomenon).
(16) seda a�-a zoq’we-n=e� ʡoloqanaw dibira

one:obl village-in be-unwit=rep young mullah
In one village there was a young mullah (they say).
(17) seda huneho t’ot’er-iš=e�, k’ox t’ot’er-iš=e�,

one:obl time count-past=rep twice count-past=rep
ɬox t’ot’er-iš=e�
thrice count-past=rep
They counted once; they counted twice; they counted thrice (they say).
As with verbal evidentiality, there is a restriction on the co-occurrence of the reporta-

tive enclitic with first person. Both first person and the reportative enclitic can only
occur together when the speaker wants to express a lack of control or unintended,
unconscious participation. For instance, in (18) a speaker described how he fell from a
high tree and almost died, until somebody undertook mouth-to-mouth resuscitation
and saved his life. He probably learned about this event from other people’s reports.
(18) hoboži Aytalo-y hut-�’o hut=no gor-no,

now Aytalo-erg mouth-spr mouth=and put-convb
ħuħel �eše-n, hibago�’o kur-iš=e� de

breath tear-convb at.that.time throw-past=rep 1sg:erg
muši
breath
Now Aytalo put his mouth on my mouth, breathed and at that time I
(also) breathed (they say).
Another context where the reportative enclitic can occur with a first person is in
reported speech constructions (§4).
2
It seems to be quite common for reportative/narrative markers to have scope over more than one sen-
tence (cf. Greed, Chapter 3; Skribnik and Seesing, Chapter 7; Overall, Chapter 11).
3 Lexical expression of information source

In addition to the grammaticalized evidentiality system Hinuq makes use of a number
of lexical devices that refer to the information source. Lexical and grammatical devices
can be freely combined to enhance the information content.
Hinuq has a number of lexical verbs that express knowledge, perception, doubt,
inference, and assumption. The most important verbs of this class are qebaː- ‘seem’,
urɣezi -iq- ‘think’, pikru -uː- ‘think’,3 boži -iq- ‘believe’, šak(ɬezi) -iq- ‘doubt’, rok’e -aq’e-
‘be sure’, -eq’i- ‘know’, and c’aɬ- ‘get to know’. At least some of these verbs can appear as
parentheticals, not syntactically related to the rest of the utterance (19). In this sentence
the speaker reports an event in which he was personally involved, but since he did not
act consciously, the unwitnessed past is used.
But normally the verbs occur in complement constructions, triggering some formal
marking on the complement clause such as the use of the quotative enclitic as comple-
mentizer and bearing a syntactic relationship to the complement (example (20)). Com-
plement clauses marked with the quotative enclitic behave exactly like quotes in reported
speech constructions, for example, they use a finite verb in the complement that keeps all
its TAM and evidentiality values (see §4 for more information on the syntactic status of
quotes and deictic shifts). For instance, in (20) the predicate in the complement clause
of c’aɬ- ‘get to know’ bears the unwitnessed past suffix because the experiencer argument
of c’aɬ-, the khan, did not see the hunter coming. When the verbs are used as parentheti-
cals, such complementizers are not employed. Another type of parenthetical is postpo-
sitional phrases such as dižo pikru-ho-r ‘in my opinion’ (lit. 1sg:gen thought-iloc-lat).
(19) waħ, diž qebaː-s, nido de Ø-i�’i-yo

wow 1sg:dat seem-past where 1sg i-go-imperv
zoq’e-n?
be-unwit
Wow, it seemed to me, where did I apparently go?
(20) [hago čanaqan Ø-aq’e-n=�en] c’aɬ-o�’o,

this hunter:i i-come-unwit=quot get.to.know-sim
hay ɬo xan-i kekir-ho wazir čanaqan e�-ayaz

that:obl khan-erg send-pres vizier hunter call-purp
When the khan got to know that the hunter came, he sent his vizier to call the
hunter.
3
There does not seem to be any semantic differences between this verb with the meaning ‘think’ and the
previous verb. Both verbs are complex consisting of a loan word from Avar and a Hinuq light verb. Since
-iq- requests the subject-like argument to be in the absolutive, the same requirement applies to urɣezi -iq-.
In contrast, -u:- and pikru -u:- take subject-like arguments in the ergative. Judging from my corpus, urɣezi
-iq- is more commonly used.
62 Diana Forker
Apart from the above mentioned lexical verbs, Hinuq has another verb, -aši- ‘find
(out), come across’, but its syntactic status is unclear. In (21) it occurs as a parentheti-
cal and in (22) as the finite verb of the clause. It is used to express surprise (20) and/or
inferentiality (22). Similar constructions can be found in other Nakh-Daghestanian
languages, for example Archi (Kibrik 1994) and Khwarshi (Khalilova 2009: 231–7),
though in the latter language the construction has an inferential rather than a mira-
tive interpretation.
(21) seda zaman-a-ɬ had k’onc’u, y-aši-n

one:obl time-obl-conta that leg:iv iv-find-convb
diž, y-ixer-no, gor-on, y-i�’i-yo gom

1sg:dat iv-lift-convb put-conc iv-go-imperv be:neg
At one time, surprisingly, the leg that I lifted up, although I placed it (down),
it does not go.
(22) rok’o gosme aže Ø-iɬi Ø-aši-š me

root without tree i-similar i-find-past 2sg
You (male) turned out to be like a tree without roots.
Furthermore, Hinuq has two direct perception verbs that also take complements, -ike-
‘see’ and toq- ‘hear’. Complement constructions with these verbs are structurally paral-
lel to complement constructions with the verbs described above. There are no simplex
verbs such as ‘feel’ and ‘smell’.
Hinuq has two modal verbs, -ese- ‘be possible, probable’ and -aq’e- ‘must’ that
express epistemic modality, that is, propositions that have been judged to be probable
or uncertain (23). In this example, epistemic possibility is also conveyed by means of
the adjective behulew ‘possible’.
(23) hibayɬu-y=no b-e�ži kumak b-uː-n

that:obl-erg=and iii-big help:iii iii-do-convb
b-ese behulew diž

iii-be.probable possible 1sg.dat
It probably helped me a lot.
In addition to its epistemic meaning, -aq’e- ‘must’ can occasionally express inferred
evidentiality (24).
(24) [Hearing music from far away]

ɣeme-�’o berten b-ič-a b-aq’-o=�en
mill-spr wedding:iii iii-be-infin iii-must-pres=quot
There must be a wedding at the mill, (they said).
Hinuq does not have adverbs/adjectives such as apparently, supposedly, allegedly, or

reportedly. The reportative enclitic fulfils basically the same function as adverbs such
as reportedly in other languages.
The lexical means of expressing information source, knowledge, perception, and
epistemic modality can be freely combined with each other and with the grammati-
calized evidentiality system, as long as this does not lead to semantic contradictions.
Examples (21) and (22) above showed the interplay of lexical verbs and grammati-
calized non-firsthand evidentials. In (25) the verb qeba:- ‘seem’ takes a complement
clause marked with the quotative enclitic and the predicate in the complement is the
epistemic modal verb -ese-.
(25) as moči r-ese=�en qebaː-s mesed-li-š

sky earth nhpl-be.probably=quot seem-past gold-obl-gen
The sky and the earth seemed to be golden.
4 Reported speech constructions, the quotative enclitic, and evidentiality

Hinuq has two reported speech constructions. The first construction is a juxtaposition
of a main clause containing a verb of speech and a subsequent unmarked quote. The
second construction, which is far more common, makes use of the quotative enclitic
=ƛen. The quotative enclitic can perhaps be considered to be part of the grammatical
evidentiality system since it is part of a closed grammatical system of marking reported
speech and information source. However, since its primary function is the expression
of reported speech, it is treated separately. As with the reportative enclitic, the precise
origin is unclear. One may suspect that it has been grammaticalized from the narrative
converb of the verb ‘say’, eƛin, but the final stem vowel of the verb is -i and not -e. The
quotative enclitic most commonly appears on the right edge of the quote, which is very
often a verb. The enclitic alone is enough to mark a clause as a quote, but it can also be
combined with a finite verb of speech or with the narrative converb of ‘say’, eƛin, plus
additionally a finite verb of speech.
In addition to its use in reported speech, the quotative enclitic is also employed as a
complementizer with matrix verbs such as ‘think’, ‘seem’, ‘doubt’, ‘believe’, ‘get to know’,
‘let know’, ‘fear’, ‘hate’, ‘request’, ‘try’, ‘remember’, ‘hear’, and others (cf. Forker to appear).
From a syntactic point of view, reported speech constructions bear some similarity
to ordinary complement clauses, but they nevertheless form a special construction
type of their own. Furthermore, the distinction between ‘direct speech’ and ‘indirect
speech’ familiar from many European languages is not applicable to Hinuq. Rather,
64 Diana Forker
Hinuq reported speech constructions show most of the characteristics of direct speech
(e.g. no tense shift, no shift of the spatial or temporal deixis, not special verb forms;
non-indicative verbs forms such as imperatives, interrogatives, optatives, freely occur
in the quote). The only difference that can be found between quotes and ordinary main
clauses is the occasional use of reflexive pronouns to indicate coreference between
the author of the quote and some argument or adjunct in the quote (26). However,
more common than the use of reflexive pronouns is the use of ordinary personal pro-
nouns. In example (26) the reflexive pronoun in the quote could be exchanged by the
first person singular pronoun without any change in meaning. Based on this usage of
reflexive pronouns, it has been argued that reported speech constructions in Hinuq
as well as in the other Tsezic languages represent instances of so-called ‘semi-direct
speech’ (Aikhenvald 2008).
(26) zeru hardezi b-iq-iš=e� “zonde
fox:iii beg iii-happen-past=rep refl:sg:aloc
cadaq �e!”=�en
together let’s.go=quot
The fox begged ‘Let’s go together with me!’
The quotative enclitic can be freely combined with the grammaticalized and the lexical
means of expressing evidentiality. The evidentials in the quote always reflect the status
of the knowledge of the author of the quote. For instance, the first predicate in the
quote of (27) is an unwitnessed past verb form because the author of the quote, Mullah
Nasredin, has not attended the killing of his sacred crow. Then he goes on and says that
he has bought the crow by paying two donkeys, and here a neutral past occurs because
Mullah Nasredin has firsthand knowledge of this event. Example (28) is from an auto-
biographical narration. The speaker remembers an event from his childhood, thus in
the main clause the neutral past appears. The clause to which the quotative enclitic is
added represents the complement of the main clause predicate and is headed by a verb
with an unwitnessed past suffix because the event described in the complement clause
has not been witnessed by the experiencer of the main clause.
(27) Malla Rasadan kutakalda Ø-aː-n ‘di

Mullah:i Nasredin:i strongly i-cry-unwit 1sg:gen
šayix b-uher-no’=�en. ‘di haw q’ono

saint:iii iii-kill-unwit=quot 1sg:gen that 2
omoq’i=n ne�-no b-ux-iš zoq’e-s’=� en

donkey=and give-convb iii-buy-partic be-past=quot
The Mullah Nasredin cried hardly, ‘My saint was (apparently) killed. I gave
two donkeys away and bought it.’
(28) šibaw ɣwede r-iɬi c’oh r-uː-n=�en

every day v-similar steal:v v-do-unwit=quot
toq-o zoq’we-s eluz

hear-imperv be-past 1pl:dat
We heard that every day things were stolen.
The use of the reportative enclitic in quotes is not completely clear, because I have only
two corpus examples. In (29), the reportative enclitic reflects the information source
of the second speaker who claims that according to what she has heard there is no
corn at her place.
(29) ‘Allah, mezza=gon simildi gom’=�en de

Allah at.your.place=TOP corn be:neg=quot 1sg:erg
e�i-yo. ‘gom=e�’
say-pres be:neg=rep
‘Allah, you do not have corn at home’, I said. ‘No (she says.)’
Normally, the quotative enclitic and the reportative enclitic are easily distinguish-
able not only formally but also functionally, since only the former is used in reported
speech and the latter marks information based on hearsay. However, very occasionally,
the quotative enclitic is employed as a hearsay evidential marker in a similar fashion
as the reportative enclitic (30). Such a use of the quotative enclitic is typical for the
East Tsezic languages Bezhta and Hunzib, which do not have an additional reportative
enclitic (cf. Comrie, Forker, and Khalilova, Ms).
(30) busurman din goɬa bac’adaw kanɬi=�en

moslem belief be:partic clean light=quot
Allahli ne�-iš [. . .] insan-i-ž

Allah:erg give-past person-obl-dat
The clean light of the Islamic religion Allah gave to the humans, they say.
5 Cultural conventions
Apart from the correct use of evidentials, there are no special cultural conventions
to be precise or vague in the information source. The evidentiality system of Hinuq
is very salient for the speakers and they readily comment on the correct use of it (see
also Nichols 2011: 243 for similar observations on Ingush, another Nakh-Daghestanian
language). For some speech genres, the evidentiality system does not play any role
(e.g. proverbs, sayings, procedurals texts) because in these genres past tense forms
are normally not employed. In dialogues and monologues such as autobiographies,
66 Diana Forker
anecdotes, and accounts about life in former times and the history of the Hinuq people
all the described linguistic devices of marking information source occur in the expect-
ed manner. Sometimes neutral and unwitnessed forms are, in principle, both possible
for referring to events for which the speaker was not an eye-witness, but the use of
unwitnessed forms highlights the secondhand source of the information whereas the
use of the neutral past expresses that the described event is well-known knowledge
belonging to the speaker’s general sphere of information (31).
(31) Lenin Ø-uhe-n / Ø-uhe-s ʡazalno ʡač’ino bišonno

Lenin:i i-die-unwit / i-die-past 1000 9 100
qura uq’ira e�a �eba-ɬ

twenty:obl four:obl ord year:obl-conta
Lenin died in 1924.
In traditional folk stories and legends neutral past forms are, except from reported
speech, only used in combination with the reportative enclitic, and switching back and
forth between these two stylistic devices is quite common. A bare use of the neutral
past form as it has been observed for Tsez, Hinuq’s closest neighbouring language, is
not possible (Comrie and Polinsky 2007). According to my informants, the combina-
tion of the neutral past form and the reportative enclitic is a way of achieving liveliness
and making the narrative more vivid. The neutral past form implies firsthand infor-
mation, but this implication is overridden by the meaning of the reportative enclitic.
Such usage can be compared to the historical present that is also used to make past
events present. However, the reportative enclitic not only corrects the implications of
the neutral past tenses, but signals at the same time a certain narrative style and that
the text belongs to the genre of traditional narratives.
Neutral past forms are used when speakers report internal states such as emotions
or pain and when talking about dreams (cf. Greed, Chapter 3, notes the same for Tatar).
Dreams are described as being ‘seen’ during sleep.
When Hinuq speakers retell information obtained through new media (radio, TV,
audio- and video recordings, the Internet), then a number of factors plays a role: audio
vs. visual, direct vs. indirect, fiction vs. real, etc. The most important factor is directly
vs. indirectly acquired information (§2.3). With regard to other new technologies such
as telephones, there is no difference between quoting face-to-face talk and informa-
tion obtained from such direct conversations and phone conversations. The usage of
evidential forms and of reported speech constructions follow the same rules.
Comparing the Hinuq evidentiality system to the evidentiality systems of the
other Tsezic languages shows that, despite a number of similarities, the systems dis-
play interesting microvariation in form and function. The Tsezic languages can be
divided into East Tsezic, comprising Hunzib and Bezhta, and West Tsezic, consisting
of Hinuq, Tsez, and Khwarshi. Tsez, the language most closely related to Hinuq, has
verbal suffixes whose affirmative forms are cognate with the relevant Hinuq suffixes
(-s(i) and -n(o)), but they express an opposition with two marked members: past wit-
nessed vs. past unwitnessed (Comrie and Polinsky 2007). The same holds true for
Khwarshi (Khalilova 2009: 221–9) and Bezhta (cf. Bokarev 1959: 103; Khalilova 2011).
In contrast, Hunzib is more similar to Hinuq because in this language the opposition
is between periphrastic verb forms that express unwitnessed evidentials and simple
synthetic verb forms that seem to be neutral (cf. van den Berg 1995: 102–4).
With regard to the enclitic, only West Tsezic languages have a separate reportative
enclitic. The East Tsezic languages have only quotative enclitics that seem to express
both reportative evidentiality and reported speech (Khalilova 2011).
6 Summary
This chapter gives an overview of the evidentiality system of Hinuq and integrates the
system in the wider context of how speakers of this language express knowledge. In
the past tenses Hinuq verb forms can be divided into neutral past forms and forms
expressing the unwitnessed past. The use of these forms is not only governed by infor-
mation source, but also by their relation to certain speech genres (e.g. traditional folk
tales vs. reports of personal experience). Two enclitics (reportative and quotative) are
also part of the Hinuq evidentiality system. These enclitics can be freely used with all
verbal forms. Further devices for referring to information source and knowledge are
complement constructions with verbs of knowledge and perception and parenthetical
uses of the same verbs.
References
Aikhenvald, Alexandra Y. 2003. ‘Evidentiality in typological perspective’, pp. 1–31 of Aikhenvald
and Dixon, 2003.
——— 2008. ‘Semi-direct speech: Manambu and beyond’, Language Sciences 30: 383–422.
Aikhenvald, Alexandra Y. and Dixon, R. M. W. Editors of Studies in evidentiality. Amsterdam:
John Benjamins.
Bokarev, Evgenij A. 1959. Cezskie (didojskie) jazyki Dagestana. Moscow: Izd. AN SSSR.
Bybee, Joan, Perkins, Revere, and Pagliuca, William. 1994. The evolution of grammar: Tense,
aspect and modality in the languages of the world. Chicago: University of Chicago Press.
Comrie, Bernard and Polinsky, Maria. 2007. ‘Evidentials in Tsez’, pp. 335–50 in L’énonciation
médiatisée II: le traitement épistémologique de l’information: Illustrations amérindiennes et
caucasiennes, edited by Zlatka Guentchéva and Jon Landaburu. Dudley, MA: Peeters.
——— Forker, Diana and Khalilova, Zaira. ‘Reported speech in the Tsezic languages’. Ms.
Creissels, Denis. 2009. ‘Participles and finiteness: The case of Akhvakh’, Linguistic Discovery 7:
106–30.
Dahl, Östen. 1985. Tense and aspect systems. Oxford: Blackwell.
68 Diana Forker
Forker, Diana. 2011. ‘Evidentiality and new media in the Caucasus: The case of Hinuq’, pp. 101–10
in Proceedings of Conference on Language Documentation and Linguistic Theory 3, edited by
Peter K. Austin, Oliver Bond, David Nathan, and Lutz Marten. London: SOAS.
———. 2012. ‘Evidentiality and markedness’, Poster at the conference ‘The nature of evidential-
ity’. 14–16 June 2012, University of Leiden.
———. To appear. ‘Complementizers in Hinuq’, to appear in Kasper Boye and Petar Kehayov
(eds), Semantic functions of complementizers in European languages.
Friedman, Victor A. 2000. ‘Confirmative/Nonconfirmative in Balkan Slavic, Balkan Romance,
and Albanian, with additional observations on Turkish, Romani, Georgian, and Lak’, pp. 329–66
of Johanson and Utas, 2000.
Johanson, Lars. 2000. ‘Turkic indirectives’, pp. 61–87 of Johanson and Utas, 2000.
Johanson, Lars. 2003. ‘Evidentiality in Turkic’, pp. 273–290 of Studies in evidentiality, edited by
R. M. W. Dixon and A. Y. Aikhenvald. Amsterdam: John Benjamins.
Johanson, Lars and Utas, Bo. 2000. Editors of Evidentials: Turkic, Iranian and neighbouring
languages. Berlin: Mouton de Gruyter.
Khalilova, Zaira. 2009. A grammar of Khwarshi. Utrecht: LOT.
———. 2011. ‘Evidentiality in Tsezic Languages’, Linguistic Discovery 9: 30–48.
Kibrik, Aleksandr E. 1994. ‘Archi’, pp. 297–365 in The indigenous languages of the Caucasus, vol. 4,
The North East Caucasian languages II, edited by Rieks Smeets. Delmar, NY: Caravan Books.
Nichols, Johanna. 2011. Ingush grammar. Berkeley: University of California Press.
Tatevosov, Sergei. 2001. ‘From resultatives to evidentials: Multiple uses of the perfect in Nakh-
Daghestanian languages’, Journal of Pragmatics 33: 443–64.
Tatevosov, Sergei. 2007. ‘Evidencial’nost’ i admirativ v bagvalinskom jazyke’, pp. 351–97 of
Evidencial’nost’ v jazykax Evropy i Azii, edited by Viktor S. Chrakovskij. Moscow: Nauka.
van den Berg, Helma. 1995. A grammar of Hunzib. Munich: Lincom Europa.
3
The expression of knowledge in Tatar
T EIJA GR E ED
1 Introduction
Knowledge cannot be detached from reality, nor is it simply to do with conscious
subjects in isolation; it is ‘a social phenomenon, an aspect of the social relations
between people’ (Hill and Irvine 1992: 17; Aikhenvald, Chapter 1). Knowledge needs
a perceiver, a means of communication, and a recipient; also, it needs interpre-
tation. Evidentials and other expressions indicating information source serve to
aid successful communication. Their appropriate use is also crucial for natural
translation.
This chapter investigates the expression of information source in Tatar; that is, how
the speaker has gained the knowledge she is conveying, and how she expresses her
attitude to it. We will first investigate how the source of information is expressed gram-
matically, and then look at the expression of information source where words and
expressions from open classes are used. The examples used without reference to source
are from my corpus of Tatar data.
2 Tatars: cultural context and society

(Volga) Tatar is spoken by around 7 million people. Most Tatars live in the Russian
Federation, forming the second largest people group in Russia. The population centres
are in Central Russia in the Republics of Tatarstan (around 2 million) and Bashkorto-
stan (around 1 million). In addition Uzbekistan, Kazakhstan, and Kyrgyzstan have a
significant number of Tatars.
For centuries Tatars have been living at the crossroads of East and West, and their
culture has been influenced by many other cultures. Since the fall of the Soviet Union,
appreciation of the Tatar language has been growing. On the basis of the 2002 Rus-
sian Census (RC), well over 90 per cent of the Tatars living in Russia speak their
mother tongue. There is also a growing interest amongst Russians in learning Tatar
70 Teija Greed
(G. R. Mugtasimova,1 p.c.). Many Tatars are at least bilingual, and for some Tatars,
especially in urban settings, Russian is the first language.
3 General characteristics of Tatar

3.1 Typological comments
Tatar is a Turkic language belonging to the Kipchak-Bulgar (Western) group (Comrie
1981: 46). Morphologically, Tatar is an agglutinating language, with suffixes denoting
grammatical categories. Tatar belongs to one of the four most common types of word
order (see Comrie 1989: 87–96): its neutral constituent order is AOV/SV, it has postpo-
sitions (Po), the genitive (G) and the adjective (A) precede their head, as does the rela-
tive clause (Rel) (see example (1)). Auxiliary verbs follow the main verb (example (2)),
and the standard of comparison precedes the comparative. The overall word order
pattern for Tatar is thus AOV/SV/Po/GN/AN (also RelN). Tatar is a dependent–head
type language.
(1) Alsu Marat jaz-gan xatny uqyj

Alsu Marat write-partic.past letter:acc read:pres.3sg
Alsu reads the letter that was written by Marat.
(2) Bala inde uqyj belä

child already read:convb know:pres.3sg
The child can already read. (Greed 2006: 244.)
3.2 Word classes

Noun, verb, and adjective–adverb are the open lexical word classes in Tatar that can ‘be
recognized on internal grammatical criteria within the language’ (Dixon 2010b: 59).
Nouns inflect for singular and plural and six cases. As for grammatical roles, Tatar
follows the regular nominative–accusative system where the subject argument is in
the nominative, and the direct object is in the accusative. The indirect object usually
occurs in the directional case.
The Tatar verb also forms an open class, except for a group of some thirty auxiliary
verb forms fulfilling an actional function in a compound verb. The verb agrees with the
subject in person and (usually) number, and inflects in TAM, except for the indicative
mood whose markers are fused with tense markers (see example (1)). A finite verb2
marked with a suffix indicating person and number forms the grammatical nucleus
of the utterance.
1
I am greatly indebted to G. R. Mugtasimova of Kazan State University and to Nailä Valitova for their
invaluable help in the preparation of this chapter.
2
A finite verb is understood as a verb form which carries information on the person, number, and TAM
of the predicate.
Tatar expresses syntactic subordination with non-finite verb forms—converbs and

participles—and conjunctions play a minor role. Participles form relative construc-
tions, and converbs code various subordinate relations.
Adjective and adverb do not form their own clear-cut classes. Tatar has a smaller
number of adjectives and adverbs than Indo-European languages (Tatarskaja gram-
matika (TG) 1993: 11); additionally, the noun can have a modifying function. In (3),
jaxšy modifies the noun and functions as an adjective, whereas in (4) it modifies the
verb, being an adverb.
(3) Ul jaxšy jazučy
3sg good writer
S/he is a good writer.
(4) Ul jaxšy jaza

3sg good write:pres.3sg
S/he writes well. (Examples modified from TG 1993: 10.)
The main grammatical word classes in Tatar are pronouns, demonstratives, and inter-
rogatives; particles, clitics, conjunctions, and postpositions. They are all closed classes.
Pronouns are divided into personal, reflexive, and reciprocal pronouns. Personal pro-
nouns have three persons in the singular and plural. Literary Tatar does not have rela-
tive pronouns. The numeral ber ‘one’ functions as an indefiniteness marker in certain
contexts, for example, when a new participant is introduced.
The structure of the Tatar verbal word is: root+derivation+inflection. It can express
morphologically the following morphosyntactic operations (see Payne 1997: 62, 172):
agreement in person and number; causative; reflexive, reciprocal, middle, passive;
tense-aspect-mood (TAM); verb negation and subordination/nominalization.
4 Grammatical expression of information source in Tatar

Tatar has four grammatical markers expressing evidentiality: the verb suffixes -GAn3
and particles ikän, dip, and di. In addition, -GAn is contrasted with the verb suffix -DY,
which is overall neutral with regard to evidentiality, but can also receive an evidential
interpretation. Of the markers, only the particles dip and di fulfil Aikhenvald’s (2004: 3)
definition of evidentials proper: their primary meaning is that of evidentiality. How-
ever, I include in my discussion both the evidentials proper, and the markers which
acquire evidential meanings as an extension of another grammatical category, since
this creates a more holistic view of Tatar evidentiality.
In general terms, Tatar evidential markers fit Aikhenvald’s (2004: 25) type A1
‘Firsthand and Non-firsthand’. However, due to the neutral evidentiality status of the
3
Capital letters indicate archiphonemes: different allophones of the same phoneme which vary as a result
of vowel harmony.
72 Teija Greed
suffix -DY in many cases, the evidentiality opposition can be described more appro-
priately as ‘Firsthand/Neutral and Non-firsthand’. The other semantic parameters
relevant to Tatar evidentiality are assumed, reported/hearsay, and quotative. Aikhen-
vald’s sixth parameter, inferred, does not get a separate marker in Tatar; it is one of
the contextual readings of -GAn, and is also one of the possible interpretations of the
particle ikän. Of the parameters, firsthand/neutral stands alone, whereas the remain-
ing four parameters are to do with non-firsthand sources of information.
The scope of evidential meanings expressed with verb suffixes is the sentence. Non-
finite clauses cannot have their own evidentiality status independent of that of the
main clause. The scope of the quotative particle dip extends to the whole speech report
that it is marking, whereas for the hearsay/reported particle di the scope can range
from a word or phrase (especially in spoken register) to a whole text/discourse, for
example, a folk tale (see Hyslop, Chapter 5, for a hearsay clitic in Kurtöp, which has
scope over a whole narrative). The scope of the particle ikän is the sentence.4
Tatar evidentiality occurs mainly in the indicative, but ikän is also used in interroga-
tive clauses. The particles dip and di are independent of other grammatical categories.
4.1 Tatar evidentiality of past tense suffixes

4.1.1 Extension of tense: witnessed vs. non-witnessed; reported and inference Tatar
evidentiality expressed with verb suffixes pertains to the past tense only.5 Tatar
has several past tense forms, the major ones being the so-called categorical past
form -DY, the resultative6 (perfect) past form -GAn, and two distant pasts formed
periphrastically: -GAn ide and -GAn bulgan.
The basic meaning of -DY is ‘the expression of an evident, whole, single event in
the past’ (TG 1993: 106–8), while the meaning of -GAn focuses on result. Thus -GAn
combines the temporal meaning of past and the aspectual meaning of resultative. The
contrast between -GAn and -DY is shown in the following examples:
(5) Rišat qajt-qan.
Rišat return-res.past.3sg
Rišat returned/has returned.
(6) Rišat qajt-ty.

Rišat return-past.3sg
Rišat returned.
4
One of the functions of ikän is to mark conditional, its evidential reading being that of assumption. In
this function, the scope of ikän is the conditional subclause.
5
Tatar has two basic future tense forms which differ from each other epistemically, in that the so-called
categorical future has an inherent nuance of categorical intention, certainty of the event expressed taking
place in the future. The indefinite future tense is neutral in this area.
6
The resultative/perfect aspect is understood as describing ‘a past action which is completed but still has
present relevance’ (Dixon 2010a: 153).
In (5) -GAn shows the result of the action of the lexical verb qajtu ‘return’. As for -DY
in (6), it places the accent on the action itself. This contrasting pair can also
receive an evidential interpretation: in (5) Rišat himself, or at least some signs of
his return, may be visible to the speaker. Thus the speaker has drawn an inference
concerning Rišat’s return, without seeing the action itself. In (6) the speaker has
witnessed the event of return, seeing, for example, Rišat coming through the door
(G. M., p.c.).
The pluperfect markers -GAn ide and -GAn bulgan share the same tense, but form
an opposition with regard to evidentiality: -GAn ide codes the neutral/firsthand mean-
ing, whereas -GAn bulgan expresses non-firsthand evidentiality (see Isxakova et al.
2007: 492).
For the past tense suffixes, the key parameter is to do with whether the event
expressed was witnessed or not.7 The following example is from an article where the
journalist describes events she had not witnessed:
(7) Qadaq suyryšyrga awyldan Marat isemle
nail pull.out:infin village:abl Marat with.name
abyj järdämgä kil-gän.

uncle help:dir.dat come-nwit/rep
An older man called Marat came from the village to help pull out the
nails. (Wälievä 2007.)
Here the suffix -GAn codes also reported information.

4.1.2 Mirative overtones of -GAn in first-person contexts Aikhenvald (2004: 394)
defines ‘mirative’ as ‘grammatical marking of ‘unprepared mind’, including unexpected
and also surprising information’. The verb suffix -GAn can have mirative overtones
in contexts with a first person participant. This meaning applies to both first person
singular and plural verbs of mental or physical states, and indicates that the speaker
was not aware of her state. A Tatar speaker told me about how she and her daughter
were watching Eurovision:
(8) Bez axyryna qadär qaryj al-ma-dy-q,

we end:poss:dir.dat until watch:convb be_able-neg-past-1pl
bez joqla-gan-byz.
we sleep-past.res.nwit-1pl
We were not able to watch till the end, we slept.
7
The Tatar term for the evidential function of -GAn is küzätelmägänlek ‘non-observedness’ (Zäkiev
1999: 247).
74 Teija Greed
She then continued, and by using -GAn showed that she became aware of her falling
asleep only when waking up:
(9) Min üzem dä siz-mä-gän-men,
I self:1sg emph feel-neg-past.res.nwit-1sg
min joqlap kit-kän-men.

I sleep:convb inch-past.res.nwit-1sg
Even I had not felt it, I had fallen asleep.
In first-person contexts, the meaning of ‘unprepared mind’ is connected with the

speaker’s lack of control. Forker (Chapter 2) discusses a similar first-person effect
in a Nakh-Dagestanian language Hinuq, and Creissels (2008) in Akhvakh. Creissels
refers to the concept of ‘assertor’s involvement’.8 In Tatar -GAn marks that the ‘asser-
tor’, speaker, is not involved in the event in question, and therefore -GAn receives an
extension of ‘non-involvement’.
4.1.3 Flexibility of use and interpretation of an evidential Even though Tatar has
an explicit marker for non-witnessed action, the speaker may choose not to use
it. When a Tatar speaker explained to me how her friend sewed curtains for other
people, she used the form -DY, even though she had not witnessed the actual
sewing. After another Tatar speaker queried the form, the first speaker responded
that she used -DY because ‘according to [my friend’s words] it was “tekte”—but in
actual fact it is “tekkän”’ (examples (10) and (11)).
(10) Ul tek-te
3sg sew-DY.3sg
(11) Ul tek-kän
3sg sew-GAn.3sg
She sewed.
The speaker chose not to make the non-firsthand information source explicit, because
she could rely on the original informant; it is as if she had witnessed the event her-
self. Indeed, Zakiev (1992: 189) extends the use of the suffix -DY not just to an event
witnessed by the speaker, but also to events of which the speaker heard from ‘the first
source, that is from the subject or agent’.
The verb suffix -GAn may show the speaker’s evaluation of the reliability of the
information. However, it would usually be strengthened by contextual cues, for exam-
ple, with words expressing uncertainty or doubt (see example (39)).
8
A more widely used term for this phenomenon is ‘conjunct/disjunct’, or volitionality. See also Aikhen-
vald 2004: 123ff.
5 Evidential functions of particles

The second layer of Tatar evidentiality, the three grammaticalized particles, is not lim-
ited to a particular tense/aspect. Particles dip and di with default evidential meanings
mark speech reports. As to the third particle ikän, its main function is to modify utter-
ances, both epistemically, and also by conveying an evidential meaning of assumption.
This layer of Tatar evidentiality does not fully fit within Aikhenvald’s evidentiality
types in that the particles cover three distinct meanings: quotative, reported/hearsay,
and assumed. In addition, the parameter reported/hearsay needs to be teased apart for
a more adequate description of Tatar evidentiality.
5.1 Evidentials proper in conjunction with speech reports

5.1.1 Quotative dip The primary function of the particle dip is to mark direct
and indirect speech. Originally a converb form of the speech verb dijü, ‘say’, it has
become fully grammaticalized as a quotative particle. (This form and its synonym
diep can be used in their lexical meaning too.)
In general, reports of someone else’s words using direct speech do not contain any
epistemic overtones. The expression of indirect speech is less clear-cut. The reporter
of indirect speech can insert modal words that show her attitude towards the message,
for example, jänäse ‘supposedly’.
In a direct speech construction, the particle dip is an obligatory connector between
the speech orienter and the quoted utterance, unless the speech verb is dijü ‘say’.9 The
following example shows both dijü and its synonym äjtü:
(12) Alsu: ‘Min kibetkä kerep čygar-myn’,

Alsu I shop:dir.dat enter:convb go.out:fut-1sg
– di-de / dip äjt-te

say-past.3sg / quot say-past.3sg
Alsu said, ‘I will pop in the shop.’
As for indirect speech constructions, they can be formed both synthetically and ana-
lytically (Zakiev 1992: 483). In the analytical forming of indirect speech, the particle dip
plays an important role. The form of the underlying direct speech is retained, except
for the subject, which becomes the object:
(13) Aju-ny10 kičä šähärdä küren-gän
bear-acc yesterday town:loc appear-res.past.3sg
9
Occasionally dip is left out if the speech orienter precedes the direct speech.
10
The object aju-ny can be replaced by a suffixless form aju (G. M., p.c.). The difference may be due to the
information structure of the sentence (Elena Skribnik, p.c.): aju-ny is used of an established topic, whereas
the suffixless aju would indicate that the bear in question is a new participant introduced to the discourse.
76 Teija Greed
dip söjlädelär
quot say:past:3pl
They told that a bear had appeared in the town yesterday.

Compare this with the sentence in direct speech:
(14) ‘Aju kičä šähärdä küren-gän’ ,
bear yesterday town:loc appear-res.past.3sg
– dip söjlädelär.
quot say:past:3pl
They said, ‘A bear appeared in the town yesterday.’

Quotative dip as a complementizer and marker of subjectivity In English one and
the same verb of perception can be used in connection with different complement
clauses. For example, a that clause can receive either an ‘auditory’ or a hearsay
meaning when combining with the verb hear (Aikhenvald, Chapter 1).
Tatar can form complement clauses synthetically and analytically. The quotative
dip takes part in the forming of the analytical complement clause. Overall, the two
constructions are synonymous (G. M., p.c.).
For example, with the verb of perception išetü ‘hear’, the options are:
(15) Marat Alsu qajtqan dip išette
Marat Alsu return:res.past quot hear:past.3sg
(16) Marat Alsu qajtqanyn išette

Marat Alsu return:partic.past:poss:acc hear:past.3sg
Marat heard that Alsu had returned.
In the first, analytically formed, example, the quotative dip functions as a complemen-
tizer (see also Forker, Chapter 2, for a similar use of the quotative marker in Hinuq; and
Dimmendaal, Chapter 12, for Tima) and indicates that Marat heard from someone else
that Alsu had returned. The second example can be understood synonymously with
the first one, but it can also indicate that Marat heard the process of Alsu’s return, for
example, the clinking of the keys. Of the verbs expressing emotion, qurqu ‘be afraid,
fear’ can take both complementation structures. When occurring with dip, the verb in
the complement clause is usually in the future tense form.
With verbs of cognition, the situation is somewhat different. Some verbs can com-
bine with both the analytic and the synthetic construction, but others can take only the
analytic one. This depends on the semantics of the verb. If we think of verbs of cogni-
tion on a continuum from the ‘objective’ end, where the verb indicates an objective
fact, to the ‘subjective end’, where the verb conveys a subjective epistemic meaning, the
closer the verb is to the subjective end of the objective–subjective continuum, the more
likely it is to combine only with the complement clause containing dip. For the general
verb expressing knowledge, belü ‘know’, both constructions are possible. However, the
two constructions are not identical in meaning:
(17) Min Qazan-ny zur šähär dip beläm
1sg Kazan-acc large city quot know:1sg
I know that Kazan is a big city.
In example (17) dip shows that Kazan’s being a big city is the speaker’s personal thought,
not a statement on the basis of (researched) facts. This is a logical extension of the
meaning of the quotative marker: it has extended from the concrete case of marking
external and internal direct speech to mark a cognitive process of the agent/experi-
encer. Thus dip is involved in creating subjective epistemic meanings. If the speaker
wants to express an objective fact, she can use a synthetic complement clause:
(18) Min Qazan-nyŋ zur šähär bul-u-y-n beläm
1sg Kazan-gen large city be-nomz-poss.3-acc know:1sg
I know that Kazan is a big city.
With the verb faraz itü ‘suppose, assume’ both types of complementation are possible,
but with verbs like ujlau ‘think’ and sanau ‘regard, be of the opinion’, only the construc-
tion with dip is possible, due to their more subjective nature.
The quotative dip has extended further to mark various types of semantic relations11
(see Overall, Chapter 11, for speech reports used for purpose and reason clauses in
Aguaruna). The dependent clause retains the form of the ‘original’ direct speech and
dip marks it as an argument (a complement) of the main clause, thus functioning as
a complementizer. The actual realization of the semantic role depends on the con-
text, but common to all cases is the subjective nature of the content expressed in the
dependent clause (Figure 1). Some possible semantic relations expressed by dip are
reason and goal. In example (19) dip codes the semantic relation of goal, and marks
the thought as a subjective aim of the speaker. Were the speaker wishing to express a
neutral goal, she could use the postposition öčen ‘for’ instead of dip.
(19) Tele ačyl-mas-my dip, xäl-äxwällären
tongue:poss.3 open-neg-q quot state/health:pl:poss:acc
žentekläp soraš-a
thoroughly inquire-pres:3sg
He1 inquires thoroughly about his2 life so that he2 would start talking.
(Literally ‘. . . saying, will his tongue not open? . . .’ .) (Zakiev 1992: 370.)
11
This is reflected in the label ‘postposition-conjunction word’ (Zakiev 1992: 368) given to dip in the
Tatar grammar.
78 Teija Greed
epistemic: subjective
evidential: quotative
epistemic: subjective + semantic relation
Figure 1. Quotative dip and its extensions
5.1.2 Reported/hearsay di (dilär) The other evidential marker to do with speech

reports is di ‘s/he says’, and its plural form dilär ‘they say’, formed from the verb dijü
‘say’ (cf. a similar ‘reportative particle’ in Kalmyk, see Skribnik and Seesing, Chapter 7).
In its evidential function it has become a grammaticalized particle (see Crystal 2003:
338) in that it is used only in this form, and always immediately follows the reported
speech.
Di is used in Tatar to mark reported information/hearsay. This evidential function
is restricted mainly to the spoken register, and is also used in a genre close to oral
speech—in fairy stories and folk tales (see §6.2).
In spoken language, someone else’s words can be reported by using di (or its syno-
nym die). A Tatar speaker recounted a talk she had had with her friend, initially using
the -GAn form (example (20)). This form shows the information given as background
information, and indicates that the words were reported:
(20) Menä ätise anyŋ, značit, qorban

so dad:poss.3 3sg:gen that.is [Russian] sacrifice(acc)
kiterä alma-gan . . . Tormyšlary anyŋ

bring:convb be.able:neg-past.res.nwit life:pl:poss.3 3sg:gen
awyr bul-gan
difficult be-past.res.nwit
So his father had not been able to offer sacrifice . . . His life had been hard.
She then continued by quoting (as if) verbatim her friend’s words:
(21) Menä ul, ätise öčen, . . . — ul äjtä,

so 3sg dad:poss.3 for 3sg say:pres.3
min die, ätijem isemennän die,

I rep dad:poss.1sg name:poss.3:abl rep
bagyšlyjm menä bu qorbanny die
dedicate:pres:1sg this_here this sacrifice:acc rep
So he, for his father, . . . —he says, I (die), in my father’s name (die), dedicate
this sacrifice (die).
The reported particle die is repeated after the subject, the adverbial phrase, and the
predicate including the object of the reported speech, so even in such a short utter-
ance it occurs three times. The frequent repetition of the marker di indicates that the
speaker was strongly emotionally involved (G. M., p.c.).
5.2 Multifunctional particle ikän

5.2.1 Ikän in declarative clauses In the Tatar Grammar (TG 1993: 298–9) the
particle ikän is defined as a modal word, expressing the speaker’s evaluation of the
reliability of the message. According to Zakiev (1992: 191), ikän12 is ‘a unique means
of forming forms of non-witnessed expression’. The Russian term neočevidnost’
suggests that the evidential parameter relevant to ikän is non-witnessed; however
‘non-firsthand’ is a more appropriate way to describe the wide range of meanings
of ikän. In example (22) two markers of evidentiality occur together, ikän
strengthening the non-firsthand meaning of the verb with the -GAn suffix:
(22) Alsu jaz-gan ikän
Alsu write-past.res.nwit.3sg nfirsth13
Apparently Alsu wrote/has written. (Zakiev 1992: 191)
Ikän brings to the utterance the meaning of new information, thus being connected with
mirativity. In my study of Tatar evidentiality (Greed 2009), I define ikän as having the
evidential meaning of assumption. Assumed evidential expresses ‘information source
based on conclusions drawn on the basis of logical conclusion and general knowledge
and experience’ (Aikhenvald 2004: 391). With ikän the conclusion is based on non-
firsthand information, so it has a hint of subjectivity. This nuance is especially evident
in present and future tenses where the verb does not carry a non-firsthand meaning,
but ikän alone expresses this meaning. In the following example ikän combines with
the categorical future tense marker -AčAQ which expresses a certain event in the future:
(23) Alsu jaz-ačaq ikän
Alsu write-fut.categ.3 nfirsth/assum
Alsu will certainly write, it appears. (Zakiev 1992: 191)
Here ikän adds a nuance of non-firsthand meaning to the verb, which is the meaning
of assumption: the speaker draws a conclusion on the basis of what she has learnt from
some other source.
The conclusion drawn can also be based on inference. The following example shows
a situation where you come home and hear voices. You exclaim:
(24) Ėj, bezgä qunaqlar kil-gän ikän!
oh us:dir.dat guest:3pl come-past.res.infer assum
Oh, guests have come to us!
12
Diachronically, ikän consists of i- ‘be’ and the resultative past suffix -GAn (see Zäkiev 1999: 249).
13
Since ikän is a multifunctional particle, note that it is glossed differently in different examples.
80 Teija Greed
The resultative -GAn has a non-witnessed meaning, which in this context is inter-
preted as inference. As for ikän, it expresses a mirative meaning: the meaning of unex-
pectedness is to the fore, and it is strengthened by a nuance of surprise (Nailä Valitova,
p.c.). If ikän is absent, the mirative meaning can be expressed with a rising intonation
on the focal element.
The meaning of surprise is at the fore in contexts where ikän occurs in conjunction
with a verb in the indefinite future tense (Zäkiev 1999: 100):
(25) Altmyš jaš’lek qartlar da šul qadär
sixty age(adj) old.man:pl emph that until
xezmät köne ėšl-är-lär ikän!

service day:poss.3 work-fut.indef-pl mirat
Were old men of sixty years really able to clock up so many working days!
5.2.2 Ikän in interrogative clauses According to Aikhenvald (2004: 249),

evidentials in questions often ‘acquire rhetorical value’. In Tatar questions, ikän can
strengthen the notion of uncertainty (Tatar teleneŋ aŋlatmaly süzlege (AS) 1977: 377);
it has the meaning equivalent to ‘I wonder’, which I call a speculative14 meaning. In
example (26) the speaker does not expect a factual answer, but speculates about the
possibility of the information expressed happening.
(26) Divizija bu qyjenlyqtan niček qotylyr ikän?
division this difficulty:abl how save:fut.indef.3 specul
How will the division escape from this difficulty? (AS 1977: 377)
5.2.3 Ikän: modal, mirative, evidential The particle ikän is many-faceted: it expresses
the speaker’s evaluative viewpoint, and, thus, subjectivity; it may express an assumption
based on various types of conclusion, or surprise, or new information.15 Thus, ikän
is connected with modality, evidentiality, mirativity,16 and information structure.
However, of these various semantic nuances, only the meaning connected with the
speaker’s viewpoint and epistemic stance (see Frawley 1992: 385) is present in all cases.
This feature I call ‘speaker-oriented subjectivity’. Subjectivity is to do with modality;
thus the default meaning of ikän is modal.
14
The speculative meaning is close to the meaning of the Finnish modal verb mahtaa in questions. See
Finnish Grammar online (Hakulinen et al. 2004: §1572).
15
One further nuance may be an expression of pleasure, especially in cases like example (24). This would
coincide with the fact that ikän has a softening effect to the statement expressed: it often sounds more polite
than a statement expressed with bare facts.
16
Even though evidential and mirative meanings tend to be closely connected in Tatar, there is at least
one grammatical structure which has a mirative extension, but no evidential meaning: the third-person
negative interrogative imperative. For example, Uzgan žyelyšqa qatnašmagan ide, bügenge žyelyšqa berenče
bulyp kil-mä-sen-me [come-neg-imp.3-q]! ‘Even though not participating in the earlier meeting, did he not
then come first to today’s meeting!’ (TG 1993: 143.) (Victor Friedman, p.c.).
ikän mirative: unprepared mind, surprise, new information, unexpectedness

modal: subjective
assumed – based on conclusion (in declar.); speculative (in interr.)
Figure 2. Modal ikän in declarative and interrogative clauses with extensions of mirativity and
assumption/conclusion
Ikän is used frequently in spoken language, and in dialogue, but less so in nar-
rative. This is understandable due to its subjective nature: it does not state objec-
tive facts but derives from conclusions drawn on the basis of personal observation,
reported information, and inference. Johanson (2003) describes evidentiality in
Turkic languages using the term ‘indirectivity’. In indirectivity, what is central is
not so much the source of information, but how it was received ‘by a conscious
subject’ (Johanson 2003: 274). This angle is helpful for the Tatar ikän, and also for
the inferred interpretation of -GAn, because in both cases it is the cognition of the
speaker, her reception of information that is at the fore, rather than the outside
source (Figure 2).
Aikhenvald (Chapter 1) mentions how mirative markers can be manipulated in
jokes. The following example shows ikän in its mirative meaning in the heading of a
news item:
(27) Universiadada tatar tele kiräk ikän . . .

university_games:loc Tatar language:poss.3 is_needed mirat
The Tatar language is apparently needed in the University Games . . .
(Zakirova 2012)
In the text itself the writer queries whether the information stated in the heading will
be realized. Thus ikän in the heading displays irony.
6 Evidentials in use
6.1 Radio, television, and the Internet
In Tatar, the non-firsthand form -GAn is used of events seen on television or heard on
the radio (N. V., p.c.). In the following examples the non-witnessed/witnessed mean-
ings of the past tense forms are contrasted:
(28) Bügen Mäskäüdä jaŋgyr jau-gan.

today Moscow:loc rain pour-past.res.nwit.3
Today it rained in Moscow.
(29) Bügen Mäskäüdä jaŋgyr jau-dy.
today Moscow:loc rain pour-past.wit.3
Today it rained in Moscow.
82 Teija Greed
Example (28) was spoken by someone who did not witness the rain directly; she may
have heard of it on the phone from someone who was in Moscow, or seen the rain on
television. Example (29) was spoken by someone who had direct experience of the
rain. Similarly, if something has been read or seen in the Internet, and then recounted
to others, the non-witnessed form is used:
(30) Min Internetta kürdem: Mäskäü

I Internet:loc see:past:1sg Moscow
rajonyn su basqan
district:poss:acc water flood:past.res.nwit.3
I saw in the Internet that there is flooding in the Moscow District.
Dreams and visions are recounted using the -DY form (G. M., p.c); they are regarded
as the speaker’s own eye-witnessed experience.
6.2 Evidentiality and literary conventions: tokens of a genre

In Tatar the grammatical expression of information source has an important con-
nection with literary conventions. Certain evidentials have developed into ‘tokens of
genre’ (Aikhenvald 2004: 310). The verb suffixes -GAn and -DY function in this role:
-GAn is used in genres, relating to events in distant past,17 whereas -DY is a token of
realistic fiction, for example, of narrative in novels and short stories. -DY is also the
form chosen by translators for the Biblical narrative.
In Tatar, as in many languages, legends and folk tales are told in the non-firsthand
-GAn form. -GAn is also the basic form for historical accounts, due to their clearly
reported, non-firsthand nature. However, the addressee does not confuse these two
genres, due to certain cues. One of these is the hearsay particle di. If a discourse/text
contains -GAn forms, with additions of di accentuating the hearsay, this shows that the
text is a folk tale, with information that is not necessarily reliable. On the other hand,
the absence of di in connection with the suffix -GAn indicates reliable information.
In the following excerpt from a biographic narrative, the author did not witness the
events he is describing and is therefore using the -GAn form:
(31) Ätise Čulmanda, . . . paroxodta matros bulyp

dad:poss.3sg Chulman:loc steamship:loc sailor being
ešlä-gän . . . Wäli üze bašta Čulmanda
work-past.res.nwit.3 Wäli himself at_first Chulman:loc
17
The present tense with -A is also used in this function, especially when it refers to background infor-
mation in narrative. -GAn and -A often alternate in narrative—their interplay is a topic for another study.
baržada jör-gän.
barge:loc go-past.res.nwit.3
His dad worked in a steamship as a sailor in Chulman . . . Wäli himself was at
first working in Chulman on a barge. (TG 1993: 112)
Aikhenvald (2004: 25) gives two possible labels to the evidential parameter ‘verbal
report’: reported and hearsay. Since reported is a neutral term, whereas hearsay can
have a nuance of unverifiable information, I use the former to cover the meaning of di
in spoken register, calling it diSP, (‘di in spoken register’), whereas the latter describes
di in the folk tale function, and is labelled diLEG (‘di in legends’) (Figure 3).
Table 1 shows Tatar evidentials as tokens of genre. Note that these show tendencies;
in narrative, for example, -GAn and -A are used to convey background information.
6.3 Summary of Tatar evidentiality

Figure 4 shows the semantic parameters and grammatical forms of Tatar evidentiality.
7 Lexical expression of knowledge and attitudes towards it

In addition to grammatical means, the speaker can express an information source
through lexical means. One common way of doing this is to use verbs of perception
and cognition. Tatar has a wide range of verbs of perception, for example, kürü ‘see’,
išetü ‘hear’, sizü ‘feel’ (see example (9)), and isnäü ‘smell’. The most common verb of
cognition is belü ‘know’ (see examples (17) and (18)).
di reported diSP
non-firsthand
hearsay diLEG
Figure 3. Particle di teased apart to reported and hearsay meanings, with register, and genre
restrictions
Table 1. Evidentials as tokens of genre
-GAn (-A)
Genre -DY firsthand non-firsthand diSP reported diLEG hearsay
novel, ‘realistic’ yes

fiction
history writing yes
folk tales yes yes
spoken language yes

84 Teija Greed
f irsthand/neutral
non-firsthand
-DY
non-witnessed
-GAn assumed quotative reported/hearsay
ikän dip di
inferred reported
reported hearsay
diSP diLEG
Figure 4. Tatar evidentiality: semantic parameters and grammatical forms
In addition to expressing the information source, a speaker can also indicate what
she thinks of the knowledge she is conveying. This can be shown to be anywhere
on the scale impossible—unlikely—uncertain—possible—likely—certain (Hakulinen
et al. 2004: §1556). In Tatar attitude can be expressed both lexically using open word
classes, for example with verbs expressing belief or doubt, and ‘frozen’ words with
modal meanings ranging from impossible to certain, and with particles and clitics
from closed classes.
7.1 Verbal forms and predicative words expressing epistemic meanings

Tatar has only a few simple modal verbs, the main ones being alu ‘be able’, and belü
‘be able, have the ability’ (example (32)). More common in expressing modal mean-
ings are various complex verbs. The most common auxiliary in such constructions is
bulu ‘be’ (example (33)), which carries the regular tense/person/number suffixes of the
complex verb (TG 1993: 87–8). The lexical verb preceding the auxiliary can occur in
various non-finite forms.
(32) Alsu uqyj al-a/bel-ä

Alsu read:convb be_able-pres.3sg/have_ability-pres.3sg
Alsu can read.
(33) Uqy-rga bul-a.

read-infin be-pres.3sg
It is possible to read.
(34) Alsu-ga uqy-rga kiräk

Alsu-dir.dat read-infin pred
Alsu must read.
Example (34) shows a common way to express modal meanings in Tatar, that is, by
using predicative words, such as kiräk ‘must’, and mömkin ‘be allowed’. Tatar also has
a variety of modal words originating from open word classes, which have developed
epistemic meanings. Among them are ixtimal ‘likely’ (< ‘likelihood’) and küräseŋ
‘apparently’ (< ‘you(sg) see’) (TG 1993: 298–9).
7.2 From true to unreliable information

Tatar uses postpositional constructions to express various types of epistemic meanings
and meanings to do with information source. Also the equative suffix -čA can be used
to refer to information source (example (35)), or opinion (36):
(35) gazetada jazyl-gan-ča

paper:loc write:pass-past.partic-equat
according to the paper, as is written in the paper
(36) minem-čä
I:gen-equat
in my opinion
Tatar also has particles and clitics expressing epistemic meanings connected with
information source. The main ones are -DYr, ällä and -mYni.
The modal clitic -DYr is used in the indicative, and it expresses uncertainty. It attach-
es to a verbal predicate which is in the present tense or the indefinite past -GAn form
(example (37)) (TG 1993: 357–8).
(37) Čäčäklär bezneng Qazanga da,

flower:pl we:gen Kazan:dir.dat also
mögaen, shul baqčadan küčkännär-der.
likely that garden:abl move:past.res.nwit:pl-uncert
It is likely that the flowers have moved to our Kazan from that garden.
(Example modified from TG 1993: 301)
Both the particle ällä and the clitic -mYni are used in interrogative sentences. -mYni
expresses surprise and disbelief, whereas ällä (see example (38)) expresses uncertainty
and doubt, and can also have a nuance of surprise.
(38) Ällä štab bezne bötenläj onytty-my ikän?

dub headquarters we:acc completely forget:past-q specul
Is it possible that the headquarters forgot about us completely? (TG 1993:
356–7)
86 Teija Greed
The modal particle imeš (example (39)), whose first meaning is that of reported infor-
mation, can also receive a nuance of doubt. Like ikän, imeš can be ‘manipulated’ to
express irony.
(39) Alsu kilgän imeš
Alsu come:nfirsth rep
It has been said that Alsu has come.
8 Historical development and pathways of evidentials

In the Old Turkic Runic texts of eighth to the ninth centuries an opposition is attested
with regard to evidentiality between the preterite -di and the perfect -mıš (the precur-
sor of the Tatar -GAn) (Isxakova et al. 2007: 473). However, there is disagreement
among Turkic linguists as to whether -di expressed a witnessed action or was neutral
with regard to evidentiality (see also Johanson 2000: 62).
The precursor of the modal particle ikän was attested in old written Turkic, called
Chaghatay, and the texts originate from the fourteenth to the seventeenth centuries.
The particle occurred in the form e(r)ken(dur), and expressed ‘doubt and supposition’
(Boeschoten and Vandamme 1998). The latter meaning is reflected in the current Tatar
use of ikän in its evidential function of assumption.
Aikhenvald (Chapter 1) outlines two pathways for the development of evidentiality
strategies. In the first, evidential meanings range from inference to hearsay, and they
are expressed through modalities, perfects, and resultatives. In the second pathway,
reported speech and particles derived from the verb ‘say’, with the primary meaning of
‘hearsay’, may develop meanings of inference ‘based on results or assumption’.
The Tatar verb suffixes follow the first pathway in general terms: the evidential strat-
egy based on the extension of the resultative meaning of -GAn has a range of meanings
covering inference and reported meaning. However, the direction is unclear, and, for
example, in the Tatar Grammar Zakiev (1992: 189) mentions only the reported mean-
ing as a semantic extension of -GAn.
As to the second pathway, the quotative particle dip, which can be regarded as forming
part of an evidentiality system, has developed other meanings, but these are not to do with
source of information but mark other semantic relations between clauses, such as purpose
and reason. Figure 5 is an attempt to describe the diachronic development of the Tatar dip.
lexical dijü
converb dip (di-p), with lexical meaning ‘saying’
grammaticalized quotative particle dip
dip, marker of indirect speech
complementizer dip
Figure 5. Diachronic development of the quotative particle dip
9 Conclusion
In Tatar information source is expressed both grammatically and lexically. For the
lexical expression verbs of perception and cognition, various more extensive lexical
expressions are used. To convey her attitude towards the information, the speaker
has a number of means at her disposal. In addition to full explanatory clauses, such
epistemic meanings can be expressed with modal simple and complex verbs, modal
words, particles, clitics, and predicative words.
Evidentiality in Tatar is a mixture of diverse features. In the first place, evidential
meanings are expressed both with past tense verb suffixes and grammaticalized parti-
cles. However, continuity is also displayed, with some unifying factors that extend over
the different types of evidentials. All the four evidential markers—the verb suffix -GAn,
and the particles dip, di, and ikän—come under the general semantic parameter of non-
firsthand information source. Two of these—the quotative dip and reported/hearsay
di—are involved in marking speech reports and are evidentials proper. Furthermore,
the verb suffix -GAn and the particle ikän display meanings of inference. In the area of
modal expression, both dip and ikän convey the speaker’s subjective viewpoint.
In the Tatar expression of knowledge there is interplay between evidentiality,
modality, and mirativity. The notion of ‘unprepared mind’ connected with mirativity
is important to both the verb suffix -GAn and the particle ikän, which can both express
unexpectedness and surprise. For ikän the mirative meaning is central, but for the
non-witnessed -GAn the mirative meanings occur only where the verb is in the first
person. Then the meaning of non-involvement is to the fore.
In addition to everyday communication, Tatar evidentials play an important part
in more literary genres: the verb suffixes -GAn and -DY and the particle di function
as tokens of genre. Even though they do not form a unified system, evidentials are an
integral part of meaningful communication in Tatar.
References
Aikhenvald, Alexandra Y. 2004. Evidentiality. Oxford: Oxford University Press.
Boeschoten, Hendrik and Vandamme, Marc. 1998. ‘Chaghatay’, pp. 166–78 of The Turkic lan-
guages, edited by Lars Johanson and Éva Ágnes Csató. London and New York: Routledge.
Comrie, Bernard. 1981. The languages of the Soviet Union. Cambridge: Cambridge University
Press.
———. 1989. Language universals and linguistic typology. Oxford: Blackwell.
Creissels, Denis. 2008. ‘Remarks on so-called “conjunct/disjunct” systems’, pp. 1–19 of Syntax of
the World’s Languages III, Berlin, 25–28 September 2008.
Crystal, David. 2003 (2007). A dictionary of linguistics & phonetics, 5th edition. Malden: Black-
well Publishing.
Dixon, R. M. W. 2010a. Basic linguistic theory, Vol. 1, Methodology. Oxford: Oxford University
Press.
88 Teija Greed
———. 2010b. Basic linguistic theory, Vol. 2, Grammatical topics. Oxford: Oxford University
Press.
Frawley, William. 1992. Linguistic semantics. Hillsdale, NJ: Lawrence Erlbaum Associates.
Greed, Teija. 2006. ‘On Tatar converb clauses and their argument structure’, Turkic Languages
10: 220–45.
———. 2009. ‘Evidentiality in Tatar.’ Secondary MA Thesis for General Linguistics University
of Helsinki, 12 June 2009. Unpublished manuscript.
Hakulinen, Auli, Vilkuna, Maria, Korhonen, Riitta, Koivisto, Vesa, Heinonen, Tarja Riitta, and
Alho, Irja. 2004. Iso suomen kielioppi [Finnish grammar]. Helsinki: Suomalaisen Kirjallisuuden
Seura. Available at: <http://scripta.kotus.fi/visk/etusivu.php> accessed 12 May 2012.
Hill, Jane H. and Irvine, Judith T. 1992. ‘Introduction’, pp. 1–23 of Responsibility and evidence in oral
discourse, edited by Jane H. Hill and Judith T. Irvine. Cambridge: Cambridge University Press.
Isxakova, X. F., Nasilov, D. M., Nevskaja I. A., and Šencova, I. V. 2007. ‘Ėvidencial’nost’ v tjurk-
skix jazykax’, pp. 467–518 of Ėvidencial’nost’ v jazykax Evropy i Azii, edited by V. S. Xrakovskij.
St Petersburg: Nauka.
Johanson, Lars. 2000. ‘Turkic indirectives’, pp. 61–87 of Evidentials. Turkic, Iranian and neigh-
bouring languages, edited by Lars Johanson and Bo Utas. Berlin: Mouton de Gruyter.
———. 2003. ‘Evidentiality in Turkic’, pp. 273–91 of Studies of evidentiality, edited by A. Y.
Aikhenvald and R. M. W. Dixon. Amsterdam: John Benjamins.
Payne, Thomas E. 1997. Describing morphosyntax. A guide for field linguists. Cambridge: Cam-
bridge University Press.
Russian Census (RC) 2002. Available at <http://www.perepis2002.ru/index.html?id=11>
accessed 29 June 2012.
Tatarskaja grammatika (TG). 1993. Tom II, Morfologija. Kazan: Tatarskoe knižnoe izdatel’stvo.
Tatar teleneŋ aŋlatmaly süzlege I (AS) [Explanatory dictionary of Tatar I]. 1977. Kazan: Tatarstan
kitap näšrijaty.
Wälievä, Gölčäčäk. 2007. ‘Ilfaq belän Ilsöjä jat awyldan jort aldylar’ [Ilfaq and Ilsöjä bought a
house in an unfamiliar village], Qäef niček? 39, 28 September–4 October.
Zakiev, M. Z. 1992. Tatarskaja grammatika, Tom III, Sintaksis. Kazan: Tatarskoe knižnoe
izdatel’stvo.
Zäkiev, M. Z. 1999. Tatar grammatikasy, III tom, Sintaksis. [Tatar Grammar, Vol. III, Syntax.]
Moscow, Kazan: Insan, Fiker.
Zakirova, Rämzijä. 2012. ‘Universiadada tatar tele kiräk ikän . . . ’ [The Tatar language is
apparently needed in the University Games . . . ], available at <http://tatclub.net/2012/03/
universiadaga-tatar-tele-kir%D3%99k-ik%D3%99n/>, 13 March 2012, accessed 28 May 2012.
4
The grammar of knowledge

in Saaroa1
C H I A-J U N G PA N
1 Preliminaries
This chapter examines the expression of knowledge through evidentials, and through
other means, and their correlations with types of knowledge, cultural conventions
and speech practices in the Saaroa language. Discussion on evidentials focuses on
the system of evidentiality and essential features of evidentiality. In addition to the
grammatical category of evidentiality, this chapter further investigates other means of
expressing the information source, including evidentiality strategy through reported
speech, evidential-like meanings through markers, and lexical means through verbs
of perception. Evidential systems have never been reported as a salient genetic feature
of Formosan languages.2 After thorough investigation and discussion, this chapter
will show that a closed grammatical category, evidentiality, can be attested in Saaroa.
A brief typological overview is in order. Saaroa is a moribund Austronesian lan-
guage of Taiwan. The language is spoken in Taoyuan Village and Kaochung Village,
Taoyuan District, Kaohsiung City, Taiwan. There are approximately 500 members of
the Saaroa community. At the moment, ten to fifteen people are able to speak the lan-
guage. Further details are in Pan (2012).
1 I am grateful to Alexandra Aikhenvald and R. M. W. Dixon for inviting me to participate in the Work-
shop and for providing valuable comments on earlier versions of this chapter, and to my Saaroa teachers for
helping me learn their language. I am indebted to two anonymous reviewers for their valuable suggestions
and comments. I would like to thank participants at the Workshop for sharing their knowledge of ‘the
Grammar of Knowledge’. Thanks also go to Elizabeth Zeitoun, Jackson Sun, Henry Chang, Lawrence Reid,
Hsiu-chuan Liao, Stacy Teng, Marie Yeh, Fuhui Hsieh, Amy Lee, Joy Wu, Rik De Busser, Apay Ai-yu Tang,
and Yu-ting Yeh for their discussions and comments.
Saaroa examples in this chapter come from my fieldwork, mostly based on texts from narratives with a few
examples based on limited corroborative grammatical and lexical elicitation. There were two major field trips
for collecting Saaroa texts: one from August 2008 to July 2009, and the other from February 2011 to May 2011.
2 There are a small number of studies on Formosan evidentiality (as a grammatical category), e.g. Yang
(2000a, 2000b) and Chang (2012).
90 Chia-jung Pan
The basic syllable pattern is (C)V. Underived roots carrying the basic meaning of
words usually have more than two syllables, in a (C)V.(C)V.(C)V pattern. A disyl-
labic (C)V.(C)V pattern is relatively rare. Grammatical morphemes are usually a single
syllable, for example construction markers. Saaroa distinguishes primary stress and
secondary stress. Neither is contrastive. Primary stress within a word typically falls
either on the penultimate or the antepenultimate syllable. A vowel with primary stress
is characterized by higher pitch and greater intensity.
Saaroa exhibits a rich morphology, and is synthetic and agglutinating. Usually a
word contains a largish number of morphemes (roots, affixes, and clitics) but at the
same time morpheme boundaries are clear. Prefixation is productive, whereas other
types of affixation are not. Reduplication is widely deployed. The two major word
classes are verb and noun, with rich morphology marking. Adjectival and adverbial
elements are two subclasses of verb.
The basic constituent order is VAO, if transitive, or VS(E), if intransitive. The pro-
nominal system consists of bound pronouns and independent pronouns. The bound
pronoun is a core argument either in S function or in A function, whereas the inde-
pendent pronoun is a core argument in S (when topicalized), E, A, or O function. The
bound pronouns can be divided into two sets: nominative pronouns, marking argu-
ments in S function, and genitive pronouns, marking arguments in A function and
possessor function. The case system includes core, oblique, and genitive. The core case
covers arguments in S, A, and O functions. The oblique case marks extended argu-
ments (i.e. E function) and peripheral arguments, for example location. The genitive
case is used to encode possessor function.
There are four verbal clause patterns in Saaroa: (i) Pattern 1: monovalent intransi-
tive clauses; (ii) Pattern 2: bivalent intransitive clauses; (iii) Pattern 3: bivalent transi-
tive clauses; and (iv) Pattern 4: applicative clauses. Patterns 1 and 2 take Actor voice
(AV), marked by um-/<um>/u-/m- or zero-marked; Pattern 3 takes patient voice (PV),
marked by -a or zero-marked; Pattern 4 takes locative voice (LV), marked by -a(na).
The definiteness effect plays a role in determining the manifestation of voice in an
independent clause. The manifestation of voice in independent clauses plays a role in
determining grammatical subjects.
In Pattern 1, monovalent predicates can be stative or dynamic. These two types of
predicate are marked by an Actor voice, and have the same argument structure. The
intransitive clause may optionally take a peripheral location argument marked by the
oblique case.
(1) t<um>a-taa-tangi [a 'a'ai]S3

irr<av>-redup-cry core baby
The baby is crying.
3 Orthography follows IPA except where ' = glottal stop, c = unaspirated voiceless alveolar affricate,
ng = velar nasal, l = alveolar flap, lh = alveolar lateral fricative, and u
– = high central unrounded vowel.
PATTERN 1: um-/<um>/u-/m-V (a/ka) N ((na) N)

Intransitive CORE OBL
S
Actor/patient
PATTERN 2: um-/<um>/u-/m-V (a/ka) N (na) N ((na) N)
Extended intransitive CORE OBL OBL
S E
Actor patient
PATTERN 3: V-a (a/ka) N (a/ka) N ((na) N)
Transitive CORE CORE OBL
A O
Actor patient
PATTERN 4: V-a(na) (a/ka) N (a/ka) N ((na) N)
Applicative CORE CORE OBL
A O E
Actor location patient
Figure 1. Verbal clause patterns and argument structures in Saaroa
In Pattern 2, the verb in an extended intransitive clause is bivalent and has a noun
phrase in S function and a noun phrase in E function. The noun phrase in E function is
indefinite, and is downgraded to oblique status. The extended intransitive clause may
optionally take a peripheral location argument marked by oblique case.
(2) um-a-urapu– [a lhaamaama]S [na 'u

–vu
–cu
–ngu
–]E
av-irr-sow core old.person obl millet
The old people will sow millet.
In Pattern 3, the transitive clause has two core arguments: a noun phrase (Actor) in
A function and a noun phrase (patient) in O function. The noun phrase (Actor) in
A function can be a genitive pronoun, a free independent pronoun or a common
noun phrase. The transitive clause may optionally take a peripheral location argument
marked by the oblique case.
(3) lhi-pai-pu–ku
–l-a=cu [a ilhaku]A
perv.asp-action.involving.hands-br-pv=cos.asp core 1sg.indep.pn
[a tangusulhu=na]O kiira
core rice.cake=def yesterday
I finished molding the rice cake yesterday.
In Pattern 4, the applicative clause consists of three arguments: a noun phrase (Actor)
in A function, a noun phrase (patient) in E function and a noun phrase (location) in
O function. The noun phrase (patient) in E function is demoted from core status to
92 Chia-jung Pan
oblique status. The noun phrase (location) in O function is promoted from oblique
status to core status. It is not compulsory for the noun phrase (patient) in E function
to be expressed overtly if it can be inferred or retrieved from the context. The most
typical example is verbs of ‘give’. The other alternative in Pattern 4 is that the applicative
clause just consists of two arguments: a noun phrase (Actor) in A function and a noun
phrase (location) in O function.
(4) racu'u salia ia, italuailipi-a[=lhamu] [parana]O

bamboo house top relax(in.a.cool.place)-lv=1pl.exc.gen place
As for the bamboo house, we relax in this place.
2 Organization of the evidential system

In Saaroa, every sentence may contain an indication of how the information was
acquired by the speaker: whether they saw the event happen, or heard it, or know
about it because somebody else told them, etc. This can be achieved by evidentials or
by other means, such as reported speech, markers, and lexical expressions of percep-
tion and cognition.
Evidentiality in Saaroa is a grammatical category in its own right, and not a sub-
category of epistemic or some other modality, or of tense-aspect. The system of evi-
dentiality is A3. reported versus ‘everything else’, following the classification of
evidentiality systems proposed by Aikhenvald (2004: 25). The reported term is marked,
and the non-reported (‘everything else’) term is not marked. The system with just one,
reported, evidential covers information acquired through someone else’s narration.
(5) maaci ka kiariari a ucani ka uka'a=mana

when link past link one link neg=imperv.asp
ka kiariari lhi-pu'a na kana sikamu– ia,
link past perv.asp-buy obl pf mat top
ma-aru=ami a tapau
–=na mu–u
–mu –a tu-sikamu –
av-exist=evid core Shell.flower=def also make-mat
In the past, when (people) had no money to buy mats, it is said that the Shell
flower could also be used to make mats.
The reported evidential is conceptualized as a genre marker, or a token of narratives. It

is widely used when the speaker tells a folk tale. Even when the speaker makes up his
own original stories from life experience or from reading books or looking at pictures,
he marks them with the reported evidential.
The function of the reported evidential in Saaroa is to signal where information
comes from. Employing the reported evidential implies the necessity of being precise
about the information source. This can be regarded as a requirement for successful
Table 1. Grammatical characteristics of the reported evidential
Grammatical characteristics Reported evidential
Grammatical means—§3.1 enclitic =ami
Omission—§3.2 possible but maybe awkward
More than once in a clause—§3.3 Yes: contrastive focus
Scope—§3.4 whole clause or sentence
Time reference—§3.5 distinct from the time reference of the clause’s

predicate: realis or irrealis
Epistemic extension—§4.1 objectivity and shifting responsibility
Mirative extension—§4.2 new, unusual, and surprising information
Nature of observer—§5.1 ‘I’ (the speaker)
First person effect—§5.2 overtones of lack of control and subsequent

surprise on behalf of the speaker
Origin—§6 a grammaticalized verb of speech
communication, and a way of avoiding possible misunderstandings. Thus, in a sen-

tence with the reported evidential, asking ‘how do you know this?’ does not make any
sense.
The grammatical characteristics of the Saaroa reported evidential, which will be
examined in the following sections, are summarized in Table 1.
3 Expression of reported evidentials

3.1 Grammatical means for reported evidential marking
In Saaroa, the grammatical means employed for expressing the reported evidential
is through encliticization =ami, usually attaching to the right of the first clause or
sentence constituent. It is a clitic in that it does not shift the primary or secondary (if
any) stress of its host; in addition, it is not selective to its host. The reported evidential
most frequently attaches to the verb which typically occurs in the clause-initial posi-
tion, as in (6).
94 Chia-jung Pan
(6) saa-lhivur-a=ami a alhaina=na

3.gen-stab-pv=evid core woman=def
It is said that he stabbed the woman.
In addition to attaching to the verb, as shown in examples (7), (8), (9), and (10), the
reported evidential =ami can attach to other types of host, for example the conditional
subordinator maaci ‘if ’, the negator ku, the ‘concerning’ clause maacu, and the quanti-
fier rianu
– ‘all’.
(7) maaci=ami kana'a m-u-sala m-ari-a-'intavangu

– ia, . . .
if=evid pf av-motion.on.foot-road av-hand.motion-irr-taro top
It is said that if (one) went to dig taros, . . .
(8) ku=ami pai-ta-tu

–alhu
– isana m-u-sala
neg=evid find-redup-achi 3.indep.pn av-motion.on.foot-road
m-aki<ka>kua n kana 'u-'ukui-a
av-action.toward.location<redup> obl that redup-goat-A
It is said that (they) went to that place where goats gathered but couldn't find
them.
(9) maacu=ami alhaama kiariari ia, auniini='ai=iau

concerning=evid ancestor past top like.this=infer=mir
lhi-angalhu
– ='ai vuvulungaa rumalhau
– saa-maruka-a
perv-from=infer mountain when 3.gen-stray-pv
Concerning ancestors in the past, what happened is like this. When they came
back from the mountains, they got lost.
(10) rianu u
–=ami alu
–mu–lhu– ka ma-maini=isa akuisa lh<um>ivuru
all=evid wild.boar ka redup-small=3.gen when stab<av>
isana ka ta-turua=isa
3.indep.pn ka redup-cousin=3.gen
It is said that when her children turned out to be all wild boars, her cousin
stabbed them.
The reported evidential does not always attach to the right of the first constituent in a
clause or sentence, and yet it has scope over the whole clause or sentence. Example (11)
shows that the reported evidential is encliticized to the locational noun vuvulungaa
‘mountain’ which occurs in sentence-final position. The speaker was so surprised at
those people who are not living in the mountains, because they are not aware of the
existence of kiwi fruit in the mountains. It is the focused constituent that the reported
evidential attaches to.
(11) ku tararianu
– kana cucu lika'a kaaiu ma-aru
neg hear that person outside far.there av-exist
luuvi=ta vuvulungaa=ami
kiwi.fruit=1pl.inc.gen mountain=evid
It is said that those outsiders didn’t hear that we have kiwi fruit in the
mountain.
3.2 Omission of reported evidentiality

If a language has obligatory evidentiality, omitting an evidential gives rise to a gram-
matically awkward ‘incomplete’ sentence (cf. Valenzuela 2003: 34). In Saaroa, the
reported evidential is not obligatory, so leaving it out does not result in ungrammati-
cal sentences, as shown in (12).
(12) saa-lhava-a m-alhu-kua salia paa-paci um-u um-aala um-u

3.gen-bring-pv av-get.to-get.to home caus-die av-eat av-take av-eat
He brought (something) home, killed (it) to eat, and took (it) to eat.
However, it is likely that if the addressee wants to know the exact information source,
he might ask the speaker ‘how did you know it?’, ‘did you see it?’, or ‘who told you?’.
This omission of evidential could be considered a language obsolescence phenomenon
in Saaroa.
3.3 A reported evidential occurs more than once in a clause

When Saaroa speakers tell folk tales and traditional stories (except for introductory
and concluding remarks), they tend to use the reported evidential once per sentence.
Not uncommonly, the reported evidential may occur more than once in the same
clause. Typically, the reason for this is pragmatic. In example (13), the reported eviden-
tial appears to be repeated to mark contrastive focus: first on the temporal expression
kiariari ‘past’ and then on the verb. The focus is on an unusual and unexpected occur-
rence time frame ‘past’: they would dig (the ground) under a bed and bury (the body)
under the bed. This is the reason why the enclitic occurs on the temporal expression.
It is also repeated on the verb, to focus on the event itself.
(13) maaci m-iungu=cu aari=isa ia, kiariari=ami ka

if av-arrive=cos.asp day=3.gen top past=evid link
k<um>a-kalii=ami n kalalhapa taaru
– a-calhumu=isa=ami
irr<av>-dig=evid obl under bed irr-bury=3.gen=evid
kalalhapa taaru –
under bed
If his time came, it is said that in the past they would dig (the ground) under
a bed and bury (the body) under the bed.
96 Chia-jung Pan
3.4 Scope of reported evidentiality

In Saaroa, the reported evidential usually has the whole clause or sentence within its
scope. In (14), the evidential enclitic is added to a negation word, meaning ‘There are
no chickens over there, I heard it’.
(14) uka'a=cu=ami ka kana turukuuka m-aa isana

neg=cos.asp=evid core pf chicken av-be:loc/temp there
It is said that there are no chickens over there.
The evidential itself cannot fall within the scope of negation. To say ‘Ancestors had a
love affair with a wild boar, I didn’t hear it’, one has to use an independent lexical verb
‘hear’.
(15) ku lhi-timalha=ku na alhaama kiariari n

neg perv.asp-hear(pv)=1sg.gen obl ancestor past link
kana m-uritalhivau– n alu
–mu –lhu–
pf av-have.a.love.affair obl wild.boar
I didn’t hear that ancestors had a love affair with a wild boar.
3.5 Time reference of reported evidentiality

Saaroa does not have the grammatical category of ‘tense’; instead, it distinguishes real-
ity status, including irrealis, which is expressed by -a or by reduplication, and realis,
which is zero-marked (Pan 2012: 190–3).
In Saaroa, when expressing the reported evidential, its time reference can be distinct
from that of the clause’s predicate. That is, the reference to the time of the utterance
can be different from the reference to the time when the information was acquired
from a particular source.
In example (16), the time when the action took place is the same as the time when
the speaker acquired the information.
(16) m-u-sala=ami a cucu=isa=na u-kiri-kirimi

av-motion.on.foot-road=evid core person=3.gen=def av-redup-search
isana
3.indep.pn
It is said that their people went to search it.
As shown in (17), the time reference of the reported evidential does not have to coin-
cide with that of the event. When a reported action is said to happen in the future, the
verb is marked with an irrealis marker, while the reported evidential refers to the ‘past’
time when the speaker acquired the information.
(17) um-a-cu –ka=ami ka kana tamu=isa lhi-calhum-a

av-irr-get.up=evid core pf grandparent=3.gen perv.asp-bury-pv
n kana kalalhapa taaru
– rumalhau–, k<um>a-kita=ami
obl pf under bed when irr<av>-look=evid
ka kana lhamu=isa kana tamu=isa
core pf grandchild=3.gen pf grandparent=3.gen
It is said that when their grandparent who has been buried under a bed gets
up, their grandchildren will guard their grandparent.
4 Semantics of reported evidentiality

The reported evidential is typically used for secondhand and thirdhand information.
Whether the information is secondhand or thirdhand is known from the context only.
In addition to reflecting on the information obtained from someone else, the reported
evidential could have epistemic extensions and mirative extensions.
4.1 Epistemic extensions of reported evidentiality

In Saaroa, the reported evidential makes an implicit reference to the speaker’s attitude
toward the information acquired from someone else. There are two reasons that the Saar-
oa speaker may opt to employ the reported evidential. The first reason is to show his or her
objectivity; that is, the speaker is not an eyewitness to the event and knows about it from
someone else. The second reason is as a means of shifting responsibility for the informa-
tion and implying that related facts may have a connotation of unreliable information.
Examples (18) and (19) illustrate the point. Saying example (18) is just stating a fact
(the sentence is evidentiality neutral: one is not saying how one knows it).
(18) akuisa lh<um>ivuru civuka=isa m-utu-pulhu
when stab<av> belly=3.gen av-move.toward-come.out
a ma-maini=isa riane alu
–mu –lhu
–
core redup-small=3.gen all wild.boar
When (he) stabbed her belly, the children came out, and all (children turned
out to be) wild boars.
Recasting this statement with a reported evidential in example (19) implies that the
speaker simply cites someone else’s report, and thus disengages himself from respon-
sibility for its veracity. Since this sentence was based on a Saaroa folktale, it is unlikely
that the speaker had been an eye-witness to this event.
(19) akuisa lh<um>ivuru civuka=isa m-utu-pulhu=ami
when stab<av> belly=3.gen av-move.toward-come.out=evid
a ma-maini=isa rianu
– =ami alu
–mu–lhu
–
core redup-small=3.gen all=evid wild.boar
When (he) stabbed her belly, they said that the children came out, and all
(children turned out to be) wild boars (but I wouldn’t vouch for it).
98 Chia-jung Pan
Example (20) is slightly different from (19), and was uttered by a male speaker who
was talking about a traditional Saaroa wedding ceremony. On the day of the wedding
ceremony, the man (husband-to-be) together with his family had to visit the woman’s
(wife-to-be) family, and give them hand-made sticky rice as a gift. If the man (husband-
to-be) had cousins, they had to give a bullet to the woman’s (wife-to-be) brother and
cousin as a gift. Although the speaker had been an eye-witness of this event while he
was young, he still used the reported evidential to indicate that he did not claim direct
responsibility. Nowadays, most young people appear not to have witnessed this kind
of traditional wedding ceremony. If the speaker did not use the reported evidential,
the hearer might ask him ‘why is it different from the current wedding ceremony?’ or
‘are you sure what you said is true?’ By employing the reported evidential, the speaker
did not vouch for the information he provided.
(20) maaci ma-aru ka turu-turua-isa lhalhusa ia,

if av-exist core redup-cousin-3.agr man top
m-alhava=ami n ripasu –
av-bring=evid obl bullet
If the man (husband-to-be) had cousins, it is said that (they) brought a
bullet.
4.2 Mirative extensions of reported evidentiality

In Saaroa, the reported evidential might extend to cover new, unusual, and surpris-
ing information; that is, it develops mirative overtones. Example (21) shows that the
reported evidential implies new knowledge and surprise for the speaker, in that, to the
speaker’s surprise, it is unusual that the third person participant had a chance to have
encountered a bear in the mountain.
(21) amilh-a amalhu –=na lhi-k<um>ita=ami vuvulungaa

say-pv male.name=def perv.asp-see<av>=evid mountain
n cumi'i
obl bear
Amalhu – said he saw a bear in the mountain (surprisingly).
5 The reported evidential and person

5.1 The reported evidential and nature of the observer
Evidentials mirror the way in which information was obtained. Typically, the report-
ed evidential in Saaroa occurs in sentences with an explicitly or implicitly specified
‘third’-person participant in S or A function, as in example (22). The implicitly speci-
fied participants can be retrieved from the context, as in example (23).
(22) [saa-]Apala-va-vililh-a=ami
3.gen-stealthily.follow-redup-stealthily.follow-pv=evid
[ka kana cucu]O [salia=isa]E k<um>ita aunaana=iau
core pf person house=3.gen look/see<av> like.that=mir
rumalhau – m-uritalhivau – [na alu
–mu–lhu
–]E
when av-have.a.love.affair obl wild.boar
It is said that he stealthily followed the person to her house and had a look.
What happened is like that. (He saw her) have a love affair with a wild boar.
(23) puaili=cu [ka lhalhusa]S m-ia<laa>langu – m-ianu– =ami

return=cos.asp core man av-prepare<redup> av-pound=evid
After the men returned (to home), it is said that (they) prepared to pound
(rice).
The narrative corpus I collected is extensive, and can be considered representa-

tive of the language. Since only a small number of language materials are based on
participant-observation, many examples come from narratives (in the third person).
However, it does not mean that someone reports the information to a ‘third’ person;
instead, it indicates someone reports the information to ‘I’ (the speaker). In other
words, using the reported evidential presupposes that the observer/perceiver is ‘I’ (the
speaker), rather than ‘you’ (the hearer or the audience) or someone else, a ‘third’ person.
5.2 ‘First-person’ effects in the reported evidential

Aikhenvald (2004: 219) states that evidentials may develop additional semantic over-
tones in the context of first-person participant. In my corpus, there are only two exam-
ples consisting of the reported evidential and the explicitly specified ‘first’ (I) person.
The rarity can be attributed to the nature of the narratives in my corpus which is mostly
based on third person participants.
In Saaroa, when the speaker ‘I’ is talking about something I myself did or am going
to do, the reported evidential acquires additional overtones. Example (24) indicates
that the reported evidential with a first-person participant implies overtones of lack
of control and subsequent surprise on behalf of the speaker. The speaker was partly
surprised at his own knowledge of the story, and partly surprised at the addressee who
made a request that he (i.e. the speaker) should tell him the story about traditional
Saaroa marriage.
(24) ki-a-lha-lhamu=aku=ami kani'i lhalhusa maaci um-aala

tell-irr-redup-tell=1sg.nom=evid this man if/when av-take
n alhaina
obl woman
I am going to tell this (story) about men marrying women.
100 Chia-jung Pan
Since it combines lack of control with the speaker’s surprise at something he was not
going to do but is doing, it is likely that this has a mirative overtone. Although this is
just one of the two examples of a reported evidential with the first person that I have
in the corpus, its semantics is reminiscent of the first-person effect.
6 Origins of the reported evidential

Markers of reported evidentiality and quotative evidentials often come from a gram-
maticalized verb of speech (Aikhenvald 2004: 271). This has been reported in Lezgian
(Haspelmath 1993), Tauya (MacDonald 1990), Tibetan (Tournadre 1994: 152), Marico-
pa (Gordon 1986: 86), Akha (Thurgood 1986: 221), Cora (Casad 1992: 154–6), Kavalan
(Hsieh 2012), and many others.
In Saaroa, the verb of ‘saying’ is the source for the reported evidential =ami. The
emergence of the reported evidential involved:
(i) a biclausal quotation or reportative construction consisting of a matrix clause
with the verb ‘say’ and a complement clause;
(ii) reanalysis of a complement clause of the verb ‘say’ as the main clause (becom-
ing a single clause), and reanalysis of the verb ‘say’ used in the matrix clause as
an enclitic in the newly-formed clause; and
(iii) reinterpretation of the enclitic from the verb ‘say’ with the meaning of ‘report-
ed’ information.
Examples (25) and (26) illustrate the above three distinct developments. The reported
speech complement clause is in square brackets.
(25) amilh-a pakiaturua, [ku-a-tumulhu=kia lhatu –ngu– paa-m-a-vacangu –

say-pv teacher eat-irr-a.lot=pr vegetable caus-av-stat-good
tilha'alhu
–]!
body
Teachers said, please eat more vegetables (because it) makes the body healthy!
(26) m-u-sala=ami a kana cucu=isa=na puaili

av-motion.on.foot-road=evid core pf person=3.gen=def return
k<um>ita ta'alhara=isa
look<av> chicken.coop=3.gen
It is said that their people returned to guard their chicken coop.
The grammatical behaviour of the verb ‘say’ in Saaroa displays some peculiarities.
The loss or absence of these grammatical features indicates that the Saaroan verb
‘say’ is in the process of grammaticalization. First, the verb ‘say’ ami in Actor voice
(zero-marked) cannot be used in speech reports, whereas the verb ‘say’ amilh-a in
patient voice form can, as in example (25).
Secondly, when the verb ‘say’ ami in Actor voice co-occurs with the polite request
marker =kia in imperative constructions, the irrealis marker is not affixed to the root.
Usually, in Saaroa, if a verb in Actor voice co-occurs with the polite request marker
=kia in imperative constructions, the irrealis marker is required to occur.
(27) ami=kia!
say(av)=pr
Please say (something)!
(28) kaamu, tu-a-puru=kia!

father sit.down-irr-br(av)=pr
Father, please sit down!
Thirdly, the verb ‘say’ appears not to co-occur with any bound pronouns. Typically,
in Saaroa, a verb in Actor voice may co-occur with nominative pronouns, and a verb
inflected with patient or locative voice may co-occur with genitive pronouns.
(29) um-au-a-u=amu uuru

av-redup-irr-eat=1pl.exc.nom rice
We are eating rice.
(30) lhi-aala=mu vutukulhu

perv.asp-take(pv)=2pl.gen fish
You have caught the fish.
7 Reported speech as an evidentiality strategy

Reported speech is another means through which knowledge and the speaker’s atti-
tude to it can be expressed.
7.1 Marking reported speech

In Saaroa, reporting someone else’s speech may involve a word-for-word quotation,
which is accompanied by a verb of speaking. Unlike the languages of the Ethiopian
Plateau (Longacre 1990; Gueldemann 2001: 329), such a quotation in Saaroa does not
have its own constituent order. Besides, it is not a special clause type, and behaves like
any other complement clause. A quoted utterance has to be followed by the verb ‘say’
in patient voice. The reported speech complement clause is in square brackets.
102 Chia-jung Pan
(31) Repeated from example (25)

amilh-a pakiaturua, [ku-a-tumulhu=kia lhatu –ngu – paa-m-a-vacangu –
say-pv teacher eat-irr-a.lot=pr vegetable caus-av-stat-good
tilha'alhu
–]!
body
Teachers said, please eat more vegetables (because it) makes the body healthy!
7.2 Reported speech and reported evidential: division of labour

Although reported speech and reported evidentials do essentially the same job; namely,
they indicate that the information was acquired from someone else, in Saaroa, report-
ed speech and reported evidential differ in their semantic nuances, and their function
and usage. The reported evidential simply shows that the ultimate source of informa-
tion is someone other than the speaker. A direct or indirect speech complement with a
verb of speech is employed in order to specify the exact ‘author’ of the information. In
other words, if one uses the reported evidential, the exact ‘author’ of the information
is left unknown (hence the English translation with an indefinite ‘they’).
(32) ma-aru=ami n alhalhapa-isa

av-exist=evid obl above-3.agr
m-i<a><tu>tungusu=na m-i<pi>pilhilhi
av-Ritual.of.Sacred.Shells<irr><redup>=def av-<redup>fly
They said (they) kept flying above the Ritual of Sacred Shells.
If a speaker chooses to specify who said what, a direct or indirect speech is used.
(33) Repeated from (21)

amilh-a amalhu –=na lhi-k<um>ita=ami vuvulungaa
say-pv male.name=def perv.asp-see<av>=evid mountain
n cumi'i
obl bear
Amalhu – said he saw a bear in the mountain.
8 Evidential-like meanings through special markers

An information source can be encoded via non-evidential categories. In Saaroa,
evidential-like meanings are expressed through special markers. Their main exten-
sions are summarized in Table 2.
The experiential lhi- refers to an event or state that occurred at least once prior to
the moment of speech, and the information has directly (through seeing) or indirectly
(through hearing, smelling, tasting, or feeling) been experienced by the speaker. To
Table 2. Evidential-like meanings through markers
Evidential-like meaning Information source
Experiential Events for which the speaker takes responsibility and/or has a
personal involvement
Inferential Information which encodes inference based on reasoning or

observable evidence
Dubitative Information which mirrors speaker’s uncertain attitude to

knowledge
Mirative Information which covers sudden discovery, surprise, or unpre-

pared mind to the speaker
describe the information they saw and heard, the experiential lhi- is employed to talk
about their personal involvement.
(34) ki-a-lha-lhamu ka lhaamaama kiariari n

talk-irr-redup-talk core old.person past obl
lhi-tima-timalha=isa lha mu –u
–mu–a lhi-kita-kita=isa
expe-redup-hear(pv)=3.gen and also expe-redup-see(pv)=3.gen
The old people in the past would talk about what they had heard and also seen.
The inferential ='ai encodes information obtained by reasoning or common sense

through observing evidence of an event or a state without directly experiencing it
or through observing direct evidence of an event or a state. The inferential ='ai may
have overtones of doubt or speculation. The following example shows that the speaker
infers the exact number of speakers who are still able to speak the Saaroa language,
based on his experience of living in the same village.
(35) maacu a m-a-ca-calhia=mana

concerning link av-stat-redup-be.able.to=imperv.asp
m-asi-lha'a-lha'alua ia, umara-maalhi=cu='ai ka
av-speak-redup-Saaroa top human-ten=cos.asp=infer link
m-a-calhia m-asi-lha'a-lha'alua n kani'i
av-stat-be.able.to av-speak-redup-Saaroa obl this
kaa-ru–lhu
–cu
–=na
person.of-place.name=def
Concerning (those who are) still able to speak Saaroa, (to the speaker’s
inference) ten people of Ru
–lhu–cu– are able to speak Saaroa.
104 Chia-jung Pan
Unlike the reported evidential, the dubitative =maanai in Saaroa does not encode the
source of knowledge or the mode of knowing. In fact, they simply reflect the speaker’s
attitude to knowledge, that is, the lack of certainty on the part of the speaker or the
speaker’s reservation with respect to the accuracy of his or her statement. Very often,
the dubitative =maanai co-occurs with the inferential ='ai. The example below shows
that the speaker is not sure about the exact number of the population, thus providing
an approximate number which is based on his inference. The accuracy of his statement
is reserved.
(36) maacu a paiciana maaci iap-a ia,

concerning link ethnic.community.name if count-pv top
taia=mana='ai=maanai mata-ma-upatu – -lhu
– mu –u–mu–a
approximate=imperv.asp=infer=dub human-tens-four-tens also
As for Paiciana, if counting (its population), (to the speaker’s inference)
perhaps (it) also still approximates forty.
DeLancey (1997, 2001) defines mirativity as conveying information which is new or

unexpected to the speaker, with overtones of surprise. Aikhenvald (2012) shows that
evidentiality and mirativity are different categories. In Saaroa, expressions of mirativ-
ity may have some grammatical connection to evidential systems. As shown in exam-
ple (24), the reported evidential exhibits overtones of lack of control and subsequent
surprise on behalf of the speaker when it co-occurs with the first person participant.
The mirative =iau can occur with the full paradigm of person. However, the range
of mirative meanings, for example sudden discovery, surprise, or unprepared mind,
typically refers to the speaker, rather than the audience (or addressee) or the main
character. As shown in the following example excerpted from a traditional Saaroa
folktale, to the speaker’s surprise, a magic object which can provoke an earthquake
still exists in the Earth.
(37) ma-aru=mana=iau ka saa-saru

–u
–-ana
av-exist=imperv.asp=mir core redup-soil/dirt-loc.nomz
m-aa-'ulutii
av-be:loc/temp-a.magic.object.that.provokes.an.earthquake
Surprisingly, a magic object provoking an earthquake still exists in the Earth.
9 Verbs of perception as lexical means to express knowledge

Having lexical means for optional specification of the source of knowledge is probably
universal (Aikhenvald 2003: 1). In Saaroa, the source of knowledge can be expressed
through lexical means. This can be achieved by verbs of perception.
Saaroa speakers are appreciative and well aware of the speech explicitly specifying
the direct visual access to evidence when they narrate stories about people’s life experi-
ences. If a speaker was an eye-witness to an event involving people’s life experiences,
he would relate his firsthand visual evidence by employing the first person singular
form with the visual perception verb kita ‘see’.
(38) uka'a=cu naani ka turukuuka lhi-kita=ku

neg=cos.asp here core chicken perv.asp-see(pv)=1sg.gen
The chicken that was seen by me is not here.
To describe events or states which the speaker has heard but not seen, the verb of
auditory perception timalha ‘hear’ is used. The use of the verb of auditory perception
usually implies less certainty than the use of the verb of visual perception.
(39) aunaana ka lhi-timalha=ku na alhaama kiariari

like.that link perv.asp-hear(pv)=lsg.gen obl ancestor past
n kana m-uritalhivau – n alu
–mu–lhu
–
link pf av-have.a.love.affair obl wild.boar
That is what I heard from ancestors in the past (about the story of) having a
love affair with a wild boar.
10 Evidentiality, communication, and cultural knowledge

The function of evidentials and evidential-like meanings is to encode the source of
information. Having just one grammatical evidential, reported, in Saaroa does not
prevent speakers from maintaining communicative efficiency. Efficient communi-
cation relies on cooperation between speakers and addressees. When the reported
evidential is used, the addressee is not supposed to ask ‘how do you know this?’ This
question makes little sense.
Speakers of the Saaroa language with the reported evidentiality think it necessary
to express it with lexical means when they have to switch to a lingua franca with no
grammatical evidentiality. When Saaroa speakers use Mandarin Chinese to report
something they heard from someone else, the source is always included. This can be
considered a way of expressing Saaroa speakers’ metalinguistic awareness of reported
evidentiality.
In Saaroa, evidentials and evidential-like meanings can be paraphrased and re-
inforced with lexical items corresponding to the information source. For example,
speakers of Saaroa may rephrase the reported evidential with the verb ‘hear’. The
evidential-like meaning =maanai (i.e. dubitative) can be rephrased with the adverb
ausi ‘possibly’.
106 Chia-jung Pan
To make their story more vivid and colourful, Saaroa narrators may finish a story by
explicitly stating the source that they had learnt it from. A story containing reported
evidentials often ends with ‘this is what old people told me’ or ‘this is what I heard from
old people’. The narrators do so to make sure their addressees realize that the story
reported to them was obtained from the most reliable source.
11 Summary
Saaroa has a closed grammatical category, evidentiality. The system of evidentiality is
A3. reported versus ‘everything else’. The reported evidential formed through
the enclitic =ami is conceptualized as a genre marker, or a token of narratives. The
reported evidential might occur more than once in the same clause for pragmatic
reasons. When expressing the reported evidential, its time reference can be distinct
from that of the clause’s predicate. The reported evidential may develop mirative over-
tones, extending to cover new, unusual, and surprising information. Similarly, it may
have additional semantic overtones in the context of first-person participant (i.e. first-
person effect). Usually, the reported evidential has the whole clause within its scope,
but the evidential itself cannot fall within the scope of negation. The verb of ‘saying’ is
the source for the reported evidential.
In addition to the reported evidential, Saaroa has further means to express
evidential-like meanings. Reported speech is employed to encode information
obtained from someone else. Although reported speech and reported evidentials do
essentially the same job, they differ in their semantic nuances, and in their function and
usage. Four markers are used to encode information based on personal involvement,
reasoning, uncertainty, and surprise to knowledge. Verbs of perception are employed
as a lexical means of encoding information which has direct visual or indirect auditory
access to evidence when they narrate stories about people’s life experiences.
References
Aikhenvald, Alexandra. 2003. ‘Evidentiality in typological perspective’, pp. 1–31 of Studies
in evidentiality, edited by Alexandra Aikhenvald and R. M. W. Dixon. Amsterdam: John
Benjamins.
———. 2012. ‘The essence of mirativity’, Linguistic Typology 16: 435–85.
Casad, E. 1992. ‘Cognition, history and Cora yee’, Cognitive Linguistics 3: 151–86.
Chang, Anna Hsiou-chuan. 2012. ‘Evidentials in Paiwan’, Tajen Journal 40: 115–29.
DeLancey, Scott. 1997. ‘Mirativity: the grammatical marking of unexpected information’, Lin-
guistic Typology 1: 33–52.
———. 2001. ‘The mirative and evidentiality’, Journal of Pragmatics 33: 369–82.
Gordon, L. 1986. ‘The development of evidentials in Maricopa’, pp. 75–88 of Evidentiality: The
linguistic coding of epistemology, edited by Wallace Chafe and Johanna Nichols. Norwood,
NJ: Ablex Publishing Co.
Gueldemann, T. 2001. Quotative constructions in African languages: A synchronic and diachronic

survey. Leipzig: Habilitationsschrift, Leipzig University.
Haspelmath, Martin. 1993. A grammar of Lezgian. Berlin: Mouton de Gruyter.
Hsieh, Fuhui. 2012. ‘On the grammaticalization of the Kavalan say verb zin’, Oceanic Linguistics
51(2): 464–89.
Longacre, R. 1990. Storyline concerns and word order typology in East and West Africa. Los
Angeles: The James S. Coleman African Studies Centre and the Department of Linguistics,
UCLA.
MacDonald, L. 1990. ‘Evidentiality in Tauya’, Language and Linguistics in Melanesia 21: 31–46.
Pan, Chia-jung. 2012. ‘A Grammar of Lha'alua, an Austronesian Language of Taiwan’. PhD
thesis, Language and Culture Research Centre, Cairns Institute, James Cook University.
Thurgood, Graham. 1986. ‘The nature and origins of the Akha evidentials system’, pp. 214–22
of Evidentiality: The linguistic coding of epistemology, edited by Wallace Chafe and Johanna
Nichols. Norwood, NJ: Ablex Publishing Co.
Tournadre, N. 1994. ‘Personne et médiatifs en tibetain’, Faits de langues 3: 149–58.
Valenzuela, P. 2003. ‘Evidentiality in Shipibo-Konibo, with a comparative overview of the cat-
egory in Panoan’, pp. 33–61 of Studies in evidentiality, edited by Alexandra Y. Aikhenvald and
R. M. W. Dixon. Amsterdam: John Benjamins.
Yang, Gloria Fan-pei. 2000a. ‘The semantics-pragmatics of the hearsay evidential nana in Tsou’,
National Taiwan University Working Papers in Linguistics 3: 69–86.
———. 2000b. ‘Tsou case markers as evidentials’, National Taiwan University Working Papers in
Linguistics 3: 41–67.
5
The grammar of knowledge

in Kurtöp: evidentiality, mirativity,
and expectation of knowledge1
G W E N D O LY N H Y S L O P
1 Introduction
Kurtöp, a Tibeto-Burman language of Bhutan, utilizes a rich, grammaticalized system
of verbal affixes and clitics that encodes a speaker’s source of knowledge, expecta-
tion of knowledge, and expectation of others’ knowledge. This system of evidentiality,
mirativity, egophoricity, and epistemicity is part and parcel of the verbal system; these
categories are not optional; rather they are required in any finite clause, including
questions. The aim of this chapter is to present the Kurtöp data as a case study in the
grammar of knowledge.
In §2 I provide the background information about Kurtöp necessary to under-
stand the analysis presented in this chapter. The following five sections present the
grammar of knowledge in Kurtöp, with §3 being devoted to perfective aspect, §4
presenting imperfective aspect, §5 illustrating contrasts made in future tense, §6
presenting the copulas, and §7 discussing verbal enclitics. §8 summarizes the analy-
ses and examines the system in terms of knowledge. Terminology is defined as it is
introduced.
1
I gratefully acknowledge the support of many individuals and institutions in the development of this
chapter. Research on Kurtöp has been generously funded by the Endangered Languages Documentation
Project, the National Science Foundation, the University of Oregon, and the Association for Asian Studies.
Much of the analysis presented here has benefited from lengthy discussion with Scott DeLancey, Karma
Tshering, Kuenga Lhendup, Pema Chhohyel, Tshering Penjor, Kezang Uden Penjor, and Avery Andrews.
The workshop organized by Alexandra Aikhenvald and R. M. W. Dixon proved to be the ideal place to
discuss the Kurtöp system as a whole, and I am grateful to the organizers and other presenters for their
comments. Of course, I alone am responsible for any errors.
5 The grammar of knowledge in Kurtöp 109
2 Background
Kurtöp is a Tibeto-Burman (East Bodish) language of Bhutan, spoken by approxi-
mately 15,000 people in the northeastern region indicated in Figure 1. East Bodish
languages are considered to be linguistic cousins of Classical Tibetan (e.g. Bradley
1997) but have been so heavily influenced by Tibetan that the actual phylogenetic
relationship may be obscured (cf. Hyslop in press).
Bhutan is a very small country (estimated population of 700,000) and not a country
of strangers. When meeting for the first time, people are quickly able to find com-
mon ground through a shared relation or friend. In addition, Bhutan is a highly-
multilingual society and the average Bhutanese speaks at least three to four languages,
if not more. In the east, many people speak Tshangla (lingua franca of the east), Dzong-
kha (national language), their own language, and often neighbouring languages as
well. Most Bhutanese also speak some Nepali, Hindi, and English. Tibetan is another
common second language, especially in the north. Classical Tibetan, as the liturgical
language, enjoys great prestige and has no doubt had tremendous influence in the past
and continues to influence the languages.
The data presented in this chapter have come primarily from texts which were col-
lected during fieldwork from 2006 to 2011. In total, the corpus consists of 19 tran-
scribed and translated texts from 15 different speakers representing both female and
male individuals who range in age from their early 20s to mid 70s with a broad range
of education and employment levels, social and marital status, and who reside in and
outside the village. In total, the corpus consists of over 500 pages of transcribed and
translated texts. The methodology for the analysis also included elicitation, especially
Tibet
Kurtöp Dzala
Bumthap
Dakpa
Nepal Bhutan Hengke Chali
Khengkha
India
Bangladesh
Figure 1. Kurtöp language area within East Bodish context in Bhutan

110 Gwendolyn Hyslop
to cross-check speaker intuitions and check for negative data. All elicited examples
have been checked with at least two different speakers.
2.1 Grammar
Like most Tibeto-Burman languages and most languages of South Asia, Kurtöp has
SV/AOV constituent order. As is the case in other languages of South Asia, light verbs
(e.g. friend do = ‘help’, swim do = ‘swim’; eye do = ‘see’, etc.) are a common construction.
Constituent order in the NP is usually Dem Gen N Adj Num with case markers attached
to the end of the phrase as an enclitic. Kurtöp has ‘pragmatic ergativity’ (see Hyslop 2010
for an analysis) in which an ergative marker may not be required on a bivalent verb and
can sometimes occur on a monovalent verb; the ergative is often (though not exclu-
sively) associated with unusual pragmatic factors. Kurtöp is more agglutinating than
isolating, with a verb usually consisting of a monosyllabic stem and up to four syllables
of tense/aspectual/mood/evidential material as suffixes/enclitics. Verbal arguments are
often omitted from discourse. The basic Kurtöp verb is shown in Table 1.2
Nominalization is an integral part of Kurtöp grammar, as is common in Tibeto-
Burman. The important role of nominalization in Tibeto-Burman languages has been
noted at least since Matisoff (1972) and since then several articles, theses, and descrip-
tive grammars have continued to note the central role nominalization has in scores
of Tibeto-Burman languages. Among the most influential have been Noonan (1997),
DeLancey (2002), Bickel (1999), Genetti et al. (2008), and many others. DeLancey
(2011) shows how nominalizations often lead to innovative main clause grammar, dia-
chronically, which is also happening in Kurtöp (Hyslop 2011a). Interestingly, some (if
not all?) of the Kurtöp nominalizers have been borrowed from Tibetan; these include
Table 1. Syntagmatic diagram of the Kurtöp verb
Negation Stem Suffix
ma-, me-, -TAME
mi- -SUBORDINATOR
má-, mé-,
mí-
2
Negation is encoded as a prefix to the verb stem and verbal suffixes encode both finite properties (such
as tense, aspect, modality, evidentiality, etc.) and subordinate properties (such as encoding that a clause is
conditional, infinitive, or subordinate in other ways). See Hyslop (2011a) for a more detailed analysis of the
Kurtöp verb and verb phrase.
the denizen -pa, imperfective -khan, and possibly the perfective -pala (discussed
below), though a detailed study of the origin of the nominalizers is well beyond the
scope of this chapter.
Not surprisingly, Kurtöp basically has two clause types: one that is formally nomi-
nalized (as in (2)) and one that is not (as in (1)).3,4
(1) khit ge-shang

3.abs go-perv.ego
He went.
(2) khit gewala wenta

khit ge-pala wenta
3.abs go-nomz:perv cop.eq.mir
He went indeed!
Another important facet of Kurtöp grammar is clause-chaining, whereby at least one

converb5 is suffixed with the non-final suffix (-si) and (eventually) followed by a final,
finite verb. A simple example can be seen in example (3), where the non-final marked
verb and the final verb encode two separate events. In practice, the Kurtöp Clause-
chaining Construction can be used for a wide variety of functions, one of which is
temporal sequencing. Clause-chaining is also widely reported in Tibeto-Burman lan-
guages (e.g. Genetti 2005; Coupe 2007).
(3) tsheni iguthe cozi boi bishang

tsheni igu-the co-si boi bi-shang
then letter-def make-nf 3.erg give-perv.ego
Then after making the letter, they gave (it).
Regardless of whether a finite clause ends with a copula or a tensed verb, any of three
possible verbal enclitics may be used as well. A verbal phrase level enclitic6 may attach
3
Data in this chapter are presented using three or four lines of analysis as needed. If a given form or
forms are associated with allomorphy in Kurtöp, four lines of analysis will be used. In these cases, the first
line indicates pronunciation in natural speech while the second line indicates the abstract representation
of the morphemes. In cases where the pronunciation and abstract representation are the same, the second
line is omitted; that is, only three lines of analysis are presented: pronunciation/morphological analysis,
glossing, and translation.
4
Data symbols correspond to the IPA as follows: <k> [k], <kh> [kh], <g> [g], <ng> [ŋ], <c> [c], <ch>
[ch], <j> [ɟ], <ny> [ɲ], <tr> [ʈ], <thr> [ʈh], <dr> [ɖ], <t> [�], <th> [�h], <d> [� ], <p> [p], <ph> [ph], <b>
[b], <m> [m], <ts> [ts], <tsh> [tsh], <sh> [ç], <zh> [ʝ ], <s> [s], <z> [z], <l> [l], <lh> [�], <r> [r], <a> [ɑ],
<e> [e], <i> [i], <o> [o], <u> [u], <ö> [ø], <ü> [y], <’CV> high tone on following vowel.
5
I am using ‘converb’ in the sense of (Haspelmath 1995: 4); that is, a subordinate verb which has the
primary function of marking adverbial subordination.
6
These are separate from enclitics that occur on nominal elements; see Hyslop (2011a) for more details
about other clitics.
to the end of a tensed verb or a copula (i.e. right edge of a clause). The forms and
functions of these enclitics are summarized in Table 2. The reported speech clitic may
actually attach to any clause type, not just finite clauses.
The following examples show the reported speech enclitic attached to a finite verb
in (4), and to a copula in (5) and (6).7
(4) [khit ge-shang]=ri

3.abs go-perv.ego=rep
(I heard that) he went
(5) khit gewala wentari

khit ge-pala wenta=ri
3.abs go-nomz:perv cop.eq.mir=rep
(I heard that) he went indeed!
(6) khit chorten kora thungzi nawalri

khit chorten kora thung-si
3.abs stupa circumambulation do-nf
nawala=ri
cop.exis=rep
(I heard that) s/he keeps circumambulating the stupa
A clitic may attach to an already cliticized verb or clause, so that something like (7),
with a negative prefix and two enclitics is possible. In elicitation, speakers accept exam-
ples such as (7) readily, and any combination of up to two enclitics appears to be pos-
sible; that is, gewalamiri, gewalarisa, gewalamisa, gewalasami, and gewalasari are also
possible utterances, though occur rarely in the corpus. There are no examples in the
Table 2. Verbal enclitics
Enclitic Value
=ri Reported speech
=sa Counter-expectation
=mi Tag
=wu Tag
7
A ‘chorten’, or ‘stupa’, is a Buddhist monument, perhaps best construed as a small temple, that houses
one or more sacred relics. Buddhist followers may walk around these chortens counter-clockwise in order
to gain religious merit and thus increase one’s chances for enlightenment, or at least a more noble rebirth
in the next life.
textual database with all three clitics used together and speakers do not accept such
combinations as being possible Kurtöp utterances.
(7) magewalarimi
ma-ge-pala=ri=mi
neg-go-perv=rep=tag:exc
I heard he went, right?
In addition to the grammatical systems presented here, Kurtöp has other means to
encode certainty, source of knowledge, etc. For example, the verbs bran ‘know’, ta
‘watch’, mik thung ‘see’, lap ‘say’, ko ‘hear’, khot ‘tell’, go ‘need; want’ are interrelated
with certainty, source of knowledge, and speaker expectation, though should not be
considered evidential, mirative, egophoric, etc. markers as such. Another interesting
fact about Kurtöp, which warrants further discussion elsewhere is the difference in
the auxillaries tshuk and ’lot. Both can be glossed as ‘be able’ but the difference is one
of expectation; tshuk is used when the actor would be expected to complete the action
but ’lot is used when the actor would not normally be expected to complete the action.
Other related forms not further discussed in this chapter are the quotative ngaksi and
the parenthetical ngai tamo (1-gen see-ctm) ‘In my opinion’.
3 Perfective
Kurtöp makes a five-way contrast in perfective aspect. Figure 2 summarizes the con-
trasts made between whether the speaker is certain or uncertain, whether the speaker
has personal knowledge or not, whether the knowledge was unexpected or not, and
whether the speaker expects the interlocutor to share the knowledge or not. This sec-
tion illustrates the difference between the five forms.
Epistemic Value
+ Certainty – Certainty
-para
+ Personal – Personal
Knowledge Knowledge
-mu
+ Unexpected – Unexpected
-na
+ Shared Experience – Shared Experience
-pala -shang
Figure 2. Kurtöp perfective aspect suffixes marked for epistemic value
3.1 Egophoric -shang

Kurtöp -shang is used to encode perfective aspect with direct evidential value when
the speaker has direct evidence of the experience and there is no expectation that
another speech-act participant would have direct evidence. I am using this term in
a sense similar to that of (Tournadre 2008: 295) for whom ‘egophoric’ expresses per-
sonal knowledge or intention on the part of the actual speaker. Tournadre (2008: 297)
describes a contrast between the narrow and broad scope of egophorics. The use here
in Kurtöp (where expectation of others’ knowledge is relevant) is more similar to nar-
row scope, though the Kurtöp category appears to be slightly different than the Tibetan
category. This is usually used with the first person, as in (8).
(8) khici mengya zhit ge-shang da

3.gen name-also forget go-perv.ego now
I also forgot his name now.
However, egophoric -shang can also be used with the third person when the interlocu-
tor is not expected to have direct knowledge of the event, as in (9), which was said by
a person who had worked closely with the given lama, had direct knowledge of the
lama’s death, and knew that his interlocutor was not already aware of the event.
(9) zheng phep-shang

heaven arrive.hon-perv.ego
(the lama) passed away (lit. arrived in heaven).
In addition to these two contexts, the egophoric can also be used with the second
person, if the speaker does not expect the interlocutor to have the knowledge, for
example in the case of a doctor telling a patient news about the patient’s condition.
While -shang is canonically used with the first person, it is not used when speakers are
narrating their dreams, in which case either the mirative -na or inferential -mu would
be used. See Hyslop (to appear) for more examples and a fuller analysis of egophoric-
ity in Kurtöp.
3.2 Unmarked -pala ~ -wala ~ -sala

The form -pala is diachronically composed of nominalizer -pa + -la (possible copula;
cf. (Hyslop 2011a)). Kurtöp -pala encodes expectation that someone else has direct
knowledge of the event. This form usually occurs with a third-person reference, as in
(10), which was drawn from a short story in which a character had entered a town and
ordered a woman to be locked in her house for one week. After six days, before the
time was up, the villagers decided to open the door. The narrator uttered this example,
relaying a fact known to the villagers (and the interlocutors who had been listening)
at the time.
(10) dutshot matshangwala

dutshot ma-tshang-pala
time neg-be.complete-perv
(as people knew) The time wasn’t up.
However, the form -pala can also occur with the first person if the interlocutor also
has direct knowledge of the event, as in (11).
(11) tshachu bang niye ’yaura drâ ngâ lappala ngai

tshachu bang ni-le ’yau=ra drak
hot.springs bathe stay-imp dem:up=emph be.good
ngaksi lap-pala ngai
quot tell-perv 1.erg
‘Stay in the hot springs up there; it is good’, I said (as you know).
Example (11) was drawn from a conversation between two speakers who had grown up
in the same village, were similar in age, and very familiar with details of each other’s
lives. They shared many experiences together, including a trip to local hot springs,
which was the topic of conversation in this portion of the conversation.
Perfective -pala is also commonly used in second-person statements and related
to the perfective question marker -pa, which shares the same allomorphy as -pala
(-w before -r, -k, -ng, old -l stems). It is important to note that -pa is used as the per-
fective yes/no question marker when the speaker expects the interlocutor to have the
answer, canonically as in second-person questions. Hyslop (2011a and to appear) pro-
vides more examples of the use of perfective -pala and Hyslop (2011a) discusses ques-
tion formation in general.
3.3 Indirect -mu

The suffix -mu is of uncertain etymology but it is possibly etymologically related to
the Marphatan Thakali ‘copula’ which encodes an event located in the past but with
present relevance (Georg 1996: 118–22). In Kurtöp, the perfective -mu encodes indi-
rect evidence, or that the knowledge was inferred. As -mu encodes grammaticalized
source of information, this form is the only true evidential form in perfective aspect
(cf. Aikhenvald 2004, Chapter 1; see also Chapter 3, Chapter 7, and Chapter 2 of this
volume, for systems with evidentiality).
Consider example (12):
(12) tshe khit gimu

tshe khit ge-mu
dm 3.abs go-perv.infer
Then he left.
This example was drawn from a story about a popular mythological figure in Bhutan
who is known for his unusual behaviour. In this particular narration he had come to
a village and caused many strange events to take place. One day quite suddenly, the
villagers noticed he was gone and the unusual events had stopped. They did not see
him leave and only had indirect evidence that he had left. Thus, the form -mu is used.
The inferential perfective -mu is most often used with the third person in cases such
as in (12), when the speaker (or actor in the case of narration) has indirect evidence
from which s/he infers his/her knowledge. However, -mu can be used with the first
or second person should the unusual context arise in which the speaker or actor has
gained their knowledge through inference.
3.4 Mirative -na

The mirative perfective copula -na is probably a recent grammaticalization of the exis-
tential copula /nak/. Synchronically, it marks mirativity, or that the completion of an
event was unexpected to the speaker. DeLancey (2001: 371) defines mirativity as ‘lin-
guistic marking of an utterance as conveying information which is new or unexpected
to the speaker’ and my own use here is similar, wherein I assume mirativity to encode
that the knowledge or event was unexpected to the speaker.
Examples are shown in (13)–(14):
(13) palang=gi jedo thilathe darnari

palang=gi je=to thila-the
bed=gen top=loc thumb/big.toe-def
dar-na=ri
remain-perv.mir=rep
On the bed remained a toe! (it is said)
(14) chak-na
land-perv.mir
It landed!
The data in (13) are drawn from a story about Drukpa Künle, also called the Divine
Madman, a mythological figure known for his unusual behaviour in Bhutan. In the
story from which (13) was drawn, he had locked an elderly woman in a room for
one week and instructed the villagers not to open the door. After six days, however,
the villagers began to question their decision to obey him and opened the door. Her
body had been in the process of being transmitted into heaven and, since the entire
week had not completed, her big toe still remained. The presence of only a toe was
clearly unexpected from the point of view of the character in the story and hence the
mirative is used.
Example (14), also with third person, was uttered by a child watching a paraglider.
Villagers had gathered on top of a hill to watch a group of paragliders take off and soar
through the valley. A group of children was watching one particular paraglider as he
floated down, behind a group of trees, and out view of the children. The next the chil-
dren saw he had landed on the valley floor. Although this action was not unusual or sur-
prising in the same way as (13), it was unexpected; hence the choice to use the mirative.
While the mirative perfective tends to be used with third-person actants, it can also
be used with first or second person, if the speaker had not expected the event. Hyslop
(2011b, to appear) offers more detailed examples and analyses of the Kurtöp mirative.
3.5 Assumptive -para ~ -wara ~ -sara

The assumptive perfective is diachronically composed of nominalizer -pa plus -ra,
itself likely derived from the verb ra ‘to come’. The form -para encodes that the speaker
is not certain of their knowledge, generally because they are assuming or presuming
something to be the case.
For example, in (15) the actors in this story are looking for the king’s hunting dog
and speculate that the dog had gone up a particular hill. They have no evidence for
this, they are just postulating an idea; it is not a fact, which is underscored by the use
of the adverbial yangna ‘perhaps’.
(15) yangna woyenang gewara ngaksi

yangna woye=nang ge-para ngaksi
perhaps dem:up=loc go-perv.assum quot
‘Perhaps (the hunting dog) has gone up there’ (they) said
Other examples in the text show this form used in other contexts as well, for example
when the speaker has indirect evidence for an event but is not certain, or when a
speaker suddenly doubts that something may be the case.
4 Imperfective aspect
Kurtöp marks a two-contrast in imperfective aspect between clauses which are mira-
tive, or unexpected to the speaker (or actor) and those that are not unexpected. This
contrast is illustrated in Figure 3.
Epistemic Value
-ta -taki
Figure 3. Kurtöp imperfective aspect suffixes
4.1 Mirative -ta

The mirative imperfective marker is possibly a grammaticalization from the verb tak
‘become’. Mirative -ta tends to be used more often for third-person actants, as in (16):8
(16) Rimpoche nango tun-ta tshe khira=ya
Rimpoche loc show-imperv.mir dm 3.refl=also
me-bran-ta
neg-know-imperv.mir
They showed (the pictures) to Rimpoche, and even he (Rimpoche) didn’t
know (the pictures)
Example (16) shows the mirative imperfective in two clauses. The first clause has a
third-person plural A argument and Rimpoche as the O. The A argument here refers
to a group of westerner Buddhists who were visiting the Rimpoche and, rather than
focus on listening to the teachings of the Rimpoche, were showing pictures to the
Rimpoche. This is an unexpected event as in Bhutanese society when one visits a Rim-
poche for teaching one should be quiet and unassuming, listening to the Rimpoche,
and not instigating new actions. In this context, however, the visitors began to show
pictures to the Rimpoche, and thus the speaker, who is reporting on this experience
he witnessed, uses the mirative imperfective. In the second clause in this example, the
point of view is shifted to the Rimpoche, but the mirative is kept. The pictures were of
the Rimpoche—the western visitors had seen the Rimpoche previously and taken his
picture—but the Rimpoche had not seen them. Again, the mirative is used as it was
unexpected that the Rimpoche would not be familiar with pictures of himself.
There is a wealth of examples of the mirative imperfective in the corpus and it occurs
most frequently with third-person actants. However, there are examples of the mira-
tive with the first and second person in instances when the speaker or actor had not
expected the knowledge. The mirative imperfective can be used in questions when
the speaker does not expect that the interlocutor would be certain of the answer, for
example when the speaker asks the interlocutor about a third person with whom the
interlocutor is not particularly close. Further examples and discussion of the mirative
imperfective in Kurtöp can be found in Hyslop (2011b).
4.2 Non-mirative -taki

For non-mirative contexts, or when the event was not unexpected for the speaker,
the imperfective form -taki is used, which is possibly derived from the verb tak
‘become’ plus nominalizer -ki. -taki is used if the event is not new knowledge, for
example if the speaker is self-reporting or reporting knowledge they previously
acquired. Examples are given in (17–18). The first example is from a story in which
8
A Rimpoche, or Rinpoche, is a high-level Buddhist practitioner, perhaps similar to a cardinal in the
Catholic church.
the king’s servants are looking for the king’s hunting dog and reach a couple’s house in
the wilderness. The couple tell the servants that there is no hunting dog with them and
that, in fact, no one ever comes to visit them; since they are reporting on old knowledge
they use the non-mirative.
(17) neci tshô khep miyang methraktaki

neci tshô khepo mi=yang me-thrak-taki
3.pl.gen here foc people=also neg-arrive-imperv
No one comes to our place.
In (20) the non-mirative is used for third-person self-reporting:
(18) tshe darung khit na-taki ngaksi khit

dm again 3.abs be.sick-imperv quot 3.abs
se-taki ngaksi
die-imperv quot
And again shei said that shei was sick, that shei was dying . . .
While the non-mirative tends to be used most commonly with first-person actants, it
can also be used with second- and third-person if the knowledge was not unexpected,
for example if the speaker is reporting on knowledge they gained previously or is
self-reporting. If the speaker has knowledge for a second- or third-person actor, the
form -taki can also be used in those cases as well. When asking imperfective questions
-taki is used if the speaker expects that the interlocutor already has the knowledge, for
example if the speaker is asking the interlocutor questions about him-/herself or about
someone with whom the interlocutor is very close.
5 Future tense
Kurtöp makes a two-way contrast in the future tense with regard to certainty; the suffix
-male is used if the speaker is certain about a given event in the future, while the verb
stem is unmarked if the speaker is uncertain. This contrast is illustrated by Figure 4.
Epistemic Value
-male -Ø
Figure 4. Kurtöp future tense
5.1 Certain -male

Though -male is usually found in texts in subordinate clauses, it also occurs as finite
verbal morphology marking future tense with speaker certainty. It is most often used
with first person statements and in the following greeting/common question:
(19) ’ au ge-male
where go-fut
Where are you going?
Future suffix -male is also commonly used for first-person statements about future
actions but can also be used for third-person statements if the speaker is certain about
an event, as in (20), which was drawn from a narration about life in the village and
reports on what future events will take place later in the year. As a person from the
village, the speaker is familiar with local events and speaks with authority about which
events will occur in the future.
(20) mi gap dangsa tatmung tazi nimale

mi gapo dangsa tatmung ta-si ni-male
person pl.foc all show watch-nf stay-fut
All the people will be watching the show
5.2 Uncertain -ø
If the speaker is unsure of the event, a verb can be left unmarked as way to encode the
uncertain future. In elicitation without further context, this form will be given with
second- and third-person actors. Consider (21):
(21) Rinzin tsikpa za

Rinzin anger become
Rinzin will (probably) become angry.
Example (21) was drawn from a conversation between two people discussing a third
person they both knew, but were not intimately associated with.
6 Copulas
Kurtöp has a rich set of copulas making a contrast between existential/equational
and affirmative/negative while also encoding a variety of knowledge-related catego-
ries. In addition to performing the canonical copular functions, such as possession,
equation, location, etc., the copulas are integrally entwined in main clause grammar,
predicating clausal nominalizations and occurring with the clause-chaining construc-
tion to encode durative aspect (cf. §2.1). As such, the copulas also encode a wealth of
knowledge-related contrasts. The four copular bases are shown in Table 3.
The equational copulas /wen/ and /min/ are used to predicate equative clauses while
/nak/ and /mut/ are associated with existential clauses, location, and possession. Each
of these ‘bases’ has a rich array of possible forms representing various knowledge-
related categories. Rather than illustrate all the contrasts here, the forms will simply
be discussed.
Table 3. Copular bases
Form Function Polarity
/wen/ Equational Positive
/min/ Equational Negative
/nak/ Existential Positive
/mut/ Existential Negative
6.1 Existential copulas

Kurtöp makes a four-way contrast amongst affirmative copulas and a five-way contrast
among negative existential copulas, with regard to various epistemic values, as shown
in Table 4.
In affirmative contexts the contrast is between mirativity and non-mirativity on
the one hand, and certainty of event or result on the other. The form nawala is used
for contexts when the speaker has direct, ingrained evidence and is certain while nâ,
Table 4. Existential copulas. Affirmative forms begin with na- while negative
forms begin with mu-. A greyed cell indicates that the particular contrast is not
relevant, while ‘+’ indicates a form positively encodes a given contrast and ‘–’
indicates a given form negatively encodes a given contrast.
Source Certainty
Form Direct Indirect Expectation Event Result
nawala +
mû
nâ –
mutna
nawara –
mutpara
mutle +
mutlera + –
Epistemic Value
+ Personal – Personal Presumption Doubt

Knowledge Knowledge nawara naki
mutle mutla
nâ nawala
mutna mû
Figure 5. Kurtöp existential copulas. Forms beginning with /n-/ are affirmative copulas while
forms beginning with /m-/ are negative copulas
the mirative form, is used if the speaker recently acquired the knowledge or if the
knowledge was unexpected. Like nâ and nawala, naki can be used if the speaker has
direct evidence of something, but unlike nawala and nâ, use of naki indicates the
speaker is no longer certain. Finally, unlike the previous three, nawara encodes that
the speaker is not sure of the entire proposition, either because they are postulating
something, speculating, or have other reasons to be doubtful. An additional form
that does not fit into the knowledge-related categories is naksho, which is used for
emphatic contexts.
In negative contexts the contrasts are somewhat different. The contrasts between
unmarked, certain, direct mû (counterpart to nawala), mirative mutna (counterpart
to nâ) and assumptive mutpara (counterpart to mawara) appear to be identical to that
for the affirmative counterparts. However, an additional evidential contrast is made;
mutle encodes indirect evidence, or inference, and mutlera encodes indirect evidence
combined with doubt. That is, mutlera marks that the speaker speculates his/her
knowledge of the absence of something would be gained through inference (Figure 5).
6.2 Equational copulas

There are five affirmative and negative equative copulas, showing contrasts similar, but
not identical, to those reported for the existential copulas (Table 5).
In the affirmative and negative paradigms, a contrast is made in certain, direct evi-
dence contexts between non-mirative (wen; min) and mirative (wenta; minta) con-
texts. The non-mirative forms are used for old knowledge for which the speaker is
certain while the mirative forms are used when the knowledge was new or unexpected
to the speaker. Further, the affirmative mirative wenta is commonly used in story-
telling, ending a clausal nominalization. Speakers report that the use of mirative wenta
rather than wen in these cases help make the story ‘more interesting’.
Table 5. Equational copulas. Affirmative forms begin with we-, while negative
forms begin with mi-. A greyed cell indicates that the particular contrast is not
relevant, while ‘+’ indicates a form positively encodes a given contrast and ‘–’
indicates a given form negatively encodes a given contrast. A ‘?’ indicates that it
is not as of yet entirely clear what the core contrast is.
Source Certainty
Form Direct Indirect Expectation Event Result

wen +
min
wenta –
minta
wenpara –
minpara
minle +
minlera + –
weni – (?) – (?)
wenim
There are three additional affirmative forms which encode lack of certainty: wen-
para encodes assumption, wenim encodes less certainty, though the speaker may still
have some evidence for the claim, and weni entails that the speaker is not at all certain.
This three-way contrast in (un-)certainty is not found elsewhere in the language, and
it is as of yet unclear how this fits into the system as a whole.
The negative equational copulas, on the other hand, have a distribution iden-
tical to the negative existential copulas; minpara encodes that the speaker is
uncertain about or assuming the absence of something; minle encodes that the
speaker gained their knowledge of the absence of something through inference,
and minlera combines these last two categories into a complex bundle which
marks that the speaker assumes there is indirect evidence for the absence of
something.
7 Particles
In addition to encoding evidential, mirative, egophoric, and epistemic contrasts
throughout the main clause grammar, Kurtöp has particles and clitics that are part of
the grammar of knowledge. In this section I present first a question particle which is
used when the speaker does not expect the hearer to have the answer, as well as three
enclitics which encode the source of knowledge or expectation of knowledge. These
are summarized in Table 6.
Table 6. Kurtöp particles and enclitics encoding knowledge. A greyed cell indi-
cates that the particular contrast is not relevant, while ‘+’ indicates a form posi-
tively encodes a given contrast and ‘–’ indicates a given form negatively encodes
a given contrast.
Expectation of knowledge
Form Grammatical Function Indirect source Context Interlocutor
shu Question particle – –
=ri Reported speech +
=sa Counter-expectation –
=mi Tag –
=wu Tag +
7.1 Unexpected shu

Kurtöp question formation is a complicated phenomenon, a full description of which is
beyond the scope of this chapter (though see Hyslop 2011a for a brief analysis). In short,
a combination of verbal morphology and sentence-final particles are used in polar and
information questions, depending on the tense/aspect and evidential/epistemic/mira-
tive value of the clause. However, if the speaker does not expect the hearer to have the
answer, the form shu may be used, either in replacement of a question particle (in case
the syntactic context requires it) or following the required question form of the predicate.
Consider (22) and (23):
(22) zhunggi . . . dazin ngako matshunani zhâ ngaksi bretakyo ngaksi

zhung=gi dazin ngak-to ma-tshuk-nani
government= erg care do-infin neg-be.able-cond
zhâ ngak-si blek-taki yo ngaksi
what do-nf keep-imperv qp quot
The government . . . says ‘if you aren’t able to care (for the dogs) then why
keep (them)?’
(23) ngai=ta zhâ lap-male shu da

1.erg=emph what speak-fut qp.dbt now
Now what should I talk about?
The first example, drawn from a conversation between two friends, involves a context
in which a government official comes to someone’s home to fine them for not taking
care of their dogs. The government official would use the question marker yo, expect-
ing that the hearer (the dog owner here) would know the answer.
This context can be contrasted with (23) which was drawn from a recording of an
interview about rice cultivation. Prior to the interview the recorders discussed the
topic of the recording and prepared the speakers for the conversation. As the recorder
was turned on and the first question was asked, the speaker uttered (23). The use of shu
here indicates that the speaker did not expect the interlocutor(s) to have the answer;
rather, it is a hypothetical question, almost akin to what we would consider thinking
out loud in English.
7.2 Reported speech =ri

Of the four optional enclitics a speaker may use at the end of a verb phrase, one encodes
an oral source of information. It is common in story-telling for at least one clause to be
marked with =ri, as in example (24):
(24) palang=gi jedo thilathe darnari

palang=gi je=to thila-the
bed=gen top=loc thumb/big.toe-def
dar-na=ri
remain-perv.mir=rep
On the bed remained a toe! (it is said)
In this example, and in other cases where =ri is used in story-telling it seems as though
the clitic has scope over the entire discourse. At other times, the scope of =ri is limited
to the clause. The scope of =ri can be even narrower; for example, =ri can attach to
any word, indicating direct speech. In this way, =ri is similar to the quotative ngaksi,
though a full analysis of the difference between the two in quoted speech is beyond
the scope of this chapter9 (cf. Chapter 6 for a discussion of a quotative as an evidential
strategy).
7.3 Counter-expectation =sa

The clitic =sa can occur at the end of any finite clause as a means to encode that the
action or event is counter to the speaker’s expectation. The scope and detailed use
possibilities are a matter of continued investigation, but two examples suffice to show
the use of =sa as encoding that something is counter to expectation. In (26) =sa marks
9
Note that quotative ngaksi is identical in form to the non-final-marked converb ngak-si ‘do-nf’. The
quotative, however, shows a greater amount of phonological reduction than the converb, commonly occur-
ring as ngak, ngâ, or simply ng.
that the event (a state) is counter to (cultural) expectations while in (25) =sa encodes
that the result of the event is counter to expectation.
(25) ’au nawori . . . ngai koshangsa

’au nak-po=ri ngai ko-shang=sa
where cop.exis-qp:perv=rep 1.erg hear-perv.ego=cexp
Where (did I hear) (he) was? I thought I heard (to self).
(26) gatasa khwi gap

gata=sa khwi gapo
enjoy-imperv.mir=cexp dog foc.pl
The dogs really enjoy (life here)
The use of =sa as counter-expectation in these examples can be understood as follows.

In (25), although the speaker heard, he cannot remember; the result is counter to the
expectation. Example (26) is best understood in the cultural context of Bhutan, where
dogs generally have poor living conditions, most often as street dogs, and are consid-
ered a low form of life. The dog’s life in the USA, on the other hand, is quite lavish.
Upon seeing a dog while he was engaged in conversation with a friend in the USA,
the speaker of (26) uses the counter-expectation particle, noting that his observation
is counter to what would be expected.
7.4 Tags
The final two verbal particles are tags which, in addition to bringing the speaker into
the discourse in typical tag function, also evidence a contrast between speaker expec-
tation of hearer knowledge. The form =mi is used when the speaker does not expect
the hearer to have the knowledge while the form =wu is used if the speaker thinks the
hearer shares the knowledge.
The exclusive10 tag =mi is commonly used in conversations between two speakers
who do not know each other well, when one speaker does not expect the other to share
the knowledge. An example is:
(27) tshe khit nya thungmo nya zhiknami

tshe khit mya thung-mo mya
dm 3.abs arrow do-ctm arrow
zhik-na=mi
be.hit-perv.mir=tag.exc
While playing archery he was hit by an arrow, right
10
In order to capture the difference between the two tags, I use the terms ‘inclusive’ and ‘exclusive’. The
‘inclusive’ term indicates that the speaker expects the interlocutor to share the knowledge, while the ‘exclu-
sive’ tag indicates the speaker does not have this expectation of shared knowledge.
The exclusive tag is often common in story-telling and procedural texts; in both cases
it can be assumed the interlocutor is hearing something for the first time. If the inter-
locutor is already familiar with the knowledge—for example if the hearer and speaker
shared an experience or if the speaker is talking about a second person—the use of
=mi may be seen as sarcastic and rude.
Like the exclusive tag, the inclusive tag =wu brings the hearer into the discourse,
but the expectation that the speaker does not share the knowledge is no longer there.
Rather, the use of =wu is limited to contexts when the speaker expects the hearer to
have the knowledge; usually this is in conversations between two people who know
each other well and are already familiar with many shared events, or those who already
have a common base for other reasons. Consider (28):
(28) khwi gap le daknawu

khwi gapo le dak-na=wu
dog pl.foc sins be.cleansed-perv.mir=tag.inc
The dogs are cleansed of their sins, right
This example, drawn from the same conversation as (27), shows the inclusive tag being
used despite the fact that the speakers are still getting to know each other and have not
been closely associated with the details of each other’s lives. However, they are both Bhu-
tanese and both living in the USA. They share a strong cultural bond and know the dif-
ference between a dog’s life in the USA and a dog’s life in Bhutan. They are further both
Buddhists, and believe that the difference between the two scenarios must be due to previ-
ous actions from the dogs’ former lives. Bhutanese dogs have a miserable life due to sins
they committed in a previous life, while American dogs must have been cleansed of their
sins in order to be granted such a comfortable life in the USA. This knowledge is intui-
tively shared between the speakers, and thus the inclusive tag =wu is used in this context.
8 Putting it together
The Kurtöp ‘knowledge’ system comprises a rich, complex set of forms that make up
the required heart of main clause grammar. It is impossible to utter a finite clause in
Kurtöp without keeping track of the source of knowledge, expectation of knowledge,
certainty of knowledge, or, often, a combination of some or all of the above. Perfec-
tive aspect contrasts five separate forms, while two are made in imperfective aspect
and future tense each. Copulas, which occur commonly in clause nominalizations and
clause-chaining, also contrast several forms which encode various shades of evidenti-
ality, mirativity, egophoricity, and epistemicity. In addition, in questions a contrast is
formally marked between questions to which the speaker expects the hearer to have the
answer and those to which the speaker does not expect the hearer to have the answer.
Finally, there are two phrasal enclitics which can optionally be used to encode an oral
source of information (evidentiality) or counter-expectation, and two additional tags

which encode a contrast between whether or not the speaker expects the hearer to
have shared the information. These forms are summarized in Tables 7 and 8. Clearly,
in Kurtöp, the source of knowledge, expectation of knowledge, and certainty of knowl-
edge are integral to the grammar.
Tables 7 and 8 summarize the grammar of knowledge in Kurtöp, according to tense/
aspect, copular contrasts, and particle/clitics.
Table 7. Kurtöp evidential and related forms in the non-copular verbal domain.
A greyed cell indicates that the particular contrast is not relevant, while ‘+’ indi-
cates a form positively encodes a given contrast and ‘–’ indicates a given form
negatively encodes a given contrast.
Source Expectation Certainty
Interlocutor’s
Structural Category Form Direct Indirect Context knowledge Event Result
Perfective -shang + –
-pala +
-na + –
-para –
-mu +
Imperfective -ta –
-taki
Future -male +
-ø –
Particle shu –
Enclitics =ri +
=sa –
=mi –
=wu +
Table 8. Kurtöp evidential and related forms in the copulas. A greyed cell indicates
that the particular contrast is not relevant, while ‘+’ indicates a form positively encodes
a given contrast and ‘–’ indicates a given form negatively encodes a given contrast.
Source Expectation Certainty
Interlocutor’s
Structural Category Form Direct Indirect Context knowledge Event Result
Affirmative Existential nawala +

Copula
nâ –
nawara –
naki –
Negative Existential mû +
Copula
mutna –
mutle +
mutlera + –
mutpara –
Affirmative Equational wen +

Copula
wenta – –
wenpara –
weni –(?) –(?)
wenim –(?) –(?)
Negative Equational min + +

Copula
minta +
minle +
minlera + –
mutpara –
References
Bickel, Balthasar. 1999. ‘Nominalization and focus constructions in some Kiranti languages’,
pp. 271–96 of Topics in Nepalese linguistics, edited by Warren Glover and Yogendra P. Yadava.
Kathmandu: Royal Nepal Academy.
Bradley, David. 1997. ‘Tibeto-Burman Languages and Classification’, pp. 1–72 of Tibeto-Burman
Languages of the Himalayas, Papers in Southeast Asian Linguistics 14, edited by David Bradley.
Canberra: Pacific Linguistics.
Coupe, Alexander R. 2007. A grammar of Mongsen Ao. Berlin: de Gruyter.
DeLancey, Scott. 2001. ‘The mirative and evidentiality’, Journal of Pragmatics 33: 369–82.
———. 2002. ‘Nominalization and relativization in Bodic’, pp. 55–72 in Proceedings of the 28th
Annual Meeting of the Berkeley Linguistics Society, Parasession on Tibeto-Burman Languages
and Southeast Asian Linguistics, Vol. 2, Tibeto-Burman Linguistics, edited by Julie Larson and
Mary Paster. Berkeley: Berkeley Linguistic Society.
———. 2011. ‘Finite structures from clausal nominalization in Tibeto-Burman’, pp. 343–59 in
Nominalization in Asian languages: Diachronic and typological perspectives, edited by Foong
Ha Yap, Karen Grunow-Hårsta, and Janick Wrona. Amsterdam: John Benjamins.
Genetti, Carol. 2005. ‘The participial construction in Dolakha Newar: Syntactic implications of
an Asian Converb’, Studies in Language 29(1): 35–87.
Genetti, Carol, Coupe, Alexander R., Bartee, Ellen, Hildebrandt, Kristine, and Lin, You-Jing.
2008. ‘Syntactic aspects of nominalization in five Tibeto-Burman languages of the Himala-
yan area’, Linguistics of the Tibeto-Burman Area 31(2): 97–143.
Georg, Stefan. 1996. Marphatan Thakali: Untersuchungen zur Sprache des Dorfes Marpha im
oberen Kāli-Gan.d.aki-Tal Nepal. München/Newcastle: LINCOM Europa.
Haspelmath, Martin. 1995. ‘The converb as a cross-linguistically valid category’, pp. 1–56 of
Converbs in cross-linguistic perspective: Structure and meaning of adverbial verb forms—
adverbial participles, gerunds, edited by Martin Haspelmath and Ekkehard König. Berlin/
New York: Mouton de Gruyter.
Hyslop, Gwendolyn. 2009. ‘Kurtöp Tone: A tonogenetic case study’, Lingua 119: 827–45.
———. 2010. ‘Kurtöp case: The pragmatic ergative and beyond’, Linguistics of the Tibeto-Burman
Area 33(1): 1–40.
———. 2011a. ‘A Grammar of Kurtöp’. PhD Dissertation, Eugene, OR: University of Oregon.
———. 2011b. ‘Mirativity in Kurtöp’, Journal of South Asian Languages 4(1): 43–60.
———. in press. ‘A preliminary reconstruction of East Bodish’, in Transhimalayan Linguistics,
edited by Nathan Hill and Thomas Owen-Smith. Berlin: Mouton de Gruyter.
———. to appear. ‘On egophoricity and mirativity in Kurtöp’, in Egophoricity, edited by E lisabeth
Norcliffe, Lila San Roque, and Simeon Floyd. Amsterdam: John Benjamins.
Matisoff, James. 1972. ‘Lahu nominalization, relativization, and genitivization’, pp. 237–57 in
Syntax and semantics 1, edited by J. Kimball. New York: Seminar Press.
Noonan, Michael. 1997. ‘Versatile nominalizations’, pp. 373–94 in Essays on language function
and language type: Dedicated to T. Givón, edited by Joan L. Bybee, John Haiman, and Sandra
A. Thompson. Amsterdam: John Benjamins.
Tournadre, Nicolas. 2008. ‘Arguments against the concept of ‘conjunct’/‘disjunct’ in Tibetan’,
pp. 281–308 in Chomolangma, Demawend und Kasbek: Festschrift für Roland Bielmeier zu
seinem 65 Geburtstag, edited by Brigitte Huber, Marianne Volkart, and Paul Widmer. Halle:
International Institute for Tibetan and Buddhist Studies GmbH.
6
The expression of knowledge in Ersu1
SIHONG ZH A NG
1 The language
Ersu is an endangered language (Bradley 1997; Moseley 2010) with about 25,000
speakers (Wang 2010: 6). According to Sun (1982, 1983) and Liu (1983: 462–500), the
language has three dialects—the eastern dialect Ersu, the central dialect Tosu, and
the western dialect Lizu. Sun (1982, 1983) hypothesized that Ersu with the three dia-
lects should be classified as a subgroup of the southern Qiangic branch in the Tibeto-
Burman language family, though Chirkova’s recent empirical studies imply that ‘the
Qiangic hypothesis remains problematic’ (Chirkova 2012).
The three dialects are all spoken in the seven counties in the south-west part of
Sichuan Province, China. More specifically, the eastern dialect, Ersu, is spoken in the
counties of Ganluo, Yuexi, Hanyuan, and Shimian; the central dialect, Tosu, is spoken
in the county of Mianning, and the western dialect, Lizu, is spoken in the counties of
Mianning, Muli, and Jiulong (Sun 1982, 1983; Liu 1983: 462–500; Wang 2010: 3).
In this chapter, the name, ‘Ersu’, will refer to the eastern dialect rather than the lan-
guage as an entirety. The language is a typical ‘topic–comment’ (Huang 2004: 248–63)
and AOV/SV language with a strong isolating tendency. Like Yongning Na (Lidz
2007), constituent order, lexical choice, and discourse context cooperate to express
grammatical relations. Gender and number agreement is not found in predicates and
ellipsis is frequently observed, especially in narratives or long conversations.
1
I am grateful to all three members of my supervisory panel, Professor Alexandra (Sasha) Aikhenvald,
Professor R. M. W. Dixon, and Dr Mark Post for their dedicated supervision during the whole process of
my PhD research. This chapter could not have been produced if Professor Aikhenvald had not offered me
heart-and-soul supervision, helpful and useful comments. Gaps, faults, or errors in this chapter are all my
own responsibility. Special thanks go to Dehe Wang and Shibu Huang for leading me to the Ersu communi-
ties, to all my Ersu friends in Lajigu for providing valuable firsthand Ersu corpora, to the Ersu couple, my
brother Zhongquan Wang and my sister Aguo Huang for accepting me at their home in Lajigu for almost
one year, to my Ersu brother Amu Wang, for his outstanding ability to interpret his mother tongue and also
for his long-time companionship when I was in the field. Heartfelt thanks also go to the two anonymous
referees for their instructive comments. This work was funded by James Cook University IPRS scholarships.
2 The data
All the data for this chapter have been obtained through ‘immersion fieldwork’ (Dixon
2007) in an Ersu village—Lajigu (28°79'77"N, 102°57'85"E). Examples presented here
are mainly extracted from notes taken through participant observation, or audio-
recordings of long conversations or folkloric, mythological, biographical, and pro-
cedural narratives recorded in a natural way. Examples obtained through elicitation
are also used. However, this ‘elicitation’ is inspired by similar examples previously
found in the language, none of them based on ‘prescriptive frameworks’ mentioned
by Aikhenvald (§2.2 of Chapter 1).
3 An overview
Knowledge in Ersu can be expressed through evidentiality. Besides this, the informa-
tion source can also be conveyed through other means such as lexical verbs, evidential
strategies, epistemic strategies, demonstratives, and directional terms.
Evidentiality, a linguistic term for the expression of information source, is a
‘closed and restricted’ grammatical category attested in many languages in the world
(Aikhenvald 2004; §2 of Chapter 1). Evidential systems have been reported as a salient
genetic feature of Tibeto-Burman languages (Sun 1993; LaPolla 2003; Aikhenvald and
LaPolla 2007; etc.). Many languages in this area show multiple evidential markers, for
example, Lhasa Tibetan (Delancey 1985, 1986, 1990, 1992), Amdo Tibetan (Sun 1993),
Qiang (LaPolla 2003) and especially the Puxi dialect (Huang 2004: 195–7), Baima
(Chirkova 2008a), Yongning Na (Lidz 2007; 2010: 476–500), and many other adjacent
languages that are not listed here.
Ersu, as a member of the Tibeto-Burman language family, is not an exception.
Evidentials are also found in the language, appearing as a C3 evidential system with
four choices, that is, direct, inferred, reported, and quotative (Aikhenvald 2004:
51–60). Ersu evidentials occur after tense-aspect morphemes and occupy a clausal- or
sentential-final position. Occasionally, a particle n� that functions as a topic, a focus,
or a pause marker may follow an evidential. Similar to Yongning Na (Lidz 2007),
pragmatics also plays an important role in the Ersu evidential system. In a larger dis-
course context such as a long narrative, an evidential is often ellipsed, depending on
the speaker’s discourse style especially when an overt evidential has already been used
earlier in the narrative. In other words, evidential marking in Ersu may seem not to
be obligatory in each sentence on the surface. But in essence, it is obligatory. If an
evidential cannot be understood from the context, a misunderstanding may occur.
Suppose a dialogue is taking place between A and B. A is asking B where a person went
yesterday. If B did not see this but got the information from someone else, a reported
evidential d�� can never be omissible, as in (1). Otherwise, A would take it for granted
that B personally saw the event.
134 Sihong Zhang
(1) a. thə yɑ-ȵo khɑ du�, nə xɑsὲ?

3sg.prest last-day:yesterday inter:where go:past 2sg understand
Where did he go yesterday, do you know?
b. kuɑʂ du�=d��
mc:town go:perv=evid:reported
(He) is said to have gone to the town.
Evidentials in Ersu have a clausal or sentential scope. Direct evidence of informa-

tion source is unmarked, while others are marked. §4 presents all the evidentials in
Ersu and discusses their semantic implications. The choice of evidential markers is
often determined by speech genres and person (speech act participant vs. non-speech
act participant). This is discussed in §5. §6 discusses the co-occurrence of evidential
markers. §7 shows evidential strategies and epistemological markers that have seman-
tic extensions of evidentiality. This includes some verbs, demonstratives, parentheti-
cals, etc. §8 summarizes this chapter.
4 Evidentials and their semantics

Direct evidentials in Ersu may refer to information acquired by seeing, hearing and
any sensory perception (Aikhenvald 2004: 54), and generic knowledge. This is for-
mally unmarked. Ersu does not distinguish inference and assumption, both marked
with =pà. Consequently, =pà may denote inference based either on ‘visible or tangible
evidence or result’ or on ‘logical reasoning’ (Aikhenvald 2004; §2.1.2 of Chapter 1).
Both d�� and d�iдə are used ‘for reported information with no reference to who it
was reported by’ (Aikhenvald 2004; §2.1.2 of Chapter 1). Quotative evidential marking
has several variants that include d��, d�iдə , d�à, and some others. They are all used
for ‘reported information with an overt reference to the quoted source’ (Aikhenvald
2004; §2.1.2 of Chapter 1). However, ‘an overt reference’ is often ellipsed in larger dis-
course contexts.
4.1 Direct evidence

According to Aikhenvald (2004: 73, and Chapter 1), direct perception tends to be
cross-linguistically less marked than other types of evidence. This is the case in Tibeto-
Burman languages such as Amdo Tibetan (Sun 1993) and Yongning Na (Lidz 2007).
Direct evidence in Ersu is also formally and functionally unmarked. It is the preferred
evidential found in nearly all text genres. Both the speaker and the listener in Ersu
can intuitively understand that the information conveyed in a clause or a sentence
without an inferential marker, a reported marker, or a quotative marker is based on
direct evidence. That is, they know that the information is either based on the speaker’s
seeing (2), or hearing (3), or feeling (4), or smelling (5), or generic knowledge (6).
(2) sipu=tɕ hò loə ɹ tə dzo

tree=loc.supress turtledove indef:one exist
There is a turtledove on the tree.
(3) tʂho tə əɹ=�ə

dog indef:one bark=prog
A dog is barking.
(4) tɑ-ȵo mεtɕo dzolo dɑ-tshɑ
present-day:today sky over pref:upward-be hot
It is very hot today.
(5) thə ŋu�-ʂɿ`=bὲ tsε də-xə

dem:this ox-meat:beef=pl really pref:upward-smell-good
This beef really smells good.
(6) ɑɹ ə ɹsu=bὲ tə-ȵo-ȵo vùt-ɕò

1pl.slf prn:Ersu=pl one-day-day:every day head-bind:turban
də-tsu=�ə
pref:upward-wear=prog.npast
We Ersu people are wearing a turban every day.
Examples (2) to (6) are all direct evidentials and they are formally unmarked. Oth-
erwise, one of the following markers, pà (inferential), d�� or d�iдə (reported) and
quotatives such as d�à and others should be used. For example, if a speaker gets the
information from others, that is, ‘hearsay’, it is obligatory for them to use (7) rather than (2).
(7) sipu=tɕ hò loə ɹ tə dzo=d��
tree=loc.supress turtledove indef:one exist=evid:reported
It is said that there is a turtledove on the tree.
4.2 Inference
The reading of inference or assumption is realized through the clausal- or sentential-
final =pà which is obligatorily used. Otherwise, the native speaker would view the
information as if it were based on direct evidence. The inferential marker is seldom
attested in narratives, but quite frequently found in daily conversations. =pà can be used
in a future context. In this situation, it has epistemic overtones, indicating the speaker’s
uncertainty about the occurrence of an event, as in (8). The speaker’s inference is either
from ‘visible or tangible evidence or result’ (8) or from ‘logical reasoning’ (9).
(8) mεtɕ o su-ȵo thə -phu=�ə =p�
sky next-day:tomorrow pref:away-change=pros=evid:inferential
The weather is going to change tomorrow.
(The speaker makes this inference based on such evidence as the changes in
the clouds, temperature, wind, etc.)
136 Sihong Zhang
(9) thə nbò tʂɑ du�=p�.

3sg.prest horse search go.perv=evid:inferential
He might have gone to search for his horse.
(The speaker went to someone’s house, and found that the person was not at home at a
particular time, for example, 5:00 pm when an Ersu often goes to find a horse. They have
this logical reasoning according to their general knowledge rather than evidence.)
4.3 Reported
Oral transmission as an information source in Ersu is realized through d�� or d�iдə.
Similar to quotative markers, d�� and �� can be seen to be grammaticalized from the
verb d�i ‘say’ (§4.4).
Ersu does not distinguish between ‘secondhand’ and ‘thirdhand’ information sourc-
es. In other words, both d�� and d�iдə are applicable to all non-firsthand information.
They show no semantic or functional differences in practical uses. The data show that
d�� occurs more frequently than d�iдə in daily conversations as an evidential.2 As
long as a piece of information is reported neither from direct evidence nor from infer-
ence or direct quotation, either d�� or d�iдə is an obligatory component in a clause or
sentence. However, in a larger context, it could be omissible (§3). For example:
(10) thə yɑ-ȵo kuɑʂɑ du�=d��
3sg.prest last-day:yesterday mc:town go.perv=evid:reported
It is said that he went to the town yesterday.
4.4 Quotative
Quotative evidentials appear to be quite complex in Ersu. There are several inter-
changeable variants used for quotative information sources. They are: d�à, d��, th ə-
ɑ-d�à, thə-ɑ-d��, and thə-ɑ-d�iдə. According to Katia Chirkova (2012, p.c.) and Dehe
Wang (2012, p.c.), d�à, d��, and d�iдə all contain a verb d�i, ‘say’ in the Ganluo variety
of Ersu (§4.3).3 Later, my language consultants Zhifu Huang, Zhongquan Wang, Amu
Wang, Aguo Huang, etc. in Lajigu also confirmed that d�i possibly means ‘say’, but
only through elicitation. Consequently, d�à à might be derived from the verb d�i, ‘say’
and the perfective aspectual marker à through vowel fusion and tonal variation. d��
derives from the verb d�i, ‘say’ and an unknown attachment, �, also through vowel
fusion. d�iдə consists of the verb d�i, ‘say’ and the progressive/prospective aspectual
marker дə without any phonological reduction during the course of grammaticaliza-
tion. It should be noted that the дə in d�iдə does not have any aspectual implications.
thə-ɑ-d�à, thə-ɑ-d��, and thə-ɑ-d�iдə might contain a demonstrative thə ‘this’ and the
2
Both in narratives and conversations, d�iдə can also be used as a particle that functions as a discourse
organizer. It is often used to denote the sequence of the development of an event that can be translated as
English ‘subsequently’, ‘consequently’, or ‘and so’.
3
This variety and the Lajigu variety are mutually intelligible. However, it is more conservative in nearly
all the aspects from phonology to lexemes.
meaning of ɑ here is not known. It is also observed that the three, thə-ɑ-d�à, thə-ɑ-d��,
and thə-ɑ-d�iдə have a free-standing position in the utterance depending on the need
for discourse organization. In addition, whenever d�à, d��, thə-ɑ-d�à, thə-ɑ-d��, or
thə-ɑ-d�iдə are used after a direct/overt quotation, no verb with the meaning of ‘say’
is used in the context. This further proves that the five quotative evidentials might be
derived from the verb d�i, ‘say’ with some phonetic changes.
However, none of my consultants could use d�i, ‘say’ well even through elicitation.4
Thorough investigation into the data obtained through my ‘immersion fieldwork’
(Dixon 2007) shows that d�i, ‘say’ has never been used as a free verb in narratives and
daily conversations. The data also demonstrate that thə-ɑ-d�à, thə-ɑ-d��, and thə-ɑ-
d�iдə are almost exclusively found in folkloric and mythological narratives in which
the speech style is more conservative than in daily conversations. Meanwhile, d�à,
d��, and d�iдə occur much more frequently than thə-ɑ-d�à, thə-ɑ-d��, and thə-ɑ-d�iдə
even in narratives. Of these, only d�� and d�iдə are also used for the reported infor-
mation source with high frequency (§4.3) though some of the speakers are observed
to use other quotative markers for the reported information source quite occasionally.
All this shows that d�à, d��, d�iдə, thə-ɑ-d�à, thə-ɑ-d��, or thə-ɑ-d�iдə are in the pro-
cess of grammaticalization. It is most likely that d�à, d��, and d�iдə might develop
into completely grammaticalized evidentials in future, with the first one used for quo-
tative and the other two used for the reported information source. Consequently, I
view them all as evidential markers instead of evidential strategies in this chapter. If
not, they are at least ‘quasi-evidential markers’ at the present stage.5 It should be noted
that a ‘stable and mature’ quotative evidential has not fully grammaticalized. It is thus
not surprising that the five forms can be used interchangeably. For example:
(11) ɑ-wɑ=nὲ, ‘khɑ-lɑ ŋə-dzɿ=�ə.'

kin.pref-grandma=top pref:inward-plough pref:outward-eat=pros.npast
=d��
= evid:quotative
The old lady said like this: ‘(The ox should be fed) to plough fields and provide food.’
ɑ-pu=nὲ, 'nɑ-kɑ ŋə-dzɿ=ə.'

kin.pref-grandma=top pref:downward-kill pref:outward-eat=pros.npast
=d�à.
=evid:quotative
The old man said like this: ‘(The ox should) be killed and eaten.’
4
A number of other words have a meaning similar to ‘say’. These include də-fusε ‘say, tell’; də-xi ‘say, tell’;
dɑ-khɑt ho ‘say, tell’; dɑ-lɑ ‘shout’; də-koyi ‘call’, etc.
t ə-ɑ-d�à, t h ə-ɑ-d��, and t h ə-ɑ-d�iдə might be more suitably interpreted as evidential strategies since
5 h
they have a free-standing position in contexts. A narrator often repeats them in narratives and then moves
to a new topic or clause/sentence. In this situation, a particle n� often follows them.
138 Sihong Zhang
Example (11) has been extracted from a traditional folkloric story which tells at the
very beginning that there is an old couple who feed an ox. The old lady wants to keep
the ox so as to plough fields and provide food while the old man wants to kill the ox and
eat the beef. The two sentences occur in succession in the same context. The first one
employs the quotative d�� and the second one uses the quotative d�à. My language
consultant’s intuition indicates that if the two evidentials were used in a reverse order,
no semantic and functional difference would arise. Consequently, the two evidentials
can be used interchangeably. In addition, (11) shows that there is no verb with the
meaning of ‘say’ used in the context.
5 Evidentiality and other grammatical categories

Ersu verbs display few restrictions on the use of evidentials. This is unlike some other
Tibeto-Burman languages, for example Amdo Tibetan (Sun 1993), in which verbs
of different semantic types function to determine the choice of different evidentials.
Moreover, evidentials in Ersu can occur together with any type of aspect markers.
Further details are in Zhang (2013). However, the choice of evidentials in Ersu is closely
associated with speech genres and speech act participant vs. non-speech act partici-
pant. Above all, it should be noted that all these factors are interrelated and intertwined
with each other in how they determine the various uses of evidentials.
In this chapter, speech genres can be classified into narratives and conversations, in
which different types of evidentials show different frequencies of occurrence.
In terms of ‘speech act participant’ and ‘non-speech act participant’, a narrator, or a
speaker of a narrative is a speech act participant and all the others are non-speech act
participants. This applies to dialogue structures in narratives, in which non-speech act
participants include first and second person. In a conversation, a speech act partici-
pant includes first person and second person (human beings) and a non-speech act
participant refers to third person (either human or non-human).
5.1 Evidentiality and speech genres

Different types of evidentials may be determined by different speech genres. The infer-
ential evidential =pà is only occasionally found in autobiographical narratives,6 not
in mythological, folkloric, procedural narratives, and narratives about creation. This
might be so because the information source in a narrative comes from the speaker’s
internalized memory, seldom from their inference. However, reported and quotative
evidentials appear quite frequently in narratives though they are sometimes ellipsed
in contexts. In conversations, all the four subtypes of evidentials are frequently heard.
6
Autobiographical narratives share a lot of similarities with daily conversations in using evidentials. This
is because autobiographical narratives, similar to conversations, are also closely linked to the real world. In
this chapter, the term ‘narrative’, other than when clearly stated, refers to mythological, folkloric narratives
and narratives about creation.
Among them, unmarked direct evidence occurs the most and the reported evidential
d�� is second to it. As discussed in §5, ‘person’ differentiation is closely linked to speech
genres in Ersu. Consequently, it is necessary to discuss the correlations between evi-
dentiality and speech genres together with the opposition between speech act partici-
pant and non-speech act participant (§5.2).
5.2 Evidentiality and speech act participant vs. non-speech act participant
‘In Tibetan, the category of person constitutes an important factor which determines
much of the verbal morpho-syntax’ (Sun 1993). Hale (1980) pointed out that there is
a ‘conjunct vs. disjunct’ person distinction in Kathmandu Newari, in which verbal
marking shows a first-person vs. second-/third-person distinction in statements in
the same way as a second-person vs. first-/third-person distinction is marked in ques-
tions. Following Hale, DeLancey (1986, 1990, 1992) employs the terms to distinguish
persons in Lhasa Tibetan. However, Tournadre (1991, 2008) holds the opinion that
it is more appropriate to use the term ‘egophoric’ to denote the person distinction
in Tibetan. Sun (1993: 955–6) prefers ‘a referentially fluid dichotomous distinction
between self person and other person’ in Amdo Tibetan since in the language, ‘self
person sentences are marked as utterances produced by oneself ’. He further states that
the term ‘self person’ is appropriate for first person statements, second person ques-
tions and some quotes. Yongning Na has a system ‘similar to a conjunct/disjunct sys-
tem’ that has a ‘self ’ and ‘other’ distinction, and ‘other’ further distinguishes between
speech act participant, with information source appearing in a question, and non-
speech act participant with information source as either a marked inference of internal
state or an unmarked observable state (Lidz 2007). In the Lizu dialect of Ersu, there
is an ‘egophoric vs. other person’ distinction. ‘Egophoric utterances express personal
knowledge or intention on the part of the speaker (the first person). . . . Other person
utterances are consequently linked to the non-first person’ (Chirkova 2008b: 28).
Similar to the above mentioned Tibeto-Burman languages, Ersu also has a person
distinction. However, it is more appropriate to categorize ‘person’ as ‘speech act partic-
ipant’ vs. ‘non-speech act participant’ in relation to speech genres, as shown in Table 1.
5.2.1 Speech act participant vs. non-speech act participant in narratives Table 1
indicates that in narratives, Ersu distinguishes a speech act participant, first person
ɑ, also the speaker her/himself and a non-speech act participant, first person yò,
often a role in a narrative.7 Information source concerning the speech act participant
7
Some of the speakers, especially the younger Ersu, often use the two different first persons interchange-
ably even in the same narrative. This is frequently seen as being ‘incorrect or inappropriate’ by the older Ersu.
It is also observed that one of my language consultants, Ms Wang Azhi who is illiterate, is always sensitive
to the differences between ɑ and yò and maintains consistency in differentiating the two in every one of her
narrations. In addition, many children in Lajigu did not know the meaning of yò while I was testing them
in the field. This implies that, first, the language is tending to become simplified; secondly, school education
with the instruction media of Mandarin Chinese is endangering the language to some extent.
140 Sihong Zhang
Table 1. Person, evidential, and speech genre in Ersu
Speech act participant Non-speech act participant
Narrative1 the speaker /1.slf(а) 1.otr (yò), 2 3
Evidential direct quotative reported
Conversation addresser (1.slf), addressee (2) 3
Evidential direct/inferential direct/inferential/reported/quotative

1
The term 'narrative' here refers to mythological, folkloric, procedural narratives, excluding autobiographical
narratives in which person distinction and evidential marking is similar to conversations. This is because auto-
biographical narratives, similar to conversations, are also closely linked to the real world.
ɑ is considered as direct evidence and unmarked as in (12a), while the non-speech

act participant yò is always associated with an overt quotation and a quotative
evidential is obligatory (12b), with the exception of a retrievable ellipsis in contexts.
(12) a. ɑ yὲ-so-xɑ tə sɿ` də-xi=�ə

1sg.slf last-before-time:ancient times one cl:bit pref:upward-tell=pros.npast
I will tell an ancient story. lit: I will tell ancient times.
b. 'yò tʂhə tə ʂɿ` nɑ tɕ hi=�ə.’=d�� =nὲ

1sg.otr rice one cl:litre 2sg.acc give=pros.npast=evid:quotative=part: pause
‘I will give you a litre of rice.’ (His elder brother) said like this.
In (12a), ‘I’ refers to the speaker himself who is going to tell a story. ‘I’ myself is a speech act
participant and the information source is viewed as direct evidence. In contrast, ‘I’ in (12b)
refers to a role in a narrative. Consequently, the quotative evidential is obligatorily used.
In a narrative, both second person and third person are treated as non-speech act partic-
ipants. Information about the second person appears either in a question or in an impera-
tive or prohibitive statement and is marked with a quotative evidential. For example:
(13) 'nə tɑ-ȵo ɑ-ndə nə-ŋu, nə khɑt ho
2sg present-day:today inter-what pref:downward-do 2sg speak
�=�ə =ὲ ?’=d��=nὲ
inter=pros=inter=evid:quotative=part:pause
‘What you did today, will you tell (us)?’ (Her nine brothers) said like this.
(14) 'nə yò yibɑyì ŋə-nbε=nὲ, nə dɑ-t hɑ-

2sg 1sg.otr armpit pref:outward-crawl=top 2sg pref:upward-prohib-
fusε . . .’=d�à
speak . . . =evid:quotative
‘You hide in (crawl into) my armpit and you do not speak . . . (She) said like this.
Information about the third person, also a non-speech act participant in narratives,
is most frequently marked by reported evidentials, as in (15). This might be because
in Ersu, a language without written records, people can only ‘hearsay’ narratives like
myths, folkloric stories, etc. that are transmitted generation by generation in oral form.
(15) thə αʂɿ̀ ʂɿ̀ =nὲ, xuɑ=yi tə t hə-pu
dem:this prn:person name=top bird=dim indef:one pref:away-become
=�=d��
=perv=evid:reported
It is said that Ashishi became a small bird.
5.2.2 Speech act participant vs. non-speech act participant in conversations In
conversations, especially in daily short conversations, the speech act person includes
first person (addresser) and second person (addressee). Both unmarked direct
evidential and marked inferential evidential are attested (see Table 1). However,
since oral information transmission between the first and second person is always
in a direct way, reported and quotative evidential are not applicable in this situation.
Information relevant to the second person often appears in a question, as in (16)
with a direct evidential and in (17) with an inferential marker pà.
(16) nə khɑ ɿ=�ə=ὲ?
2sg inter:where go.imperv=prog=inter
Where are you going?
(The question in (16) occurs frequently in daily conversations. It pragmatically func-
tions as a greeting, similar to English ‘How are you?’. This conversation occurs when
two people meet each other along the way. Consequently, the information is visual
and the unmarked direct evidential is used here.)
(17) nə yɑ-ȵo �o-də-nd�ə=p�
2sg last-day:yesterday thorax-pref:upward-tremble:be angry=evid:inferential
You must have been angry yesterday.
In addition, a second person statement is seldom used in daily conversations. How-
ever, it is heard when a first person gives an order, a command, or a request to a second
person, such as to assign a task at a meeting or to allocate a seat at a ceremony, etc.
For example:
(18) nə t hə=kə nə-zɿ

2sg dem:this=loc.iness:in:here pref:downward-sit
You sit here.
A non-speech act participant in conversations refers to the third person. All eviden-
tials can be used with a non-speech act participant, depending on the source of the
information. More specifically, the choice of an evidential is dependent on whether the
speaker acquires the information through their sensory perception (direct evidence),
142 Sihong Zhang
through their inference (inferential evidential), or through oral transmission (report-

ed or quotative evidential). As discussed in §4.3 and §4.4, in conversations, the most
frequently used reported evidential is d�� and the most frequently used quotative
evidential is d�à, other evidential variants are not frequently used.
6 Co-occurrence of evidentials
The co-occurrence of different types of evidentials in Ersu has been only sparsely
attested. The inferential evidential =pà is the only one to co-occur with the reported
and quotative evidentials and =pá always precedes them. The reverse order is never
allowed. For example:
(19) thə yɑ-ȵo kuɑʂ� du�=p�=d��
3sg.prest last-day:yesterday mc:town go.past=evid:inferential=evid:reported
It is said that he must have been to the town yesterday.
(20) Amu=nὲ, ‘thə yɑ-ȵo kuɑʂɑ du�

prn:person name=top 3sg.prest last-day:yesterday mc:town go.past
=p�.’=d�ɑ
=evid:inferential=evid:reported
Amu said like this: ‘He must have been to the town yesterday.’
7 Information source conveyed through other means

In Ersu, there are several verbs with the meaning of ‘say, tell’ (see note 4) that can
be used to indicate where the information comes from. Verbs that denote sensory
perception tend also to be linked to information source (§3.2.1 of Chapter 1). Verbs of
this type found in Ersu include: kh-ə-dzolo ‘look’; thə-ndo ‘see’; khɑ-bɑȵi ‘listen, hear’;
khə-liɑ ‘touch, feel’; dɑ-ŋuɑ ‘smell’; də-ndzɿndzà ‘think’, etc. In addition, information
source can be further conveyed through evidential strategies, epistemic strategies,
demonstratives/directional terms, and parentheticals.
7.1 Evidential strategies

Aikhenvald (2004: 105) points out that ‘categories and forms which acquire secondary
meanings somehow related to information source are called evidentiality strategies’
(see also Chapter 1). In other words, categories and forms whose semantic extensions
(not primary meanings) are relevant to the information source are evidential strate-
gies. Two other types of marking might be considered evidential strategies rather than
evidentials. They are: tɕhi denoting ‘a person’s previous personal experience’ and mixɑ
‘seem’.
7.1.1 Evidential strategy tɕhi In Ersu, there is an experiential aspectual marker
tɕhi denoting that an event happened to a narrator/speaker quite a long time ago—a
semantically similar marker occurs in Mandarin Chinese (Li and Thompson

1981: 226–32). Whenever tɕ hi ‘previous experience’ is used, the information being
conveyed is, first, about the narrator or the addresser themselves and secondly, about
one’s earlier life experience. It should be noted that though one’s personal experience
belongs to one’s own knowledge, that is, direct evidence, a clause or a sentence is
often marked with tɕ hi. For example:
(21) ɑ vɑkə ɿ̀ =tɕ hi
1sg.slf prn:Ersu name of the county seat go:imperv=exp:previous experience
I have previously been to Yuexi (name of the county seat).
7.1.2 Evidential strategy mixɑ The meaning of mixa is very similar to English ‘seem’
or ‘(be) like’. It acts as a verb in Ersu, taking the whole clause in its scope. It functions
not only to denote that the information being reported is based on the speaker’s own
inference (often logic reasoning), but also to express uncertainty. For example:
(22) gu� ò=gə mixɑ
rain fall=pros seem
It seems that it is going to rain.
(The speaker may make the inference based on the dark clouds in the sky, or
stuffy air, etc.)
7.2 Epistemic strategy
Unlike an evidential strategy that may denote information source, the epistemic strat-
egy discussed here only conveys the speaker’s degree of certainty in the statement
and it ‘undergoes semantic extension to assess epistemic value’, which may undertake
some other grammatical duty (Lidz 2007). Ersu has an element that might be an idiom
functioning as an epistemic strategy, that is, là-ma-ntɕ h�. Literally, là here là seems
to be an emphatic particle that means ‘all’; ma seems to be a negative that means ‘not’.
However, though the meaning of ntɕ h� can be understood as ‘know’ in this context,
its meaning in other contexts or in isolation is attested as ‘bite’. The term for ‘know’
in Ersu is xas�. là-ma-ntɕ h� often occupies a clause-final slot and has a clear meaning
like English ‘no one knows’ or ‘who knows?’ It is used to reflect a speaker’s inner flow
of thinking, especially with uncertainty. For example:
(23) thə ŋə-dzɿ=� 1�-mɑ-ntɕhì ?
3sg.prest perf:outward-eat=perv. past emph:all-neg:not-know?
He has eaten. Who knows?
7.3 Demonstratives and directional terms
Similar to Dyirbal, an Australian language that has ‘a three-system of noun markers’
(Chapter 8 and §2.3 of Chapter 1), Ersu also has a three-system of demonstratives and
two-system of directional terms ‘which combine reference to visibility and spatial dis-
tance of the noun’ that they modify.
144 Sihong Zhang
In Ersu, if the demonstrative t h ə ‘this’, near the speaker, is added with a prefix a-, the
meaning of thə becomes ‘that’, that is, not near to the speaker. A referent modified by
a-thə may be either visible or invisible. If a- is lengthened as a long vowel /a:/, that is,
ɑ:-, a referent modified by ɑ:-t hə should be invisible, and most often, information about
the referent may only exist in one’s memory. This agrees with Haiman’s (1983: 781–819)
statement that ‘the linguistic distance between expressions corresponds to the concep-
tual distance between them.’ The three-system demonstratives are shown in Table 2.
The stem of an Ersu directional term often contains a prefix a-, for example, a-kua,
‘north’.8 A referent being modified by a-kua might be visible or invisible to the speaker,
but it is certain that it is not near to them. Similar to demonstratives, if the vowel of the
prefix in a-kua is lengthened to be ɑ:-kua, a referent is then understood to be invisible
to the speaker or only to exist in one’s memory. Examples are given in (24):
(24) a. ɑ-kuɑ nbi=tɕ hò yi tɑ kɑ

pref:distal-north mountain=loc.supress:on house one cl:generic, sticklike
xɑ
exist:have
There is a house on the mountain in the north. (The house may be visible or invisible
to the speaker.)
b. ɑ:-kuɑ nbi=tɕhò yi tɑ kɑ
pref:remote-north mountain=loc.supress:on house one cl:generic, sticklike
xɑ
exist
There is a house on the mountain in the north. (The house is invisible to the speaker
and maybe only exists in one’s memory.)
Table 2. Ersu demonstrative system
Demonstrative Gloss
thə ‘this’, visible and near the speaker
ɑ-thə ‘that’, visible or invisible and not near the speaker
ɑ:-thə ‘that remote’, invisible and often in memory
8
Ersu direction corresponds to the local community’s topography. Normally, they live in deep valleys
and there are high mountains in the east and in the west. There is no direction matching with ‘east’ and ‘west’
in daily conversations. Instead, they use a-дa ‘uphill’ and a-ȵa ‘downhill’. Consequently, the four directional
terms commonly used in Ersu are: a-дa ‘uphill’, a-ȵa ‘downhill’, a-kua ‘north’ and a-ŋua ‘south’.
7.4 Parentheticals
There are two particles that function to attract the audience’s attention. Both are quite
frequently used both in narratives and in daily conversations. Both might derive from
the verb k h ə-dzolo ‘look’, each respectively taking a syllable of the disyllabic root dzolo
with tonal variation. They are: a-dzò and a-lò and both mean ‘look/you see’. The dif-
ference between them is as follows: a-dzò denotes that the information being reported
either comes from others, or is far from the speaker, or irrelevant to her/him while
a-lò denotes that the information being reported either comes from the speaker her/
himself, or is relevant to, or is close to her/him. Consequently, more precisely, a-dzò
means ‘look there/you see there’ and a-lò means ‘look here/you see here’.
8 Summary
This chapter offers a preliminary study of the expression of knowledge in Ersu. The
language has a C3 system of evidential markers (Aikhenvald 2004: 51). Direct evi-
dence is formally and functionally unmarked. =pà is a marker for inferred information
source and the expression of reported evidence is realized through d��, and d�iдə.
Quotative evidential markers include several interchangeable variants, that is, d�à,
d��, t h ə-ɑ-d�à, t h ə-ɑ-d��, and t hə-ɑ-d�iдə. Among them, d�à is the most frequently
used. We hypothesize that all of them are grammaticalized from the verb d�i ‘say’.
t hə-ɑ-d�à, thə-ɑ-d��, and t hə-ɑ-d�iдə may have a free-standing position. All these indi-
cate that reported and quotative evidentials, similar to Yongning Na (Lidz 2007), are
still in the process of grammaticalization in Ersu. Evidentials can be omissible if they
can be recoverable through the context. Thus, they are not formally obligatory in every
clause or sentence on the surface.
The choice of evidentials is determined by such interrelated and intertwined factors
as speech genres and speech act participant vs. non-speech act participant. In narra-
tives, the inferential =pà is seldom attested. An unmarked direct evidential is applica-
ble to the speech act participant, that is, the narrator. Quotative evidentials are found
for both the first person (a role in a narrative) and the second-person non-speech act
participants. Non-speech act participant third person often occurs with reported evi-
dential d��, or d�iдə. In conversations, the unmarked direct evidential and the marked
inferential are attested for both the addresser and the addressee, that is, first-person
and second-person speech act participants. All evidentials can be used for third-
person non-speech act participants in conversations (see Table 1).
The information source can also be conveyed through other means. tɕ hi, function-
ing as an evidential strategy, denotes the speaker’s own previous experience. mixa,
another form of evidential strategy, denotes that the information is sourced from the
speaker’s inference and also their uncertainty about the information. là-mɑ-ntɕ h� is
an epistemic strategy that indicates uncertainty rather than information source and
reflects the speaker’s inner flow of thinking. The lengthening of the prefix a- of the
146 Sihong Zhang
demonstrative a-t hə and the directional terms can help illustrate whether the refer-
ent being reported is visible or invisible to the speaker or just exists in the speaker’s
memory. a-lò and a-dzò, two parentheticals, are used to denote whether the referent
being reported comes from a speaker or from others, whether it is relevant or close to
the speaker or not.
To conclude, Ersu shows a C3 system of multiple evidential markers that include
direct, inferential, reported, and quotative evidentials. The information source can
also be conveyed through other means, such as evidential strategies, epistemic strat-
egies, parentheticals, demonstratives, and directional terms. Finally, speech genres,
speech act participant vs. non-speech act participant may intersect with the eviden-
tials to denote the information source.
References
Aikhenvald, Alexandra Y. and LaPolla, Randy J. 2007. ‘New perspectives on evidentials: A view
from Tibeto-Burman’, Linguistics of Tibeto-Burman Area 30(2): 1–16.
Bradley, David. 1997. ‘Tibeto-Burman languages and classification’, pp. 1–72 of Tibeto-Burman
languages of the Himalayas, edited by David Bradley. Canberra: Pacific Linguistics.
Chirkova, Katia. 2008a. ‘Evidentials in Baima and Tibetan dialects compared’, Minority Lan-
guages 3: 36–43.
———. 2008b. ‘Essential characteristics of Lizu, a Qiangic language of Western Sichuan’, Work-
shop on Tibeto-Burman Languages of Sichuan (Nov. 21–24, 2008). Taipei: Institute of Linguis-
tics, Academia Sinica, pp 1–43.
———. 2012. ‘The Qiangic subgroup from an areal perspective: a case study of languages of Muli’,
Language and Linguistics 13(1): 133–70.
DeLancey, Scott. 1985. ‘Lhasa Tibetan evidentials and the semantics of causation’, pp. 65–72 of
Proceedings of the 11th Annual Meeting of the Berkeley Linguistic Society. Berkeley: University
of California at Berkeley.
———. 1986. ‘Evidentiality and volitionality in Tibetan’, pp. 203–13 of Evidentiality: The linguistic
coding of epistemology, edited by Wallace L. Chafe and Johanna Nichols. Norwood: Abex.
———. 1990. ‘Ergativity and the cognitive model of event structure in Lhasa Tibetan’, Cognitive
Linguistics 1(3): 289–321.
———. 1992. ‘The historical status of the conjunct/disjunct pattern’, Acta Linguistica Hafniensa
25: 39–62.
Dixon, R. M. W. 2007. ‘Field linguistics: A minor manual’, STUF—Sprachtypologie und Univer-
salienforschung 60 (1): 12–31.
Haiman, John. 1983. ‘Iconic and economic motivation’, Language 59(4): 781–819.
Hale, Austin. 1980. ‘Person markers: Finite conjunct and disjunct verb forms in Newari’,
pp. 95–106 of Papers in Southeast Asian linguistics No.7, edited by Ronald L. Trail. Canberra:
Pacific Linguistics.
Huang, Chenglong. 2004. ‘A Reference Grammar of the Puxi Variety of Qiang’. PhD thesis, Hong
Kong: Hong Kong City University.
LaPolla, Randy J. 2003. ‘Evidentiality in Qiang’, pp. 63–78 of Studies in evidentiality, edited by
Alexandra Y. Aikhenvald and R. M. W. Dixon. Amsterdam/Philadelphia: John Benjamins.
Li, Charles and Thompson Sandra. 1981. Mandarin Chinese: A functional reference grammar.
Berkeley: University of California Press.
Area 30: 45–88.
———. 2010. ‘A Descriptive Grammar of Yongning Na (Mosuo)’. PhD thesis, Austin: University
of Texas at Austin.
Liu, Hui Qiang. 1983. ‘A brief introduction to Ersu’, pp. 462–500 of Studies on Ersu Tibetan, edited
by ShaomingLi and Junbo Liu in 2007. Beijing: Ethnic Publishing House.
Moseley, Christopher. 2010. Atlas of the world’s languages in danger. <http://www.unesco.org/
culture/en/endangeredlanguages/atlas> accessed 25 February 2012.
Sun, Hong Kai. 1982. ‘A sketch of Ersu (Tosu)’, Linguistic Study 3: 241–64.
———. 1983. ‘Minority languages of the Six River Valley and their genetic classification’, Scholarly
Journal of Nationalities 18(3): 99–274.
Sun, Jackson T.-S. 1993. ‘Evidentials in Amdo Tibetan’, The Bulletin of the Institute of History and
Philology Academia Sinica 63(4): 945–1001.
Tournadre, Nicolas. 1991. ‘The rhetorical use of the Tibetan ergative’, Lingusitics of the Tibeto-
Burman Area 14(1): 93–107.
———. 2008. ‘Arguments against the concept of “conjunct”/“disjunct” in Tibetan’, pp.
281–308 of Chomolangma, Demawend und Kasbek. Festschrift für Roland Bielmeier zu
seinem 65. Geburtstag, edited by Huber Brigitte, Marianne Volkart, and Paul Widmer. Halle
(Saale): International Institute for Tibetan and Buddhist Studies GmbH.
Wang, Dehe. 2010. Studies on Ersu Tibetan cultures. Chengdu: Sichuan University Press.
Zhang, Sihong. 2013. ‘A Comprehensive Grammar of Ersu, a Tibeto-Burman Language Spoken
in the Southwest of China’. PhD thesis, Cairns: James Cook University.
7
Evidentiality in Kalmyk1
EL ENA SK R IBN IK A N D OLGA SEE SI NG
1 Kalmyk
Kalmyk is an endangered2 Western Mongolic language spoken in the Republic of
Kalmykia (Russian Federation, lower Volga region). The number of native speakers of
Kalmyk is ca. 180,000 (census 2010).
Kalmyk3 is an agglutinative language with AOV/SV basic constituent order, post-
fixes and postpositions, and vowel (backness and rounding) harmony. The nominal
system comprises nine cases and possessivity (both personal and reflexive). The verbal
system includes ten moods: indicative, precative, voluntative, optative (both simple
and expanded), benedictive, concessive, permissive, dubitative, and potential (Bläsing
2003: 241); in the indicative, eight simple (synthetic) temporal, aspectual, and eviden-
tial forms are used. The category of voice has derivational character and expresses the
passive with the marker -gd-, causative with -Ul4-, reciprocal with -ld- and sociative
with -lc-.
Negation involves several negative particles: the prepositional prohibitive bičä with
imperative forms, the postpositional contrastive biš with nominal predicates, and
prepositional es and postpositional uga with participles and converbs. Indicative finite
forms, with one exception, cannot be combined with negation, and instead special con-
structions based on participles and converbs with uga are used (asymmetric negation).
1
We express our sincere gratitude to the Fritz-Thyssen-Stiftung for its generous support of this research
project (Az 40.10.0.023). Special thanks go to our consultants, especially to A. D. Lidzhiev and T. D. Chemi-
dova, who patiently answered all of our sometimes odd questions, to Rogier Blokland and Christianna
Stavroudis for their comments on the first version of this chapter, and last but not least to Sasha Aikhenvald
for the valuable discussion and support at all stages of this project.
2
According to the UNESCO Atlas of the World’s Languages in Danger, Kalmyk is identified as an endan-
gered language (Atlas of the world’s languages in danger. 2010. Edited by Moseley, Christopher, 3rd edn. Paris:
UNESCO Publishing. Online version: <http://www.unesco.org/culture/en/endangeredlanguages/atlas>).
3
The following brief grammatical sketch of Kalmyk is based mainly on the grammatical descriptions of
Sanžeev (1983), Pyurbeev (1977), Bläsing (2003), and Say, Baranova, and Serdobol’skaya (2009).
4
The capital letters in the affixes stand for vowel-harmonic variants.
Table 1. Verbal predicates (terms after Bläsing 2003: 244)
Affix Function Example Negation
-nA present-future ir-nä ‘(he) comes’ ir-x biš (> ir-x-š)
-v terminative ir-v ‘(he) came’ ir-sn uga
-lA confirmative ir-lä ‘(he) came’ ir-sn uga bilä
-ž resultative ir-ž ‘(he) came’ ir-ž uga
-sn perfective participle ir-sn ‘(he) came’ ir-sn uga
-x future participle ir-x ‘(he) will come’ ir-x uga (> ir-š-go)
-dg habitual participle ir-dg ‘(he) usually comes’ ir-dg uga (> ir-d-go)
-A imperfective participle ir-ä ‘(he) is still on his way’ ir-äd uga
Clausal subordination in Kalmyk is based on an extensive system of non-finite

forms: ten converbs and seven polyfunctional forms traditionally labelled as
participles. Subject agreement of finite predicates is based on the personal pre-
dicative paradigm derived from personal pronouns (with the exception of the
unmarked third person); with non-finite dependent predicates possessive affixes
are used (personal possessive in different-subject sentences, reflexive-possessive
in same-subject ones). Direct speech constructions use the quotation verb gi- ‘say’;
some of its forms are grammaticalized as complementizers, others as evidentials
(see §§8, 9, 11).
Verbal predicates can be simple (synthetic) and complex (analytic). Simple predi-
cates consist of the stem and one of the affixal forms, four specialized finite and four
participial (Table 1).
Complex predicates with directional, temporal, aspectual, modal, and evidential
meaning consist of the main verb plus one or more analytic constructions with
either auxiliary verbs or auxiliary elements of nominal character. Auxiliary verbs
have one of the eight affixes listed in Table 1. Such constructions belong to three
classes:
(a) The main verb with one of the converbial markers (only three converbs out of
ten can be used here) and an auxiliary verb (there are more than 30 different
auxiliaries); the meanings thus expressed belong mostly to directional and
aspectual spheres, e.g.:
150 Elena Skribnik and Olga Seesing
– directivity/benefactivity
(1) güü-ž od-

run-convb.imperv go away
run away (from the deictic centre);
(2) har-č ir-

go out-convb.imperv come
go out (towards the deictic centre);
(3) xuld-ž ög-

buy-convb.imperv give
buy (for somebody else);
(4) xuld-ž av-

buy-convb.imperv take
buy (for oneself);
– aspect/aktionsart (here perfective and continuative)
(5) kel-äd ork-

say-convb.perv put away
having said;
(6) kel-ž bää-

say-convb.imperv be
saying;
– modality
(7) kel-ž čad-

say-convb.imperv be able
be able to say;
(8) kel-ž jad-

say-convb.imperv be unable
be unable to say.
(b) The main verb with one of the participial markers (out of seven participial forms
only four are used here) and two auxiliary verbs bää- ‘be’ and bol- ‘become’.
Our preliminary inventory5 of participial analytic constructions (Skribnik and

Seesing 2012) shows that the majority of forms are highly idiomatized (includ-
ing negation and the finalizing finite or participial affix) with almost exclu-
sively evidential (see Table 2) and modal meanings.
(c) The auxiliary verb of quotation gi- ‘say’ is used in only two analytic construc-
tions: one with the modal converb in -n (‘having almost done something’), the
other with the future participle in -x (‘going to do something’):
(9) kel-n gi-

say-convb.mod say
having almost said;
(10) kel-x gi-

say-partic.fut say
be about to say.
Nominal elements in analytic constructions are grammaticalized abstract head nouns

with participial relative clauses6 reanalysed as particles following the participial predi-
cate. There are about twenty such elements (Mulaeva 2011) conveying different nuances
of modality and evidentiality, for example zövtä (< etymologically zöv ‘right; obliga-
tion’ in the comitative case) and zövgo (zöv plus the negative particle uga) expressing
deontic modality:
(11) kel-x zövtä-v

say-partic.fut mod-1sg
I have to say;
(12) kel-x zövgo-t

say-partic.fut mod.neg-2pl
you do not have to say.
Analytic constructions are highly combinatory; for example (13) contains three (-ž ir-
for directivity, -dg bol- for a phasal change leading to a habitual event and -Ad bää- for
perfectivity), the last auxiliary taking the finite past affix:
5
The preliminary description was based on the corpus compiled by ourselves (ca. 3,000 clauses); for this
study we were able to use the Kalmyk National Corpus (KNC) which was officially launched in January
2012 (<http://web-corpora.net/KalmykCorpus/search/?interface_language=ru>). This corpus comprising
ca. 800,000 words enabled us to reassess our first research results and revealed new evidential forms and
strategies, as well as differences in the use of evidentials across text genres.
6
Such nominal elements/particles take personal predicative suffixes, but no temporal marking; to
express the past and future tenses they appear with auxiliary verbs bää- ‘be’ and bol- ‘become’ in a temporal
form. The absence of an auxiliary denotes the present tense. Predicative suffixes instead of possessive ones
signal the final stage of grammaticalization.
(13) ükr-müd orat-ž ir-dg

cow-pl be late-convb.imperv come-partic.hab
bol-ad bää-v
become-convb.perv be-past
The cows have started to come home later [Pyurbeev 1977: 124].
In this chapter we will analyse two synthetic and eighteen analytic markers of eviden-
tiality, as well as four constructions based on the quotation verb (see §§8, 9).
2 Organization of the evidential system
Kalmyk has a complex evidentiality system with seven evidential terms expressed by
structurally different markers: Direct and Indirect Unspecialized on the one hand, and
Specialized Inferred, Assumed, Prospective, Reported, and Common Knowledge on
the other (Table 2). Evidential marking is not strictly obligatory; it is used only when,
in the speaker’s opinion, it is relevant where the information comes from. Marking
of the information source takes place on two levels: first, the speaker can signal that
the event was directly or indirectly witnessed with the most frequent etymologically
opaque verbal affixes. Secondly, the speaker can specify the indirect source of informa-
tion (inference, assumption, prediction, report, or common knowledge) using differ-
ent idiomatic analytic constructions.
Three evidential categories based on logical operations—inferred, assumed, and
prospective—are characterized by the interaction between temporal characteristics of
the input (e.g. perceived visual evidence) and the logical operation/utterance. In the
case of Inferred we distinguish between current and previous visual evidence (cf. ‘cur-
rent evidence’ observed at the time of speaking vs. ‘previous evidence’ observed before
the time of speaking, San Roque and Loughnane 2012: 118). Assumed and Prospec-
tive distinguish between current, previous, and expected states-of-affairs. Additional
aspectual distinctions are possible.
3 Direct evidential
Direct perception in Kalmyk is encoded in the synthetic form -lA (negation: -sn uga
bilä) and two analytic constructions with -lA-marking on the auxiliary bää- (bilä),
both negated by the postpositional particle uga. The most frequently used lA-form
(1,539 occurrences in KNC) refers mainly to the third-person sg/pl (1,001 occurrences)
and the first-person sg/pl (495 occurrences); it denotes recent events that the speaker
witnessed or took part in:
Table 2. A multi-term evidentiality system in Kalmyk
Evidential Epistemic/mirative
meaning TAM/realization overtones Marker
Direct Past -lA

(neg.) -sn uga bilä
Past Habitual -dg bilä
Past Continuative (rare) -A bilä
Indirect Past mirative -ž
Pluperfect mirative -sn bää-ž
Past Habitual mirative -dg bää-ž
Inferred Present uncertainty -dg bäädltä

(current evidence)
certainty/mirative -dg bol-ža-na
Past uncertainty -sn bäädltä

(current evidence)
certainty/mirative -sn bol-ža-na I
Pluperfect certainty/mirative -sn bol-ža-na II

(previous evidence)
Future uncertainty -x bäädltä

(current evidence)
Assumed Present -dg bol-x
Past certainty -sn bol-x
Prospective Future-in-the-past -x bol-v I

(previous evidence)
Near future certainty -x bol-(ža-)na

(current evidence)
Remote future -x bol-x

(expected evidence)
continued
Table 2. Continued
Reported Past -ž ginä
Past based on reliable -sn bilä

information source
Future based on personal -x bol-v II

interviews
Common — -dg ginä

knowledge
(14) xürm özkldür ekl-lä

wedding yesterday begin-wit.evid
The wedding began yesterday [KNC].7
In principle, analytic constructions of different types behave very differently concern-

ing their combinations with -lA. Converbial constructions show no constraints and
the meaning of the resulting complex predicate is compositional, -lA retaining its
original meaning. Among participial constructions only five combinations with -lA
are attested: where only two (-dg bilä and -A bilä) retain the original meaning of -lA,
one builds a separate evidential category (-sn bilä) and two (-x bilä and -x bolla) are
idiomatized as modal constructions with irreal and deontic meanings, respectively.
Thus the direct evidential subsystem includes three values with tense-aspect-mood
(TAM) oppositions: -lA direct evidential past, -dg bilä direct evidential past habitual,
and -A bilä (very rare) direct evidential past continuous.
Example (15) illustrates a witnessed habitual past event (-dg bilä, 715 occurrences):
(15) mana eež xal'mg zää-g jahž čan-dg

our grandmother Kalmyk tea-acc how cook-partic.hab
bilä?
be:wit.evid
(You surely remember) How did our grandmother usually make Kalmyk tea?
[KNC]
7
References to the example sources are organized as follows: The letters refer to the initials of the text
author (e.g. BA for Badmin Aleksey, with the exception of KNC which refers to the Kalmyk National Corpus
and XŰ which refers to the newspaper Xal’mg Űnen); the letters are followed by the page number within
the textbook or the date of the newspaper issues. All examples without explicit references are taken from
personal interviews with informants.
Example (16) shows a witnessed continuous event occurring in a definite time period
in the past (-A bilä, less than ten occurrences in the KNC):
(16) čama-g iigž kel-x giž küläh-ä

you-acc so say-partic.fut cmplz wait-partic.pres
bilä-v
be:wit.evid-1sg
I have been waiting for you to say so [KNC].
Through negation the imperfective participle is replaced by the converb in -Ad (-Ad
uga bilä).
To sum up, the core of the subsystem of direct evidentiality in Kalmyk comprises
the simplest and most frequent marker -lA and its negative counterpart -sn uga bilä;
in addition there are two analytic constructions with aspectual characterization of the
witnessed event.
4 Indirectivity
Indirectivity is expressed by one simple synthetic marker -ž (the most frequent—851
occurrences) and two analytic constructions, all of them with past time reference
and a mirative extension. The source of information is not specified—it can mark the
information obtained through hearsay (17, 21) or inference (18), it can also denote an
uncontrolled event (19) or a dream (20).
(17) Namsa, soŋs-v-či? Dorž-in gölm-d-n' söö-nä

N. hear-past-2sg D.-gen fishing net-dat-poss.3 night-gen
arat tor-ad xon-ž.
fox stick-convb.perv spend a night-indir.evid
Däkäd ükr-n' ikrl-ž. Ün-är-ij?
also cow-poss.3 give birth to two calves-indir.evid truth-inst-q.part
Teŋgr en. Min' oda itkltä kün kel-v
heaven this just now trusty person say-past
Namsa, have you heard? (Surprisingly) a fox spent the night stuck in Dorž’s
fishing net. And (surprisingly) his cow gave birth to two calves.—Really?—
Really. Just now a reliable person told me about it [BB-102].
(18) čirä-d-än zahan pomad zövär zuzanar türk-ž

face-dat-refl white powder very thickly spread-indir.evid
Her face is unnaturally white;(she) will have spread white powder on
her face very thickly [KNC].
(19) gee-čk-ž-v… Xama uŋh-a-sn

lose-compl-indir.evid-1sg where fall-caus-partic.perv
bol-x-v-i?
become-partic.fut-1sg-q
(To my surprise) I have lost it . . . Where could I have dropped it? [BB-123]
(20) Čüüčä, bi söö-nä neg aal'ta züüd üz-ü-v.

Č. I night-gen one strange dream see-past-1sg
Neg örün öndr šavr bääšŋ deld-gd-äd
one morning high clay castle build up-pass-convb.perv
xon-ž. Bi tana ax tüšml
spend.a.night-indir.evid I your senior councillor
bol-sn bol-ž jov-ž-v
become-partic.perv aux-convb.perv aux-indir.evid-1sg
Čüüčä, I had a strange dream this night. One morning (to my surprise) a tall
clay castle had been put up during the night. I was there as your senior coun-
cillor [BB-183].
As (20) shows, this marker can be freely combined with analytic converbial construc-
tions (aspect, benefactivity, etc. plus indirectivity); it can be negated by the postposi-
tional particle uga (21):
(21) vaxtjor jum üz-ž uga

porter something see-indir.evid neg
The porter did not see anything / claimed not to have seen anything (from a
police report about a theft) [XŰ-02/02/02].
The analytic construction -dg bääž adds aspectual meaning of habituality:
(22) mini eež ik lam-nr-in xuvz-ig altn utz-ar

my grandmother big monk-pl-gen clothes-acc golden thread-inst
keerül-ž uj-dg bää-ž
decorate-convb.imperv sew-partic.hab be-indir.evid
(I was told that) My grandmother used to decorate the clothes of important
monks with embroideries made with a golden thread [XŰ-06/03/29].
The second construction, -sn bääž, refers to an event located in the ‘more remote past’,
that is, in relation not to the utterance/narration time, but to another event in the past
(pluperfect indirectivity). Often this event is named in the same sentence or paragraph
and represents one of the first two links in the chain ‘movement – perception – cognition –
communication’ (‘veni, vidi . . . ’):
(23) ger-t-än ir-xlä, šaltgta avh-m sän

house-dat-refl come-convb.cond sick uncle-poss.1sg well
bol-ad serg-äd bol-sn bää-ž
become-convb.perv cheer up-convb.perv become-partic.perv be-indir.evid
When (I) came home, (it turned out that) my sick uncle had (already) cheered up,
feeling better [KNC].
To sum up, in the indirectivity zone there is an opposition between simple past indirec-
tivity (with an additional construction to mark habituality) and pluperfect indirectivity.
5 Inferred evidential
The constructions based on current evidence are further opposed by an additional
epistemic meaning ‘degree of certainty’: if the speaker is certain of his/her conclusion,
the constructions with the auxiliary bol- ‘become’ are used, and a less certain conclusion
is encoded in the series with the particle bäädltä (etymologically bäädl ‘look, appear-
ance’ + comitative in -tA). The ‘less certain’ inferentials differentiate past, present, and
future. The ‘certain’ inferentials have no future form. Previous evidence is employed in
just one pluperfect construction with the epistemic overtone ‘certainty’ (see Scheme 1).
(a) If the current evidence is interpreted more or less simultaneously with the
speech act and the speaker is sure of their conclusion, they can use two con-
structions: -dg bol(ža)na signals that the inferred event still continues at the
speech moment (present time reference, absolute in speech act or relative in
narrations, see 24), and -sn bol(ža)na shows that the inferred event took place
previous to the perception of the current evidence (25):
(24) ä ugah-ar tiigäd bič-äd dür-äd

sound neg-inst so write-convb.perv put away-convb.perv
jov-dg bol-ža-na
go-partic.hab become-prog-pres
(Look, Bata’s suitcase under his bed is full of his poems!) Without telling any-
one, he writes (poems) and puts them away [KNC].
current evidence online processing inferred event:
→ utterance + certainty: present/past
– certainty: present/past/future
previous evidence remote processing reinterpreted evidence:
→ utterance pluperfect + certainty
Scheme 1. Inferred evidentials in Kalmyk

(25) xulxač-nr ödr-är xöö-d-ig košar-t bär-sn

thief-pl day-inst sheep-pl-acc shed-dat hold-partic.perv
bol-ža-na, asxn bürül-in aldnd tedn-ig
become-prog-pres evening dusk-gen at they-acc
ač-sn bol-ža-na.
load-partic.perv become-prog-pres
Ter saam-la-n’ Bad mir-sn bol-ža-na
that time-comit-poss.3 B. come-partic.perv become-prog-pres
(The sheep owner sees traces of his sheep and the body of his murdered assistant
Badm in an empty shed.) The thieves must have kept the sheep in the shed during
the day, and then in the evening at dusk they must have been loading the sheep.
That is when Badm must have come [DzA-55].
(b) If the current evidence is interpreted more or less simultaneously with the
speech act but the speaker is less certain in his conclusion, s/he uses construc-
tions with the particle bäädltä. The inferred event can refer to the past (-sn
bäädltä), present (-dg bäädltä) and future (-x bäädltä):
(26) en xoir-in negn-d-n’ möŋg es giž posylk ir-sn

this two-gen one-dat-poss.3 money or parcel come-partic.perv
bäädltä
part
(On watching two students cooking a big dinner) Probably one of these two
has received money or a (food) parcel (from the parents) [KNC].
(27) mini kövün-ä dun bol-dg bäädltä

my son-gen voice become-partic.hab part
(Hearing a baby crying) Probably it is the voice of my son [KNC].
(28) xur or-x bäädltä

rain enter-partic.fut part
(Look, a big cloud is coming) Probably it will rain [KNC].
(29) ör zää-tl gii-x bäädl uga

dawn glimmer-convb.term hold-partic.fut part neg
(The rain is drizzling everywhere) Probably it will not hold till sunrise [KNC].
(c) The previous (remembered) evidence (re)interpreted by the speaker has no separate
construction; it is the second meaning of the past inferred evidential in -sn bolžana.
The fact that it is about recaptured evidence is understood from the context. The logi-
cal operation here is not actually inference of some event causally connected with the
state-of-affairs perceived as evidence, but the (re)interpretation of this evidence:
(30) ter-n' digtä Dzhek avtomat-as žiŋnül-sn

that-poss.3 exactly D. phone booth-abl call-partic.perv
bol-ža-na
become-prog-pres
(Before I opened the door for you, the phone rang, but the caller hung up
without speaking. Now I think that) It must have been Dzhek calling from the
telephone booth [TA-54].
In brief, the specialized inference in Kalmyk demonstrates two main oppositions: cur-
rent vs. previous evidence, certainty vs. uncertainty. The current evidence allows the
differentiation of present, past, and future events inferred with uncertainty and pre-
sent and past events inferred with certainty. The previous evidence allows inferring
only pluperfect certain events.
One question is how the specialized inferential construction with past time refer-
ence differs from indirectivity constructions, whose reading is also inferential (see §4).
According to our informants, the specialized inferential construction presupposes a
more intensive deduction process.
6 Assumed evidential
The assumed evidential (Scheme 2) is used in utterances motivated by some cur-
rent input, sensual or informative, whereby the given state-of-affairs is interpreted/
explained on the basis of the speaker’s knowledge. This evidential distinguishes past
and present forms, -sn bolx (31) and -dg bolx (32); both forms have no fixed epistemic
overtone (depending on context, different examples can have different degrees of cer-
tainty—or none).
(31) oln hazr-ar or-sn bol-x-č

many place-inst enter-partic.perv become-partic.fut-2sg
oln küün-lä xarh-sn bol-x-č
many man-comit meet-partic.perv become-partic.fut-2sg
You must have visited many places. You must have seen many people [DB-19].
(32) gertks-m nama-g külä-žä-dg bol-x

family-poss.1sg I-acc wait-prog-partic.hab become-partic.fut
(I will go home now) My family is probably waiting for me [DzJ-25].
current state-of-affairs online processing assumed event: present /past
+ knowledge → utterance
Scheme 2. Assumed evidentials in Kalmyk

The assumed evidentiality has no future form of its own: assumptions for the future
are covered by a separate category Prospective.
7 Prospective evidential
As this evidentiality category (see Scheme 3) exclusively refers to the future, the question aris-
es whether it is a separate category—or the future realization of either Inferred or Assumed
evidentiality (cf. ‘The Prospective resembles the Inferential in that it implicitly refers to an
additional situation (E), which serves as the source of information’ (Maslova 2003: 225). We
consider it a separate category with the following arguments. First, it has common features
with both Inferred and Assumed evidentiality. As in Inferred, logical operations are used
to postulate some event that is different from the input state-of-affairs. As in Assumed, the
input state-of-affairs is not necessarily visual/sensual and the logical operations deal with
general knowledge. There are also features that are specific to the Prospective evidential: the
possibility to use as input a future, expected state-of-affairs and the fact that the temporal
characteristics reflect both the moment of speech and the state-of-affairs.
There are three prospective forms in Kalmyk: based on the current (-x bol(ža)na),
the previous (-x bolv) and the expected state-of-affairs (-x bolx).
(a) The speaker predicts an event that takes place in the near future as a logical
consequence of the current state-of-affairs; this state-of-affairs often presup-
poses the set of social norms and rules, so that the utterance gets a strong
necessitive reading and the epistemic overtone of certainty:8
(33) noolda-hi-n‘ tadn ke-x bol-ža-na-t

fight-acc-poss.3 you do-partic.fut become-prog-pres-2pl
bidn bolxla, zug zal’vr-x bol-ža-na-vidn
we become-convb.cond only pray-partic.fut become-prog-pres-1pl
You (young people) will have to continue the fight. As for us, we (being old now)
will have only to pray [DzA-7].
current state-of-affairs online processing predicted event: near future
→ utterance
previous state-of-affairs remote processing predicted event: future-in-the-past
→ utterance
expected state-of-affairs processing predicted event: remote future
→ utterance
Scheme 3. Prospective evidentials in Kalmyk
8
Cf. with the deontic modality constructions (see examples (9) and (10)) which contain a reference to known
social norms instead of a reference to another situation which serves as an impulse for a certain conclusion.
(b) The form -x bolx codes that the speaker predicts an event for the remote future
on the basis of the expected state-of-affairs. By contrast, this evidential is char-
acterized by a low degree of certainty:
(34) bi terün-lä xarh-ž čad-š-go
I he-comit meet-convb.imperv be able-partic.fut-neg
bol-ža-x-m-b?
become-prog-partic.fut-aff-q
(He intends to leave for the city.) Would I not (be able to) see him? [DzA-145]
(c) The form -x bolv codes that the speaker makes a prediction based on a previous
state-of-affairs: the situation in the past (-v) was such that its logical conse-
quence (-x) could only be the event named in the sentence:
(35) zug xal’mg kel-är ög-gd-sn zakvr-mud-yg

but Kalmyk language-inst give-pass-partic.perv command-pl-acc
en šinäs das-x bol-v
this anew learn-partic.fut become-past
(The dog I found knew commands in German and Russian.) But it had to
learn anew the commands given in Kalmyk [XY-06/07/29].
This construction has developed one more meaning (see §8).
Summing up, the three prospectives designate a prediction based on informa-
tion sources in the present, past, and future; by contrast, the simple future form in -x
denotes a hypothesis without any reference to an information source.
8 Reported evidential
The category of Reported in Kalmyk includes the construction -ž ginä and two other
constructions, -sn bilä and -x bolv, reanalysed as Reported, both characteristic of the
modern newspaper style. Besides the Indirective, -ž can have hearsay as one of its
possible readings.
(a) The quoting strategy of Kalmyk, as in all other Mongolic languages, consists
of using the quotation verb gi- alone or as an auxiliary in combination with
other verbs of speaking. Gi- introduces direct speech in its original form and
the speaker as its first argument; as an independent quotation verb, it takes
finite forms and personal marking: gi-nä-v ‘I say’, gi-nä-č ‘you say’, gi-nä ‘he/
they say’, gi-vü-v ‘I said’, gi-lä-č ‘you said (witnessed)’ etc.:
(36) či jun gi-v-č?- ezk-ek-äs zövšäl sur-na-v
you what say-past-2sg parents-abl permission ask-pres-1sg
gi-lä-v
say-wit.evid-1sg
What did you say?—I said, I will ask my parents for permission [KNC].
The most frequently used form gi-nä, the present form 3sg/pl of gi-, is grammatical-
ized as the reported evidential particle ginä; it therefore represents a widespread phe-
nomenon in which the complement clauses of verbs of quotation are reanalysed as main
clauses, and the verbs themselves as evidential particles (Aikhenvald 2004: 123). In the
process of grammaticalization, ginä has changed its properties considerably: it reports
instead of quotes, there is no slot for the exact speaker as there is just one complex predi-
cate instead of two clauses, the main and the complement, and there are restrictions on
verbal forms that encode the reported information. Compare the independent form
gi-nä in a direct speech construction with a cited form of the direct evidential (37) and
ginä as a reportative particle with the only possible form of the indirective in -ž (38):
(37) en-tn xuld-ž av-la-v gi-nä

he-poss.2pl buy-convb.imperv aux-wit.evid-1sg say-pres
(Interpreter:) He says he bought it (lit.: I bought it, he says) [KNC].
(38) Batah-as bičg ir-ž ginä, ünn-ij?

B-abl letter come-indir.evid rep truth-q
Is it true that (as they say) a letter from Bata has come? [KNC]
Ginä combined with the habitual participle in -dg is grammaticalized to a different

evidential with the meaning of common knowledge (see §9).
(b) By the construction -sn bilä the speaker signals that the information was
directly witnessed or experienced by another person (often named in the con-
text), that is, obtained through hearsay, but is considered reliable:
(39) xud-lgo, özkldür kövü-d sad dotr naad-ž

lie-convb.neg yesterday boy-pl garden inside play-convb.imperv
jov-ad, nemšin maši üz-sn bilä
aux-convb.perv German car see-partic.perv be:wit.evid
That is no lie; the boys who played in the garden yesterday saw a German car [KNC].
This construction is stylistically specialized: the majority of the examples in the KNC
are from newspapers (only 10 out of 161 occurrences come from fictional texts; there
were none in our initial self-collected corpus), so this form has become genre-specific
(= a journalist always has reliable firsthand witnesses):
(40) en xö-d Zugäräsän hääxül-d čign

this sheep-pl All-Russian exhibition-dat even
üzül-gd-sn bilä
show-pass-partic.perv be:wit.evid
These sheep were shown even at the All-Russian exhibition [XŰ-06/07/12].
A negated form of this construction is not attested; there does exist a formal counter-
part -sn uga bilä, but semantically this corresponds not to the analytic form itself, but
to the simple form in -lA: it denotes the direct witnessing by the speaker himself, not
the reliable witnessing by another person:
(41) nan-la ädl surhul' sur-sn uga bilä

I-comit like sciences study-partic.perv neg be:wit.evid
He did not study (as much) as me [KNC].
(c) To the Reported category we assign the prospective evidential -x bolv in its sec-
ond meaning realized exclusively in newspaper language,9 where it has become
restricted to reported information about planned future events obtained in per-
sonal interviews with their planners, so that often the information source is the
subject of such a sentence (or is introduced in constructions like According to . . .):
(42) en tör-müd-ig xahllhn-d ministr on’g-an ög-x

this topic-pl-acc study-dat minister care-refl give-partic.fut
bol-v
become-past
The minister will pay attention to the study of these questions [XŰ-06/09/02].
Summing up, the two central reported evidentials describe past events and differ in
the type of information source: either an unspecified person or a reliable person (often
mentioned in the context). An additional construction covers future events reported
by their planners (also often mentioned in the context). The last two constructions
show that the modern, actively developing newspaper style does not just use the exist-
ing evidentials, but creates new meanings.
9 Common knowledge
An evidential construction -dg ginä (habitual participle in -dg + grammaticalized form
of the quotation verb gi-) occurs mainly in proverbs, referring to common knowledge
(cf. Lidz 2007: 60–3):
(43) äämtxä-g asr-ž tus uga bol-d-mn ginä

coward-acc care-convb.imperv use neg become-partic.hab-aff rep
Taking care of a coward is useless (as everyone knows) [XŰ-06/08/10].
9
We found this reading only in the newspaper part of the KNC; there were no examples of it in our self-
collected corpus of Kalmyk fiction.
10 Evidentials in discourse
In Kalmyk there are three established strategies of past tense narration encoded by evi-
dentials, one for the witnessed past and two for the non-witnessed (historic descrip-
tions, folk tales etc.).
(a) A story witnessed by the speaker is marked by the synthetic marker (-lA) and
analytic constructions of direct evidentiality (-dg bilä, -sn bilä, and -A bilä):
(44) Ik när tedn-ä-d bol-la. Tüünd či bas

big party they-gen-dat become-wit.evid there you also
ir-lä-č. Ger-in haza oln bahčud zuglr-la.
come-wit.evid-2sg house-gen outside many youth gather-wit.evid
Dun dombr xoir-in ais soŋs-gd-la
song dombra two-gen melody hear-pass-wit.evid
Tiigxd bi čini duuh-i-čn soŋs-la-v,
at that time I your singing-acc-poss.2sg hear-wit.evid-1sg
čamag jahž biil-dg-i-čn üz-lä-v.
you-acc how dance-partic.hab-acc-poss.2sg see-wit.evid-1sg
Bi kesgtän čini nüdn gerd-lä-v
I long your eye stare-wit.evid-1sg
There was a big party at their place. Also you went there. Many young people
gathered in front of the house. There was dombra [a string instrument] music
and singing. At that time I heard your singing, I saw how you danced. I stared
very long into your eyes [IL-21].
(b) In traditional narrative stories and folk tales the indirectivity marker -ž is used in
the beginning, indicating a non-firsthand information source (cf. Bläsing 1984). A
typical opening formula kezänä bääž (long ago be-indir.evid) ‘once upon a time’
sets the framework of the fairy tale genre. With the non-firsthand framework
established, the following narration contains the unmarked past tense form -v
expressing a dynamic development of a story or -nA as the historical present:
(45) Kezänä bää-ž. Hurvn küük-tä emgn övgn xoir

long ago be-indir.evid three daughter-comit old woman old man two
bää-ž. Edn jir taalta bül bää-ž.
be-indir.evid they very harmonious family be-indir.evid
Küük-d-n' soŋsvrč, xoorndan eeltä bää-ž.
daughter-pl-poss.3 obedient between friendly be-indir.evid
Edü-tedü žirh-žäh-äd, küük-d-in ek-n'
some time be happy-prog-convb.perv daughter-pl-gen mother-poss.3
gemt-äd öŋgr-ž od-na

take sick-convb.perv die-convb.imperv aux-pres
It was long ago. There was an old woman and an old man with their three daughters.
They were a very harmonious family. The daughters were obedient and friendly to each
other. Having lived happily for some time, the mother becomes sick and dies [XT-4].
(c) One more narration strategy found in traditional stories and fairy tales is based
on the historical present, which makes the presentation more vivid. In this case
a standard beginning in the folkloric texts contains the analytical construction
-dg bolna:
(46) Kezänä neg bain kün kövün-d-än ger av-č

long ago one rich man son-dat-refl house take-convb.imperv
ög-dg bol-na. Ödr ir-väs ber
give-partic.hab become-pres day come-convb.cond daughter-in-law
du har-l uga atyh-ad bää-dg bol-na
sound emit-convb.neg be sad-convb.perv be-partic.hab become-pres
Long ago a rich man arranged (lit.: arranges) a marriage for his son. Day by day the
daughter-in-law is getting sadder and sadder without uttering a sound [EK-343].
The stylistic possibility of evidential markers in creating humorous effects is illustrated

by the use of the direct evidential in the rhymed folktale Daln xoir xudl (‘Seventy-
two lies’):
(47) ezk-äs-n türül-ž har-ad,

father-abl-refl be ahead-convb.imperv come-convb.perv
elnzg ezk-in-n’ adu xärül-lä-v
great-grandfather-gen-poss.3 herd look after-wit.evid-1sg
Being born before my father, I looked after a herd of horses of my great-
grandfather [Xudl-8].
11 Evidentiality strategies: complementation

There are two competing clause complementation types in Kalmyk: a non-finite (par-
ticipial complement with a case marker) and a finite with an introducing complemen-
tizer giž (< imperfective converb of the previously discussed quotation verb gi-, see
§8). We argue that these two types distinguish meanings related to the information
source. Our analysis of the corpus indicates that not every predicate taking a clausal
complement can take both these types; in the following we will show the distribution
of complement types with verbs of immediate perception and cognition.
The verb soŋs- ‘hear, listen’ takes both kinds of complement clauses; a participial
complement with the accusative case implies direct perception (actual hearing), while
a finite complement clause with the complementizer giž marks information obtained
through hearsay:
(a) Verbs of immediate perception (soŋs- ‘hear’, üz-, xälä- ‘see’)
(48) zal-d bää-sn uls Kugultinov šülg-üd-än

auditorium-dat be-partic.perv people K. poem-pl-refl
umš-s-ig soŋs-v
read-partic.perv-acc hear-past
The people sitting in the auditorium listened to Kugultinov reading his poems
[XŰ-06/06/20].
(49) xalx Moŋhl-yn političesk boln olna üüldäč-nr dund

Khalkha Mongolia-gen political and social leader-pl among
dörvd jas-ta uls oln bilä giž soŋs-la-v
Dörbet bone-comit people many be:wit.evid cmplz hear-wit.evid-1sg
I heard that there were many Dörbet people among Khalkha Mongolian political
and social leaders [XY-06/05/18].
The verbs üz- ‘see’, xälä- ‘see’, indicating direct visual perception, only take a participial
complement and never take a finite clause complement with the complementizer giž:
(50) özkldür zetkrtä jovdl učr-s-ig

yesterday dangerous accident happen-partic.perv-acc
[*učr-v giž] üz-lä-v
happen-past cmplz see-wit.evid-1sg
Yesterday I saw a dangerous accident happen [XŰ-06/02/16].
(b) Verbs of cognition (med- ‘know’, san- ‘think’)

The verb med- ‘know’ allows both types of complements with the difference that the
participial complement with accusative case encodes direct perception of a situation
(49), while the giž-complement indicates that the speaker’s knowledge is indirect (50):
(51) xaalh-s mu bol-ža-x-i-n' bidn evrän en

road-pl bad become-prog-partic.fut-acc-poss.3 we self this
jovud-t med-ü-vidn
trip-dat know-past-1pl
We have understood ourselves during this trip (we have taken part in) that the
roads are becoming poor in quality [XŰ-06/02/16].
(52) en kövü-g iigž zov-x giž

this boy-acc like this suffer-partic.fut cmplz
med-sn uga-v
know-partic.perv neg-1sg
I did not know (at that time) that this boy would suffer like this [KNC].
Commonly held views are expressed with a giž-complement:
(53) kü allhn ik kilnz giž med-x kergtä

man killing great sin cmplz know-partic.fut necessary
One should know that it is a great sin to kill a man [XY-06/09/07].
With the verb san- ‘think’ only a finite complement clause with giž is possible, probably
because this verb denotes complex logical operations like evaluations, etc. and not a
direct reflection of events:
(54) tiim mu zaka sol’x kergtä giž san-gd-na

such bad law change-partic.fut necessary cmplz think-pass-pres
I rather think that this bad law needs to be altered [XY-06/04/22].
Therefore, the choice of complement types by verbs of cognition and perception in

Kalmyk depends on the information source: the non-finite (participial) comple-
ment signals a firsthand information source, and the finite clause complement with
the complementizer giž indicates a non-firsthand source (hearsay or complex logical
operations).
12 Evidentiality strategy: demonstratives and interpretation of evidence

There are two idiomatic constructions, gidg en (60 occurrences) and gidg ter (29 occur-
rences), related to the evidence/knowledge marking system. They are grammaticalized
on the basis of complex structures with demonstratives en ‘this’ and ter ‘that’ as for-
mal main predicates plus subject clauses built by habitual participles in -dg from the
quotation verb gi- (‘this/that is what they would call . . . ’). Both constructions denote
that the speaker generalizes the evidence and gives it a complex interpretation or an
evaluation; that is, these constructions are close to both inferential and assumptive
evidentials, but are still specific. In the case of en the current evidence is interpreted
(55), and in the case of ter the previous evidence is like remembered facts (56):
(55) iigäd edgd-nä gi-dg en

this way recover-pres say-partic.hab this
(This is what they call) such a recovery [KNC].
(56) nan-as zul-ža-na gi-dg ter

I-abl avoid-prog-pres say-partic.hab that
(She) is avoiding me (I’d say remembering her behaviour) [KNC].
Very often proverbs occur as ‘labels’ or formulaic means of interpretation:
(57) xojr uul biš, xojr kün xarh-na gi-dg en

two mountain neg two man meet-pres say-partic.hab this
Two mountains do not meet, but two men do (this is what they say) [KNC].
Compare also (58) with the mirative particle attached on the demonstrative (see §13
for further discussion):
(58) yosn tuul-in orn gi-dg-tn en-ž

real rabbit-gen burrow say-partic.hab-poss.2pl this-mir
(Just look, this is, for your attention) a real rabbit burrow! [XM-26]
13 Mirativity
In addition to the verbal indirective marker -ž that can have a mirative extension,
Kalmyk has a special mirative particle -ž which is homonymous with the indirective
marker and etymologically related to it: historically it is an indirective form of the
old copula verb *a- (cf. Sanžeev 1983: 290). These two markers are in complementary
distribution: the mirative marker accompanies nominal predicates (59), existential
negation (60), and participles functioning as finite predicates (61, 62). The central
meaning of the particle is mirative, though with participial predicates it can also
convey indirectivity; separate mirative values (Aikhenvald 2012) are not differenti-
ated, the most common readings being surprise, newly acquired and unexpected
information:
(59) Kermn gem-tä-ž-ij?

K. illness-comit-mir-q
Is Kermn ill? [TA-12] (new unexpected information)
(60) aav eež xoir ger-t-än uga-ž

grandfather grandmother two house-dat-refl neg-mir
Grandfather and grandmother are both not at home (surprise, contrary to
expectations) [Sanžeev 1983: 290].
(61) xurl xaa-x-m-ž

temple close-partic.fut-aff-mir
The temple will be closed (contrary to expectations) [EK-75].
(62) Dzek zavod-t ködl-dg-ž

D. factory-dat work-partic.hab-mir
(Surprise, sudden discovery:) Dzek is working in a factory [TA-41].
14 Summary
Kalmyk has an optionally used seven-term evidential system covering Direct, Indi-
rect, Inferred, Assumed, Prospective, Reported, and Common Knowledge evidentials
organized on two structurally different levels. On the first level the primary synthetic
forms -lA and -ž are in opposition: the marker -lA indicates direct perception, and -ž
encodes information obtained indirectly (inference, assumption, and hearsay as well
as mirativity are represented as contextual readings).
The second level is shaped by idiomatic analytic constructions of different character
with highly specialized evidential meanings. These can be roughly divided into ‘Logi-
cal operations’ and ‘Communicated information’.
In the first group we find five Inferred, two Assumed, and three Prospective eviden-
tial forms. These are opposed through the temporal reference to the state-of-affairs
that serves as input or evidence (previous or current, for Prospective is also expected).
In the case of Inferred they are also opposed by epistemic overtones (certainty vs.
uncertainty, whereby the ‘certain’ inference has no future form—this option is sepa-
rately elaborated in prospectives).
‘Communicated information’ includes Reported and Common Knowledge. Report-
ed information is encoded in the three constructions -ž ginä, -sn bilä and -x bolv, with
the last two being exclusively used in newspaper texts encoding information about the
past or future obtained from a reliable first witness or the planner. The construction of
common knowledge -dg ginä indicates by its form that common knowledge is com-
municated (ginä from the quotation verb gi-).
Evidential distinctions are also encoded in complement clauses with verbs of per-
ception and cognition: non-finite clauses with case affixes as connectors denote the
information that is directly perceived, and finite complement clauses with the com-
plementizer giž (one more form of the quotation verb gi-) represent their content
as indirect information (e.g. hearsay or the results of logical operations). One more
evidentiality strategy of Kalmyk is based on demonstratives combined with another
form of gi-; two constructions with proximal and distal demonstrative pronouns, gidg
en and gidg ter, respectively encode online or postponed interpretation or evaluation
of evidence (‘this/that is what you would call . . . ’). This strategy is close to both infer-
ence and assumption, but seems to be something different.
The indirectivity marker -ž as well the inferred constructions may have mirative
extensions; a pure mirative meaning in Kalmyk is conveyed by the particle -ž, etymo-
logically connected with the verbal indirectivity marker -ž (the old copula verb *a- in
indirective form), but accompanying nominal predicates.
Text corpus
BB Basŋga Baatr. 1981. Bumbin orn. Elista: Xal’mg degtr harhač.
DB Doržin Basŋ. 1981. Ezn. Elista: Xal’mg degtr harhač.
DzA Dzhimbin Andrey. 1990. Küünä žirhl iim. Elista: Xal’mg degtr harhač.
DzJ Dzhambin Jaroslav. 1983. Mu uga—sjan uga. Elst: Xal’mg degtr harhač.
EK Ernžänä Konstantin. 1979. Halan xadhl. Elista: Xal’mg degtr harhač.
IL Inžin Liž. 1972. Ol’dan küükn. Elista: Xal’mg degtr harhač.
KNC Kalmyk National Corpus <http://web-corpora.net/KalmykCorpus/search/
index.php?interface_language=ru>
TA Tačin Anža. 1986. Buurldan ünr. Elista: Xal’mg degtr harhač.
Xudl Budzhalov Egor. 1990. Editor of Daln xoir xudl. Elista: Xal’mg degtr harhač.
XM Xon’na Mixail. 1974. Či medxmč, Smolenskin hazr. Elista: Xal’mg degtr harhač.
XT Xal’mg tuuls. 2010. Elista: Xal’mg degtr harhač
XY Xal’mg Űnen (Kalmyk national newspaper, examples taken from the KNC)
References
Aikhenvald, Alexandra Y. 2004. Evidentiality. New York: Oxford University Press.
———. 2012. ‘The essence of mirativity’. Linguistic Typology 16: 435–85.
Bläsing, Uwe. 1984. Die finit-indikativischen Verbalformen im Kalmückischen. Untersuchung
ihrer Anwendung und ihrer Abgrenzung voneinander. Stuttgart: Steiner-Verlag Wiesbaden.
———. 2003. ‘Kalmuck’, pp. 229–47 of The Mongolic languages, edited by Juha Janhunen. Lon-
don: Routledge.
Area 30(2): 45–87.
Maslova, Elena. 2003. ‘Evidentiality in Yukagir’, pp. 219–36 of Studies in evidentiality, edited by
Alexandra Y. Aikhenvald and R. M. W. Dixon. Amsterdam: John Benjamins.
Mulaeva, N. M. 2011. Modal’nost’ vozmožnosti, neobxodimosti i dostovernosti (na materiale
sovremennogo kalmyckogo jazyka). [Modality of possibility, necessity, and certainty in con-
temporary Kalmyk]. Elista: NPP Džangar.
Pyurbeev, G. Z. 1977. Grammatika kalmyckogo jazyka. Sintaksis. [The grammar of Kalmyk.
Syntax]. Elista: Kalmgosizdat.
San Roque, Lila and Loughnane, Robyn. 2012. ‘The New Guinea Highlands evidentiality area’,
Sanžeev, G. D. 1983. Editor of Grammatika kalmyckogo jazyka. Fonetika i morfologija. [Grammar
of Kalmyk. Phonetics and morphology]. Elista: Kalmgosizdat.
Say, S. S., Baranova, V. V., and Serdobol’skaja, N. V. (eds)2009. Issledovanija po grammatike
kalmyckogo jazyka. [Studies in grammar of Kalmyk]. Acta Linguistica Petropolitana V,
Vol. 2. Sankt-Peterburg: Nauka.
Skribnik, Elena and Seesing, Olga. 2012. ‘K opisaniju evidenzial’nosti i mirativnosti v kalmy-
ckom jazyke’. [Towards a description of evidentiality and mirativity in Kalmyk]. Voprosy
jazykoznanija 4: 39–72.
8
The non-visible marker in Dyirbal
R . M . W. D I XO N
1 Introduction
In the Dyirbal language community of north-east Australia there was a convention—
nay, a requirement—of being at all times maximally specific. If a snake is referred to,
one should say which type of snake—whether the deadly brown snake, bayi walguy,
or the black snake, bayi gajamay, or whatever. There is, it is true, a generic term bayi
wadam ‘snake’ but this is only used if the identity cannot be determined; say, if only
the shadow of a snake is seen, or just its tail. Everyone in the community was familiar
with the various snake species (some highly dangerous, others harmless). Identifica-
tion should always be made, and be stated.
A verb ‘know’ is lacking from Dyirbal, simply because it would be too vague.
Whereas a speaker of English can say just I know where the money is hidden, in Dyirbal
there is a requirement to say how one knows this—perhaps ‘My father told me where
the money is hidden’ or ‘I saw the money being hidden’. (There are, however, adjectives
meaning ‘don’t know’: ŋañum ‘not familiar with a person or place’ and juru ‘don’t know
where someone is, never heard a particular story, etc.’ .)
In keeping with the need for precision, there is a set of grammatical markers which
accompany nouns and verbs, indicating whether the referent is ‘there and visible’
(markers commencing with ba-), or ‘here and visible’ (ya-), or ‘not visible’ (ŋa-)—this
is an evidentiality system (see §2.3 of Chapter 1). This chapter explains the meaning
and function of the markers, paying particular attention to the non-visible ŋa- forms.
These may describe something only known from its noise, something which has just
passed out of sight but is still audible, something neither visible nor audible, or some-
thing remembered from the past. In addition, spirits are always described with the
non-visual marker, even if they can be seen.
All the examples in this chapter are taken from texts (rather than being construct-
ed). Dyirbal has free ordering of words not only within a clause but also within a
sentence—see example (13). It is also highly elliptical. The reader will thus need to
172 R. M. W. Dixon
be attentive in studying examples, but this should provide reward through an under-
standing of the modus operandi of this wonderful language.
The language has just 16 phonemes—3 vowels (i, a, u) and 13 consonants. There is
a stop and nasal at each of four places of articulation: bilabial (b, m), apico-alveolar
(d, n), lamino-palatal (j, ñ) and dorso-velar (g, ŋ). Liquids comprise apico-alveolar
lateral l and two rhotics: apico-alveolar trill rr and apico-postalveolar continuant r.
And there are two semi-vowels: lamino-palatal y and bilabial-velar w.
All roots and words have at least two syllables, except for interjections ŋa ‘yes’ and
ŋu ‘alright’, and short forms ban, bam, ŋan, ŋam of absolutive noun markers balan,
balam, ŋalan, ŋalam (see §3.1, §3.3). The sign ‘/’ is used in transcribed texts to separate
intonation groups; an utterance could end at any place marked by ‘/’.
2 Grammatical background
Dyirbal shows mixed ergative/accusative morphological marking. Nouns, adjectives
and noun markers have ergative case suffixes for transitive subject (A) function, and
absolutive case, with zero realization, for intransitive subject (S) and transitive object
(O). In contrast, for first- and second-person pronouns the nominative form covers A
and S functions and the accusative is used for O function.
Despite this morphological split, there is homogeneity in syntax. Dyirbal oper-
ates with an exclusively S/O pivot (grammaticalized topic). Two clauses can only be
adjoined to form a sentence if they have a shared argument which is in S or O function
in each (it can be omitted from a non-initial clause in the pivot chain). As illustrated
in (13), an anti-passive derivation is available to place an underlying A argument into
surface S function, so that it can take part in a pivot chain.
Basic grammatical points relevant to the discussion of noun markers are summa-
rized here.
2.1 Case and genitive suffixes on nouns and adjectives

Nouns and adjective select one term from the following system:
Core syntactic cases:
1 absolutive (S and O functions): zero realization.
2 ergative (A function). Allomorphs include -ŋgu after a disyllabic stem ending
in a vowel, -gu after a longer stem ending in a vowel, -du after n, -bu after m,
and -u after r.
Peripheral syntactic cases:

3 instrumental: identical to ergative. The two cases are distinguished syntac-
tically: the anti-passive derivation affects ergative and leaves instrumental
unchanged, while the applicative derivation affects instrumental and leaves
ergative unchanged.
4 dative: gu.
Peripheral locational cases:
5 allative: -gu. This has the same form as dative; they are distinguished by the
fact that a dative noun or adjective is accompanied by a dative noun marker—
as in (13), (17), (30b), (38), and (39a)—while an allative noun or adjective is
accompanied by an allative verb marker—as in (19).
6 ablative: -ŋunu.
7 locative: identical to ergative but with final -a in place of -u.
Cases mark the function of an NP within a clause. This system of nominal suffixes also
includes one suffix marking function within an NP.
8 genitive: -ŋu after a vowel, liquid or semi-vowel, -u after a nasal.
The genitive is essentially a derivation. A genitive modifier within an NP takes the
same case ending as the noun it modifies; for example yara-ŋu (man-gen.abs) guda
(dog.abs) ‘man’s dog’ in S or O function, yara-ŋu-njin-du (man-gen-linker-erg)
guda-ŋgu (dog-erg) ‘man’s dog’ in A function. Note that the genitive is used only for
alienable (including kinship) possession, as in (7). A whole–part relationship (‘inal-
ienable possession’) is shown just by apposition, as in (6).
2.2 Case and genitive forms of pronouns

There are singular, dual, and plural numbers for first- and second-person pronouns.
Their paradigm is:
(1) nominative accusative

(S and A functions) (O function) dative genitive
1sg ŋaja ŋaygu-na ŋaygu-ngu ŋaygu
2sg ŋinda ŋinu-na ŋinu-ngu ŋinu
1du ŋali ŋali-na ŋali-ngu ŋali-ŋu
2du ñubala ñubala-na ñubala-ngu ñubala-nu
1pl ŋana ŋana-na ŋana-ngu ŋana-ŋu
2pl ñuray ñura-na ñura-ngu ñura-ŋu
The plurals and duals show regular suffixes added to the nominative (with final y
omitted from 2pl): accusative -na, dative -ngu, and genitive -ŋu after a disyllabic and
-nu after a longer form. For the singulars, accusative -na and dative -ngu are added to
the genitive.
An important point is that there is no 3sg pronoun. As will be seen below, noun
markers carry some of the functional load of 3sg pronouns in other languages. There
are 3du and 3pl pronouns, which differ between dialects and are on a different pattern
from 1st and 2nd persons.
174 R. M. W. Dixon
2.3 Verbs
There is a strict division between transitive verbs (taking A and O arguments) and
intransitives (taking an S argument).
There are also two conjugations, which are independent of transitivity. The main
verbal inflections are:
• past tense: -ñu ~ -n.
• future tense: -ñ.
In southern dialects, the past tense also covers the present, and is glossed ‘non-future’.
In northern dialects, the future tense also covers the present, and is glossed ‘non-past’.
• imperative: zero realization.
• purposive: -ygu ~ -li.
A sequence verb1-tense verb2-purp means that the action of verb1 is carried out
so that the activity referred to by verb2 should eventuate. If the first verb in an utter-
ance is inflected with the purposive, this means ‘should do, want to do’, as in (39a)
and (39b).
• Relative Clause (RC) suffix -ŋu. This is followed by the case inflection in the
main clause of the common argument which is shared by the main and relative
clauses; see (30b).
The most common verb in the language, yanu- ‘go’ is slightly irregular. Its past tense
form is just yanu, when *yanu-n would be expected.
2.4 Noun classes

Each noun belongs to one of four noun classes (or genders). These are shown only by
the form of a noun marker which accompanies the noun in its NP. The rough semantic
contents of the classes are:
M human masculine, moon, rainbow; non-human animate;
F human feminine, sun; water; fire; fighting;
E edible vegetables and fruit;
N all else (neuter)—including trees with non-edible parts and ‘place’.
There are also a number of principles for shifting noun class membership and, of
course, some assignments that are irregular and cannot currently be explained.
3 Noun markers
A noun is generally accompanied by a ‘noun marker’ showing the location of its refer-
ent, case (or genitive), and noun class. By far the most common marker begins with
ba-, meaning ‘there and visible’. It also has a default function, used when distance and
visibility are not relevant. Noun markers reflect the core, syntactic peripheral and geni-
tive forms of inflections on nouns and adjectives.
3.1 Noun markers with ba- ‘there’ (and default form)

The ba- paradigm is:
(2) Noun absolutive (S ergative (A function)

class and O functions) and instrumental dative genitive
M bayi ba-ŋgu-l ba-gu-l ba-ŋu-l
F bala-n ~ ba-n ba-ŋgu-n ba-gu-n ba-ŋu-n
E bala-m ~ ba-m ba-ŋgu-m ba-gu-m —
N bala ba-ŋgu ba-gu ba-ŋu
The absolutive M form bayi is irregular; it would be expected to be *bala-l. For F

and E absolutive, long and short forms bala-n and ba-n and bala-m and ba-m bear no
semantic difference and are in free variation. There is no genitive form in the E noun
class because a fruit or vegetable cannot be an alienable possessor. (The form baŋum
is actually a verb marker ‘from there’; see §4.)
Added to root ba(la)-, the paradigm is (apart from bayi) completely regular, with
case suffixes zero, -ŋgu-, -gu-, and -ŋu-, followed by noun class endings -l for M, -n for
F, -m for E, and zero for N.
The root appears to be bala- for absolutive and ba- elsewhere. In fact, longer forms
based on bala- are occasionally used in place of those in the last three columns. They
carry a meaning of indeterminacy. For example, one may hear bala-ŋgu-l ‘man (A
function) somewhere or other there’, in place of the much more frequent ba-ŋgu-l
‘man (A function) there’.
We can now illustrate an absolutive noun marker in an NP. (Noun and verb markers
are printed in bold throughout.)
(3) /[bayi yirriñjila]O/ ŋajaA ŋinu-ngu buwa-ñu/

there.abs.m dragonfly.abs 1sg.nom 2sg-dat tell-non.fut
The dragon fly, I told about it to you
The NP in O function for (3) has head noun yirriñjila ‘dragonfly’ which, being non-
human animate, belongs to the M noun class. It is accompanied by the absolutive M
form of the ‘there’ and default noun marker, bayi.
Example (4) also includes bayi in the O noun phrase, and noun class F ergative
form ba-ŋgu-n in the A NP, alongside head noun yabu ‘mother’. All three nouns in
(4) bear derivational suffix -jarran ‘a pair of ’. (Note the discontinuous A NP, its two
components separated by the verb.)
176 R. M. W. Dixon
(4) /[bayi muwarri-jarran/ ñalŋga-jarran]O

there.abs.m initiated.youth-pair.abs child-pair.abs
ba-ŋgu-nA ñinay-ma-n / yabu–jarran-duA/
there-erg-f sit-applic-non.fut mother-pair-erg
The pair of mothers camped (lit. sat) with the pair of children,
(newly) initiated youths
An NP most often includes an appropriate noun marker, but there are many excep-
tions. In speaking Dyirbal, it would be pedantic always to include one, and infelicitous
never to do so. Example (5) is from George Watson’s life story, recounting how one
of his kinsmen helped a policeman to locate him in the forest, for transportation to a
penal settlement. Here the A NP includes a noun marker ba-ŋgu-l. The O NP, whose
head is bulijiman ‘policeman’ (a loan from English) could have included bayi but in
this instance does not.
(5) /[ba-ŋgu-l Munudiyur-u]A bulijimanO bani-ma-n

there-erg-m Sandy.King-erg policeman.abs come-applic-past
ŋaygu-ngu ñimal-ŋay-gu/
1sg-dat grab-anti.passive-purp
Sandy King brought (lit. came with) the policeman to catch (lit. grab) me
A pivot chain links a number of clauses which share an argument that is in S or O

function in each. Typically, the pivot NP is stated in full at the beginning with just its
noun marker being repeated later in the chain. The noun marker is functioning like
an anaphoric third-person pronoun. For example, when Bessie Jerry was explaining
how a bark blanket used to be made, in the olden days, she began:
(6) /balaO nudi-n magurraO/ gugaO guñji-n/

there.abs.n cut-non.fut fig.tree.abs bark.abs peel-non.fut
guñji-n balaO/ ginbi-n/ budi-li/
peel-non.fut there.abs.n peel.top.layer-non.fut carry-purp
A fig tree (Ficus variegata) was cut down, the bark was peeled off it, it was
peeled off, just the top layer was peeled off, to be carried away
This illustrates the S/O pivot (ergative syntax) of Dyirbal. In (6) there are five transitive
verbs whose common O argument is bala magurra ‘fig tree’, expanded to bala magurra
guga ‘fig tree bark’. No NP in A function is stated for any of the verbs. There may have
been a single agent all through, or different agents for the various operations—felling the
tree, peeling off its bark, separating off the top layer of bark, and carrying it away (to be
made into a blanket). The focus here is on the tree and its bark, and what happens to it.
The possibilities for inclusion in an NP are wide. For example, in (7)—from a story
by Daisy Denham—the discontinuous S NP includes pronoun, noun marker, and head
noun, plus an embedded alienable possessor phrase which consists of noun marker
and head noun.
(7) /ñubalaS wiyama-ñ/ [ba-n bulgu-jarran

2du.nom do.what-fut there.abs-f wife-pair.abs
[ba-ŋu-l mugirray-ŋu]POSSESSOR]S/
there-gen-m.abs elder.brother-gen.abs
What are you two, pair of wives of the elder brother, going to be doing?
3.2 Noun markers with ya- ‘here’ and giyi/giña- ‘this’

The paradigm for noun markers commencing with ya- is:
(8) Noun (absolutive ergative (A function)

class form) and instrumental dative genitive
M (yayi) ya-ŋgu-l ya-gu-l ya-ŋu-l
F (yala-n) ya-ŋgu-n ya-gu-n ya-ŋu-n
E (yala-m) ya-ŋgu-m ya-gu-m —
N (yala) ya-ŋgu ya-gu ya-ŋu
Noun markers with ya- from the ergative/instrumental, dative, and genitive columns
behave exactly like ba- forms. An example where the referent of the A NP is ‘here’ and
that of the O NP is, contrastively ‘there’ is:
(9) /[ya-ŋgu-n bajinjila-gu]A maŋga-n

here-erg-f spangled.drongo-erg pick.up-non.fut
[ba-n buni]O/
there.abs-f fire.abs
The spangled drongo bird (Dicrurus hottentottus) here picked up the fire
there
This comes from Andy Denham’s recounting of the legend of the origin of fire. The
spangled drongo snatched the only fire in the world from the clutches of the rainbow
snake, and this is why the bird is in F noun class, the same as fire.
Ya- markers mean ‘here’. They are not demonstratives and do not mean ‘this’.
Also they cannot be used in S or O function. What we get instead is a separate set
of demonstrative markers; they only have absolutive form and only occur in S or O
function:
178 R. M. W. Dixon
(10) Noun class Demonstrative (S and O functions)

M giyi
F giña-n
E giña-m
N giña
As usual, suffixes -n, -m, and zero are used for the F, E, and N noun classes, and the M
form giyi is again irregular.
In his autobiographical narrative, George Watson (whose birth was due to the rape
of his Aboriginal mother by a white man) tells how his grandfather kept him from a
police round-up of part-bloods, saying:
(11) /ŋayi/ ŋajaA giyiO wabu-ŋga buyba-ñ/

yes 1sg.nom this.m forest-loc hide-non.past
Yes, I’ll hide this [masculine one] in the scrub
And in a story from traditional religion describing how ancestral beings called Jujaba
created the world:
(12) /[Jujaba-gu ba-ŋgu-l]A [giña [ŋana-ŋu]POSSESSOR mija]O

Jujaba-erg there-erg-m this.n 1pl-gen.abs place.abs
wayu-ma-n/ ñinay-ma-li yala-y/
transformed-caus-past sit-applic-purp here-loc
Jujaba created these places of ours, to be lived in here (by us)
Demonstratives only occur in S or O function. But what if we want to have deictic ref-
erence to an argument in transitive subject function, A? An anti-passive derivation is
applied, putting the argument which is in underlying A into surface S function. When
Tommy Warren related the story of ancestral being Girugarr, who travelled through
the land naming places, he employed the transitive verb manja- ‘point out by shout-
ing’. Girugarr would be in A function for verb manja- and bala mija ‘places’ would be
in O function. Anti-passive, shown by -ŋa- on the verb, puts Girugarr into S function,
referred to by demonstrative giyi. And the erstwhile O argument, bala mija ‘places’ is
now in dative form, ba-gu mija-gu.
(13) /jañja ba-ŋum [ba-gu mija-gu]

now there-abl there-dat.n place-dat
manja-manjal-ŋa-yarray-gu giyiS yanu-ñ/
redup-point.out-anti.passive-begin-purp this.m go-fut
Now from there this [masculine one] goes to point out by a shout all the places
Example (13) illustrates the exceptional freedom of word order in Dyirbal. In temporal
sequence, the first clause has the verb yanu-ñ ‘will go’ and the second has the verb in
purposive inflection manja-manjal-ŋa-yarray-gu ‘in order to start to point out a lot’
(verbal reduplication indicates ‘do a lot’ and derivational suffix -yarray- is ‘start to do’).
In fact, the verbs occur in reverse order, their logical relationship being clear from the
inflections they bear.
A verb marker with ya- ‘here’ has never been heard in the same NP with a ba- ‘there’
marker. However, a demonstrative may co-occur with an absolutive ba- form. This is
illustrated by Chloe Grant’s story of how a man transmogrified into the spirit Jigubina,
who can be seen as a shooting star streaking across the sky. In (14) the first clause has a
discontinuous S NP, with bayi giyi preceding the verb wayu-bi-n and mija-ŋunu yara
following it. Then in the second clause the pivot NP is recapitulated by bayi ‘he’.
(14) /[bayi giyi]S wayu-bi-n

there.abs.m this.m transformed-become-non.fut
[mija-ŋunu yara]S/ bayiS Jigubina-bi-n/
camp-abl man.abs there.abs.m Jigubina-become-non.fut
This man there from the camp became transformed; he became a Jigubina
Putative absolutive forms of ya- noun markers were included in parentheses in (8).
They cannot occur in S and O function. However there are some suffixes which may
be added to the absolutive forms of noun marker, and these do involve yayi, yalan, etc.
For instance, with aversive (‘for fear of ’) -ñaŋga, we can get bayi-ñaŋga ‘for fear of M
referent there’ and yayi-ñaŋga ‘for fear of M referent here’. (Note that there are no short
forms of yala-n, yala-m similar to ba-n and ba-m.)
3.3 Noun markers with ŋa- ‘non-visible’

The third variety of noun markers begins with ŋa- ‘non-visible’. Its paradigm is identi-
cal to that of ba, in (2), with initial ba- replaced by ŋa-:
(15) Noun absolutive (S ergative (A function)

class and O functions) and instrumental dative genitive
M ŋayi ŋa-ŋgu-l ŋa-gu-l ŋa-ŋu-l
F ŋala-n ~ ŋa-n ŋa-ŋgu-n ŋa-gu-n ŋa-ŋu-n
E ŋala-m ~ ŋa-m ŋa-ŋgu-m ŋa-gu-m —
N ŋala ŋa-ŋgu ŋa-gu ŋa-ŋu
Chloe Grant told of sitting around the camp fire when she was a girl and hearing a
noise which sounded like talking, but was so far off that one couldn’t make out what
was being said (this is described by the noun mulgu). The old people said that it must
be made by the frightful female spirit Dambun.
180 R. M. W. Dixon
(16) /ŋala-nS ŋanda-ñu mulguS/

non.vis.abs-f call.out-non.fut indistinct.noise.abs
She [Dambun] was heard calling out with an indistinct noise
In (16), absolutive F form ŋala-n refers to something which is heard but not seen.
In (17), dative E form ŋa-gu-m is used for something remembered from the past
(and not currently visible). Ida Henry recounted how a man told the mother of two
boys:
(17) /ñalŋga-jarranO/ ŋajaA munda-ñ [bayi daman-jarran]O

child-pair.abs 1sg.nom lead-fut there.abs.m son-pair.abs
[gayŋgal-gu/ ŋa-gu-m-guŋgarri]/
finger.cherry-dat non.vis-dat-e-north
I’ll take the pair of children, your two sons, for finger cherries [Rhodomyrtus
macrocarpa], remembered to be (in a place) to the north
In §3.1, longer forms of ba- noun markers were mentioned; for example bala–ŋgu-l—
indicating indeterminacy, ‘somewhere’—rather than ba-ŋgu-l. No such longer forms
have been encountered for ya- or ŋa- markers.
Ya- and ba- markers cannot co-occur, whereas demonstrative and ba- forms may
feature in the same NP. And so may ŋa- and ba- forms; this is illustrated in (27), (31b),
(34), and (36).
4 Verb markers
Whereas a noun in a core or syntactic peripheral case, or in the genitive, may be
accompanied by a noun marker (showing its noun class, and agreeing with it in case),
a noun in a peripheral locational case—allative, ablative, or locative—may be accom-
panied by a verb marker (or verb modifier). This agrees with the noun in case but does
not show noun class. The full paradigm of verb markers is:
(18) ‘there’ ‘here’ ‘non-visible’

allative of place ba-lu, bala-rru ya-lu ŋa-lu
allative of direction ba-li, bala-rri ya-li —
ablative ba-ŋum, bala-ŋum ya-ŋum ŋa-ŋum
locative bala-y yala-y ŋala-y
The first entry in each row for the ba- column is the most common form—ba-lu ‘to a
place there’, ba-li ‘in a direction there’, ba-ŋum ‘from there’ and bala-y ‘at there’. As with
non-absolutive ba- noun markers, the longer forms indicate indeterminacy: bala-rru
‘to some place or other over there (possibly also involving a circuitous route)’, and so on.
The nominal suffix -gu is used for both the syntactic peripheral case dative and
the locational peripheral case allative. Exemplifying with the noun mija ‘place, camp,
house’, we can have dative mija-gu with noun marker ba-gu ‘for the house (e.g. to build
it)’. Or allative mija-gu can co-occur with either of the allative verb markers—balu
mija-gu ‘to the place there’ or bali mija-gu ‘in the direction of the place there’. Similarly
with ya- and ŋa- noun and verb markers.
Verb markers commencing with ŋa- appear just to have the ‘remembered’ sense.
There is no allative of direction form in the ŋa- column. That is, one may say ŋa-lu
‘to a place remembered from the past’—as in (22)—but not *ŋa-li ‘in a direction
remembered from the past’. In fact, there is a commonly-occurring pronoun, first-
person dual ŋali, used in (21); and it is a feature of Dyirbal that it generally avoids
homonyms.
There is just one verb marker which also has a temporal meaning: ba-ŋum can be
either ‘from there’—as in (13)—or ‘and then’—as in (22).
In (19), the verb marker ba-lu co-occurs with a noun in allative case, in (20) bala-y
is used with a noun in locative case, and in (21) ya-ŋum accompanies one in ablative
case.
(19) /[bayi yara]S

there.abs.m man.abs
[ba-lu-bawal buluba-gu] yanu/
there-all(place)-long.way corroboree.ground-all go.past
The man there went to a corroboree ground a long way off there
(20) /ŋaygu-naO yabundi-guA yuba-n/

1sg.acc mother-erg put.down-past
[bala-y Dabagiyam-ba]/
there-loc Top.Camp-loc
Mother gave birth to me (lit. put me down) at Top Camp [an 1890s
mining camp where Aborigines worked for the miners]
(21) /ŋaliS banaga-ñ ya-ŋum burbi-ŋunu/

1du.nom go.home-fut here-abl half.way-abl
We’ll go home from here, from half-way
182 R. M. W. Dixon
Verb markers are frequently used on their own, without any accompanying noun. We
have yala-y ‘at here’ in (12), ba-ŋum ‘from here’ in (13), and reduplicated bali-bali ‘in
that direction’ in (35). An example with ŋa-lu ‘to a place remembered from the past’ is:
(22) /baŋum bayiS/ bani-ñu /

and.then there.abs.m come-non.fut
ŋa-lu-guya bagaga-ñu bayiS /
non.vis-all(place)-other.side turn.off.path-past there.abs.m
Then he came; he turned off the path to a place remembered on the other side
5 Bound forms
As will have been noticed from the examples thus far, noun and verb markers may be
followed by bound forms indicating location:
• -gala ‘up’, -gali ‘down’—in (32) and (35)—and -galu ‘out in front’—in (27), (31a/b),
(33–34), and (37).
• -bawal ‘long way (in any direction)’—in (19) and (29).
• -guya ‘on the other side (of river, path, etc.)’—in (22).
• suffixes indicating long, medium, and short distances uphill and downhill, and
medium and long distances upstream and downstream. Those appearing in
examples here are -dayi ‘short distance uphill’—in (36)—-daya ‘medium dis-
tance uphill’—in (39a)—and -dawulu ‘long way upstream’—in (38)
• -guŋgarri ‘north’—in (17)—and -guyŋgurru ‘south’.
6 Interrogative noun and verb markers

Parallel to ba-, ya-, and ŋa- forms, there are interrogative noun markers, set out in (23),
and interrogative verb markers, in (24), both based on wuñja- and meaning ‘where’.
Once again, the absolutive M noun marker is irregular, being wuñjiñ, where *wuñja-l
would be expected. Note that the allative verb markers end in -rri and -rru, as do the
longer indefinite forms of the allative ba- markers, in (18).
(23) ergative
Noun absolutive (S (A function) and
class and O functions) instrumental dative genitive
M wuñjiñ wuñja-ŋgu-l wuñja-gu-l wuñja-ŋu-l
F wuñja-n wuñja-ŋgu-n wuñja-gu-n wuñja-ŋu-n
E wuñja-m wuñja-ŋgu-m wuñja-gu-m —
N wuñja wuñja-ŋgu wuñja-gu wuñja-ŋu
(24) allative of place, ‘to which place’ wuñja-rri

allative of direction, ‘in which direction’ wuñja-rru
ablative, ‘from where’ wuñja-ŋum
locative, ‘at where’ wuñja-y
Unlike ba-, ya-, and ŋa- forms, interrogative noun markers do not co-occur with a
noun, but make up a whole NP, as in (25), where Yimanu is a verbless clause subject
and wuñja-n is a verbless clause complement:
(25) /YimanuVCS wuñja-nVCC/

Dinah.abs where.abs-f
Where is Dinah?
Interrogative verb markers provide direct qualification of a verb, without any accom-
panying noun in the same case.
Interrogative noun markers can co-occur with ba- and especially with ŋa- markers.
If something cannot be seen, it is natural to enquire where it is. For example wuñja
ŋala ‘where is that (noun class N) noise?’.
7 Meanings of ŋa- markers

We now survey the range of meanings of non-visual noun and verb markers, com-
mencing with ŋa-. Noun markers occur in all senses, verb markers only with the
‘remembered’ sense, (e).
(a) Only known from its noise

A ŋa- noun marker may occur in an NP with a noun referring to a type of noise. In
(16) we had ŋala-n with mulgu ‘noise that sounds like talking but the words cannot be
made out either because it is too far off or because it is mumbled’. In (26) ŋayi is in an
NP with wawan ‘noise made by a person or animal walking over dry leaves or grass’.
(26) /gaji ŋayiO ŋamba

try non.vis.abs:m listen.imp
[wawan / yanu-ŋu]RC]O /
noise.of.walking.abs go-rc.abs
Listen to the tramping noise of some [M thing] who is going!
The third component of the O NP in (26) is the relative clause yanu-ŋu, ‘who is going’.
This involves a relative clause (RC) suffix -ŋu followed by the case appropriate to the
NP it is in—here, absolutive, with zero realization.
In (27) the non-visible noun marker is followed by an onomatopoeic representation
of the sound of running d-d-d-d-d-d:
184 R. M. W. Dixon
(27) /ŋayi-n-galu bayi/ d-d-d-d-d-d/

non.vis.abs.m.-linker-out.in.front there.abs.m onomatopoeia
He could be heard going out in front, d-d-d-d-d-d.
Example (28) comes from Chloe Grant’s story of a man in a fight. He holds up his shield
and buum represents the sounds of spears hitting it. Bigin ‘shield’ is in noun class F and
thus the F noun marker ŋala-n is used.
(28) /jumba-n ba-ŋgu-lA biginO/ buum/

hold.up-non.fut there-erg-m shield.abs onomatopoeia
ŋala-nS banda-ñu/
non.vis.abs-f impacted-non.fut
He held up his shield. Buum. It (the shield) was heard being hit.
In (26–28) something is known only from its noise. But people often want to know
what it is making a noise. When a bird (F noun class) is heard calling out in a particular
way, it indicates that some person or animal is approaching. Two sisters heard such a
bird call and one told the other to go and look to see what it was:
(29) /gaji bura ŋa-n-bawalO/

try see.imp non.vis.abs-f-long.way
Go on, look for (the sources of) the noise a long way off!
A similar example comes from a story told by George Watson, again involving two
sisters wishing to discover what a bird was singing about:
(30) a. /ba-ŋgu-nA giga-n [bala-n midi]O/ buŋa-li/

there-erg-f tell-past there.abs-f small.abs go.down-purp
She [the big sister] told the small one to go downhill:
b. gaji ŋindaS buŋa/
try 2sg.nom go.down.imp
[ŋa-gu-l miña-gu/ [ñiyi-ŋguA/ baya-ŋu-gu]RC]/
non.vis-dat-m what-dat bird.noise-erg sing-rc-dat
‘Go on, you go downhill, to [see] what (M thing) it is the birds are singing
about!’
In (30b) the long dative NP describes what the younger sister is going downhill for. Its
head is miña ‘what’ (which inflects like a noun), accompanied by a non-visual noun
marker ŋa-gu-l and a relative clause consisting of transitive verb baya- ‘sing about’ and
its A NP which is noun ñiyi ‘noise bird makes when it is drawing attention to some-
thing moving along’. Dative case suffix -gu follows relative clause marker -ŋu on baya-.
(b) Previously visible but now just audible

A curlew (noun class F) was seen flying, with ‘there’ noun marker ba-n. Then it moved
out of vision—non-visible noun marker ŋa-n—but was known to be still flying since
its distinctive call continued to be heard:
(31) a. /ba-n-galuO/ bura-n ŋajaA

there.abs-f-out.in.front see-non.fut 1sg.nom
[guyibarra [yambi-ŋu]RC]O/
curlew.abs fly-rc.abs
I saw a bush stone curlew [Burhinus magnirostris] out in front, which was
flying
b. giyu/ giyu/ giyu/ ŋa-n-galu ba-n
onomatopoeia non.vis.abs-f-out.in.front there.abs-f
yambi-n/
fly-non.fut
‘Giyu, giyu, giyu,’ she could be heard (calling as she) flew out in front but
couldn’t be seen
In a story told by Tommy Springcart, a man sees a fruit hanging from a tree and knocks
it down with a thrown stick. In the first clause of (32), the O NP for verb minba- ‘hit
with a long rigid implement which is thrown’ is ‘there’ noun marker bala-m since he
can see the fruit on the tree. In the second clause, the non-visible noun marker ŋala-m
is used as the S NP for baji- ‘fall’ since the man’s vision is obscured and he cannot actu-
ally see the fruit falling.
(32) /minba-n bala-mO wirrga-ŋgu/

hit-non.fut there.abs-e stick-inst
ŋala-m-galiS baji-gu/
non.vis.abs-e-down fall-purp
(He) threw a wirrga stick (a nulla nulla) at the fruit, and as a conse-
quence it is heard falling down
(c) Neither visible nor audible

In a story concerning the early years after the European invasion, one Dyirbal man
escaped a white tyrant by crawling beneath a tree, where he was no longer visible (nor,
of course, audible):
186 R. M. W. Dixon
(33) /yanu bayiS guyñan-da-rru

go.non.fut there.abs.m Moreton.Bay.tree-loc-through
juyma-n ŋira-ŋga-rru
crawl-non.fut underneath-loc-through
ŋayi-n-galu jañja/
non.vis.abs.m-linker-out.in.front now
And he went crawling underneath a Moreton Bay hardwood tree [Eucalyptus
tessellaris] and now he was no longer visible
In (33) there are two verbs making up a discontinuous serial verb construction, yanu
juyma-n ‘went crawled’. The discontinuous locative NP consists of the noun guyñan
‘Moreton Bay tree’ and the adjective ŋira ‘underneath’, both with a locative case suffix
followed by -rru ‘along, through’.
The text collection includes several descriptions of sunset, when the sun itself is no
longer visible but its light remains. Sentence (34) consists of just an NP, with its three
words in absolutive form, plus the time word jañja ‘now’.
(34) /ŋa-n-galu jañja ba-n garri /

non.vis.abs-f-out.in.front now there:abs-f sun.abs
The sun is out in front now and not visible
There is one textual description of a sunrise, also including a non-visible noun marker.
The sun has not yet appeared and so, interestingly, we get the N noun class form ŋala.
It takes the suffix -gali ‘down’, indicating that the dawn light is just above the horizon.
(35) /ŋala-gali bali-bali ganda-ñu/

non.vis.abs.n-down redup-there.all(direction) burn-non.fut
Something non-visible is burning down in that direction
(d) Spirits are described as non-visible

Throughout Australia, there is a spirit which can take the form of a rainbow or a snake.
In one tale told by Chloe Grant, two women climb a tree to capture starlings and
their climbing vine falls down. They call out for someone to help them descend. The
rainbow appears and suggests that they should climb down him (they decline). They
can see the rainbow, but since he is a spirit the non-visible noun marker is employed:
(36) /[ŋayi-n-dayi bayi]S

non.vis.abs.m-linker-short.way.uphill there.abs.m
bandu.bana-n/
bend.over-non.fut
He [the rainbow], a little way uphill, bent over up there (suggesting that they
climb down him)
One day Bessie Jerry explained how Dambun spirits can take on human form. You
might see what appear to be two women and start following them, and then you won-
der who you are following, that they are probably spirits.
(37) /bambay ŋindaA banja-ñ

unknowlingly 2sgA follow-non.fut
[ŋa-n-galu wañuña]O/
non.vis.abs-f-out.in.front who.acc
You might follow them unknowingly and wonder who you’re following
(e) Remembered
Example (17) used the non-visible dative E noun marker ŋa-gu-m to describe a man
taking his grandsons to gather finger cherries, whose location he remembers. In (22)
the non-visible allative of place verb marker ŋa-lu describes travelling towards a
remembered place.
In a story told by Ida Henry, two girls have killed a black goanna and dug an earth
oven to cook it. They go to a place where they remember seeing the right kind of leaves
to cover the earth oven:
(38) /yanu [ba-n nayi-jarran]S/

go-non.fut there.abs-f girl-pair.abs
[gubu-gu ŋa-gu-dawulu]/
leaf-dat non.vis-dat.n-long.way.upstream
The two girls went for leaves in a place remembered from the past a long way
upstream
In another of Ida Henry’s narratives, an old man realizes that his life is drawing to a
close, and plans to travel to a place he remembers, his father’s father’s conception site,
where he wishes to die:
(39) a. /[ŋa-gu-l-daya bulunja-gu]

non.vis-dat-m-mid.distance.uphill father’s.father-dat
añja ŋajaS banagay-gu /
new.action 1sg.nom return-purp
And I want to go back to my father’s father remembered place a medium
distance uphill
b. añja ŋajaS bala-y guyi-bi-li/ jañja/
new.action 1sg.nom there-loc spirit-become-purp now
I want to die (lit. become a spirit) there now
188 R. M. W. Dixon
8 Conclusion
Dyirbal assigns importance to distinguishing between things which can be seen and
are real, and everything else. And, when something is visible and real, whether it is
near the speaker—the noun and verb markers commence with ya- ‘here’—or not near
the speaker, or where location is irrelevant—the noun and verb markers commence
with ba- ‘there’. Noun markers commencing with ŋa- are used of spirits, which may
appear to be visible but are not real. And for anything within the context of speaking
which cannot be seen—it may be audible or neither visible nor audible. In addition,
a person or place or direction remembered from the past is referred to by a ŋa- noun
or verb marker.
All this is consistent with the requirement for precision, mentioned at the very
beginning of this chapter. If possible, one should always employ a specific term, rather
than a vague or general one. However, it should not be inferred from this that the Dyir-
bal speech community lacks, or does not use, generic terms (that they are incapable of
conceptualizing in general terms).
When, in traditional times, one was speaking in the presence of a classificatory
mother-in-law, son-in-law, father-in-law, or daughter-in-law, it was required to
employ a special speech register called Jalnguy (dubbed ‘mother-in-law language’ by
Chloe Grant). Jalnguy has the same phonology, phonetics, morphology, and syntax as
the everyday language style, but every single lexeme (except for the four grandparent
terms) is different.
Whereas the everyday style has a profusion of specific terms, Jalnguy operates
entirely with generic terms. The everyday register has names for a score of species of
dangerous snakes, but Jalnguy simply has the general term bayi jumbiñ ‘venomous
snake’. Similarly for other nouns, for adjectives, and for verbs. Jalnguy is an avoidance
style used to mark a relationship of taboo. While a high value is placed on precision in
normal speech, it is considered appropriate to be deliberately vague in an avoidance
situation.
There you have it—different levels of specificity and generality, each in its proper
place.
Appendix
Just the main points of Dyirbal grammar have been outlined here—those necessary
for understanding the discussion of noun and verb markers. A full account is in Dixon
(1972), including details of dialect differences. See especially pages 38–58, 222, 254–64,
306–11. The S/O pivot is also described and exemplified in Dixon (1994: 9–18, 160–72).
All examples in this chapter are from texts; some have been lightly edited, omitting
repetitions and the like.
Examples (5), (11–12), (19–20), (25), and (30) are from stories told by George Watson
in Mamu, a northern dialect. The remaining examples are from recordings in Dyirbal,
a southern dialect, by Chloe Grant, Ida Henry, Bessie Jerry, Daisy Denham, Andy
Denham, Tommy Springcart, and Tommy Warren. I thank these friends and teach-
ers, now all returned to the land of spirits, for teaching me their wonderful language.
Non-visible markers recur in the poetic style employed in Dyirbal songs. The 174
songs documented in Dixon and Koch (1996) includes six ŋa- noun markers—in
songs Gama-E (pp. 81–2), Gama-Y (p. 100), Gama-AY (pp. 124–5), Marrga-K (pp.
201–2), Jangala X (pp. 255–6), and Jangala-Z (pp. 258–9).
References
Dixon, R. M. W. 1972. The Dyirbal language of North Queensland. Cambridge: Cambridge Uni-
versity Press.
———. 1994. Ergativity. Cambridge: Cambridge University Press.
——— and Koch, Grace. 1996. Dyirbal song poetry: The oral literature of an Australian rainforest
people. St Lucia: University of Queensland Press.
9
The grammar of knowledge in Maaka

(Western Chadic, Nigeria)
A N N E S T O RC H A N D J U L E S J AC Q U E S C O LY
Knowledges or knowledge traditions should be imagined in the plural. . . . Knowl-

edges may be divided into explicit and implicit (or tacit), pure and applied, local
and universal. Although histories of skills are rarely written, ‘Knowing how’ clearly
deserves a place alongside ‘knowing that’. In similar fashion, dominated or subju-
gated knowledges (savoirs asujettis) deserve a place alongside rather than underneath
dominant ones. There is a political aspect to the question, ‘what is knowledge?’ Who
has the authority to decide what is knowledge?
—Burke (2012: 5)
1 Introduction
In Maaka, a language of north-eastern Nigeria, the notions of knowledge and truth
can be expressed in various ways, depending on the context. Indeed, the very concept
of knowledge is rather complex in Maaka, and requires that we appropriately define
and describe several principles first of all. For example, know could be framed as voli-
tional and agentive, and may have semantic extensions into the domain of control and
possession. Other possibilities include the conceptualization of know as being corre-
lated to a perceptual process, which needs to be further specified in terms of how and
under which circumstances knowledge was achieved and information gathered. This
specification is articulated by means of evidential markers and epistemic and modal
verbs, which help to estimate the reliability of the reported event, of the informant, or
of one’s own cognitive potential in terms of grasping inherent contextual information.
There are various types of evidential and epistemic markers in Maaka, which either
relate to a speaker’s knowledge and general attitude towards the truth of a proposition
and refer to the source of information—highlighting eye-witness, intuition, and so
9 The grammar of knowledge in Maaka 191
on—or express the speaker’s certainty or doubt about the reliability of the information,
regardless of the information source. Other epistemic markers encode modality and
type of information, such as inferential predictability of an event or action, indirect
evidence for a completed action, or joint perception.
Those constructions that refer to perception and the source of information stand in
a kind of binary opposition to those that highlight certainty or doubt about the reli-
ability of information, and that express cognition rather than perception. This oppo-
sition is also found in the predicative expression of perception and cognition, where
a specialized verb expressing know can be distinguished in terms of its evidential
semantics from a set of verbs encoding various modalities of perception.
Table 1 illustrates how Maaka articulates the grammar of knowledge.
Maaka stands out somewhat among Chadic languages in having such a complex
and developed repertoire of evidential markers and epistemic modalities. These have
not been described for other languages in the group, with the exception of the rather
common doubt-in-truth modality (Frajzyngier 1996: 180 ff.). However, explaining the
emergence of the varied system in Maaka remains problematic, as its formatives do
not derive from a common source, but have developed out of a variety of grammatical
morphemes and lexical items. Still, there is no doubt that the indication of information
source is particularly important in Maaka and that, at the same time, speakers tend to
highlight certainty of truth and type of knowledge in other contexts.
The present contribution presents a detailed analysis of the semantics and con-
struction types of the individual epistemic expressions in Maaka, and provides con-
textual information about the social history and cultural context of the grammar of
knowledge in this language. All data stem from in-depth fieldwork and extensive text
corpora that have been collected in the frame of a four-year interdisciplinary docu-
mentation project by the present authors and their team members.1
1.1 Classification, location, and sociolinguistic context

Maaka is a West Chadic (Afroasiatic) language of the Bole-Tangale subgroup. The
Maaka are a community of some 10,000 people living in northern Nigeria not far
from the town of Gombe (Map 1). They inhabit two main villages, Bara and Gulani,
and several hamlets around these. In principle, their ethnolinguistic identity is based
on their shared Maaka language. However, many speakers are not very competent in
Maaka and rather speak other Chadic languages such as Bole, Bura, Hausa, or Kanuri,
an unrelated Saharan language. Hence, Maaka is not the specific language of an eth-
nic group, but the socio-culturally marked part of a highly multilingual community’s
1
We are grateful to Jibril Jatau Bara, Musa A. Baba, and the people of Bara for sharing their knowledge
with us, and to Herrmann Jungraithmayr, Rudolf Leger, Johannes Harnischfeger, Alessandro Suzzi-Valli
for their many insightful comments, as well as to two anonymous peer-reviewers. We owe heartfelt thanks
to Sascha Aikhenvald and Bob Dixon for their inspiring comments and interest in our work. Research on
Maaka has been generously funded by the German Research Foundation.
192
Anne Storch and Jules Jacques Coly
Table 1. Maaka evidential and epistemic markers
Source of information Certainty in truth Control over knowledge Scope
Noun suffixes -mú vision speaker certain none topicalized NP

-dìyà joint perception speaker/hearer certain none topicalized NP
-kà assumption speaker certain none topicalized NP
Reported speech marker nà heard speaker certain by speaker clause
Complementizers kònò reported doubt-in-truth none clause or NP

kóŋ witnessed speaker certain exclusively by speaker sentence
Adverbial yàayé inferred doubt-in-truth none clause
Verb suffixes -ntí inferred doubt-in-truth by speaker VP

-râ intuition speaker certain exclusively by speaker sentence
to be right
Verbs kìn ‘be unable, possess speaker/hearer certain by speaker or by third person clause
mystical knowledge’
nòn ‘know’ speaker/hearer certain by community sentence
repertoire: using Kopytoff ’s (1987) model, we can characterize Maaka as the official
language of a hybrid and fragile frontier society, whose various segmental groups,
such as kin-groups for example, also speak several other languages. This particular
pattern of multilingualism is widespread in the area. The two main dialects, Maaka and
Maha, serve as markers of local identity and speakers’ association with the respective
hegemonies of Bara and Gulani. In our chapter, we will focus on the language of Bara.
Most families in the village of Bara are mixed, with family members stemming
from various ethnolinguistic backgrounds. Very often, people claim different ethnic
and linguistic identities in different contexts, so that ‘being Maaka’ is a fluid and con-
stantly negotiated concept. This situation is characteristic of societies who dwell at the
margins of large empires, where such multi-ethnic and multi-lingual groups emerge
(Kopytoff 1987). They represent part of the ‘African Frontier’, whereby they distinguish
themselves from adjacent polities, while at the same time copying their nation build-
ing strategies, for example the enthronement of sacred kings. Groups like the Maaka,
who constituted themselves as marginal groups at the fringes of the large empires
(namely the Sultanate of Borno, the Kingdom of Kororofa, and the Hausa Emirates),
are precarious ethnolinguistic communities whose members have elusive and diverse
Yobe
Kanuri
Gongo
la
Bole Maaka
Bara Borno
Gulani
Kupto
Maha
Biu
Dadin
Kowa Bura
Reservoir Tera
Kwami
Bauchi
NIGER CHAD
Gombe
i
Pa
Gombe
NIGERIA
Waja
Hone Tangale
CAMEROON
Map 1. The Maaka area

194 Anne Storch and Jules Jacques Coly
cultural and linguistic identities. This extremely multilingual and fragile group is uni-
fied and made cohesive by a shared but imaginary political unity. This is important
to consider in the context of our topic, since despite the pronounced multilingual
practices of the Maaka, there are specific epistemic features that are unique to Maaka
that are not shared by its neighbours and cohabitants.
1.2 Typological profile

Maaka is a tone language with two tone levels (high, low) and two contour tones (falling,
rising), in which tone has a salient lexical role, but also fulfils grammatical functions.
Maaka can be characterized as a weakly agglutinating language, with a tendency to
compensate the absence of Chadic morphology by innovative compounding strategies
(Coly forthcoming). Very much like other languages of the Gongola-Benue linguistic
area, Maaka makes use of syntactic structures in order to make grammatical relations
discernible, as salient morphological devices such as case marking are absent. The
constituent order AVO/SV consequently is rather inflexible.
In the noun morphology, the historical development of morphological simplifica-
tion has resulted in gender levelling in favour of the feminine (Leger 1998). Besides
such phenomena of areal convergence, there are various contact-induced innovations.
For example, the strong influence of Kanuri can be seen in the use of derivational
suffixes such as -kùr (< Kanuri kәr-; note that there is no convergence in terms of the
morpheme order), as illustrated in the following example:
(1) Maaka Kanuri

máy-kùr ‘chieftaincy’ kər-mâi ‘kingship’
chief-der der-king
The verbal morphology in turn exhibits much more conservative and typical Chadic
features (Jungraithmayr 1970, Schuh 2003), such as the nominalizer -áayò, which con-
structs the citation forms of transitive verbs (cf. 2). Derivational morphology, such as
the construction of pluractional verbs is quite productive as well: either by infixing -k-,
or by partial stem reduplication:
(2) tèɗ-áayò, pl tè-k-ɗ-áayò ‘beat s.th.’

lòɓ-áayò, pl lò-lòɓ-áayò ‘beat with the help of s.th.’
Verbs may be either transitive or intransitive, whereby intransitive constructions may

be marked by an intransitive copy pronoun (icp) in specific contexts. This pronoun
then stands after the verb and has the same features for number, gender, and person
as the subject-participant of the clause (as in example (35) below). Icp constructions
are widely found in Chadic, but also in Benue-Congo contact languages. In Chadic,
they seem to derive from middle voice markers (Leger and Zoch 2011), which may also
play a role in Maaka. In some verbs that have cognitive meanings, for example ‘forget’,
the use of the icp and the indication of middle semantics reveal that some domains
of the grammar of knowledge have to do with middle, auto-benefactive semantics.
This again seems to fit into the (areal) Chadic picture, where a similar situation has
been observed in Mupun (Bole-Angas branch of Chadic) by Frajzyngier (1993), who
mentions that the deponent verbs ‘forget about something’ and ‘ignore something’
and obligatorily require a suffixed pronoun that encodes the fact that the subject-
participant also has the role of a patient or experiencer.
The affectedness of the subject in cognitive experiences and perceptual actions gen-
erally plays a role in the grammar of knowledge in Maaka, which is exemplified in the
next sections.
2 Evidentiality, information source, and truth

As already indicated above, Maaka generally converges to neighbouring languages in
terms of its morphological and syntactic structure. But in contrast to this tendency, it
clearly exhibits considerable divergence in the domain of cognition and perception.
In Maaka, and specifically not in other Chadic languages of the area, it is important to
indicate the speaker’s own attitude about the certainty and probability of a proposition,
and this is achieved by a stunning variety of means. The evaluation of truth usually
entails an indication of the information source, and there is a tendency among speak-
ers to correlate certainty with vision. This mode of perception indeed appears to be
the most authoritative information source in Maaka, which also exhibits the means for
describing a gradation of vision. Other evidential markers refer to non-visual sensory
information sources, such as the speaker’s intuition, and to inferred information.
What makes the definition of the choice types within the system rather problem-
atic is that evidential markers occur with different parts of speech. The means used
for encoding information source and certainty of knowledge are nominal suffixes,
verbal suffixes, verbs, complementizers, adverbial expressions, and reported speech
markers (besides dedicated perception verbs and other means to which we will refer
in more detail below). This implies that evidential markers can have various scopes,
such as speaker, patient, action or place, and that they can frame both knowledge and
perception.
2.1 Evidential markers coding information source

Maaka has a complex system of demonstratives, of which one modified sub-set has
developed salient evidential meanings and has largely turned into a group of bound
markers indicating information source. The perceptual meanings of these nominal
suffixes basically cover the range of an evidential system with three choices (Aikhen-
vald 2004: 42 ff., 130 f.), whereby the use of the suffixes referring to information source
does not correlate with the syntactic function of the NP. Moreover, this system is
restricted to specific semantic roles and to topical NPs: these evidential markers spec-
ify what the speaker knows about a particular participant, whereby such participants
are hardly core participants, but rather topicalized peripheral participants that moti-
vate an action or event.
2.1.1 Vision The suffix -mú occurs in indicative clauses where it marks topicalized
participants. It implies certainty on the part of the speaker who is absolutely sure
about the inclusion of the marked participant in the event, having eye-witnessed the
occasion.
The suffix may be combined with a definite marker, as in (3), which refers to nego-
tiations in which the speaker participated. In (4), the speaker has eye-witnessed the
life-story of the topicalized participant, and in (5) the speaker has already seen the
butterfly to which the addressee is directed.
(3) tô báayà-à-mú mìnè-ʔákkó ɓà máy=ʔàŋgùwà

top <name>-def-vis 1pl-do:perv conj chief=<title>
As for Baaye [eye-witnessed], we dealt with Mai Anguwa.
(4) yáayà círòmà-mú nín-nì gùu=ɓálɓìyá

<name> <title>-vis mother-poss:3sg:masc person=<toponym>
tà-lòwó gàamôɗí bòɲcéttí

3sg:fem-deliver:perv once dem.ref
Yaaya Ciroma [eye-witnessed]: her mother is from Balbiya town, she once
gave birth there.
(5) móy p�rp�r ním mákkà-mú

see.imp butterfly rel Mecca-vis
See the butterfly from Mecca!
2.1.2 Joint perception Reliable knowledge and truth can also be expressed by means
of the suffix -dìyà which indicates that both speaker and hearer know or see the
participant in question. Very similar to -mú, the suffix marks patients, but these are
conceptualized as being even more reliably known because there are two witnesses
to the event, namely speaker and hearer, as in (6, 7).
(6) làa nàmáa-dìyà sáy mìnè-póɗí-ní gè-gòrkù-wà

child this-joint:vis must 1pl-remove:tel2-obj:3sg:masc loc-village-def
This child [whom we can both see], we must chase him from the village.
2
The derivational morpheme -(d)í constructs telic verbs in the imperfective aspect. This form also has
directional semantics, compare the ‘Destinativ-Ventiv’ in Bole-Tangale, e.g. Kwami (Leger 1994: 212 ff.).
(7) ʔáa-kè-díɓɓò zùlúm-tò-dìyà

cond-2sg:masc-crush:perv anus-poss:3sg:fem-joint:vis
tà-kwáadà-ntí-mìnê gè-ʔámmà-à
3sg:fem-throw:tr-assert-obj:1pl loc-water-def
If you crush her anus [that we can both see] she will definitely throw us into
the water.
Even though it has been integrated in a new context where it has developed slightly
different semantics, -dìyà can be identified as a loan from Kanuri, stemming from the
postpositional clausal conjunction dìyé ‘surely, entirely, only’ (Schuh 2011: 137 f.). See
§4.2 for more on Kanuri loans.
2.1.3 Assumption In contrast to the two former suffixes, the evidential -kà refers
to the speaker’s intuition about an event and the location where it takes place. Even
though the speaker may not have any firsthand information, s/he is sure that there
is a specific place where, or a causator through whom, the action takes place. In
(8), the speaker utters a strong supposition about the addressee’s intended action,
and in (9), it is considered that a natural phenomenon will certainly affect the core
participants at a given point.
(8) ʔáa-kè-góm gè-gòmà-à-kà ( . . . )

cond-2sg:masc-go loc-market-def-assum
If you go to that very market [I know you go there] ( . . . )
(9) ʔáa ɓà làmbà-kà mìnè-móy tàríyá-á-ɓá

cond conj cloud-assum 1pl-see moon-def-neg
When there are those very clouds [I know they will be there], we cannot see
the moon.
2.2 Coding the certainty of truth

This section deals with various markers, including a quotative that carries overtones
of certainty, a reported marker, and two epistemic markers.
2.2.1 Speech reports In many Chadic languages, including Hausa, Mupun, and
other members of the Western branch, embedded clauses may have epistemic
modalities. The most common one expresses the speaker’s attitude towards the
reliability of information given in the proposition. Frajzyngier (1996: 180 ff.) has
explored this in some detail, remarking on the phenomenon:
The markers of doubt in the truth of the proposition have sometimes been included in the literature in
the class of ‘evidentials’. . . . Unlike most of the evidentials, however, doubt-in-truth markers in Chadic
do not encode the source of knowledge or the mode of knowing. They simply indicate lack of certainty
on the part of the speaker. The doubt-in-truth markers in Chadic languages do not constitute a homo-
geneous grammatical category, nor can they be traced to the same historical source. Instead they
appear to be innovations in particular languages or language subgroups. (Frajzyngier 1996: 180)
In principle, this also holds true for Maaka. Like other Chadic languages, Maaka uses
a grammaticalized speech act verb to introduce reported speech, which also intro-
duces indirect evidence. But unlike the examples provided by Frajzyngier (1996), in
Maaka the reported speech marker nà expresses principled certainty on the part of the
speaker. Consider the following examples:
(10) ʔálí nà sù-wókkó músá kò máytà
<name> quot 3sg:masc-see:perv <name> do vomit:intr.vn
Ali said he saw Musa vomiting.
(11) sù-pòró nà sú kò tíiná

3sg:masc-say:perv quot 3sg:masc do eat:intr:vn
Hess said that hess is eating.
Doubt in truth can be expressed through the complementizer kònò, which indicates
that the reported information relates to a rumour, but not to an event that had been
reliably witnessed. Hence, the sentence in (10) can be reframed as a doubt-in-truth
construction as follows:
(12) ʔálí nà kònò sù-wókkó músá kò máytà
<name> quot rep 3sg:masc-see:perv <name> do vomit:intr:vn
There is a rumour that Ali said he saw Musa vomiting.
The scope of kònò is reflected in its position in a clause: if it is placed in clause-initial position
(but following nà), its scope is the whole clause (or sentence). But when it is placed after the
NP, its scope is limited to that NP. Here, the complementizer can also be used in contexts
other than reported speech. It stands after the respective phrase about which the speaker
expresses uncertainty in terms of correct information or a reliable witness. Hence, kònò may
indicate unsure information concerning the agent or object, as in examples (13) and (14):
(13) s�-ndée ɓà líimó-wà kònò ɓà láà-n-tò
3sg:masc-come:narr conj camel-def rep conj child-link-poss:3sg:fem
He came with the reported camel and with its calf.
(14) shèní ndókò-n-nì-yà s�-ŋgèró jàmmb�làm

year arrival-link-poss:3sg:masc-def 3sg:masc-tie:perv bundle
dùɓú kònò
thousand rep
In the year of his very arrival he tied a reported thousand millet bundles.
Standing at the beginning of a clause, kònò indicates doubt about the truth of the entire
utterance, as in (15):
(15) kònò dóoshé mínéé-gòm ʔáshàakà
rep tomorrow 1pl:fut-go <name>
Rumour has it that tomorrow we will go to the Ashaka cement factory.
Moreover, kònò can be used in questions in combination with the anti-potential evi-
dential marker râ. Here, even without an interrogative morpheme, both rhetorical
interrogation and surprise are expressed. In this context, the speaker has more infor-
mation about the reported, repeated event which is framed as being unlikely to happen
again.
(16) kònò-râ-zì
rep-evid.npot-dur
Again? (It’s impossible!)
2.2.2 Epistemic markers A feeling of uncertainty, or weak doubt, is expressed by

means of the adverbial expression yàayé. It refers to information one hesitates to
believe in confidently, and to the speaker’s doubts in the validity of the utterance. In
most cases, this doubt-in-truth marker occurs in clauses with ʔálá ‘maybe’.
(17) ʔà kùtà-gókkò kérwá-à ʔálá pàɗɗú yàayé
then 3pl-find:perv fish-def maybe four dub
Then they found the fish, maybe four (of them), but I doubt this is correct.
(18) ʔálá sù-mòɗó yàayé

maybe 3sg:masc-return:perv dub
Maybe he has returned, but I have my doubts.
There is also a marker that expresses the opposite of doubt in truth, namely the speak-
er’s explicit certainty about the truth of the proposition. This is the suffix -ntí, which
constructs inferential verb forms. It indicates that the speaker is absolutely sure about
his/her assumptions, whereby the source of the information is irrelevant—similar to
the egophoric markers in Kurtöp (Hyslop, this volume). In most of the examples,
-ntí constructions refer to logical reason and occur in consecutive constructions that
denote a series of actions, for example (7) above and examples (19) and (20):
(19) ʔìnndá mmù ʔà mìnè-ndéré ɓáyà mòo-yá-dìyà
stand:imp 1du then 1pl-run:narr otherwise people-def-joint:vis
ʔà dùkà-ntí-mìnê
then kill:tr-assert-obj:1pl
Stand up! We both then run, otherwise the people we both see/know will
definitely kill us.
(20) ʔà tá nà tày nàmá ʔíɲɲcé lée mày ɓà páadàa

then 3sg:fem quot food det.dem today even chief conj cabinet
wáɲcè-n-nì tíinà-ntì-râ
dem:pl-link-poss:3sg:masc eat:tr-assert-evid.npot
Then she said that as for this food today, even the chief and his advisors will
definitely not eat (it).
Such forms may also have overtones of control over information, as the inferential
verb form highlights the speaker’s own knowledge of the context of the event. As we
shall see in the following section, evidentials in Maaka may explicitly refer to control
over knowledge and ownership of information.
2.3 Coding knowledge and control

The mode of perception is a peripheral meaning of evidential markers that specifically
express one’s ownership and control of knowledge. Such usages of evidential markers
indicate that the speaker has precious information, which he/she shares either with
the addressee or with the person referred to in the utterance. Whether this informa-
tion has been obtained through seeing, hearing, or hearsay, is not directly indicated.
Such egophoric (or ‘ego-modal’) concepts of controlled information are indicated by
means of anti-potential verb forms which are constructed with the suffix -râ. These
forms express that an action cannot be performed or that an event has very little prob-
ability of happening, whereby the speaker possesses more, as yet undisclosed context
information (e.g. in (20), where the narrator knows more than us about the chief ’s
food). Such meanings are related to those of mirativity pronouns in neighbouring
Jukun languages (Storch 2011a), which express that an agent miraculously performs
an action in spite of an obstacle.
(21) sì-gìɬlà rìinà-râ
3sg:masc-be_able:imperv enter:intr:vn-evid.npot
He will never ever be able to enter.
The complementizer kóŋ encodes a similar egophoric concept, of the speaker’s con-
trol over information, albeit with a different view on the event. This marker expresses
the completed transformation of a situation or a referent, and the speaker’s exclusive
knowledge of this fact. The speaker implies that even though every participant in the
speech act knows something about the discussed event or referent, only the speaker
knows that their information is now outdated. In examples (22), (23), and (24), the
speaker’s view of change and transition as well as the hearer’s unprepared mind are
referred to:
(22) káanàm s�-ʔáwákkó bòɲcé bòɲcé béggè nàmàa-mú ɓà

like_this 3sg:masc-open:compl:perv dem dem room dem.det-vis conj
kéekée kò-s-sú ŋgàánù kóŋ

bicycle head-link-poss:3pl seven evid.mod
As he opened this very room, here on this visible spot there were seven individual
bicycles [and this is no longer the case].
(23) mò nàmá kóŋ kútá gè-ʔágrik kálá ʔálá mà ʔúl-sú

people dem evid.mod 3pl loc-agric but maybe people take-obj:3pl
táshà
station/car.park
These people were in the Agricultural Department [and this is no longer the
case], maybe people took them to the car-park (‘fired them’).
(24) mày-yá pór-ná-t-tò nà kóŋ zùtí-nì

chief-def say-obj:1sg-link-obj:3sg:fem quot evid.mod heart-poss:3sg:masc
dàmmbìkínà ɓà ʔínà bòktì kóŋ nì-bòo-búrì

spoil:partic conj obj:1sg because evid.mod 1sg-hab-lie:intr
The chief told me it [the matter] that he was very angry with me [no longer the
case] because I used to lie [no longer the case].
Maaka has another evidential that encodes exclusiveness of knowledge possessed by

the speaker, but in this case, there is no reference to the speaker advancing informa-
tion, but rather to a particular skill or capability. This ability singles the speaker out
among other participants in the speech act, whose inability to cope is emphasized. This
is achieved by the middle verb kìn ‘be unable’ as in examples (25a) and (25b):
(25) a. kéè-kínèy ʔàwày

2sg:masc:fut-be_unable:imperv:tr open:tr:vn
You will be unable to open it [because I’ve set a password].
(25) b. kè-kìnàkkó ʔàwày

2sg:masc-be_unable:compl:perv open:tr:vn
You were unable to open it [I saw that you tried, but I’ve set a password].
When this verb is used as a participle it can expresse a person’s control over mystical
and taboo knowledge. This is not, however, the case when the S-participant is a child or
a person still in the process of maturing and learning, as in examples (26a) and (26b):
(26) a. adult S/controlled knowledge

làatù-wà kìnèykínà
person-def be_unable:tr:partic
The person has mystical knowledge.
b. puerile S/no controlled knowledge

làa-yà kìnèykínà
child-def be_unable:tr:partic
The child is stubborn.
Example (27) refers to a blacksmith’s knowledge (which has to do with the transfor-
mation of substances and concepts, such as turning ore into iron, burying the dead,
curing the sick, and so on; cf. Van Beek 1992; Storch 2011b). It demonstrates that inspi-
ration and ownership of knowledge are conceptualized according to cultural practice
and taboo.
(27) s�-kìnèykínà pátí gìsò
3sg:masc-be_unable:tr:partic toward blacksmithing
He is talented in blacksmithing.
In all of examples (26) and (27), the verb is constructed as a participle. Such a form,
with a resultative meaning that could develop non-firsthand overtones, is found as
a secondary evidential in various languages cross-linguistically (Aikhenvald 2004:
112 ff.). Here, this aspect of the development of evidentials may be marginal. What
is really interesting is that the verb now expresses ‘be talented, gifted, enlightened’
instead of incompetence. And this particular ability— talent and enlightenment—is
framed as the type of knowledge one cannot acquire through simple learning and imi-
tating, but that is rather an innate part of one’s character. This knowledge is an inalien-
able possession, either inborn or achieved through spiritual experience. In Maaka, a
mystical, illuminated person is therefore called bòbóŋŋgrá ‘reversed/metamorphosed
person’, that is a person within whom inability and ignorance turn into wisdom.
3 Transmission of knowledge
The achievement of wisdom and knowledge has changed considerably through the
implementation of Islamic and Western education, which only happened a few gen-
erations ago. Since social change has been rather abrupt and rigid, most Maaka either
no longer know much about how mystical knowledge and practical skills were for-
merly taught, or as modern citizens and Muslims, they do not want to disclose infor-
mation that they now associate with paganism. The verbs of teaching and learning,
however, reveal that there exist different concepts of the transmission and acquisition
of knowledge.
Leger (1993) observes that in Bole-Tangale languages, the verb ‘learn’ is polysemic,
also encoding ‘teach, fear, try, imitate’. (For this reason, a more adequate translation
might be ‘participate in transmission of knowledge’.) Moreover, there is no uniform
root for this verb in these otherwise close languages, and none of the roots Leger gives
for Kwami, Kupto, Widala, and Piya are cognate with the Maaka root. In all of these
languages, ‘learn’ is a middle verb that in principle frames the agent as patient. However,
in Maaka there are slight differences. This language has two verbs for ‘transmit knowl-
edge’, whereby only one—bùd—has full semantic resemblence with the forms Leger
describes. The verb bùd frames ‘learn’ as an agentive, active, explicitly volitional prac-
tice that could involve imitation and exposition, but not secret actions. In a clause with
an animate transitive object, bùd translates as ‘teach’. Compare the following examples:
(28) a. ní kó bùdàadí
1sg do learn:tel:intr:vn
I’m learning.
b. ní kó bùdàadà rúbù
1sg do learn:tel:tr:vn writing
I’m learning to write.
(29) a. ní kó bùdàadà-n-nì
1sg do teach:tel:tr:vn-link-obj:3sg:masc
I’m teaching him.
b. ní kó bùdàadà-n-nì krà
1sg do teach:tel:tr:vn-link-obj:3sg:masc reading
I’m teaching him to read.
The second verb is kér and refers to the transmission of a different type of knowl-
edge, namely special skills, taboo knowledge, and mystical insights. Here, knowledge
is tantamount to an inalienably possessed capacity. Teaching is directed and planned,
involving work and immersion, and may disclose secrets. Outside the contexts of ritual
learning, in the sense of wrong register, kér translates as ‘guide; pretend, challenge to
teach’. Compare examples (30), (31), and (32):
(30) sú kó kér gìsò
3sg do teach blacksmithing
He is teaching blacksmithing.
(31) ké kérà-n-nó bòk=máakà?

2sg:masc teach:.tr-link-obj:1sg mouth=Maaka
Do you pretend to teach me Maaka?
(32) ní kérà-n-kó bòk=màakà!

1sg teach:tr-link-obj:2sg:masc mouth=Maaka
I challenge you, I can teach you Maaka!
As we have seen, information source is explicitly referred to in only a few evidentials.

At the same time, different types of knowledge and different ways of transferring that
knowledge may also relate to different modes of perception playing a role in the respec-
tive cognitive processes. The next section will therefore present an analysis of the various
predicative devices for expressing perceptual and cognitive actions and experiences.
4 Lexical expression of perception and cognition

In many languages, perception verbs such as ‘see’ and ‘hear’ develop salient evidential
and cognitive overtones (see Chapter 1). In Maaka, such verbs rarely refer to knowl-
edge, while a specialized cognition verb is used to encode the source of information.
This is interesting, because Maaka has borrowed this verb from Kanuri.
4.1 Perception
Maaka has a number of terms for see. Each of these terms covers a range of meanings:
vision as observation, perception, fantasy, illusion, imagination, and so on. Some of
these terms are well-attested Chadic roots, while others stem from different contexts.
So far, we have recorded the following lexemes denoting see:
• móy, vision as a non-deliberate, non-telic action;
• w�k, vision as a volitional and more telic action;
• yèt ‘discern, reason, deduce’;
• kàal, volitional action that involves motion of the agent;
• gòkk ‘spot, find’;
• jìnd ‘unexpected sighting’.
We have found only one more perception verb, namely sòl ‘perceive’ which encodes
smell, hear, feel, experience, obey, depending on the object-participant and the
context (examples 33–36):
hear
(33) nàsàrà-à sòló làbàr
European-def perceive:perv news
The Europeans hear news.
experience, feel
(34) kè-sòló gè-kúzùm
2sg:masc-perceive:perv loc-hunger
You experience a famine.
smell
(35) s�-sòló ɗàŋ kwám-yà
3sg:masc-perceive:perv smell cow-def
He smells the cow.
obey
(36) kè-sòló ɓá dàashí kè-dèyó-n-kò
2sg:masc-perceive:perv neg alright 2sg:masc-sit-link-icp
mbòkùm
blind
You didn’t obey, so you went blind.
This contrasts to other languages of the area, which code auditory, olfactory, and gus-
tatory perception by means of specialized verbs. For example, the verb ‘hear’ tends to
have semantic extensions into emotive experience, cognition, smell, and taste (Jaggar
and Buba 2009). The verb tùnn, which encodes ‘touch’, is hardly used as a perception
verb, but as an expression of proximity. However, when a particular dish or taste is
used as a complement, ‘touch’ also expresses ‘taste’.
4.2 Cognition
Instead of preserving the widespread Chadic root pVn (e.g. Kwami fònd) for ‘know’ or
using a polysemic perception verb to encode cognition, Maaka here uses nòn, which
is borrowed from Kanuri nongîn ‘get to know’ (Cyffer 1994: 102). In Maaka, this verb
encodes a wide range of cognitive meanings, such as ‘remember, be aware of, know,
recognize, be able’ (examples 37–40):
(37) nì-nònóyà gàlà=yékkú gàlà=dálá mà

1sg-know:perv:tr <title> <title> also
I recall them, Gala-Yeku and Gala-Dala.
(38) kè-nòn-ntì sháàní ngè má=bàarà

2sg:masc-know-assert habit conj folks=<toponym>
You surely are aware of the habit of the Bara people.
(39) s�-wà-nònò ɓá kàu kè-nòn-ntì

3sg:masc-neg-know:perv neg conj 2sg:masc-know-assert
màaká-n-cé
Maaka-link-pl
He didn’t know [them], [but] you then for sure recognized the Maaka folks.
(40) nì-wà-nònò krà ɓá

1sg-neg-know:perv reading neg
I don’t know how to read.
This verb can also express common knowledge as the source of information. In rhe-
torical questions, it evokes implied answers about shared wisdom. Such idiomatic con-
structions always indicate that the speaker is sure of what has happened. In examples
(41) and (42), the speaker refers to a habit that is common knowledge and therefore
true:
(41) wònnjì lái má nònò njéɗí máará wài
formerly inter folks [know:perv pay:tel]svc farm:intr:vn inter
Formerly, who would know how to pay for farming? [implied answer: clearly
none of us].
(42) wònnjì lái nònò kùnyènngí wài

formerly inter know:perv measuring_bowl inter
ʔáa-kè-sòltò ʔá mà pòró nà ɗèlá námáá sáa

cond-2sg:masc-hear:caus:perv then people say:perv quot corn immed ideo
Formerly, who knew measuring bowls? [implied answer: clearly none of us] When
you listened, people said: corn is like [the sound] saa . . .
This is the only instance in Maaka where a cognition verb expresses the source of infor-
mation. However, like most of the other devices to do with the grammar of knowledge,
these constructions do not highlight a source of information, but rather emphasize
possession and control of knowledge, exclusively shared insights, and the common
experience of truth.
5 Conclusions
Epistemic systems of languages seem to be very flexible, and loans such as nòn probably
reflect changes in how people perceive and explain the world. Despite the various socio-
cultural influences that motivated change in its grammar of cognition and perception,
the conceptualization of knowledge and truth in Maaka remains unique and doesn’t
converge areally. In sociolinguistic terms, the multilingual speakers of Maaka are able
to distinguish different ways and constructions of coding truth and wisdom, obviously
keeping apart singular worldviews and epistemes. And even though areal convergence
is characteristic of many properties of the languages shared by this multilingual com-
munity, the speakers of Maaka prefer to keep their distinctive grammar of knowledge.
In Maaka, knowledge and truth can be conceptualized as:
• the result of direct observation and affectedness (vision, joint perception, intuition);
• belief and conviction;
• possession and ownership of epistemic resources;
• control over information;
• shared insight and common experience.
We have argued that the notion of control and ownership prevails over perceptual
connotations. This is somewhat idiosyncratic with regards to the more widespread
semantics and functions of evidential markers and epistemic grammar cross-

linguistically. Why is this so, and why has the grammar of knowledge in Maaka
developed in this direction?
We suggest that the answer lies in the profile of the Maaka as a multilingual and
socio-culturally hybrid community. They inhabit a small part of a vast zone that has
known centuries of convergence. Group identities and boundaries are retained through
exclusive ideologies, secrecy of rituals and ritual language, and the conceptualization of
wisdom (enshrined in language and cultural praxis) as power. Under circumstances
of high social and economic pressure, the protection of potentially powerful knowl-
edge and wisdom possessed by such a group guarantees internal social coherence and
superiority over possible competitors. Therefore, the perpetuation of their unique
epistemic grammar and their specific conceptualization of knowledge as a means of
controlling and dominating other people is a salient, decisive strategy of being Maaka.
The other side of the coin may also be relevant. The Maaka live together with similar
multilingual people from various ethnolinguistic backgrounds, and use the individual
languages of their repertoire in specific contexts, so that truth or knowledge encoded
in Maaka has a different status than knowledge expressed in Hausa or Bole (Lüpke and
Storch 2013). These multilingual speakers in the villages of Bara and Gulani are not
only rich in linguistic choices and means of expressing themselves in different envi-
ronments and contexts, but they are also rich in ways of perceiving the world, mentally
organizing knowledge, and evaluating truth: multilingual people have several pairs of
glasses through which they can look at life.
References
Burke, Peter. 2012. A social history of knowledge. Vol. II. Cambridge: Polity.
Coly, Jules Jacques. forthcoming. ‘Some considerations on the structure of the Maha noun phrase’,
in Fading delimitations, edited by Johannes Harnischfeger, Rudolf Leger, and Anne Storch.
Cyffer, Norbert. 1994. English-Kanuri dictionary. Cologne: Köppe.
Frajzyngier, Zygmunt. 1993. A grammar of Mupun. Berlin: Reimer.
———. 1996. Grammaticalization of the complex sentence. Amsterdam and Philadelphia:
Benjamins.
Jaggar, Philip and Buba, Malami. 2009. ‘Metaphorical extensions of “eat” [OVERCOME] and
“drink” [UNDERGO] in Hausa’, pp. 229–51 of The linguistics of eating and drinking, edited by
J. Newman. Amsterdam and Philadelphia: Benjamins.
Jungraithmayr, Herrmann. 1970. ‘On root augmentation in Hausa’, Journal of African Languages
9: 83–88.
Kopytoff, Igor (ed.). 1987. The African frontier. Bloomington: Indiana University Press.
Leger, Rudolf. 1993. ‘Die Verben “lernen”, “wissen” und “vergessen” in einigen Bole-Tangale-
Sprachen (Nordnigeria)’, pp. 149–56 of IX. Afrikanistentag, edited by J. W. G. Möhlig,
Siegmund Brauner, and Herrmann Jungraithmayr. Cologne: Köppe.
———. 1994. Eine Grammatik der Kwami-Sprache. Cologne: Köppe.

———. 1998. ‘Grammatical gender in some southern Bole-Tangale languages—Kwami, Kupto,
Kushi and Piya’, pp. 204–16 of Africa: Society, culture and languages, edited by Institute of
African Studies. Moscow: IAS.
——— and Zoch, Ulrike. 2011. ‘Intransitive copy pronouns in Chadic’, pp. 11–46 of Copy pro-
nouns, edited by Anne Storch, Gratien G. Atindogbé, and Roger M. Blench. Cologne: Köppe.
Lüpke, Friederike and Storch, Anne. 2013. Repertoires and choices in African languages. Berlin:
Mouton de Gruyter.
Schuh, Russell G. 2003. ‘A comparative study of West Chadic verb suffixes’, pp. 71–86 of Selected
comparative-historical Afroasian linguistic studies in memory of Igor M. Diakonoff, edited by
M. L. Bender, Gabor Takacs, and D. L. Appleyard. Munich: Lincom.
———. 2011. ‘Grammatical influences of Kanuri on Chadic languages of Yobe State’, pp. 137–54
of Kanuri, Borno and beyond, current studies on the Lake Chad region, edited by Doris Löhr,
Eva Rothmaler, and Georg Ziegelmeyer. Cologne: Köppe.
Storch, Anne. 2011a. ‘Copy pronouns and pragmatics in Northern Jukun’, pp. 81–102 of Copy pro-
nouns, edited by Anne Storch, Gratien G. Atindogbé, and Roger M. Blench. Cologne: Köppe.
———. 2011b. Secret manipulations. New York: Oxford University Press.
Van Beek, Walter E. A. 1992. ‘The dirty smith: Smell as a social frontier among the Kapsiki/Higi
of north Cameroon and north-eastern Nigeria’, Africa 1: 38–58.
10
Expression of information source

meanings in Ashéninka Perené
(Arawak)
ELENA M IH AS
1 Introduction
Ashéninka Perené is an endangered South Arawak language spoken by about 1,000
people in the tropical rainforest of the Andean eastern foothills of the Chanchamayo
Province, Junín Region, Peru. The language is closely related to and has various
degrees of mutual intelligibility with Kakinte and other Ashéninka/Asháninka vari-
eties, notably Ashéninka Pichis, Ashéninka Pajonal, Ashéninka Ucayali, Ashéninka
Apurucayali, and Asháninka Tambo-Ene, all of which constitute the Northern branch
of the Kampan (Campa) subgrouping of Arawak (Michael 2008: 218). The past social
history of Ashéninka Perené hunters-horticulturalists is characterized by mobility
of nucleated families, no real sense of community due to the atomized structure of
residence, lack of centralized power networks, and enmity towards the outside world.
Currently, the native population resides in thirty-five native communities scattered in
the Upper Perené River valley. Farming is the main occupation: citrus fruit, pineap-
ple, bananas, and coffee are grown for sale, and vegetables are cultivated in gardens
for individual consumption.
Data for this research come from four field trips to the Chanchamayo Province
of Peru, conducted as part of the 2009–2012 Ashéninka Perené language docu-
mentation project.1 The collected materials include over thirty hours of audio- and
1
Many thanks are due to the Ashéninka Perené community for their participation in this research. I am
grateful to the granting agencies, the Firebird Foundation for Anthropological Research, National Science
Foundation (DDIG #0901196), and Endangered Languages Documentation Programme (SG 0002) for their
support. I also thank Sasha Aikhenvald and Bob Dixon for the helpful comments on the chapter, and mem-
bers of the International Workshop on Grammar and Knowledge, 16–21 July 2012, The Cairns Institute, for
the stimulating ideas and suggestions.
210 Elena Mihas
video-recordings of conversations, narratives, discussions, public speeches, songs,

jokes, and incantations, with approximately a third of all collected materials tran-
scribed and translated.
The Ashéninka Perené language is highly polysynthetic, head-marking, predomi-
nantly suffixing, with a very complex verbal and nominal morphology. It has an essen-
tially nominative–accusative system of grammatical alignment. The basic constituent
order is VOA and VS. Open classes are nouns, verbs, and derived adjectives; under-
ived adjectives form a small class of thirteen members. Verbal categories include per-
son, number, valence-adjusting/preserving, manner, direction, degree, aspect, reality
status, mood, modality, and past tense categories. Nominal categories include case
(although core constituents bear no case marking, there is a locative case -ki with a
diffuse spatial meaning ‘at/on/in/to’); morphologically expressed gender (masculine
-ri, non-masculine -ro); optionally expressed plural number; possession; definiteness
(encoded by the demonstrative enclitics); and past tense.
The language lacks a grammatical category of evidentiality but it has multiple evi-
dentiality strategies with a range of evidential-like meanings. Evidential strategies
involve evidential extensions of non-evidential categories and forms (Aikhenvald
2004: 105). The forms which serve as the source of evidential strategies include (i)
the marker of dubitative modality =ma which typically encodes inference, based on
traces of visual evidence (see §2.1); (ii) the bipartite unit =ratya, formed from the affect
modality marker =tya ‘mild preoccupation’ and demonstrative =ra ‘not here, not in
the speaker’s interactional space’, which has an assumptive (speculative) meaning ‘I
suppose’ (see §2.2); (iii) the bipartite unit =tyami which combines counterfactual con-
ditional modality marker =mi with the affect modality marker =tya, with the resultant
assumptive sense ‘probably’, ‘presumably’, ‘as far as I can tell’, when the speaker makes
an assumption on the basis of general expectations and common knowledge (see §2.3);
and (iv) the intuitive suspicion marker -mampy ‘have suspicions, misgivings’ which
refers to intuition as an information source (see §2.4). The marker’s diachronic source
is hypothesized to be the verb amampy ‘be in the shade’.
Other means of the expression of information source include verbs of perception
and cognition, used to express the following evidential-like meanings (see §3): ny ‘see’
refers to visual sensory evidence, kim ‘hear, smell, taste, feel’ describes non-visual
experiences, yoshiry ‘sense, intuit’ refers to own intuition as an information source,
and kimi ‘be like, resemble’ encodes the inferential meaning of ‘seem’.
In addition, there are parenthetical verbs which come from the verb of vision ny
‘see’, cognition verb kony ‘appear’, and the verb of speaking kant ‘say’. The parenthetical
(pi)nyaakiro ‘you’ve seen it’ conveys a peculiar meaning of the hearer having access to
visual evidence, typically shared with the speaker; konyaaro ‘appear clearly in a vision’
is applied to visions induced by drugs, or experienced in a sleep or an awakened state;
the parenthetical verbs ikantziri/akantziri ‘they/we say’ express the meaning of hear-
say (reported); and the parenthetical verb nokantziro ‘I say’ is self-reportative (see §4).
10 Information source meanings in Ashéninka 211
In Ashéninka Perené society, it is expected to be explicit in the identification of

information source. The sensory modality (visual, acoustic, tactile, ‘gut feeling’),
engaged in the process of the lived sensory experience, is invariably specified, by
essentially lexical means, when the speaker reports on his experience. In the case of
absence or paucity of sensory experiences, a variety of other devices are utilized such
as the verb kimi ‘seem’, inferential =ma, expectational =tyami, and assumptive =ratya.
The modal enclitics =tyami and =ratya express information sources which are not
directly experienced by the speaker who is relying on general knowledge of the world
or mere conjecture, respectively, to make an assumption.
2 Information source extensions of modal operators and of the verb

amampy ‘be in the shade’
2.1 Inferential and assumptive uses of the dubitative =ma~=taima
The basic meaning of =ma~=taima (in free variation) is associated with doubt. The
dubitative clitic typically attaches to the first constituent in the clause. It occurs in a wide
variety of clauses including declarative, imperative, interrogative, and negated clauses. It
is common in conversations and in narratives, especially when direct speech is quoted. A
suitable example of its dubitative use is in greetings, when someone visits the community
after a long period of being away. The local, who may have difficulty identifying the visi-
tor, will say in doubt: Aviroka=ma? [you=dub] ‘Is that you?’ In (1), when the speaker is
told that her sister was taken to the hospital, the speaker—who knows that the sister was
fine a day before—reacts by expressing doubt that this indeed happened. Example (2)
also illustrates a display of doubt in the context of an argument about which way is best
to go home. In both cases, the speakers use =ma to express doubt.
(1) y-a-ai-t-an-ak-e-ro=ma?
3masc.a-take-impers.pass-ep-dir-perv-real-3nmasc.o=dub
She was taken [to the hospital]?
(2) kaari=ma ironta

neg=dub dem
No, that [trail] is not the one [I doubt that that trail is the right choice].
The enclitic =ma~=taima is commonly used to express inference, when the speaker
ventures a guess about a state or event relying on observed results and applying the
‘cause–effect’ logical reasoning. In situations when observable evidence is present, the
ma-marked clause renders an interpretation of ‘seems that’ or ‘apparently’. Example
(3) is a comment made by a language consultant about his fellowmen who often make
a false inference, based on their observation of his whiskers which he inherited from
his Machigenka father (his mother was Ashéninka). His fellowmen think that he is not
an Ashéninka, because Ashéninka tend to lack facial and upper body hair. Example (4)
212 Elena Mihas
is a statement made by a warrior called Pitzikekentsi (the Coward), cited from a story
about ovayeripaye ‘warriors’. Pitzikekentsi is complaining about his sickness which he
attributes to a combat wound.
(3) pi-shipatona-t-ant-a-ri, aviroka=taima virakocha
2s-whiskers-ep-applic.reas-real-rel you=dub white.person
Because you have whiskers, you are apparently a white person.
(4) i-kint-ai-t-ak-i-na iro-t-ak-i=taima

3masc.a-shoot.with.arrow-impers.pass-ep-perv-real-1sg.o 3nmasc-ep-perv-real=dub
no-katsi-t-ant-a-ri
1sg.s-be.in.pain-ep-applic.reas-real-rel
I was shot with an arrow; this seems to be the reason why I feel pain.
The enclitic also encodes pure speculation, with the meaning ‘I am guessing’ or ‘prob-
ably’, that is, it is used when no sensory evidence is available. Due to the conjectural
nature of the ma-marked statements, the speaker does not vouch for the accuracy of
such speculative utterances. The speculative meaning of the dubitative marker =ma is
illustrated with examples (5)–(6). Example (5) is taken from a folk-tale. It cites the inner
thoughts of a protagonist, a fisherman who is followed by the murderous bone-breaking
bird shashintzi. The fisherman makes a tentative guess that he could escape shashintzi,
if he hides underneath his raft while going downriver. In (6), a narrator speculates that
manitzi (the Jaguar), a protagonist of the narrated folk-tale, may have a conjugal relation
to pakitsa (the Hawk), another famous character from Ashéninka folk-tales.
(5) no-shiy-aty-e=ma ironyaaka no-man-aty-eya

1sg.s-escape-prog-irr=dub now 1sg.s-hide-prog-irr
o-tapinaki=nta no-shintzipaa-ri-ki=nta
3nmasc.poss-back=dem 1sg.poss-raft-poss-loc=dem
I might escape, [if] I hide underneath the raft.
(6) iri=taima o-ime=ra antar-ite

3masc=dub 3nmasc.poss-husband=dem big-aug
pakitsa-t-atsi-ri=ra
hawk-ep-stat-nomz=dem
I am guessing, he [the jaguar] was the spouse, of that big hawk.
A non-evidential extension of the dubitative marker is associated with mirativity,

found in contexts when the speaker is stunned or perplexed by a discovery. A mirative
sense is often manifest in the ma-marked content interrogatives, as seen in (7). This is
a line from the Ashéninka story about a birth of the legendary cultural hero Naviriri
which includes a subplot of his pregnant mother’s entrapment. The mother is asked
by a cannibal woman to go up the ladder to fetch a basket of peanuts. The cannibal
woman wants to push the pregnant woman off the ladder into a big pot with boiling
water. The supernatural powers of Naviriri, who is in his mother’s womb, are such that
he makes the young woman unable to climb the ladder. After the mother makes a few
unsuccessful attempts, she says, in great surprise, expressing her bewilderment at a
totally unexpected event:
(7) tsika=ma o-kant-aty-a-ri=ka?
what=dub 3nmasc.s-happen-prog-real-rel=inter
How come?
In coordinated clauses, the dubitative marker has an emphatic adversative sense, when
it attaches to the topical personal pronoun in the subject function in the second clause.
Example (8) is from a text about the origin of sorcery whose female protagonist assures
her grandson that she has the power to perform sorcery, whereas her grandson doesn’t.
(8) aviroka airo pi-kan-tz-i-ri naaka=ma

you neg.irr 2a-be.able-ep-real-3masc.o I=dub
no-kant-ak-e-ro
1sg.a-be.able-perv-irr-3nmasc.o
You won’t be able to do it, but I will be.
In content interrogatives, =ma~=taima sometimes has a rhetorical emphasis, trans-

lated as ‘what the heck’, ‘what the hell’, ‘what in the world’. This emphasis is evidenced
in the higher than usual pitch and increased loudness of the voice. Example (9) cites
the speakers’ reaction to an offensive smell which began to disturb villagers. In (10),
after a visitor openly lays claim to the host’s cache of coca leaves, the host responds by
making a ma-marked emphatic statement.
(9) tsika=ma o-tzim-atz-i=ka yoka, anta shitzivari?

what=dub 3nmasc.s-be-prog-real=inter this over.there stinking
What in the world is this, the stinking [thing] over there?
(10) tsika=ma opaita=ka pi-nint-tz-i=ka?

what=dub cop=inter 2s-want-ep-real=inter
What the hell do you want?
The dubitative marker =ma~=taima can also occur in emphatic rhetorical questions
which presuppose a negative response. For example, a young woman reports to an
elderly fellow-speaker that in her house she has seen two kamitzi, blood-thirsty,
dwarf-like miniature people whose sighting forebodes ill to the Ashéninka house-
hold. The elderly woman responds: Pi-nint-atz-i=ma ironyaaka i-n-kitore-e-mi?
[2s-want-prog-real=dub now 3masc.a-irr-break.off-irr-2o] ‘Do you want them
to kill you?’
214 Elena Mihas
2.2 Affect modality marker =tya, combined with absentive demonstrative =ra:
assumptive =ratya
Before I discuss the ratya-strategy, I will briefly address the functions and distribution
of the clitic =ra which is found to function on three basic syntactic levels, namely noun
phrase, predication, and (subordinate) clause. Its host is either the head of a noun
phrase, or the verb predicate. With nouns, the demonstrative clitic has an absentive
meaning ‘absent/not in the speaker’s interactional space’, that is, the referent is outside
the speaker’s reach, up to a few meters away, for example eentsi=ra [child=dem] ‘those
children’ in (20), or it may be used to refer to an absent referent, located outside the
speaker’s interactional space but presumed to be close by, within the perimeter of the
house. In addition, =ra is commonly used in participant-tracking to refer to any iden-
tifiable (given) referent which is either active (i.e. the current focus of the addressee’s
consciousness), accessible (i.e. not in the current focus but inferentially and textu-
ally available), or inactive (i.e. neither in the focus nor periphery of the addressee’s
consciousness). The pervasive tracking use of =ra in narratives is found within a dis-
course node (i.e. a paragraph), at a discourse node boundary, and across a discourse
node boundary. Examples (11), (22), (38), and (46) illustrate the participant-tracking
use of =ra which refers to active referents in shirampari=ra [man=dem] ‘that man’,
intsompai-ki=ra [inside-loc=dem] ‘in that enclosure’, nihaa-ki=ra [water-loc=dem]
‘in that water’, mapi=ra [stone=dem] ‘that stone’, respectively.
In simple affirmative declarative sentences with a verb predicate, =ra functions as an
adverbial modifier of place, for example Ara=ha o-kant-a-kintsi-t-ap-ar-a=ka [dem.
adv=emph 3nmasc.s-be-ep-neck-ep-dir-perv-real=adv]‘Here, the incline of the
hill is here’.
In negated declarative clauses and imperative sentences, =ra displays a modal sense.
In the syntactic contexts involving constituent and clausal negation, =ra has a fixed
position, being always inserted between the negative particles airo~eero ‘irrealis nega-
tor’ or te ‘realis negator’ and the verb. In (11), which is a statement made by a female who
was forced to marry the man whom she didn’t like, the negated clause is interpreted
as containing a strong emphatic assertion, and =ra is translated as ‘never’, ‘absolutely’,
expressing a modal sense of the speaker’s certainty about the truth of the proposition.
In the imperative construction in (12), =ra is attached to the verb predicate, having
a sense of potential modality, or possibility/ hypotheticality of a future action, car-
ried out by the addressee. In (12), the mother is asking her son not to play with a cut
sugar cane stalk whose sharp edge can cause severe bleeding, if handled carelessly.
The language has no dedicated imperative paradigm, and commands are commonly
expressed via irrealis marking, often in combination with modal morphology, used to
modify the force of command.
In subordinate clauses, =ra always attaches to the verb, having either a temporal
‘when’ or a locative ‘where’ subordinating function. In (13), =ra occurs on the sub-
ordinate verb in the subordinating adverbial function, expressing the ‘where’ sense.
(11) te=ra no-nint-aty-e-ri shirampari=ra!

neg.real=mod 1sg.a-like-prog-irr-3masc.o man=dem
I sure never liked that man!
(12) airo pi-nyaatsa-vai-t-a-ro-tsi=ra?

neg.irr 2s-play-dur-ep-real-3nmasc.o-neg=mod
You won’t be playing with it, right?
(13) i-py-ak-i-ri ironyaaka

3masc.s-disappear-perv-real-3masc.o now
atsin-ak-ap-ak-i-ri=ra
crush-caus-dir-perv-real-3masc.o=sub
He disappeared, where they made the stone crush him.
The second component of =ratya, the affect marker =tya, plays various roles depend-
ing on distinct syntactic contexts: in imperative sentences, it expresses impatience
and urgency; in interrogative sentences, it conveys surprise; in declarative affirma-
tive sentences, =tya carries a sense of mild preoccupation; and in negated clauses, a
sense of alarm. The affect marker =tya is common in everyday discourse but is rarely
encountered in narratives outside of quoted speech. The clitic attaches to verb predi-
cates, arguments, question words, and negative particles, and can occur multiple times
in a clause, as seen in (14), which is an urgent command to begin the meal, issued by
the elder brother to his siblings. In (15), the female speaker is surprised that her belly
has grown so much. In the question, addressed to her mother, her surprise is signalled
by =tya, marked on the question word. Example (16) expresses the speaker’s alarm
when she informs her brother that a valuable item was accidentally dropped from her
tsampiro~champiro ‘wide, diagonally worn cloth for carrying loads’.
(14) tsame=tya a-ye-t-e a-v-av-ak-ya=tya!

come.on=aff 1pl.s-serve-ep-irr 1pl.s-eat-recip-perv-irr=aff
Let’s serve [the food] and eat!
(15) paita=ka=tya ant-ak-i-na-ri=ka no-motya-ki?

cop=inter=aff make-perv-real-1sg.o-rel=inter 1sg.poss-belly-loc
What happened to my belly?
(16) te=tya no-kim-i-ro tsika o-kant-a

neg.real=aff 1sg.a-sense-irr-3nmasc.o how 3nmasc.s-be-real
o-pary-ant-an-ak-a-ri
3nmasc.s-fall-applic.reas-dir-perv-real-rel
I didn’t feel it fall down.
The affect modal marker =tya exhibits assumptive values in a restricted syntactic
environment, namely when it is combined with the absentive demonstrative =ra with
216 Elena Mihas
the basic meaning ‘not here’, ‘not in the speaker’s interactional space’. The assumptive
meaning of =ratya ‘probably’, ‘I suppose’ overlaps with that of the dubitative marker
=ma ‘I am guessing’, ‘probably’. The primary characteristic of =ratya is the speaker’s
reliance on mere conjecture in lieu of any sensory evidence or logical reasoning as a
source of information, used to assess a probability of a given situation. Under such cir-
cumstances, the speaker’s statements carry an overtone of uncertainty about the stated
proposition. The composite morpheme =ratya tendentially cliticizes to the first clausal
constituent and is found in affirmative declarative, negated, and interrogative clauses.
Example (17) is an inquiry made by a language consultant about his female neigh-
bour who—the speaker speculates—is at her house at the moment of speaking. Exam-
ple (18) is cited from a narrative about the narrator’s encounter with a jaguar in the
midst of the jungle, which he survives unscathed (miraculously, the jaguar does not
attack him and simply retreats into the jungle). The speaker has to stay overnight in
the area, alone. Before falling asleep, he comforts himself with a speculative thought
that the jaguar is not stalking him and is unlikely to come. In (19), a hunter, who failed
to encounter any prey, presumes that a tapir might have passed the area. Alternatively,
the dubitative =ma ‘probably’, ‘I am guessing’ can be used here rather than =ratya,
signalling an overlap in the meanings of the two modal operators.
(17) saik-atsi=ratya o-vanko-ki?

be.at-stat=assum 3nmasc.poss-house-loc
Might she be in her house?
(18) airo=ratya i-pok-i manitzi=ranki

neg.irr=assum 3masc.s-come-real jaguar=abst
That jaguar is unlikely to come.
(19) y-avis-an-ak-i=ratya kemari

3masc.s-pass-dir-perv-real=assum tapir
A tapir might have passed.
When =ratya occurs in content interrogatives, it indicates a mirative sense of surprise

due to an unexpected discovery combined with reference to pure speculation as a
source of information. In (20), the speaker is wondering who the parents of the two
little children might be. These strange children are playing unattended on the river
bank. The speaker has never seen the children before and is puzzled.
(20) tsika=ratya ipaita=ka? ninka=ratya ashi-t-a-ri-ri

what=assum cop=inter who=assum own-ep-real-3masc.o-rel
eentsi=ra?
child=dem
What could it be? Who could be the father of those children?
2.3 Counterfactual modality marker =mi, combined with affect modality marker =tya:
expectational =tyami
Before I discuss in detail the tyami-strategy, I will briefly characterize the role of the
modal clitic =mi in Ashéninka Perené grammar. As a clause linking device, the coun-
terfactual clitic =mi marks verbs of both predicates to denote an action that never
eventuated. In single-clause statements, it either functions as an avertive marker to
indicate that an event was closely missed, as seen in (21), or as a marker of frustrative
modality in (22) to signal that an action failed to succeed or was performed in vain,
contrary to the speaker’s or the other participant’s expectations.
(21) iro-t-ak-i i-m-pasa-t-av-ak-i-ri=mi,

3n-masc-ep-perv-real 3masc.a-irr-hit-ep-dir-perv-real-3masc.o=cntf
monty-an-ak-a intatzikiro=nta
cross-dir-perv-real other.side=dem
They nearly hit him, [but] he crossed to the other side.
(22) intsompai-ki=ra ari no-saik-i=mi kameetsa
inside-loc=dem there 1sg.s-be.at-real=cntf well
Inside that corral, there I was supposed to stay [as a good boy, according to
my parents’ expectations].
The counterfactual conditional clitic =mi exhibits an information source extension
when it is coupled with the affect modal clitic =tya ‘mild preoccupation’. The result-
ant composite sense is translated as ‘probably’, ‘presumably’, ‘as far as I can tell’. It has
an ‘expectational’ meaning, having to do with the speaker’s reliance on their general
knowledge of the world and expectations of ‘similar situations, regular patterns, or
repeated circumstances’ (Mithun 1999: 184). The meaning of mild preoccupation of
the affect clitic =tya (which in this context displays a caring, non-confrontational atti-
tude toward the addressee) and the expectational sense of the counterfactual clitic =mi
contribute to the general assumptive meaning of the combination of =tya and =mi. It
is typically found attached to the verbal predicate or the verbal negative particle. The
clitic combination is not as tightly knit as the impenetrable speculative =ratya which
disallows any other clitic to separate the two components. In the case of =tyami, the
assertive emphasis clitic =kya can be inserted between the two, indicating a looser
integration of the two elements.
The tyami-marked examples reveal the speakers’ reliance on general knowledge of
the routinely expected turn of events, yet they all indicate the speaker’s uncertainty
about the veracity of the stated proposition. In (23), the speaker makes an assumption
about the expected rise of the water level in the local river. Common knowledge is that
the river floods every year during the rainy season, so the speaker’s statement is coded
by =tyami. Example (24) is from a funeral song, sung by a grief-stricken mother whose
toddler son died due to an unidentified illness. Using the tyami-strategy, the mother
218 Elena Mihas
expresses her belief that they (the family members) will not see the dead boy again,
since the dead do not come back. In (25), a female speaker makes a tyami-marked
commentary on her younger sister’s intention to punish her teenage girl. The older
sister doesn’t believe that the sister will do it because she knows that her sister always
threatens to administer severe punishment but never follows through.
(23) p-amavint-ya-ro parenini, osheki

2s-be.wary.of-irr-3nmasc.o river much
antaro-tz-i=tya=mi shintsini
big-ep-real=aff=cntf quickly
Be careful of the river, its level rises quickly, as far as I can tell.
(24) eero=tya=mi a-ny-ah-i-ri

neg.irr=aff=cntf 1pl.a-see-regr-real-3masc.o
We won’t see him anymore, I presume.
(25) pi-kant-ako-nent-ak-a-ro, eero=tya=mi
2a-do-applic-little-perv-real-3nmasc.o neg.irr=aff=cntf
pi-pasa-tz-i-ro
2a-punish-ep-real-3nmasc.o
You dote on her, you won’t punish her physically, I presume.
In addition, =tyami is used in imperative sentences to signal affective requests (pleas

and entreaties), illustrated by (26).
(26) no-tyomi, pi-n-tzin-ah-a-it-e=tya=mi

1sg.poss-sonny 2s-irr-get.up-regr-ep-incom-irr=aff=cntf
Sonny, get up, please.
2.4 The bound intuitive suspicion marker -mampy ‘have suspicions, misgivings’
The intuitive suspicion marker might have developed from the second compound
component via the verb+verb compounding technique. Compounding is quite pro-
ductive and often results in the production of non-compositional meanings, for exam-
ple I-ny-a-pinkatsa-atz-i-ri [3masc.a-see-ep-obey-prog-real-3masc.o] ‘He showed
him respect.’ The intuition marker -mampy is hypothesized to be diachronically relat-
ed to the verb mampy ‘be in the shade’, commonly used in discourse, for example
Amampy-av-ak-ya=ta [1pl.s. be.in.the shade-recip-perv-irr=opt] ‘We wish to be in
the shade’. The grammaticalized compound component shows signs of phonological
erosion, as seen in (27), where -mampy is reduced to -mpy.
(27) iroka mayempiro o-mishi-mpy-ak-a-na

dem loose.woman 3nmasc.a-dream-intuition-caus-real-1sg.o
This loose woman made me see her in my dream, it bodes ill, I suspect.
The bound marker -mampy ‘have suspicion, misgivings’ occurs in declarative affirma-
tive statements reporting the speaker’s past experiences, when the speaker relied on his
or her gut feeling as the information source to make an assumption about the future
course of events. The intuitive suspicion marker is typically used with the first-person
subjects of the experiential predicates mishi ‘to have dreams’, kaini ‘to have an itch or
some sort of inflammation’, py ‘to get lost, disappear’, pary ‘to fall’, when the experiencer
is a human participant who undergoes an involuntary mental or physical state, arising
from an external stimulus. In example (28), a female describes her mental state during
the initial phase of her pulmonary illness as being ‘lost’ between life and death, sensing
that this disease could be lethal. Another example (29) illustrates the speaker’s intui-
tion about her bad fall off a steep hill, when she nearly broke her spine.
The verb marked by -mampy does not take a complement clause and is found on verb
predicates in simple, as shown in (27)–(29), (31), and subordinate clauses, as seen in (30).
(28) no-py-mampy-an-ak-a
1sg.s-get.lost-intuition-dir-perv-real
I was lost [lost my heart], intuiting that something bad was going to happen.
(29) no-pary-a-mampy-an-ak-a, iro-t-ak-I

1sg.s-fall-ep-intuition-dir-perv-real 3nmasc-ep-pefv-real
no-katsi-t-ant-ak-a-ri
1sg.s-be.in.pain-ep-applic.reas-perv-real-rel
I fell, sensing that something bad would happen, that’s why I am in pain.
The marker -mampy is also attested in the second and third persons’ statements. In (30),
the mother chastises her daughter who walks within the perimeters of the house, disap-
pears from sight, and then the mother discovers that some items are missing. Every time
when the mother sees her daughter disappear, she suspects that something will be stolen.
(30) pi-ha-tz-i pi-koshi-tz-i nokantziro, iro=ma
2s-go-ep-real 2s-steal-ep-real I.say 3nmasc=dub
pi-py-a-mampy-ant-a-ri
2s-disappear-ep-intuition-applic.reas-real-rel
You are walking and stealing, I say, perhaps, that’s why you disappeared, it
bodes ill, I suspect.
Both the speaker’s and the protagonist’s viewpoint can be encoded by -mampy, the
former being illustrated by (27)–(30). Example (31) expresses the protagonist’s gut
feeling that killing a deer would get him into trouble. The hunter’s suspicion is rooted
in Ashéninka beliefs that a deer is an animal to be avoided, and shooting a deer may
cause a grave illness and even death. The intuitive suspicion marker in (31) occurs on
the agentive predicate whose subject argument is a human participant, consciously
and voluntarily causing a change of state of another participant.
220 Elena Mihas
(31) i-tonk-a-mampy-ak-a-ri maniro, ironyaaka

3masc.a-kill-ep-intuition-perv-real-3masc.o deer now
mantsiya-t-ak-i
get.sick-ep-perv-real
He killed a deer, having misgivings, and he got sick.
3 Other means of expressing evidential-like meanings
3.1 Perception and modal verbs
3.1.1 Visual perception verb ny ‘see’ Among Ashéninka Perené, disclosed information
has more credibility if it has been personally witnessed. Speakers are especially mindful
of the convention to explicitly indicate their direct visual access to evidence when they
relate stories about other people’s experiences. If the speaker witnessed an event involving
other people’s affairs, she will reference her firsthand visual evidence by using the first
person singular form of the perception verb ny ‘see’. For example, a narrative about the
persecution of sorcerers in the village of Nasaraiteki, witnessed by a female narrator in
1961, concludes with the line which contains nonyaakiri ‘that I have seen’, as cited in (32).
(32) irotaki no-kinkitsa-tz-i-ri ironyaaka naari okaratzi
foc 1sg.a-tell-ep-real-3masc.o now I.foc all
no-ny-ak-i-ri naari
1sg.a-see-perv-real-rel i.foc
‘What I told is all that I have seen.
3.1.2 Perception verb kim The perception verb kim is a default verb, used to describe
all sorts of non-visual sensory experiences: auditory in (33), olfactory in (34), tactile
in (35), and gustatory in (36).
(33) hear
no-kim-atz-i i-mo-tz-i manitzi hee hee.
1sg.s-hear-progr-real 3masc.s-roar-ep-real jaguar ideo ideo
I heard the jaguar roar hee hee.
(34) smell
no-kim-i-ro o-sankaa-ni inchatyaki,
1sg.a-smell-real-3nmasc.o 3nmasc.poss-scent-poss flower
o-kasanka-t-ak-i
3nmasc.s-have.nice.aroma-ep-perv-real
I smelled the scent of the flower, it has a nice aroma.
(35) feel
no-kim-i-ro o-kant-ak-i-na tsinkari
1sg.a-feel-real-3nmasc.o 3nmasc.s-do-perv-real-1o ideo
I felt that it did to me tsinkari [a momentary action of spreading warmth or cold
inside the body].
(36) taste
pi-n-kim-e-ro poshiniri-t-ak-i=rika ovarentsi
2a-irr-taste-irr-3nmasc.o well.done-ep-perv-irr=cond food
Taste it [to see] if the food is well-done.
3.1.3 Intuition verb yoshiry ‘sense’, ‘intuit’ To report ‘gut feelings’ experienced in the
past, the speakers use the generic intuition verb yoshiry ‘sense’, ‘ intuit’. The verb is
hypothesized to consist of yo ‘know’ and shiri(tsi) ‘soul’. The verb does not have negative
connotations and functions as a complement-taking verb. In (37), the speaker reports
on her intuition that the linguist would come to visit with her.
(37) no-yoshiry-ak-i-ro pi-m-pok-e

1sg.a-sense-perv-real-3n-masc.o 2s-irr-come-irr
I sensed that you would come.
3.1.4 Comparative verb kimi ‘be like’, ‘resemble’ In its basic sense ‘be like’, ‘resemble’,
the verb kimi ‘be like’, ‘resemble’ occurs with an NP complement, for example Kimi-
t-ak-a pamoro [be.like-ep-perv-real dove] ‘It is like a pamoro [dove species]’.
The verb kimi is found to express an inferential sense ‘it seems that’, ‘presumably’,
manifest in complement clause constructions which consist of two juxtaposed
verb predicates. When the verb has an inferential reading ‘seem’, it can be taken to
refer to non-firsthand evidence as a source of information, basically the speaker’s
logical reasoning (Dixon 2005: 204). The verb’s information source meaning in
(38)–(39) indicates that the speakers are able to make inferences based either on
sensory tactile information or a verbal report. In (38), the speaker comments on his
childhood practice of taking early morning baths in the river, together with his baby
sister, when the water seemed to feel warm if they stayed in the river long enough.
Example (39) is cited from a story about Pichootzi and his son-in-law Tziso who lied
to his father-in-law about his fishing and hunting prowess. When the father-in-law
finds out from a fellow villager that Tziso simply picks up rotten fish along the river
banks, he confronts his son-in-law, accusing him of deception.
(38) arika a-saik-i nihaa-ki=ra, kimi-t-ak-a

when 1pl.s-be.at-real water-loc=dem seem-ep-perv-real
o-saav-a-nint-atz-i
3nmasc.s-be.warm.liquid-ep-dim-prog-real
When we stay in that water, it feels warm.
(39) kimi-t-ak-a p-amatavi-t-ak-i-na antaro-ite

seem-ep-perv-real 2a-deceive-ep-perv-real-1sg.o big-aug
It seems that you have deceived me badly.
222 Elena Mihas
3.2 Parentheticals: (pi)nyaakiro ‘you’ve seen it’, konyaaro ‘appear clearly in a vision’,
ikantziri/akantziri ‘they say’, nokantziri ‘I say’
3.2.1 (Pi)nyaakiro ‘you’ve seen it’ (Pi)nyaakiro ‘you’ve seen it’ has an informa-
tion source meaning, having to do with the addressee’s firsthand access to visual
evidence. It is used in a situation when the speaker appeals to the addressee’s
direct visual experience by using the second person form of the verb ny ‘see’, pi-
ny-ak-i-ro [2s-see-perv-real-3nmasc.o] ‘you’ve seen it’. This reanalysed verb of
perception is very common in casual everyday speech (which might explain the
habitual elision of its first syllable, the second person subject prefix pi-) and is
considered to be part of a socially acceptable conversational behaviour. The form
nyaakiro is especially prominent in contentious familial exchanges. It is also
found in narratives in quoted direct speech. It typically appears in the clause-
initial slot but is also found either clause-finally or clause-internally.
The full translation of nyaakiro renders a rather cumbersome ‘you know it is
true because you have witnessed it yourself ’, interpreted here as a way to hedge
one’s position and to back up a statement, in order to make it sound more cred-
ible, with an explicit reference to the addressee’s own visual experience. In (40),
the use of nyaakiro gives more credibility to the wife’s words when she reproaches
her husband for unjustly punishing their daughter. When a character from a folk-
tale explains to his adult relative why he was abandoned by the fishing party on
the way home in (41), he uses nyaakiro ‘you’ve seen it’ in order to sound more
convincing. During a demonstration of the healing steam-bathing procedure, a
female healer uses nyaakiro multiple times, every time a new phase of the steam-
based herb treatment is shown to the audience. After the procedure is finished, the
healer inserts nyaakiro in the final statement in (42). This appeal to the addressee’s
visual experience aims to hedge the healer’s assessment of the patient’s future path
to recovery.
(40) nyaakiro, pi-pas-a-vai-shi-t-ak-a-ro

you’ve.seen.it 2sg.a-punish-ep-dur-applic.int-ep-perv-ep-3nmasc.o
a-shinto
1pl.poss-daughter
You’ve seen it, you punished our daughter for no reason.
(41) ani, nyaakiro, pi-tsaink-i=ranki

brother-in-law you’ve.seen.it 2s-disobey-real=anterior
Brother-in-law, you’ve seen it, you disobeyed [prior to having been abandoned].
(42) i-n-tekatsi-t-an-ah-e=ta, iyora mantsiyari=ra, nyaakiro

3masc.s-irr-recover-ep-dir-regr-irr=opt dem sick=dem you’ve.seen.it
He will recover, that sick [person], you’ve seen it.
3.2.2 Konyaaro ‘appear clearly in a vision’ The verb konyaaro ‘appear clearly in a
vision’ enjoys full positional freedom in a sentence. It is used in accounts of visions
either induced by hallucinogenic drugs or experienced without any stimulants in
a normal awakened state. It can be used to describe visions, seen in a dream. The
vision verb kony-a-ro [appear-real-3n-masc.o], whose diachronic source is the
verb kony ‘appear’, is invariant, although it can take a diminutive morpheme -heni~-
ini to describe a less clear vision. Konyaaro ‘appear clearly in a vision’ is attested
with either of the three verbs ny see, mishi ‘see in a dream’, as shown in (43), and
kim ‘hear’ in (44), in declarative affirmative clauses. In (44), a shaman describes
his vision of hearing an invisible female spirit maninkaro ‘the one who is hiding’ in
a remote location up in the hills. The parenthetical is also used to describe visual
sensory experiences of watching television, as shown in (45).
(43) konyaaro-heni no-ny-av-a-i-ro
appear.clearly-dim 1sg.a-see-dir-regr-real-3nmasc.o
no-mishi-t-av-ak-i-ro inaa-ni tsiteni=ranki
1sg.a-dream-ep-dir-perv-real-3nmasc.o mother-past at.night=anterior
I had a vision of my dead mother last night in a dream.
(44) konyaaro no-kim-ak-i-ro, o-saik-I
appear.clearly 1pl.a-hear-perv-real-3nmasc.o 3nmasc.s-be.at-real
okaaki-t-ap-ak-i
be.close-ep-dir-perv-real
I heard her in a vision, she is close.
(45) no-ny-ak-i-ro konyaaro shiyak-a-mento-tsi-ki=ra

1sg.a-see-perv-real-3nmasc.o appear.clearly image-ep-nomz-npossd-loc=dem
I saw it clearly on television.
3.2.3 Reportatives The hearsay ikantziri (literally) ‘they say [it] to them’ and akantziri
‘we say [it] to them’, and self-reportative nokantziro~nokantziri ‘I say [it] to them’
are common in casual exchanges which have some instructive purpose. Sometimes,
the third-person masculine object marker -ri is elided on the reportative forms. The
reportative forms function syntactically as a parenthetical, that is, as a syntactic unit
which interrupts a larger unit and enjoys positional flexibility, and whose semantic
function is ‘to prime the hearer to see . . . the reliability of our statements’ (Urmson
1952: 484). The two forms have a similar, conventionalized reportative meaning:
language consultants either ignore them in their translations or interpret them
as dicen ‘they say’. The self-reportative device nokantziri is typically translated as
semantically unreduced, le(s) digo or lo digo ‘I say to him/them’ or ‘I say it’.
The reportative forms appear to have been reanalysed as epistemic phrases which
assess the reliability of a given statement. In particular, akantziri~ikantziri ‘they say’ is
perceived to make reference to the less reliable, unidentified people’s experience, while
224 Elena Mihas
nokantziro is used to reference the concrete speaker’s personal experience, of more

credibility. In (46), an elderly male speaker, when asked about the use of the grinding
stone to produce fire, inserts the reportative ikantziri ‘they say’ in his response to a
younger male relative. By using ikantziri ‘they say’ and the dubitative marker =ma, he
signals his lack of commitment to the stated proposition.
(46) iro=ma, ikantziri, mapi=ra
3nmasc=dub they.say stone=dem
i-n-chenki-t-ak-i-ro=rika, morik-an-ak-e ampee=ra
3masc.a-irr-light-ep-perv-irr-3nmasc.o=cond burn-dir-perv-irr cotton=dem
Perhaps, they say, that stone, when it produces fire [on contact with the machete],
it ignites the cotton.
The reportative forms often co-occur with the assertive =kya. The assertive enclitic
=kya attaches word-finally to continuous topic arguments, verb predicates, or to time
adverbs, and can occur multiple times in a clause. When the clitic attaches to NPs,
it functions as a topic marker which reactivates a previously mentioned discourse
participant. When found on verb predicates, it can exhibit a negative component of
disagreement (and impatience) with the addressee who holds a contrary view.
The combination of ikantziri~akantziri ‘they say’ or nokantziri ‘I say’ with the assertive
=kya is devoid of negative nuances. When =kya marks the reportative form, it aims to
emphasize the credibility of verbal reports or self-reports. Examples (47)–(49) illustrate the
use of the kya-marked reportative form by an elderly woman who talks to a much younger
fellow speaker about old customs. She explains to the addressee the specifics of the pre-
colonial native diet and fabrication methods of garden tools. The speaker assertively uses
ikantzirikya and akantzirikya to signal a relatively reliable information source. When she
talks about her own experience of using coca leaves for therapeutic purposes, she switches
to nokantzirikya ‘I say’ to indicate that her knowledge is derived from her personal past.
(47) aisatzi a-p-i-ri, ikantziri=kya, te
also 1pl.a-give-real-3masc.o they.say=assert neg.real
a-n-kotsi-t-e-ro
1pl.a-irr-cook-ep-irr-3nmasc.o
Also, we were giving [to our children], they say, manioc which we didn’t boil.
(48) ara i-vitsik-a-vai-shi-t-a-ro, akantziri=kya

dem 3masc.a-make-ep-dur-applic.int-ep-real-3nmasc.o they.say=assert
Like this they made it [digging stick], they say.
(49) nokantziri=kya, kapicheni n-ako-ki koka=ra

I.say=assert little 1sg.poss-hand-loc coca=dem
no-v-ako-t-ak-i-ro
1sg.a-place-applic-ep-perv-real-3nmasc.o
I say, a little bit of those coca leaves I placed in my palm.
3.2.4 Direct and indirect speech The quotative verb kant ‘say’, which rigidly
precedes a direct quote, functions as a complement-taking verb. Apart from the
verb kant, there are also other verbs of speaking which introduce direct quotes:
ak ‘respond’, kaim ‘call out’, exhort’, and sampi ‘ask a question’. The argument
structure of the quotative form is typically limited to the actor participant:
ikantzi ‘he says’, okantzi ‘she says’, nokantzi ‘I say’. For example, a typical quotative
construction is as follows: I-kant-tz-i, ‘Pi-saik-e aka,’ [3masc.a-say-ep-real 2s-sit-
irr here] ‘He said, “Take a seat here”’. The quotative construction is extensively
used in narratives and is common in daily communication (cf. Michael 2008:
163 about the prevalence of direct quotations of the involved parties in Nanti
narrations of past interactions). Direct quotes serve two purposes: to meet the
expectation of being accurate and precise, and to mitigate responsibility for the
quoted statement.
Indirect speech is uncommon, and is typically attested with the speaking verbs
kamant ‘inform’, kamen~kamin ‘advise to prevent something undesirable’, kami ‘plead’,
kantzimanint ‘threaten’ and mental state verbs of thinking shiyakant ‘think, imagine’
and kinkishiri ‘believe, think’. Changes in the person category marking relative to the
original statement are indicative of indirect speech, as shown in (50)–(51).
(50) o-kame-t-ak-i-ri i-m-pina-t-e-ro kireeki
3nmasc.a-plead-perv-real-3masc.o 3masc.a-irr-pay-irr-3nmasc.o money
She pleaded with him to pay her money.
(51) i-kantzimanint-ak-i-ri y-ant-e-ri

3masc.a-speak.badly-perv-real-3masc.o 3masc.a-beat-irr-3masc.o
He threatened to beat him up.
4 Conclusions
Ashéninka Perené has multiple ways of encoding information source meanings
through the evidentiality strategies, which are largely expressed by modality markers,
and a variety of lexical means, which are finely attuned to details in expressing the
speaker’s sensory experiences. The evidentiality strategies and lexical means are com-
binable, as seen in (30), where the parenthetical nokantziri ‘I say it’ and the intuitive
suspicion marker -amampy co-occur in the sentence.
The evidentiality strategies form a ‘scattered’ system, that is, morphemes with
evidential-like meanings do not have a single syntagmatic locus (Aikhenvald 2004):
the clitics typically attach to the first clausal constituent, while the suffix -amampy
follows the verb root. The use of the bound information source markers, character-
ized by extreme pervasiveness in discourse, is pragmatically motivated, having to do
with foregrounding the salient information source or disambiguation of the discourse
context. The markers modify the entire clause and are found in declarative (affirmative
and negative), interrogative, and exclamatory sentences.
226 Elena Mihas
At the core of all information source meanings is the speaker’s viewpoint. Excep-
tions are the parenthetical nyaakiro ‘you’ve seen it’, which alludes to the addressee’s
firsthand sensory evidence, and the intuitive suspicion marker -mampy which can
express the protagonist’s premonition. The mirative extensions of the dubitative =ma
and assumptive =ratya express the surprise of the speaker, and the same situation
holds for epistemic extensions. In particular, objective or credible information sources
are those over which the speaker has exclusive authority, which are directly experi-
enced. They are encoded via the perception verbs, the parenthetical self-reportative
nokantziro, the parenthetical konyaaro ‘appear clearly in a vision’, and the intuitive
suspicion marker -mampy. In contrast, subjective or non-credible information sources
are those for which the speaker is not responsible, which are not directly experienced.
These are encoded by the inferential =ma, assumptive =ratya, expectational =tyami,
hearsay parentheticals ikantziri/akantziri ‘they say’, and the verb kimi ‘seem’.
References
Dixon, R. M. W. 2005. A semantic approach to English grammar. Oxford: Oxford University
Press.
Michael, Lev. 2008. ‘Nanti Evidential Practice: Language, Knowledge and Social Action in an
Amazonian Society’. PhD dissertation, University of Texas at Austin.
Mithun, Marianne. 1999. Languages of native North America. Cambridge: Cambridge University
Press.
Urmson, James Opie. 1952. ‘Parenthetical verbs’, Mind 61: 480–96.
11
Nominalization, knowledge,
and information source
in Aguaruna (Jivaroan)1
S I M O N E . OV E R A L L
1 Background information and typological profile

Aguaruna (known to its speakers as iiniá chicham) is a Jivaroan language spoken by
about 55,000 people in north Peru (INEI 2009). Most speakers live in areas of native
title along the Marañón River and its tributaries, in the department of Amazonas.
Aguaruna is nominative–accusative, and both head and dependent marking. The
usual clause structure is predicate final, and this is obligatory in non-finite clauses. Clause
chaining and tail-head linkage are integral to the structure of discourse and narratives.
The morphology in general is agglutinating with some cases of fusion, and it is almost
entirely suffixing, with just one unproductive prefix position for some verbs. Widespread
productive processes of vowel elision obscure the underlying regularity of much mor-
phology; see Overall (2007, 2008) for details of the phonology. Surface forms are given
in the examples, after vowel elision and other phonological processes have applied.2
1
I gratefully acknowledge the contribution of my Aguaruna teachers and consultants to this work, espe-
cially Pablo Santiak Kajekui, Segundo Cungumas Kujancham, and Jessica Danducho Yampis. I also received
valuable comments from the other authors and editors of this volume; of course any remaining errors or
shortcomings are entirely my own. The data collection and research were funded by an Australian Post-
graduate Award and La Trobe University.
2
Examples in this chapter use a modified version of the practical orthography in which the following
symbols differ from their IPA values: <g> = /ɰ/, <y> = /j/, <sh> = /�/, <ch> = /�/. Spanish words in the exam-
ples retain their original orthography. Phonemic nasality spreads within a domain of contiguous vowels
and glides; I mark it only on the underlying source. Nasal stops /m, n/ alternate with (prenasalized) oral
[b~mb, d~nd]; the alternation is not entirely predictable so I transcribe the oral forms as <b, d>. Accent is
contrastive but not relevant to the present discussion so I have left it unmarked. Verbs cited in the text are
in the standard citation form that consists of the root plus the action nominalizer -t(a); nouns are cited in
the unmarked nominative case. Grammatical suffixes are cited in their underlying form, with potentially
elided vowels in brackets.
228 Simon E. Overall
The analysis presented here has been formulated mainly on the basis of about
t wenty hours of text data collected in Amazonas, Peru, during 2004–06, augmented
by elicitation and discussion with native speakers in subsequent trips in 2008 and
2012. The texts were transcribed, translated into Spanish, and then glossed, all with the
assistance of native speakers of Aguaruna. They consist mainly of traditional stories,
with some autobiographical material.
2 Cultural practices
Ethnographic accounts of the Jivaroan people have emphasized their war-like culture,
in particular the practice of shrinking the heads of enemies killed in battle, and the
fiercely independent attitude of individuals. A second important aspect to the tradi-
tional culture is the use of plant-derived hallucinogenic drugs to gain access to the
spirit world (Harner 1973; Brown 1986; Rubenstein 2002), and this aspect has retained
its importance in modern times. All men, not just shamans, would traditionally seek
a vision of an ahutap spirit, the manifestation of ancestral warriors. The power thus
obtained ensured success in battle; nowadays visions are sought to gain insight into
future events or to help solve difficult problems. A person who has received a vision is
known as waimaku (< waima-ka-u = see.vision-perv-nomz). Traditionally a warrior
could not discuss the details of his vision until just before battle, or the spirit power
would be lost. The three most important ‘powerful plants’ are: datem (Banisteriopsis
spp.; Sp. ayahuasca, yagé); baikua (angel’s trumpet; Datura spp.; Sp. toé); and tsaaŋ
(tobacco). Tobacco is a narcotic, not a hallucinogenic, but is often used in conjunction
with the other two plants. It was traditionally consumed in liquid form, and when a
woman was teaching a younger relative magic songs she would chew some tobacco
leaves and spit the liquid into her palm for the student to suck up her nose. The narcosis
from the tobacco facilitates the learning process, and the passing of tobacco juice can
be read as symbolic of the passing of knowledge.
The spirit world can only be accessed and seen under the influence of one (or more) of
the three powerful plants; however, there is no grammatical or lexical correlate of knowl-
edge gained through visions and dreams. The same verb wainat ‘see, come to know’ is
used for waking vision, dreams, and drug-induced visions of the spirit world, and the
knowledge is treated as firsthand. Example (1) comes from a story in which a man sees
an ahutap in the form of a kinkajou (an arboreal nocturnal mammal, Potos flavus).
(1) wi-ka naa-num ... hinta tɨpɨ -sa-n,

1sg-top hesit-loc path.loc lie-dep-1sg.ss
tsaaŋku-n nampɨ -ka-n, kaŋkɨga tɨpɨ -kma-n
tobacco-acc get.drunk-perv-1sg.ss roll lie-term-1sg.ss
kuhi mina-u-n naŋkaɨm-u-n
kinkajou come.imperv-nomz-acc pass.by.imperv-nomz-acc
11 Nominalization and knowledge in Aguaruna 229
wain-ka-m-ha-i
see-perv-past-1sg-dec
Lying down in the um . . . , in the path, drunk on tobacco, as I was lying down
dizzy (lit. ‘rolling’) I saw a kinkajou come and pass by.
The prefixed causative form of the same verb, i-wainat (caus-see) ‘show’, is used in
example (2), with ahutap as the subject.
(2) imani wɨkaɨ-tatman i-wain-ka-u-ai

so.much wander-su>obj caus-see-perv-nomz-cop.3.dec
ahutap kuhi pachi-s
spirit kinkajou.acc regarding-3.dep.ss
As he was wandering so much, the ahutap spirit showed him a vision of a
kinkajou (lit.: ‘caused him to see, with respect to a kinkajou’).
3 Expressing and talking about knowledge

As one would expect, there are lexical words that express concepts related to knowl-
edge; a selection is listed in Table 1.
Table 1. Some words relating to knowledge
Verb Gloss Derived form(s)
dɨkat ‘know (saber)’ dɨkapɨt ‘know oneself, feel’
wainat ‘see, know (conocer)’ waimat ‘see a vision’ > waimaku

‘one who has gained spirit power
from seeing a vision’
i-wainat (caus-see) ‘show’
antut ‘hear, listen, understand’
hintintut ‘teach’ < hintĩ (path.1pl) ‘our path’

(with -tu vblz)
unuimat ‘learn, become accustomed to,

become tame (of an animal)’
anɨɨt ‘remember, miss, love’ > anɨntai ‘heart’ > anɨntaimat

‘think’ (with -ma vblz)
yacha (< Quechua) ‘wise, knowledgeable’

(Adj. or N.)
A B C D E
ROOT OBJECT ASPECT/POTENTIAL/DURATIVE TENSE SUBJECT MOOD/MODALITY
Figure 1. Verbal morphology
For the remainder of this chapter I will focus on the expression of knowledge
through non-lexical means, beginning with an overview of verbal morphology and
then in §4 describing the use of nominalization as a non-firsthand evidentiality strat-
egy. The basic template for a finite verb (excluding derivational morphology) is shown
in Figure 1.
Verbal morphology may be added directly to the unmarked root plus any object
marker in slot A,3 or to one of the following four stems (all consisting of root + A + B):
Imperfective characterized by stem-final -a (sg. subject) or -ina (pl. subject), and
typically used in present tense;
Perfective characterized by one of a group of suffixes selected on the basis of verbal
Aktionsart (all glossed perv ‘perfective’ here, for simplicity) and typically used in
non-present tenses and imperatives;
Potential with suffix -mai;
Durative with vowel lengthening and suffix -ma, and only used with imperative
mood.
3.1 Modality
Neither (traditional) epistemic nor deontic modality is well-represented lexically or
paradigmatically. There is no set of modal verbs, nor is there a single straightforward
paradigm of modality as a verbal grammatical category. Deontic modality tends to
be expressed through implicature, using potential verb forms or imperatives, while
epistemic meanings are scattered throughout the verbal morphology. There is a single
paradigm of mood/modality markers that form an obligatory grammatical category
for finite verbs in Aguaruna. These are listed in Table 2.
Of the thirteen types, eight are marked in slot E. The imperative/prohibitive forms
are all marked in the same slot as tense markers (C), leaving slot E empty.
Counter-expectation and speculative both express epistemic meanings. Specula-
tive relates to the speaker’s degree of commitment to the truth of the proposition
(example (3)), while counter-expectation relates more to the speaker’s assessment of
the addressee’s expectations.
(3) numi-na-ts tsupia-tai
wood-acc-spec cut.imperv.3-spec
Perhaps it’s wood that he’s cutting.
3
Only SAP objects are overtly marked on the verb, with zero marking on a transitive verb indexing
third-person object. The combination of first-person subject with second-person object is marked with a
portmanteau suffix in the subject slot (D).
Table 2. Formally marked moods/modalities
Mood Clause type Marker
Indicative Declarative -i
Counter-expectation -hama
Narrative tuwahamĩ
Speculative -tai
Interrogative Polar interrogative ka (or -Ø if marked

elsewhere in the clause)
Content interrogative suppression of apocope (clause
contains an interrogative word)
Tag question -api
Exclamative Exclamative -Ø
Imperative Imperative -ta ⎫

Jussive -ti ⎪⎪
Hortative -mi ⎬ (marked in slot C)
Apprehensive -ĩ ⎪
Prohibitive -ipa ⎪⎭
The exclamative may, like the mirative, mark an ‘unprepared mind’ (cf. DeLancey
1997), or it may assert emphatically, as in example (4) used in scolding a child who
wouldn’t stop fiddling with expensive equipment.
(4) achi-ka-ipa, ta-ha!

grab-perv-prohib say.imperv-1sg.exclam
‘Don’t touch it’, I say!
3.2 Reported speech

Speech reports are all direct quotations accompanied by the speech verb tuta ‘say’,
either alone or supporting another verb. Larson (1978) describes the importance of
speech reports for narrative structure (and cf. Beier et al. 2002 and references therein),
and there are some grammatical constructions that have arisen from speech reports.
Some complement clauses, purpose (example (5)) and reason clauses must be
formed using speech reports. Desiderative, future, and frustrative (see Overall in
prep.) markers have all developed from speech report constructions.
(5) iwi-ya-hi tɨpɨ -s-ti tu-sa

raise.hand-rem.past-1pl.dec lie.down-perv-jus say-dep.1pl.ss
We raised our hands so that it (the truck) would stop. (lit: saying ‘let it lie down’)
The narrative modality marker tuwahamĩ (see example (10) below) comes from
a form of the verb tuta ‘say’, although it is morphologically opaque synchronical-
ly. Aguaruna speakers translate it into Spanish as así decían ‘so they would say’.
Although tuwahamĩ is a separate phonological word, it clearly forms part of the same
paradigm as the bound mood/modality markers: it cannot co-occur with any other
modality marker, and it fulfils the requirement for obligatory modality marking in
finite verbs.
The fact that tuwahamĩ is a separate phonological word suggests that it has
recently grammaticalized, and further evidence comes from the fact that mor-
phologically transparent forms of tuta ‘say’ may also occasionally function as nar-
rative modality markers (e.g. ti-mĩ say.perv-past.3.dec), also in complementary
distribution with bound modality markers. These forms apparently mark a text
as having identifiable authorship, but it is not necessary to explicitly attribute
the text to a specific author. As a result, there is no sharp delineation to be made
between direct quotation and narrative modality expressing generally accepted
cultural knowledge.
4 Nominalization in Aguaruna
Like many Amazonian languages (cf. van Gijn et al. 2011), Aguaruna makes use of
deverbal nominalizations for a variety of functions, which include: relative clauses;
clause chaining; auxiliation; and heading independent clauses as a non-firsthand evi-
dentiality marking strategy. The seven nominalizing suffixes are listed in Table 3.
The nominalizers can be subgrouped according to their properties (indicated by
lines in the table). The two nominalizers in group A take an unmarked verb stem, and
are more like canonical lexical nominalizations (cf. Comrie and Thompson 1985). They
refer to habitual or inherent properties, and may be semantically unpredictable. The
pair in group B generally take an aspect-marked stem, and have a realis sense. They
both form relative clauses, and -m(a)u can also function as an action nominal. Within
each group, there is a distinction between subject and non-subject nominalizers. Note
that overt objects are marked with accusative case for subject nominalizations but not
for the non-subject ones.
The negative nominalizer -ch(a)u refers to any participant, giving a form such as
yu-chau ‘that which is not eaten’, ‘one who does not eat’. It appears to consist etymo-
logically of negative -cha plus -u.
The ‘remote past’ nominalizer -haku, like -m(a)u and -ch(a)u, has final /u/ and
may etymologically involve the subject nominalizer -u, but is synchronically not
decomposable.
The event nominalizer -t(a) is used to form complement clauses, and also abstract
nouns, for example ha-ta ‘illness’; puhu-t ‘way of life’. This nominalizer also gives the
citation form of verbs.
Table 3. Nominalizers
Suffix Stem Referent Object marking
a. -in(u) unmarked subject (S/A) acc
-taĩ unmarked non-subject (O/E/ nom

instrument/location)
b. -u aspectual subject (S/A) acc
-m(a)u aspectual non-subject (O/E/ nom

instrument/ location);
event
-ch(a)u unmarked/ S/A/O participant + acc

aspectual negative
-haku unmarked S/A participant + acc

past tense
-t(a) unmarked event nom
None of the nominalizers involves a change from clausal to NP like syntax and there
is no genitive marking of subjects.
4.1 Functions of the subject nominalizer -u

Relative clauses are formed with nominalizers -u and -m(a)u. They are frequently
headless, as in example (11) below, and this accords with the fact that modified NPs are
relatively rare and anything more complex than [N Adj] tends to appear prosodically
and structurally like multiple NPs in apposition (see Overall, to appear b.). Relative
clauses formed with -u may be reanalysed as temporal dependent clauses (Overall to
appear a.; cf. Epps 2009; Guillaume 2011).
(6) yaki wa-kã ɨkɨ -t-u-n

[above go.up-perv.3.ss sit-applic.imperv-nomz-acc]
yunum-tu-k-u-i
approach-applic-perv-nomz-cop.3.dec
(The man) approached (the boa) that had gone up and was sitting up above him.
OR
When (the boa) had gone up and was sitting above him, (the man) approached it.
In (6), the bracketed nominalization can be parsed as an NP or a temporal clause. If

the latter, the case marking on the nominalized verb functionally simulates switch-
reference, as it indexes the role of a common argument in both clauses. M
orphological
evidence of this reanalysis (actualization) can be found in occasional examples of
nominalized verbs taking dependent clause person marking.
(7) Belén wɨ-u-n wakɨt-ha-i
B. go.perv-nomz-1sg.ss return.imperv-1sg-dec
ta-wa-i
say.imperv-3-dec
He says he’s coming back from Belén (village). (lit.: ‘I’m returning, having
gone to Belén’)
We will see in §4.3.2 below that the same development seems to have been the source
of one of the dependent verb forms.
4.2 Nominalization as an evidentiality strategy

Verbs nominalized with -u may function as the main verb in an independent clause,
with or without a copula suffix. These ‘stand-alone’ nominalizations are used to mark
non-firsthand information source.
In declarative finite equative/attributive clauses with singular subject, there is not a
separate copula verb but a suffix is attached to the copula complement which is then
followed by person and mood marking. In (8) and (9), the nominalization is formally
a copula complement.
(8) tupika-kĩ, wɨga wɨga-kuã,

run-perv.3.ss redup go.imperv-repet.3.ss
hɨga-u-wai Manchumush
arrive.perv-nomz-cop.3.dec M.
puha-mu-num
live.imperv-nomz-loc
Having run, going and going, she arrived at the place where Manchumush
lives. (lit: ‘she is one who arrived. . . ’)
(9) Herodes Judea nugka apu-hĩ waha-s

Herod Judea land.gen chief-3 stand-dep.3.ss
puha-ĩ, nunu nugka-nma-ŋ,
live.imperv-1/3.ds anaph land-loc-top
yaakat Beleŋ-num, Jisus akiina-u-wai
town Bethlehem-loc Jesus be.born.perv-nomz-cop.3.dec
When Herod was King of Judea, in that land, in a town called Bethlehem,
Jesus was born. [Matthew 2:1] (Yyamajam [sic] Chicham Apajuinu 2008: 8)
This type of verb marking is standard in traditional narratives, in place of finite tensed
verbs, and example (9) shows that it has also been used in a translation of the Bible. The
nominalization functions here as a non-firsthand evidentiality strategy, in contrast to
finite verb forms that are neutral with respect to information source.
The stand-alone nominalization may take the narrative mood/modality marker.
This is the only member of the mood/modality paradigm that is not bound to the
verbal word, and when associated with a stand-alone nominalization does not require
a copula suffix. This suggests that the copula marking in (8) and (9) is required only to
host declarative mood marking.
(10) nuwa makichik naŋkai-n usupa-ŋ,

woman one fruit-acc crave-perv.3.ss
hiin-ki-u tuwahamĩ
go.out-perv-nomz narr
A woman having craved fruit, (she) went out they say.
The relationship between nominalized clauses and narrative modality is discussed

further below (§5.2).
Finally, a bare nominalization can stand as the main predicate. In (11) there is no
copula suffix and no mood/modality marking at all.
(11) nihamchi nampɨ-kã hiin-u-na-shkam

masato.acc get.drunk-perv.3.ss go.out-nomz-acc-also
hu-wa-u
take-perv-nomz
(The devil) also took (people) who left the house having got drunk on masato.
It is not entirely clear what factors allow this usage, but it happens mainly in texts
where narrative modality is used. It seems to be the case that narrative modality mark-
ing can have scope over more than one independent clause, so that an example like
(11) in context would be implicitly marked for modality. The first clause of a text can
never contain a bare nominalization.
The overwhelming majority of verbs in narratives have a third-person subject,
but there are one or two examples of nominalized verbs with a first-person subject.
Example (12) is the beginning of an autobiographical story.
(12) mina daa-hu-k Pablo-i

1sg.gen name-1sg-top Pablo-cop.3.dec
wi-ka akiina-u-ait-ha-i
1sg-top be.born.perv-nomz-cop-1sg-dec
comunidad Chikais
community Chikais
My name is Pablo. I was born in the community Chikais.
The nominalized form is appropriate because the speaker cannot be said to have
witnessed his own birth (despite having been present!). The rest of the narrative is
couched almost entirely in finite past tense forms, in keeping with the fact that the
speaker witnessed all the events. Note, however, example (13): finite verbs are used to
describe events that happened when the narrator had fainted from illness, so he can
only have been told about them later.
(13) tũha-sh waamak yutun-ha-abia-ha-i

but-also quickly get.worse-perv-past-1sg-dec
nuni-taĩ, waamkɨs medico-num
do.that-dep.1/3.ds quickly.3 doctor-loc
ɨhɨ-tu-aw-aha-amayi
take-1sg.obj-perv-pl-past.3.dec
I quickly got worse. When I did that, they quickly took me to a doctor.
This appears to be a counter-example to the hypothesis that nominalizations are used

to mark a non-firsthand information source, and it is not clear why this should be so.
More texts consisting of first-person accounts would surely shed some light; at present
the best we can say is that it presumably reflects some optionality to the use of this
evidentiality strategy.
I have only a few examples of the use of nominalized verbs in conversation. The
following two are drawn from my own interactions. The context of (14) is that I am
going to the tap in the village and I ask my neighbour Doris if the water is working.
(14) a. (Simon) yumi atsa-wa-k?

water exist.neg.imperv-3-inter
Is there any water?
b. (Doris) atsa-u!
exist.neg.imperv-nomz
There isn’t any!
In (15), I am visiting my friend Jerónimo and call out as I approach his house. As
Jerónimo is not home, his wife Florentina responds.
(15) a. (Simon) Jerónimo, puha-mɨ -k?

J.voc live.imperv-2-inter
Jerónimo, are you there?
b. (Florentina) atsa-wa-i!
exist.neg.imperv-3-dec
He’s not here!
The (b) responses form a minimal pair. In (14b), Doris has not actually been up to the
reservoir and checked the pipes, she is passing on what she has heard (or inferred),
so the nominalized form is appropriate. In (15b), Florentina is sitting right there and
knows that Jerónimo is not home. Segundo Cungumas, a native speaker consultant,
suggests (personal communication) that a nominalized form in reply to a question is
somewhat curt, in contrast to a finite form that implies more engagement in conversa-
tion. Indeed, (14b) has the effect of shutting down conversation: Doris has no more
details and cannot answer questions; whereas Florentina presumably would be happy
to discuss where Jerónimo has gone and when he will be back.
4.3 The role of nominalizations in verbal paradigms

There is evidence that other nominalizations have infiltrated the verbal inflectional
paradigms, both finite and dependent. Verbal and nominal morphology are quite dis-
tinct, in spite of a few points of overlap, and the following four morphological criteria
can be used to show that a suffix is (or was) a nominalizer:
1. case marking;
2. inability to take finite verb person markers directly;
3. possibility of taking copula suffix;
4. non-obligatoriness of any further morphology.
4.3.1 Remote past -haku The remote past nominalizer is relatively rare, appearing
in the first few lines of narratives to set the scene. It is typically followed by the third-
person declarative form of the copula suffix, as in (16).
(16) makichik muun a-haku-i
one adult exist-nomz-cop.3.dec
There was a man. [first line of a story]
The postvocalic allomorph of the copula suffix is -i, homophonous with the
declarative suffix. Evidence that this is indeed the copula suffix comes from examples
such as (17), where the separate copula verb ata is required to host the plural
marking, as the suffixed copula does not mark number.
(17) tuhã tikichi-k duik yaunchuku-k ishama-haku

but other-top long.ago long.ago-top be.afraid-nomz
a-ina-wa-i
cop-pl.imperv-3-dec
But the other people long ago were afraid.
Some examples in my corpus show -haku followed by the 1sg subordinate verb marker
-n(u). Consider the following example taken from a man’s description of the prepara-
tion for a battle he underwent as a youth. (Note that the language used here is fairly
stylized. The idiom ‘follow the path’ refers to undergoing preparation for battle, and
‘dreamed of a waterfall’ means he went to a waterfall to drink the plant preparations
and seek a vision.)
(18) wi-it-ha-i, hinta aintu-haku-n,

1sg-cop-1sg-dec path.acc follow-nomz-1sg.ss
datɨma amu-haku-n,
ayahuasca.acc drink-nomz-1sg:ss
baikua amu-haku-n,
angels.trumpet.acc drink-nomz-1sg:ss
tsaaŋku amu-haku-n,
tobacco.acc drink-nomz-1sg:ss
tuna-na-sh kaham-in a-ya-ha-i
waterfall-acc-also dream-nomz cop-rem.past-1sg-dec
It is I, following the path, drinking ayahuasca, drinking angel’s trumpet,
drinking tobacco, I was one who dreamed of a waterfall.
The same possibility exists for the nominalizer -u, as in example (7) above. So both
nominalization types can be reanalysed as dependent verbs and directly take person
marking, but can only take finite verb suffixes when mediated by the copula suffix.
4.3.2 Dependent verb markers -ma, -tatamana Two dependent verb markers
encode a switch-reference that is not of the canonical type that opposes ‘same-
subject’ to ‘different-subject’ verbs. Instead, they refer to roles of a common
argument in both clauses, as shown in Table 4.
It is suggestive that the form with final /na/ encodes an object in the control-
ling clause, as the accusative suffix has the form -n(a). It is possible that this suffix
originated in a subject nominalizer with accusative marking, which was reanalysed
Table 4. Non-canonical switch-reference markers
Role of common argument:
Suffix Stem in marked clause in controlling clause
-ma perfective or non-subject subject

imperfective
-tatamana unmarked subject object

as non-canonical switch-reference marking, the same reanalysis I suggested for -u

nominalization in §4.1.
In sum, verbal and nominal morphology are quite distinct, but some verb forms ap-
pear to have nominal origins, having passed through a ‘full circle’ of grammaticalization:
verb → deverbal nominalization → predicate nominal → verb
Verbs nominalized in -u take on the added nuance of non-firsthand evidentiality,

which is carried over into the reanalysis as a verb, adding a new function to the canon
of verbal categories.
There remains the question of why I have labelled this an evidentiality strategy,
rather than an evidential proper. Aikhenvald (2004: 105) says of evidentiality strate-
gies: ‘They are distinct from evidentials proper, whose primary—and not infrequent-
ly exclusive—meaning is information source.’ The -u nominalizations in Aguaruna
retain their full range of nominal functions, and take nominal morphology—the
information source marking is clearly not their primary function. Further, there
seems to be some choice for speakers whether or not to use the nominalized forms,
as in example (13) above where a nominalization was not used, in spite of the non-
firsthand information source. In sum, there is no reason to treat information source
as an obligatory grammatical category of Aguaruna so the label ‘evidentiality strategy’
is appropriate.
5 Why nominalize a verb if you’re going to use it to head a finite clause?

Aikhenvald (2004) reports on the use of deverbal nominalizations to mark a non-
firsthand information source in a number of languages. Why should this be so? Con-
sider Givón’s (2001) cline of finiteness, defined along the lines of verbal categories
marked. Nominalization is at the non-finite extreme of this cline: nouns do not mark
verbal categories. This is not quite the full story for Aguaruna, as nominalizations may
be marked for aspect, and case marking of arguments remains the same as in finite
clauses. But there is no marking of the obligatory verbal grammatical categories of
person, tense, and mood/modality. This reduction in verbal grammatical categories
can be seen as iconically representing the speaker’s reduced firsthand knowledge of
the events being reported.
Nominalization has been much discussed in the Tibeto-Burmanist literature (Noo-
nan 1997; Genetti et al. 2008; DeLancey 2011) and the wider South Asian context,
however much less has been said about Amazonian languages. Tibeto-Burman nomi-
nalizations express a range of modal and stance meanings, such as:
In Tamangic languages, when nominalizations appear as main clauses, the typical effect is one of
mirativity, i.e. the sense that the predication so expressed is in some sense surprising, contrary
to expectation, or in some way exasperating. (Noonan 2011: 202).
In Lahu stand-alone nominalizations give ‘a strong assertive, matter-of-fact inter-

pretation’ (Matisoff 1972), and Watters (2002) describes nominalizations used in a
mirative construction in Kham, and in narratives to mark ‘discontinuity’, that is both
backgrounded material and pivotal events on the main narrative line.
Aguaruna nominalizations clearly show some similarities to Tibeto-Burman forms,
with two notable formal differences. First, when a nominalization modifies an NP in
Aguaruna the syntax is distinct from a possessive NP, unlike the usual Tibeto-Burman
pattern. Secondly, the Aguaruna -u nominalization refers to the subject participant,
whereas Tibeto-Burman languages tend to use action nominals in stand-alone con-
structions, as in (19), which could be literally translated ‘it is the case that I have gone
to Kathmandu.’
(19) Chantyal
[kadmandu-ri ɦya-si-wa] ɦin
Kathmandu-loc go-ant-nmlz be+npast
I’ve gone to Kathmandu. (Noonan 1997: 380)
In Aguaruna, the nominalization refers to a participant and the copula must agree with
the subject. In (20), the nominalization plus copula translates literally as ‘I am (one)
who was born in Chikais’, and not ‘It is the case that I was born in Chikais.’
(20) wi-ka akiina-u-ait-ha-i comunidad

1sg-top be.born.perv-nomz-cop-1sg-dec community
Chikais
Chikais
I was born in the community Chikais.
5.1 Discourse functions of nominalizations and speech reports

Following Hopper (1979: 213), I assume that ‘in any extended text an overt dis-
tinction is made between the language of the actual story line and the language of
supportive material which does not itself narrate the main events’. In Aguaruna nar-
rative, dependent verbs contrast with independent (=finite) verbs, but this contrast
does not directly relate to grounding. DS dependent forms tend to be associated
with backgrounding while SS is associated with foregrounding. Similarly, imper-
fective stems tend to be associated with backgrounded clauses and perfective with
foregrounded (Overall to appear a). The different tense forms also appear to pat-
tern together as foregrounded/backgrounded pairs. There are four synthetic past
tense forms and they combine with the two nominalized forms already discussed
as shown in Table 4.
There are thus two parameters involved in the selection of independent verb
forms: grounding (narrative structure) and the neutral/non-firsthand distinction in
Table 5. Past tense markers and narrative functions
Foreground
Distance in time Neutral Non-firsthand Background
− -ma -ma. . .ya

-u
‘recent past’ ‘intermediate past’
‘non-firsthand’
(nominalizer)
-ama. . .ya -ya
‘distant past’ ‘remote past’
+ -haku
‘narrative past’
(nominalizer)
information source. This contrasts with Watters’ findings on stand-alone nominaliza-

tions in Kham: ‘In general, everything on the narrative event line is marked by regu-
lar finite forms, while everything subsidiary to it—background, parenthetic material,
stage s etting—is marked by nominalization’ (Watters 2002: 34).
5.2 Nominalization and narrative modality

The -u nominalizations functioning as a non-firsthand evidentiality strategy and
the narrative modality marker tuwahamĩ that has developed from a speech report
construction share the property of lowered personal involvement on the part of the
speaker. How do they differ?
Aguaruna traditional narratives fall roughly into two types, those presented as
historical events, often with a moral; and mythical stories which often include super-
natural elements and may involve some ‘creation myth’ type elements. These are not
classified as distinct genres in the indigenous tradition: all are auŋmatbau ‘things that
are told’ (Sp. cuentos típicos). However, they do appear to show distinct grammati-
cal properties. The historical stories make more use of -u nominalizations, typically
with the copula suffix and declarative mood marking. The myth type use narrative
modality tuwahamĩ as well as -u nominalizations, which then do not require the
copula suffix.
For non-literate societies, such as the Aguaruna were until recently, cultural
knowledge is embodied in traditional stories. The canon of oral literature can be
likened to Wikipedia: it is a means by which the community can maintain a body of
knowledge that is more than one person can handle alone. It is pertinent to ask what
the use of the narrative modality marker actually means to someone in this culture.
In Australia and similar legal jurisdictions, ‘hearsay’ is explicitly rejected in legal
contexts as unreliable:
The rule against hearsay means that an assertion made by someone other than a witness in the
court proceedings will generally be inadmissible as evidence, because of its lack of reliability.
(Mann 2010: 283)
However, I suggest that in the context of a traditional body of oral literature, mark-
ing a narrative as hearsay (as the narrative modality marker derived from a speech
report construction does) has precisely the opposite value: it imparts the legitimacy
of precedent to the narrative currently being related. Given this view of ‘hearsay’,
then, it is unsurprising that a dedicated narrative modality marker should be distinct
from, and independent of, the more general non-firsthand evidentiality strategy,
that also covers (less reliable) personal inference. While a number of Amazonian
languages have developed evidential marking from speech report constructions,
in Aguaruna the latter have shown a separate, parallel development into narrative
modality.
6 The areal context

The development of a nominalizer into an evidentiality marker can be viewed in the
context of wider areal tendencies. The position of the Jivaroan languages in the foot-
hills of the Andes has meant that they are in both the Amazonian and the Andean
contact areas, and they share many typological features of both the Amazonian and
Andean prototypes as defined by Dixon and Aikhenvald (1999: 8–10) and elabo-
rated by Payne (2001) and Aikhenvald (2012: 72–4). On the Amazonian side, Beier
et al. (2002) discuss a number of discourse phenomena that lead them to suggest
a greater Amazonian ‘discourse area’. Two of their proposed features clearly relate
to the Aguaruna data presented above: ‘evidential systems in discourse’ and ‘speech
reporting practices’. In the same paper the authors suggest that a preoccupation with
epistemology is a shared cultural feature in the greater Amazonian area, and that this
preoccupation may motivate the diffusion of evidentiality as a grammatical category.
On the Andean side, there is evidence for a long history of contact with Quechua, a
language with grammaticalized evidentiality (Adelaar 2004; Overall 2007; Muysken
2010). A detailed study of the contact motivations for the Aguaruna evidential strategy
remains to be undertaken, and until then we can only speculate on the possible source.
Such a study will also contribute to our understanding of the linguistic history of the
eastern slopes of the Andes more generally.
References
Adelaar, Willem F. H. (with Pieter C. Muysken). 2004. The languages of the Andes. Cambridge:
Cambridge University Press.
———. 2012. The languages of the Amazon. Oxford: Oxford University Press.
Beier, Christine, Michael, Lev, and Sherzer, Joel. 2002. ‘Discourse forms and processes in indig-
enous lowland South America: An areal-typological perspective’, Annual Review of Anthro-
pology 31: 121–45.
Brown, Michael. 1986. Tsewa’s gift: Magic and meaning in an Amazonian society. Washington:
Smithsonian Institution Press.
Comrie, Bernard and Thompson, Sandra A. 1985. ‘Lexical nominalization’, pp. 349–98 of
Language typology and syntactic description, Vol. 3, Grammatical categories and the lexicon,
edited by Timothy Shopen. Cambridge: Cambridge University Press.
DeLancey, Scott. 1997. ‘Mirativity: The grammatical marking of unexpected information’,
———. 2011. ‘Finite structures from clausal nominalization in Tibeto-Burman’, pp. 343–59 of
Yap, Grunow-Hårsta, and Wrona, 2011.
Dixon, R. M. W. and Aikhenvald, Alexandra Y. (eds). 1999. The Amazonian languages.
———and ———. 1999. ‘Introduction’, pp. 1–21 of Dixon and Aikhenvald, 1999.
Epps, Patience. 2009. ‘Escape from the noun phrase: From relative clause to converb and beyond
in an Amazonian language’, Diachronica 26(3): 287–318.
Genetti, Carol, Coupe, Alexander R., Bartee, Ellen, Hildebrandt, Kristine, and Lin, Y ou-Jing.
2008. ‘Syntactic aspects of nominalization in five Tibeto-Burman languages of the H
imalayan
area’ , Linguistics of the Tibeto-Burman Area 31(2): 97–143.
van Gijn, Rik, Haude, Katharina, and Muysken, Pieter (eds). 2011. Subordination in native South
American languages. Amsterdam/Philadelphia: John Benjamins.
Givón, Talmy. 2001. Syntax: An introduction, 2nd edition. Amsterdam/Philadelphia: John
Benjamins.
Guillaume, Antoine. 2011. ‘Subordinate clauses, switch-reference, and tail-head linkage in
Cavineña narratives’ , pp. 109–28 of van Gijn, Haude, and Muysken, 2011.
Harner, Michael J. 1973. The Jívaro: People of the Sacred Waterfalls. New York: Anchor Books.
Hopper, Paul. 1979. ‘Aspect and foregrounding in discourse’, pp. 213–41 of Syntax and semantics,
Vol. 12, Discourse and syntax, edited by Talmy Givón. New York: Academic Press.
INEI [Instituto Nacional de Estadística e Informática] 2009. Resumen Ejecutivo: Resultados
definitivos de las comunidades indígenas. <http://www1.inei.gob.pe/biblioineipub/bancopub/
Est/Lib0789/Libro.pdf>.
Larson, Mildred L. 1978. The functions of reported speech in discourse. Texas: SIL International.
Mann, Trischa (ed.). 2010. Australian law dictionary. Melbourne: Oxford University Press.
Matisoff, James A. 1972. ‘Lahu nominalization, relativization, and genitivization’ , pp. 237–57 of
Syntax and Semantics, Vol. 1, edited by John Kimball. New York: Seminar Press.
Muysken, Pieter. 2010. ‘The copula in Ecuadorian Quechua’ , pp. 191–206 of Linguistics and
Archaeology in the Americas, edited by Eithne Carlin and Simon van de Kerke. Leiden/Bos-
ton: Brill.
Noonan, Michael. 1997. ‘Versatile nominalizations’ , pp. 373–94 of Essays on language function
and language type. In honor of T. Givón, edited by Joan Bybee, John Haiman, and Sandra A.
Thompson. Amsterdam/Philadelphia: John Benjamins.
———. 2011. ‘Aspects of the historical development of nominalizers in the Tamangic languages’ ,
pp. 195–214 of Yap, Grunow-Hårsta, and Wrona, 2011.
Overall, Simon E. 2007. ‘A Grammar of Aguaruna’. PhD dissertation, La Trobe University,

Melbourne.
———. 2008. ‘On the non-phonemic status of the velar nasal [ŋ] in Jivaroan’ , LIAMES 8: 45–59.
———. To appear, a. ‘Clause-chaining, switch-reference and nominalisation in Aguaruna
(Jivaroan)’.
———. To appear, b. ‘Attributive adjectives and the internal syntax of the NP in Jivaroan’ .
———. In preparation. ‘Frustratives in Amazonian languages’.
Payne, Doris L. 2001. Review of Dixon and Aikhenvald, 1999. Language 77(3): 594–98.
Rubenstein, Steven. 2002. Alejandro Tsakimp: A Shuar healer in the margins of history. Lincoln,
Nebraska: University of Nebraska Press.
Watters, David E. 2002. A grammar of Kham. Cambridge: Cambridge University Press.
Yap, Foong Ha, Grunow-Hårsta, Karen, and Wrona, Janick (eds). 2011. Nominalization in
Asian languages: Diachronic and typological perspectives. Amsterdam/Philadelphia: John
Benjamins.
Yyamajam [sic] Chicham Apajuinu, 5th edition. 2008. La Liga Bíblica. <http://www.christusrex.
org/www1/pater/wolf3-pdf/aguaruna-nt.pdf>.
12
The grammar of knowledge in Tima
G E R R I T J . D I M M E N DA A L
1 Preliminary observations
Tima is spoken by approximately 6,000 people living mainly in the Nuba Mountains,
in Sudan; in addition, there is a community of around 1,000 speakers in the capital
Khartoum. The speakers refer to the language as �àmáá dù-mùrík, ‘the language like-
Tima’, whereas they call themselves kù-múrìk (sg), ì-múrìk (pl). Based on the pioneer-
ing survey of languages in the Nuba Mountain area by Roland Stevenson in the 1950s,
Tima was classified as a member of the Katla cluster by Tucker and Bryan (1956: 64).
Greenberg (1963) classified this cluster as a member of the Kordofanian branch within
Niger-Congo (or Niger-Kordofanian), a position also followed by Schadeberg (1981a,
1981b). The latter author excluded Greenberg’s Tumtum (Kadugli) cluster from Kor-
dofanian, and arrived at a four-way division for this Niger-Congo branch (Figure 1).
More recent comparative research by the present author on the three languages together
forming the Katla cluster, Katla, Julud, and Tima, suggests that this cluster is most closely
related to the Rashad group, with which it probably forms a genetic unit. The Heiban
and Talodi languages, however, are only distantly related to these, and probably should
be treated as a distinct, primary branch of Niger-Congo. The Katla-Rashad group on the
other hand shows a considerable degree of grammatical and lexical affinity with Benue-
Congo languages and appears to be more closely related to these (Dimmendaal to appear).1
1
Data on Tima were collected as part of a language documentation project between 2006 and 2012. Exam-
ples in the present contribution are derived from narrative discourse and other text genres such as conversa-
tions, procedural discourse, and dialogues recorded, transcribed, and annotated as part of this project. Research
on Tima was made possible through two grants from the Volkswagen Foundation, which allowed a team of
researchers from the universities of Cologne and Khartoum (Susan Alamin, Abeer Bashir, Meike Meerpohl,
Abdulrahim Mugaddam, Gertrud Schneider-Blum, and the present author) to carry out fieldwork on this fas-
cinating language. We would like to express our deeply felt gratitude to the Volkswagen Foundation for making
this research possible, and to the Tima community for their enthusiastic participation. We would also like to
thank the student-assistants in the project, Meikal Mumin and Nico Nassenstein for their support. For further
details on the Tima documentation project see: <http://www.mpi.nl/DOBES/projects/tima/>. I would also like
to express my sincere gratitude to Sasha Aikhenvald and Bob Dixon for the kind invitation to participate in the
Workshop on the Grammar of Knowledge, and to the participants for their various questions and suggestions.
246 Gerrit J. Dimmendaal
Heiban
Talodi
Kordofanian
Rashad
Katla
Figure 1. The four-way division of Kordofanian
Tima has a two-tone system with downdrift and downstep (Dimmendaal 2009:
335; Bashir 2010). In addition, it has an Advanced Tongue Root harmony system with
six [−atr] and six [+atr] vowels as well as a system of fronting harmony (see Bashir
2010 for details). It furthermore has a reduced noun class system with traces of a more
extensive former system (Alamin 2012; Dimmendaal 2013). Only one productive sin-
gular noun-class prefix (kV-) and one productive plural prefix (ı-/ i-) is found with
nouns as well as with nominal modifiers synchronically:
(1) k-ùrtú k�-mál

ncl-house agr-beautiful
Beautiful house.
(2) y-ùrtú �-mál

ncl-house agr-beautiful
Beautiful houses.
Apart from several closed sets of noun classes, there are various incorporated noun
class prefixes in Tima, as becomes clear when comparing this language with the
closely related Katla language, or more distantly related languages belonging to the
Rashad group (a phenomenon referred to as the ‘annual ring’ model in Dimmendaal
2013). Similarly, cross-reference marking for Subject (or Agent) and Object on the
verb is sensitive only to number and person, not to the noun class a particular noun
belongs to.
Constituent order varies between OVA, AVO, VAO, and AOV, depending on the
pragmatic context. Tima is predominantly head-marking at the clausal level, with up
to thirteen morphological slots for the verb. When following the verb, nominal and
pronominal Agents are marked by way of an Ergative clitic (N). The Ergative marker
procliticizes onto an Agent noun, with which it forms a phonological word; if the
postverbal Agent is pronominal, this marker plus pronoun encliticize onto the verb.2
Because the verb plays a central role in the present chapter, its formal structure is given
in Table 1.
2
In order to mark the phonologically bound status of clitics, the symbol = is used, whereas the bound
status of affixes is marked by way of the symbol -.
Table 1. The verbal complex in Tima
1 NEG proclitic
2 Aspect proclitic
3 Person (S, A) prefix
4 Tense prefix
5 Root
6 Der.1 suffix
7 Der. 2 suffix
8 Der. 3 suffix
9 ERG enclitic
10 DAT enclitic
11 Pron. O enclitic
12 Pron. A, S enclitic
13 NEG enclitic
Below, a first (non-exhaustive) list is presented of grammatical features that are

central to the grammar of knowledge in Tima. The four subsystems or grammatical
domains discussed here are:
• the temporal dimension;
• the spatial dimension;
• reference tracking;
• experiential knowledge.
2 The temporal dimension: construction-type effects

As shown in Chapter 1 of the present volume, tense-aspect marking frequently plays
a role in the grammar of knowledge. Whereas at an earlier stage of research on Tima,
this was also assumed to be the case for this language, our current understanding of
the language is that the tense-aspect-mood (TAM) system itself is not involved in
this respect. Instead, the combination of specific tense-aspect-mood forms with other
words in a sentence or clause trigger certain conversational implicatures, as argued
below.
Tima has a binary tense system with a distinction between a morphologically
marked future and an unmarked non-future.3 The future marker-dV- in Tima prob-
ably is related historically to the motion verb díy�ŋ, ‘come’, which can be divided into
a root di- plus (one of the allomorphs of) the ventive marker -ʌŋ, with the epenthetic
glide y-being inserted in order to avoid specific vowel sequences within a word; this
verbal root, however, is no longer used in isolation in Tima. Paradigmatic distinctions
within the future tense are expressed morphologically by distinct portmanteau mor-
phemes preceding the future marker. These are illustrated for the first-person plural
inclusive in example (3).
(3) é-dí-díík=n�y b��ìn
1pl:inc-fut-go-1pl:inc soon
We (incl.) will go soon. (Future)
(4) cé=dí-díík=n�y á�n�k��

imperv.1pl:inc-fut-go-1pl:inc later on
We (incl.) will be going. (Imperfective Future)
(5) k�=dí-í-díík=n�y
pot-fut-1pl:inc-go-1pl:inc
We (incl.) will (definitely) go. (Potential Future)
At an earlier stage in the analysis of Tima, labels such as Definite as against Indefinite
Future were used (Alamin 2012), or Immediate vs. Remote Future. Similar terminol-
ogy exists for the non-future tenses, for example, Remote or Indefinite vs. Recent or
Definite Past), again reflecting the fact that it is difficult to come to grips with the
semantics of these TAM forms in Tima. The corresponding current terminology for
these non-future tenses is again illustrated for the first-person plural inclusive forms
when combined with the verb ‘go’:
(6) céé-díík=n�y
imperv:pres.1pl:inc-go-1pl:inc
We are going. (Imperfective Present)
(7) cèè-díík=n�y
imperv:past.1pl:inc-go-1pl:inc
We were going. (Imperfective Past)
3
For a detailed survey of the various TAM forms, including full paradigms, the interested reader is
referred to Alamin (2012: 78–103).
(8) í-díík=n�y
perv:past.1pl:inc-go-1pl:inc
We went. (Perfective Past)
Whereas the TAM system of Tima is thus not particularly complex from a more gen-
eral, typological point of view (compare typological surveys such as Bybee, Perkins,
and Pagliuca 1994), the actual system is nevertheless somewhat intricate for a number
of reasons—as is true, presumably, for many other languages too. First, as a result of the
way relative time is expressed in a complex sentence. For example, in a context ‘when
you came yesterday, we were cleaning’, the TAM used in the matrix sentence is Imper-
fective Present, with the subordinated clause providing the background information,
and TAM marking in this latter clause providing the time of reference. Secondly, the
interaction of the various verbal affixes and clitics with each other and with the root
results in a rather complex system of morphophonemic alternations affecting both the
segmental and the tonal layer (as in example (7) vs. (8) above). Third, focus marking
in a clause also affects tense-aspect marking on the verb, in that syncretism occurs for
specific TAM forms; moreover, the formal expression of Agent and Subject marking
is affected by information packaging within a clause or sentence.
Based on our current understanding of the Tima TAM-system, the conclusion is
that there is no evidentiality marking within the actual TAM system itself. Instead,
evidentiality marking results from construction-level effects at the clause level, more
specifically from the combination of TAM forms on the verb with adverbs or adverbial
phrases functioning as adjuncts in the clause. The inferred meaning in a statement
with a focused adverb in (9) preceded by ‘since’ is that the person talked about is still
present.
(9) Kw�kw�ŋ ádáá kùl�=w� díy�ŋ
Kwʌkwʌŋ since yesterday-foc walk.ven
Kwʌkwʌŋ came yesterday (and is still here)
In the corresponding statement without the word for ‘since’ in (10), there is no implica-
tion that Kwʌkwʌŋ is still around.
(10) Kw�kw�ŋ kùl� díy�ŋ
Kwʌkwʌŋ yesterday walk.ven
Kwʌkwʌŋ came yesterday (we don’t know whether she is still around).
Also, the presence of the marker t�k, here translated as ‘really’, in the example below
taken from a trickster tale, renders a conversational implicature that the speaker does
not actually believe this to be possible.
(11) n�-m�nt-�k=nàn t�k w�r��máád�h cé=kúùn
perv:past:2pl-hear-ap-2pl part man imperv:pres-bear
Have you really heard that a man gave birth?
Nevertheless, evidentiality marking by way of specific tense-aspect morphemes is

common cross-linguistically. It is found, for example, in Nilotic languages south of the
Nuba Mountains. Storch (2006) describes this phenomenon for the Western Nilotic
language Jur Lwoo, and Miller and Gilley (2007) for the closely related language Shil-
luk. In this latter language, there is a direct evidence marker á- expressing the fact that
the speaker was an eyewitness to a certain event:
(12) dyàŋ á-‘kwālì
cow past-steal.tr
Someone (and I know who it was) stole the cow.
As pointed out by Miller and Gilley (2007), the conversational implicature triggered
by the combination of this tense-aspect marker with the verb ‘steal’ is that the speaker
saw this person do it, but is not ready or willing to make an accusation.4 This visual
aspect expressed by way of a specific tense-aspect marker in Shilluk is covered by an
alternative strategy in Tima as we shall see later, that of location and direction mark-
ing, which forms the core of the grammar of knowledge in this language.
3 The spatial dimension: the viewer-centred perspective

Tima grammar is particularly rich with respect to location marking. Apart from
adverbs of place and demonstratives as lexical categories, different grammatical strat-
egies play a role in local deixis. For example, noun classes alternate when occurring
in locative constructions (along lines which are reminiscent of locative-class marking
in Bantu).
(13) k-�cùk (The) baobab tree.
w-�cùk At the baobab tree.
y-�cùk Exactly on/at the baobab tree.
Whereas w-�cùk indicates an approximate location (glossed as loc1 in the examples

below), y-�cùk (glossed as loc2) expresses an exact connection or attachment of some
object to a specific location. In addition, there is a locative prefix lV- (glossed as loc),
as in the name for ‘the Tima area, Tima country’, lú-múrìk. A number of nouns are
inherently marked for location, for example ‘field’, lâh.
Next to prepositions derived from nouns specifying the search domain (as in (15)
below), there are two basic prepositional proclitics, a Source marker a=, and a Direc-
tion marker V= (consisting of an underspecified high vowel). But the typologically
most interesting feature of Tima grammar probably is the ‘obligatory’ marking of the
position of the speaker. Tima marks the movement or path of a Figure (in the sense of
4
Hieda (2012) shows still another strategy, namely for the Western Nilotic language Acholi, where per-
ception verbs followed by a paratactic (as against a hypotactic) complement denote direct perception.
Talmy 1985) towards the deictic centre (the speaker or the protagonist in a story) by
way of a derivational suffix on the verb, called ventive here. In addition, the position
of the speaker can be expressed on prepositional phrases (glossed as EGO in examples
below) in order to express the Ground.
(14) dí-y-�ŋ �t�=lâh
walk-ep-ven ego-field
Come to the field (where I am)!
The clitic n�(V)- forms a phonological word with (basic and prepositional) nouns, as
in ��-�hí (<��-�hí) in example (15) below. It encliticizes onto adverbs derived from
verbs, such as ày�n-�� (<ày�n-n��); see Alamin, Schneider-Blum, and Dimmendaal
(2012) for further details on location and direction marking in Tima.
The following sentence was uttered by a Tima speaker while looking at a picture
used as a stimulus:
(15) �hwáá-y=�� h�làk ��=�hí w-�cúk k�-p�rár��k
people-ep-foc stay ego-ground loc1-baobab agr-hollow
The people stay(ed) under the hollow baobab tree.
By adding the proclitic n�(�)-, the speaker expressed the fact that (s)he was present on
the occasion. Compare also the following alternative constructions:
(16) Kw�kw�ŋ àn-d�wá-y-�ŋ á=l�ŋ��
Kwʌkwʌŋ imperv:past-go.down-ep-ven prep-east
Kwʌkwʌŋ came from the east/top; Kwʌkwʌŋ descended towards me.
If, on the other hand, the speaker himself/herself were in the east or on the top of
the mountain, and the person moved away from him/her, the following construction
would be appropriate:
(17) Kw�kw�ŋ án-d�wà á=n��=ŋ��
Kwʌkwʌŋ imperv:past-go.down prep-ego-east
Kwakwang went down from the east/from the top.
Alternatively, when the speaker is neither at the starting point nor at the goal, there is
no separate marking for EGO either on the verb or on the prepositional phrase:
(18) Kw�kw�ŋ án-d�wà á=l�=ŋ��
Kwʌkwʌŋ imperv:past-go.down prep-loc-east
Kwʌkwʌŋ went down from the east/top.
In other words, if no ventive or EGO marking is used, the conversational implicature is

that the speaker (or the protagonist in a story) is not a witness to the event described.
Such utterances consequently involve a ‘quantity inference’ in the sense of Levinson
(2000): ‘what isn’t said, isn’t’.
The position of the speaker thus plays a central role in Tima discourse. Its formal
expression may be found in utterances referring to the present or past, but also to the
future, as in the following example drawn from a story.
(19) má=dɔ�wá ɲìhìn� t�n-�l-�ŋ
seq-rise erg.3pl turn-rev-ven
. . . And then they will return (to where the speaker is).
It is common in languages to develop idiosyncratic meanings from the interaction of

lexical roots and derivational affixes. This can also be observed in the following Tima
example with a ventive marker on the verb, where neither the position of the speaker
(EGO) nor movement towards the deictic centre necessarily plays a role.
(20) kìnéè c�n-d�ŋ��ŋ á=n��=lí-ŋ��

sun imperv:pres-ascend.ven prep-ego-loc-east
The sun rises in the East.
Ventive marking is extremely common in Niger-Congo, Nilo-Saharan, Chadic, and

Cushitic (Afroasiatic) languages as well as in Ubangian (treated as an independent
family in Dimmendaal 2011: 89–90), and therefore constitutes an areal feature. It occa-
sionally contrasts with itive (allative) marking on the verb. However, there is no cor-
responding itive marking in Tima. It seems that so far no author has related ventive
marking as a derivational phenomenon to evidentiality marking in these languages,
although in a language like Tima it is obvious that there is such a direct link. When a
ventive marker is used in combination with a verb, the inferred meaning or implication
is that the speaker was, is, or will be a witness of the event. Tima appears to be unique so
far (based on the existing literature on the subject) in that it also marks the position of
the speaker (or the protagonist in a story) on prepositions or nouns in locative phrases.
Tima—and in fact numerous other languages with ventive marking—manifests
another typologically interesting feature, the formal expression of two events in one
and the same verb, referred to as ‘alloying’ in Alamin, Schneider-Blum, and Dim-
mendaal (2012: 26–9). In the following example with an imperative verb, somebody is
ordered to first drink and then move towards the speaker:
(21) m��k-�ŋ
drink-ven
Drink and come (to where I am)!
Newman (2000: 663) presents a similar case of ‘alloying’ for the Chadic language
Hausa:
(22) yā say-ō nām�
3sg:perv buy-ven meat
He bought some meat and brought it back here.
Newman (2000: 663) also mentions lexicalized forms, and points out addition-
al semantic mappings such as ‘association with, involvement by, or benefit for the
speaker’ in the case of Hausa. This, again, parallels the semantic widening observable
for ventive marking in Tima, as shown by example (20) above.
4 The referential dimension: logophoricity marking as an evidential

hedging strategy
One of the better studied domains of pronominal reference marking with respect
to African languages involves logophoricity, a system also found in Tima, as shown
in Schneider-Blum (to appear). In Tima, only Subjects and Agents can be cross-
referenced by way of pronominal elements. Subjects and Agents are expressed by way
of pronominal prefixes and enclitics on the verb, depending on the tense-aspect form
or information packaging in the clause, as shown in Table 2 and as illustrated in (23)
and (24).
(23) �-kúmún-�d� Àb��r

perv:past-see-1sg Abeer
I met Abeer.
(24) cáá-dàh-��-dà d�ŋdèè . . .

imperv:past.2sg-say-ben-1sg like
You told me . . .
Table 2. Pronominal clitics
S, A ERG LOG
1sg -dʌ/-da -nʌ/-na

/-dɔ /-nɔ
2sg -ŋaŋ -ŋaŋ
3sg (p�n�) m�n� ŋùŋ
1pl: incl -nεy -nεy
1pl: excl -nin -nin
2pl -nan -nan
3pl (ìhìn�) ɲìhìn� ŋìŋ

Ergative forms as in (25) result from the phonological merger of the Ergative clitic
N= with the enclitic pronouns:
(25) ŋààŋ=á húm-áá-yáŋ=n� cíd�
2sg-foc depend-inst-loc 3sg -erg.1sg body
I depend on you (sg).
The Ergative pronoun set only occurs with transitive predications when either the
object or the transitive verb (plus unexpressed object) carries focus. The object may
consist of a focused noun (phrase), an independent pronoun or complement clause
functioning as the object of the transitive verb. The first two precede the verb, whereas
complement clauses always follow the main verb in Tima. Focus marking on the com-
plement clause is expressed by using the Ergative form for (postverbal) Agents in the
main clause; no Ergative marker is used, if the complement clause does not carry focus.
Tima distinguishes between direct speech (or quotatives) and reported speech (ver-
batim indirect speech or semi direct speech). The following examples, taken from a
trickster tale about Lion and Hyena, illustrate the first type:
(26) má=c��-w-áá l-�� má=dáh�� =wáy�nː
seq-arrive-ep-inst loc-home seq-say dat-father
pàpáŋ kwàná-�l�y=l� ú-kúún
daddy cow-1pl:incl –foc perv-bear
. . . And he reached home and said to his father: ‘Daddy, our cow gave birth’.
In the next example, the complement clause is introduced by the reported speech
marker yε. This clause, ‘it is mine that gave birth’, functions as an object of the main
verb (‘say’). Ergative marking occurs on the Agent noun phrase because the possessor
phrase in this clause carries focus.
(27) má=dáh-��=cíb�=n� �=k�ŋ�wúŋì
seq-say-ben erg-child=dem prep-hyaena
�= l��n�=l�=y�
 ú-kúún
prep-poss:1sg -foc-rep perv:past-bear
. . . And the child of the hyaena said (to him): ‘It is mine that gave birth’.
The clitic =yε (or its allomorph =ye) expresses the fact that the information following
such a complementizer verb is attributed to another speaker. The position of the reported
speech marker within a complex sentence is determined by the information structure
expressed in a complex sentence. When preceding the verb in the dependent clause, as in
example (27) above, a categorical statement is made. Categorical sentences contain a pre-
dicative base or topic entity about which some state of affairs is predicated, while thetic
sentences are statements whereby unitary information is given about the whole situation,
as argued by Sasse (1987: 511).When the reported speech marker occurs sentence-finally
in Tima, as in example (28) below, a corresponding thetic statement is made.
(28) cíbóó�nín=n� c�=ŋâh k�d�k à-mál=ŋùŋ=yè

girl-dem imperv:pres-see neck stat-beautiful-log3sg-rep
This girl considers herself to be beautiful.
It is within this context of the reported speech marker =yε that the third-person
logophoric pronouns =ŋuŋ(sg) or =ŋiŋ (pl) occur, as illustrated by example (28).
Logophoricity plays a role primarily within complex sentences. As pointed out by
Culy (1994), there is usually a hierarchy for complementizer verbs with respect
to the frequency of logophoric marking: 1. verbs of saying > 2. volitional verbs
> 3. perception verbs > 4. cognition verbs. But discourse units larger than the sentence
(i.e. paragraphs or episodes) potentially constitute the relevant domain for logophoric
marking, itself licensed by the reported speech marker, as in the following example:
(29) ŋ�ɲáŋ=l� kú-�ùúŋ.

work-foc agr-loc:log3sg
ŋk�=y� dììk-àà=ŋúŋ
cop-rep walk-inst-log3sg
(S)he has got work (to do). That’s why (s)he is going.
Through the presence of the reported speech marker as an epistemic validator, testi-
monial authority is assigned to another speaker. The use of contrasting pronominal
devices (logophoric marking as against disjunctive pronominal reference marking)
helps to reduce referential ambiguity in this respect.
(30) c�=dàh-� � =Kw�kw�ŋ c�=y�
imperv:pres-say-ht erg-Kwakwang imperv:pres -rep
�-dé-n-díy�ŋ=ŋùŋ
3-fut-3-walk.ven-log3sg
Kwakwang(i) says that she(i) will come.
(31) c�=dàh-� � =Kw�kw�ŋ c�=y� p��n�
imperv:pres-say-ht erg-Kwakwang imperv:pres-rep 3sg
�-dé-n-díy�ŋ
3-fut-3-walk.ven
Kwakwang(i) says that (s)he(j) will come.
Logophoric marking thus helps to reduce ambiguity in reference tracking for the
participants constituting the main pivot in an episode. This is particularly important
when information is attributed to another source. Dimmendaal (2001) makes the fol-
lowing observation in this respect:
Logophoric markers put important constraints on the search for relevance, when the speaker
presents his point of view concerning the mental state of others, e.g. in reporting on the mental
activities of a third person not participating in the speech event. In this sense, they are expo-
nents of the cognitive coding of epistemology, more specifically of evidential hedging strategies.
(Dimmendaal 2001: 26).
The reported speech marker introduces attributed discourse, and the logophoric
markers serve the purpose of marking the role of the reported speaker as distinct
from that of the narrator.5 Ameka (2004: 24) endorses this view, but adds ‘that the use
of these forms has a cultural basis and that a holistic explanation of logophoricity in
West Africa must take account of both the cognitive and the cultural factors’. Whether
this observation can be or should be extended to speech communities in the Nuba
Mountains or elsewhere in Central and East Africa where logophoricity is attested still
needs to be investigated.
Perception and cognition verbs also take a reportative speech marker when fol-
lowed by a sentential complement. But as argued next, these semantic classes of verbs
also play a role in another domain of the grammar of knowledge in Tima.
5 Sharing experiential knowledge: the expressive use of ideophones

In what has probably come to be one of the standard reference works on ideo-
phones, Voeltz and Kilian-Hatz (2001), various intricate and fascinating prop-
erties of these words reflecting sensory perceptions are discussed. But the role
they potentially play in the grammar of knowledge appears to be a somewhat
neglected property so far. Nevertheless, this is one of their functions in Tima.
Ideophonic words in Tima are mainly adverbial in nature, although there are
also ideophonic adjectives. Reduplication is one of the formal properties of ideo-
phonic adverbs (glossed as IDEO in the examples below), monosyllabic root
structures ending in a consonant being another formal property. Their role in the
grammar of knowledge becomes clear when looking at perception or cognition
verbs in Tima.
Schneider-Blum and Dimmendaal (2013) describe the syntactic and semantic
nature of cognition and perception verbs (the latter involving vision, audition, olfac-
tion, gestation, and tactition). The lexeme -ŋah ‘see’, for example, also covers the notion
of ‘noticing, thinking, conjecture’.
(32) �-ŋáh-�=ná cíd� k��=l�=y� �-t��

perv:past-see-ep-1sg.erg body family-foc-rep perv:past-pass
I noticed that someone passed by.
5
Probably the most detailed study to date showing how speakers use logophoric marking to distance
themselves from the discourse they are reproducing, and to signal the intrusion of another’s voice in their
own words, is to be found in Nikitina (2013) on the Mande language Wan.
By adding an ideophonic adverb to such verbs, the meaning becomes more specific,
that is, it is confined to a certain reading. Ideophones are thus used to restrict the
potential range of meanings, as in the following example, where the ideophonic adverb
cùk expresses insufficient visually obtained information:
(33) cùk=w� ŋáh-�dá w�r��máad�h
ideo-foc see-1sg man
I only saw the man from the corner of my eye.
The co-occurrence of the verb -ŋah with another ideophonic adverb, p�ŋk�lp�ŋk�l,
renders the notion of ‘looking around’:
(34) k�dáád� c�=ŋáh-ák p�ŋk�lp�ŋk�l
thief imperv:past-see-ap ideo
The thief looked around (repeatedly).
Non-visual sensory evidentials (in the typology of Aikhenvald 2004) play a role too in
this respect. For example, the lexeme -mɨn is semantically general over ‘hearing, listen-
ing’. By adding the ideophone h�dàh�dàk, a constructional meaning ‘overhearing’ or
‘eavesdropping’ emerges.
(35) �-m�n�t-�k-áá�áŋ-�=dà �àmáá=n� á=y-ìhìn�

imperv:past-listen-ap-inst:compl-ee-1sg talk-dem prep-ep-3pl
h�dàh�dàk
ideo
I overheard this conversation between them.
What is at stake when such ideophonic adverbs are used with perception verbs, for
example, is the joint perception. They express depictions of an event which are intend-
ed to invite the hearer ‘ . . . to “look” in such a way that we make believe we are actually
experiencing the scene depicted’, as Dingemanse (2011: 28) phrased it in his analysis
of ideophones in the Kwa (Niger-Congo) language Siwu. As the description of Siwu
shows, Tima is far from unique when it comes to the use of ideophones in this manner.
6 Some preliminary conclusions

Whereas cross-linguistically tense-aspect systems frequently play a role in the gram-
mar of knowledge, there is no evidence for this in Tima, as far as our present under-
standing of this language goes. Instead, construction-level effects are found, emerging
from the combination of verbs with other syntactic constituents.
‘The visual, or a combined visual and sensory, evidential tends to be less formally
marked than any other term’, as pointed out by Aikhenvald (2004: 368). The typo-
logically most striking feature involved in expressing knowledge in Tima, therefore,
is the formal marking of direct or firsthand (visual) evidence of the speaker (EGO) or
the protagonist in narrative discourse. When movement towards the speaker (or the
deictic centre) is involved, ventive marking occurs on the verb. Whether the formal
marking of EGO on constituents other than the verb, for example on prepositions, is
unique to Tima or not remains to be determined, as no data are available in this respect
(at least to the author). In addition, the grammar of knowledge in Tima is essentially
centred around verbs of cognition or perception, as shown through the marking of
logophoricity and the use of specific ideophonic adverbs in combination with these
verbs. Consequently, much of the grammar of knowledge in Tima is ‘construction-
based’ rather than being coded by way of morphological elements that are in paradig-
matic contrast with each other. The ‘scattered coding’ does not necessarily imply that
the system is less ‘grammaticalized’, as this also depends on one’s conceptualization
of grammar. Rather than having a strict division between lexical and syntactic opera-
tions, it is argued in studies such as Croft (2001) or Goldberg (2006) that construc-
tions in fact play a role from the smallest morphological unit all the way to complex
sentence structures. The grammar of knowledge in Tima would seem to support this
position.
References
Alamin, Suzan. 2012. The nominal and verbal morphology of Tima, a Niger-Congo language
spoken in the Nuba mountains. Cologne: Rüdiger Köppe.
———, Schneider-Blum, Gertrud, and Dimmendaal, Gerrit J. 2012. ‘Finding your way in Tima’,
pp. 9–33 of Directionality in grammar and discourse: Case studies from Africa, edited by Ange-
lika Mietzner and Ulrike Claudi. Cologne: Rüdiger Köppe.
Ameka, F. K. 2004. ‘Grammar and cultural practices: The grammaticalization of triadic com-
munication in West African languages’, The Journal of West African Languages 30(2): 5–28.
Bashir, Abeer. 2010. ‘A Phonetic and Phonological Study of the Tima Language’. PhD thesis,
University of Khartoum, Sudan.
Blackings, Mairi and Fabb, Nigel. 2004. A grammar of Ma’di. Berlin: Mouton de Gruyter.
Bybee, Joan, Perkins, Revere, and Pagliuca, William. 1994. Evolution of grammar. Tense, aspect,
and modality in the languages of the world. Chicago: University of Chicago Press.
Culy, Christopher. 1994. ‘Aspects of logophoric marking’, Linguistics 32: 1055–99.
Croft, William. 2001. Radical construction grammar: Syntactic theory in typological perspective.
Dimmendaal, Gerrit J. 2001. ‘Logophoric marking and represented speech in African languages
as evidential hedging strategies’, Australian Journal of Linguistic 21(1): 131–57.
———. 2009. ‘Tima’, pp. 338–55 of Coding participant marking: Construction types in twelve
African languages, edited by Gerrit J. Dimmendaal. Amsterdam and Philadelphia: John
Benjamins.
———. 2011. Historical linguistics and the comparative study of African languages. Amsterdam
and Philadelphia: John Benjamins.
———. 2013. ‘Where have all the noun classes gone in Tima?’, pp. 103–25 of In and out of
Africa. Languages in questions in honour of Robert Nicolai, Vol. 2: Contact and language
change in Africa, edited by Carole Féral, Maarten Kossmann, and Mauro Tosco. Louvain-la-
Neuve : Peeters.
Dingemanse, Mark. 2011. ‘The Meaning and Use of Ideophones in Siwu’. PhD thesis, Radboud
University, Nijmegen.
Frazjyngier, Zygmunt (ed.). 2012. The Afroasiatic languages. Cambridge: Cambridge University
Press.
Goldberg, Adele E. 2006. Constructions at work: The nature of generalization in language.
Greenberg, Joseph H. 1963. The languages of Africa. Bloomington and The Hague: Indiana Uni-
versity Press, Research Center in Anthropology, Folklore and Linguistics and Mouton.
Hieda, Osamu. 2012. ‘Complementation and evidential strategy in Acooli’ , Studies in Nilotic
Linguistics 5: 73–95.
Levinson, Stephen. 2000. Presumptive meanings: The theory of generalized conversational impli-
cature. Cambridge. MA: The MIT Press.
Miller, Cynthia and Gilley, Leoma. 2007. ‘Evidentiality and mirativity in Shilluk’, pp. 191–206
of Advances in Nilo-Saharan linguistics, edited by Doris Payne and Mechthild Reh. Cologne:
Rüdiger Köppe.
Newman, Paul. 2000. The Hausa language: An encyclopaedic grammar. Yale: Yale University
Press.
Nikitina, Tatiana. 2013. ‘Logophoric discourse and first person reporting in Wan (West Africa)’,
Anthropological Linguistics 54: 280–301.
Sasse, Hans-Jürgen. 1987. ‘The thetic/categorical distinction revisited’, Linguistics 25: 511–80.
Schadeberg, Thilo C. 1981a. A Survey of Kordofanian, 1: The Heiban group. Hamburg: Helmut
Buske.
———. 1981b. A survey of Kordofanian, 2: The Talodi group. Hamburg: Helmut Buske.
Schneider-Blum, Gertrud. To appear. ‘Pronouns in Tima’, in Proceedings of the First Nuba Moun-
tain Languages Conference, edited by Thilo C. Schadeberg and Roger Blench.
——— and Dimmendaal, Gerrit J. 2013. ‘Excite your senses: Glances into the field of percep-
tion and cognition in Tima’, pp. 217–50 of Perception and cognition, edited by Alexandra Y.
Aikhenvald and Anne Storch. Leiden: Brill.
Storch, Anne. 2006. ‘Aspect and evidentiality in Luwo’ , pp. 393–402 of Insights into Nilo-
Saharan language, history and culture, edited by Al-Amin Abu-Manga, Leoma Gilley, and
Anne Storch. Cologne: Rüdiger Köppe.
Talmy, Leonard. 1985. ‘Lexicalization patterns: Semantic structure in lexical forms’ , pp. 36–149
of Language typology and syntactic description, Vol. 3, edited by Timothy Shopen. Cambridge:
Cambridge University Press.
Tucker, A. N. and Bryan, Margaret A. 1956. The non-Bantu languages of Northeastern Africa.
London: Oxford University Press for the International Institute.
Voeltz, F. K. Erhard and Kilian-Hatz, Christa. 2001. Editors of Ideophones. Amsterdam and
Philadelphia: John Benjamins.
13
Saying, seeing, and knowing among

the Karawari of Papua New Guinea
B O RU T T E L B A N
Karawari (called anduk apianak ‘our mouth’ by its over 3,000 speakers) is spoken
in eight main villages and countless bush camps in the area of the lower Kara-
wari River, lower Konmei Creek, and upper Kangramai Creek in the Sepik River
basin of Papua New Guinea. Karawari people and other river dwellers, and their
languages, can also be called yakwaym ‘water people’. They further differentiate
themselves according to the word ‘no’ (kaya, kayak, or kayang). While Ambonwari
would say apia kaya ‘we (are people who say) kaya’ they would say about some other
Karawari people (from Konmei, Kundiman, and Manjamai), and the Yimas, mbu
kayak ‘they (are people who say) kayak’. The people of Middle Sepik would then
be grouped under the term kayang. Being one of the languages of the Lower Sepik
Family (belonging to a large group of New Guinean languages known as Papuan
or non-Austronesian) Karawari shows considerable similarities with the closely
related Yimas language (Foley 1991) although the two languages are not mutually
intelligible. I conducted ethnographic research, which began in 1990, mainly in
Ambonwari, which has over 750 inhabitants and is the largest Karawari-speaking
village.
Karawari has two major word classes, noun and verb. Nouns are divided into eight
major noun classes. The inclusion into a particular noun class depends on both seman-
tic and phonological criteria. Some nominal classes distinguish three numbers (singu-
lar, dual, and plural) and others only two (singular and plural). A few nominal classes
have paucal (marking number between three and five). Like Yimas (Foley 1991: 3),
Karawari is an agglutinative polysynthetic language, with verbs being the most mor-
phologically complex class with many prefixes and suffixes (throughout the present
text the individual verbs will be written in the present tense and not in the form of
roots or stems). Reduplication and serial verb construction are common, with usually
two or three verbs describing a single action.
13 Saying, seeing, and knowing in Karawari 261
(1) awa mɨn yakurapi-am-amɨng-ka-r yuwan

cassowary 3sg collect-redup-eat-stay-immed.past tree
sɨsɨnɨng-gi saman
seed-pl here
Cassowary has collected and eaten the seeds of a tree here.
Roots and stems (for example, am- and amɨng- ‘eat’) represent morphological bases
on which, by the use of a large number of prefixes and suffixes, the words and clauses
are built. It often happens that there is no immediate strict division between a verb,
noun, and adjective, and the same word (appropriately inflected for tense or gen-
der, for instance), can be found in all three word classes: a verb in the present tense
(angguringa-n ‘open eyes’), a noun in masculine grammatical form (angguringa-n
‘man with a light skin, white man’), and an adjective inflected for certain nominal
classes (angguringa-n ‘bright, light, white’). Lexical bases are highly productive and
are extensively used in everyday speech. Note that repetitions, reduplications, and
parallelisms form a significant part of talking, story-telling, and singing. Therefore, a
statement, an observation, a mourning song or any verbal act for that matter can be
made by using first a clause with a verb form, while in its repetitive or parallel line it
appears in a nominalized form (see examples 11 and 12 below). Not all roots and stems
can be used in this way. While people might say about a word or a clause artificially
made up by the ethnographer that they are possible, they would immediately add that
the whole expression does not really sound good.
The Karawari people, like those of other Sepik societies (Bateson 1958; Harrison
1990; Wassmann 1991; Aikhenvald 2008a), have a large repertoire of ‘proper nouns’
or names of people, spirits, and places, which represent a very significant part of their
understanding of the world and knowledge generally. Wi ‘name’, however, does not
refer only to ‘proper nouns’ but to beings and things generally. So, words for individual
animals, plants, things, and foods, as well as generic terms like ming ‘bird’, yuwan ‘tree,
shrub, plant’, saki ‘bush spirit’, or amɨng ‘food, meat, animal’ are also ‘names’. The rest
is mariawk ‘talk, story, speech’.
There are no grammatical evidentials in the Karawari language. Like the related
Yimas (Foley 1991) and the unrelated Manambu (Aikhenvald 2008a) they do not have
this grammatical category. Foley (1991: 112–13) reports that in Yimas the invisible/
visible contrast is found in the copula, a highly complex verb form with many irregu-
larities. It is only found in the singular and only with some classes. Foley (1991: 113)
does not refer here to evidentiality but to: (a) identificational use of the copula when
someone is showing something and identifying it at the same time, that is, the object is
seen by the speaker (visible form); (b) predicational use of the copula when describing
things unseen by the speaker (invisible form). I checked for the possible existence of
such invisible/visible contrast in the Karawari language while still in the field. I could
only conclude that Ambonwari rarely use a copula, though it is found in one of the
262 Borut Telban
most important and intriguing Karawari verbs aykapɨkan ‘remember, think, know’.
While grammatical expression of information source may be lacking there are other
means of referring to the source of information and expressing the nuances of perceiv-
ing and knowing. Aikhenvald (2004: 20) calls these evidential extensions ‘evidentiality
strategies’.
The aim of this chapter is to explore not only what the Ambonwari find important to
know but how do they know, or better how they do come to know (see Marchand 2010:
S7). During more than three years of fieldwork in Ambonwari and five in Papua New
Guinea, I often saw how certain practices are learned without words: carving, sharpen-
ing, cutting, sawing, and one could say all the knowledge that requires practical skills.
They are accompanied only by short interjections such as yaw ‘yes’, mba ‘enough, that’s
it’, yandamba ‘enough of this, in this way’ or mbayaw ‘all right, O.K.’, mɨndɨn ‘like that’
or mba mɨndɨn ‘that’s it, like that’, samɨndɨn ‘so, in this way’, kambandɨn ‘not like that,
not so’. All these observational expressions are part of those practices which come
under the concept of kay ‘way of doing things, habit, ritual, being’, where knowing is
doing, and where the level of knowing is recognized by the mastery of the skill. Kay
in itself does not reflect upon itself but requires wambung ‘insideness, understanding’,
which is capable of reflecting upon both kay and mariawk ‘act of speaking, speech, talk,
utterance, story, thought’ (Telban 1998). It is this last concept that I will focus on in the
first part of the chapter. I will address interdependency of knowledge, pan mariawk
‘very thoughtful speech’, and ignorance, kambra mariawk ‘empty speech, just talk’ and
how they are related to evidential strategies. I will also show how the Ambonwari are
very much aware of tangible consequences of speaking and how words can actually
‘touch’. This will lead me to discussion about issues of perception, seeing and hearing
in particular, and their relation to knowledge generally.
1 Speaking
For speaking one needs first of all a suitable anduk ‘mouth, language’. What do people
think about the difference between the local vernacular and the lingua franca Tok
Pisin? First, they say that the vernacular is anduk kwanggiak ‘long language’ and Tok
Pisin is anduk pambiak ‘short language’. The noun phrase ‘long language’ (not mean-
ing a long speech) refers to Karawari in which one can express oneself in more detail,
with a larger variety of words, inflections, and clauses and in a plurality of ways. Anduk
kwanggiak is used also for a person who talks strong and clear (and not necessarily a
lot). The noun phrase ‘short language’ refers to Tok Pisin in which one is not always
able to fully and clearly express oneself. Anduk pambiak is also used for a person who
mumbles or speaks too quietly. People also say that local vernacular is imɨnggan awi
‘fire of the village’, when people do not only argue and cross each other but also play,
joke, and laugh. Knowing anduk ‘language’, however, is not enough to make a person
knowledgeable. It is important what one says, how one says it, and when one says it,
that is, the employment of language and the proficiency in mariawk ‘speech, story,
thought’. So, when a person too often uses mariawk maman ‘bad talk: swearing, curs-
ing, abusing’ someone else will say:
(2) mi mba mi-nya-n ya-n su-ngor anduk minya-k-ngɨna
2sg enough2sg-poss-obl A-O talk-imper mouth your-Vsg-obl
Enough! You talk only with your mouth (you don’t think).
In everyday life, when making a speech is closer to creation than just to communi-
cation, talking represents an extremely important part of people’s relationship with
their environment. It is regarded not just as a speech but as a creative act, a speech act
that is situated in temporal, spatial, social, and cultural context. There are four verbs
in Karawari that can be translated as ‘talk’: mariawkusɨkan ‘talk, speak, narrate, tell
a story; sing (bird)’, mariawk sukwan ‘talk out, speak out, divulge, gossip’, sɨmɨnggan
‘talk clearly, tell, explain’, and yaykan meaning either ‘weep, cry’ when intransitive or
‘talk to, say to’ when transitive. Of all four verbs, only yaykan is used in speech reports.
1.1 Mariawkusɨkan ‘talk, speak’

In the first verb mariawkusɨkan ‘talk, speak’, the noun mariawk ‘talk, speech’ merges in
a compound with the verb (w)usɨkan, which is, in the dialect spoken in the village of
Masandanai, clearly pronounced as wusɨkan. This latter verb has several interrelated
meanings: (a) blow, make sound by blowing; (b) stir, spin, turn over; and (c) pour in,
put inside.
(3) mɨn ya-n sɨnggan mariawkus-ɨkan
3sg A-O together talk-pres
He talks with him.
Mariawkusɨkan is a generic term for talking, so when people are sitting in silence
(which in Ambonwari is not acceptable unless there is tension between people or
excessive embarrassment makes people silent) somebody will say: Mariawkusɨra!
‘Speak!’ A serial verb construction with the verb sarikan ‘grab, hold tight’ changes the
verb ‘speak’ into ‘confess’:
(4) mɨn ya-n sar-mariawkus-ɨr mɨn-ang kay mama-n
3sg A-O grab-speak-immed.past 3sg-dat way bad-IIIsg
He has confessed his sin.
In this example kay maman ‘bad habit, bad way, bad practice’ is a local translation
of sin or wrongs that someone did. In the Ku Waru area of the Western High-
lands, people use the expression ‘neutralize bad talk’ to translate the verb ‘confess’
(Rumsey 2008: 457), that is, neutralizing talk about somebody’s bad and hidden
practice. In the Karawari area to ‘confess’ is expressed by ‘grab bad practice and
speak about it’.
264 Borut Telban
1.2 Mariawk sukwan ‘speak out, tell’

The second verb used with mariawk is sukwan:
(5) mi kambra mariawk mi su-kwan
2sg nothing talk 2sg tell-pres
You are talking nonsense (you are telling lies).
(6) mɨn wasa mariawk ya-n su-r
3sg small talk A-O tell-immed.past
He has divulged the secret.
When alone, the verb sukwan means: (a) ‘kill, trash, beat’; and (b) ‘bite, sting’. Together
with mariawk it refers to ‘speak out, tell, divulge, gossip’. Gossip, rumours, hearsay stories,
and talk behind someone’s back are common features in small-scale egalitarian societies
of Papua New Guinea and malicious gossip has often been associated with sorcery (Young
1971: 135; Brison 1992: 116–20). Schieffelin (2008: 436), however, writes that speculative
talk or gossip among the Bosavi is something one does not have any right to say, rather
than being an evaluation of somebody, or talking about them in negative terms. For Kara-
wari, I would say, it covers all these situations and fully depends on the context and those
who are present. That gossip ‘touches’ one’s name holds true for both the Bosavi and the
Ambonwari, and any kind of speculative talk attached to one’s name makes them angry.
(7) mbu wi i-mbu sari-kan mɨn-aki
3pl name A-O grab/hold-pres 3sg-VIIsg
They are calling his name.
There can be other reasons for ‘talking behind someone’s back’ such as, for example,
feelings of being wronged, which occur on a daily basis. One can hear people saying
amɨndɨ wamɨri mbu kuran ‘they are mumbling behind our back about food’ (which we
did not give them when they came to our house), kurar wamɨri mbu kuran ‘they are
mumbling behind our back that they too wanted to go’ (and we did not take them with
us in our canoe when we went to town), mamanggar wamɨri mɨn kuran ‘he is talking
behind our back about paddling’ (as we did not call him when we went to check the
fishing net on the river). The expression for talking behind someone’s back, wamɨri
‘mumble, gossip’, is used with the verb kura-n ‘beat, strike, hit’.
Variations of mariawk sukwan, either in an SVC or with a prefix to the verb, are used
for gossip, rumour, and talk behind someone’s back.
(8) ipa mariawk ya-nan sari-su-kwan
2pl speech A-O grab/hold-talk-pres
You slander him (you talk ill of him).
Here a SVC includes the verb sarikan ‘hold tight, grab’, which is the main verb also in
example (7) above when it refers to the calling of a name. We can notice that tangible
aspects of speaking are explicitly present in all these examples.
Let me illustrate a culturally specific way of talking in parables when someone is dis-
satisfied or angry with somebody else. Such a person will often not take any direct step
(start a fight, for instance) but will in an indirect and figurative way talk loudly either
inside or outside of their own house and will semi-directly address some other person.
For example, there are seven houses in the area of the Bird of Paradise Clan and the
man X, who got his wife from this clan, is angry with the woman A, who married a man
from this clan. She apparently spread around that he is greedy and does not share things
which he buys in town. The angry man X does not come close to A’s house and does not
call her name but starts near his own house to call the names of women B and C, who are
A’s classificatory sisters married into two nearby houses (male members of these houses
call A’s husband either brother or father). Two features characterize his verbal expression
of anger: first, by calling other names, that is, the names of women B and C, their names
become illustrative and analogous of the culprit whom X believes is the woman A. He
uses tok piksa (‘picture talk’ in Tok Pisin), or talks, as people today explain, in parables.
(9) ya-mban sɨnggan mi-nyang suwasa-su-kwan
A-O together 2sg-dat pref-talk-pres
I am talking to you in parables.
Second, by using figurative speech, or tok antap (‘talk on the top’ in Tok Pisin), X
does not hide or obscure things but tells them in a very straightforward way. The
two women B and C, whose names were called in public, are then expected to go to
woman A, tell her that X expressed his anger in parables, and ask her what had actu-
ally happened. Together they either start a dispute or make a plan for how to comfort
the offended man.
There are several noun phrases which refer to the kind of talk mentioned in the
above story:
(a) sari-su-r mariawk
grab/hold-talk-nomz speech
Gossip (Lit. ‘grab-speaking.out speech’), tok baksait ‘talk behind someone’s
back’ in Tok Pisin.
(b) wapay-su-r mariawk
climb-talk-nomz speech
Parable, figurative speech (Lit. ‘climb-speaking.out speech’); in Tok Pisin this kind of
talk is called tok bokis ‘box talk’, tok piksa ‘picture talk’, or tok antap ‘talk on the surface’.
(c) kapak-ɨr mariawk
be.angry/scold-nomz speech
Quarrel, dispute (scolding speech).
(d) suwa-siria-r mariawk
pref-dance-nomz speech
Rumour, talk without proof, talk that ‘dances’, tok win ‘wind talk’ in Tok Pisin.
266 Borut Telban
Talk in parables, (b), can be joyful and entertaining but also despised when it is equated
with hidden talk, when a person tries to conceal the truth. Just as ‘climbing’ in figura-
tive speech should be understood as climbing on one’s arɨm ‘skin’, others of the above-
mentioned expressions are also closely associated with tactile experiences. When
somebody talks in an uncontrolled manner others will use figurative speech to describe
his way of jumping from subject to subject and not getting to the heart of the problem.
(10) sanggwa-ra sɨmɨn-ia ma-n akrisay-mbɨn ya-r
look-imp rattan-pl A-O pull.out-seq get-immed.past
Look at him, he pulled the rattans out (from the bush) and has got them.
In the example (d) suwasiria refers to ‘movement of something is indicating move-
ment of something else’: when somebody walks through the forest and we do not see
the person but only the movement of branches and leaves. It is similar with talking:
we do not hear the actual person saying something but hear only rumour about what
he or she said.
The Ambonwari already know all those who are prone to spread the word or gossip
around the village. Jocelyn was often telling what people said about a possible marriage
between her brother and a woman from the upper part of the village. Her brother got
tired of constant speculations and angry with her for ‘carrying’ this kind of stories
around all the time. He said:
(11) Jocelyn mi ya-n-ma pay-pia-r mariawk
Jocelyn 2sg A-O-upriver carry-dir-immed.past story
Jocelyn, you’ve carried the story downriver (from upriver down).
(12) mi mariawk mi pan pa-mbay-nja-r-ma
2sg story 2sg very redup-carry-cont-nomz-fem
You are the woman who carries stories (gossips) around all the time.
1.3 Sɨmɨnggan ‘tell clearly, explain’

The third verb expressing ‘talk’ is sɨmɨnggan ‘tell, talk in a clear and direct way, explain’.
(13) panbi mɨn sɨmɨng-gan
true 3sg talk-pres
He is telling the truth.
Adverb panbi ‘very so’ or pan mariawk ‘very speech’ are expressions for truth and agree-
ment. Truth actually is not something absolute but depends on those who are present
and their consensus. It often happens, especially during situations when somebody
has been wronged or has wronged someone, that different groups of people, according
to their kin relation to the person in question, confirm different truths. Plurality of
competing truths is therefore a common outcome especially when social relationships,
misfortunes, and people’s access to resources are in focus. Equally, knowledge as such
is not absolute but always in the making: questioned, challenged, and won.
1.4 Yaykan ‘cry, say, speak’

Only the transitive verb yaykan is used in direct and indirect speech reports either with
a particle bɨni ‘thus, so’ or without it. Indirect speech reports can contain the speech
report introducer bɨni.
Direct speech report:
(14) ama pɨ-ka ya-r-a (bɨni) mi karis amɨn-da
1sg A-O talk-immed.past-3sg (so) 2sg sago.pudding eat-imp
I told him: ‘Eat sago pudding!’
Indirect speech report:
(15) ama pɨ-ka ya-r-a (bɨni) mɨn karis ka-n-a-mba
1sg A-O talk-immed.past-3sg (so) 3sg sago.pudding imp-IIIsg-eat-imp
I told him that he should eat sago pudding.
In both examples a person can omit transitive verb yaykan and begin the sentence only
with ama bɨni ‘I thus, I so’ followed by direct or indirect speech.
When just thinking about something, and not actually saying it, this can also be
expressed by direct speech without the verb yaykan. This is the way of expressing
the speaker’s intention or desire (see Aikhenvald 2008b: 391 for Manambu and other
Papuan and Austronesian languages). So, the expression ama bɨni arin ama pandarin
‘I think I would like to process sago tomorrow’, can be said without the particle bɨni:
ama arin pandarin ‘I would like to process sago tomorrow’. In this case my intention
has already become a confirmed decision.
In the legend about the woman called Pingginsɨmbukmay, audio recorded by Dan-
iela Vávrová on 29 May 2011, the legendary woman was not given any meat by her
husband. Feeling offended, she went to see her father:
(16) apasɨ-mbɨn mɨn-ma kura-kia-kɨ-pia-k bɨni ama anay
go.outside-seq 3sg-upriver go-night-irr-dir-rem.past so 1sg father
ama anja anga-ra kawi karis ya-ka-ma awsa-r
1sg A-O give-imp fish sago pudding A-O-upriver put-immed.past
kambra-n wusɨ-mbɨn
nothing-IIIsg stir-seq
She came out and went up the village at night and (said) so: ‘It’s me. Father, give me
fish! I stirred plain sago pudding and put it (there).’
The Ambonwari are inclined to put words into other people’s mouths and in this way to
(re)create their intentions. This is especially the case when parents talk about thoughts,
feelings, and wishes of small children. Their customary songs are full of reported
speech where other people’s aims, desires, and feelings are either imagined or imposed
on them by the composer of the verses. It is therefore thought that direct quotation
is more an expression of opinion or presupposition (anxious, desired, assumed, false,
268 Borut Telban
or exaggerated) of the reporter than of the original speaker or the person involved in
the event. In short, as Aikhenvald argues in Chapter 1 of this volume, in the section
on speech report constructions, direct quotations can contain overtones of doubt and
lack reliability.
Self-quotation can be used with the aim of lifting the speaker’s importance and can
equally be doubted by some if not all the listeners. This depends on who the speaker
is and what the context is. It may also express the speaker’s support of certain people,
opinion, or position.
When an old man Lawrence Manjawe is hungry he does not go to his children’s
houses but to the house of his sister, where he is not ashamed of asking for food.
(17) mbu-nang pɨ-ka yay-ngor bɨni ama sipi-n-gwarɨng
3pl-dat A-O talk-imper thus 1sg sago.pancake-obl-piece:IIIsg
an-ja anga-ra
A-O give-imp
I tell them thus: ‘Give me a piece of sago pancake!’
Self-citation is used to explain or justify why somebody did what they did. In the fol-
lowing text Francis explains his movement when he went fishing in a canoe. He says
first that he wanted to smoke, so he decided to find the fire.
(18) mba ama kwasa-kia-ndukun bɨni Pita mɨ-na-n
enough 1sg get up-night-rem.past so Peter 3sg-poss-Isg
kra-r imɨngga-n awi mi-ka sa-n yaki-n imbrum
cut-nomz place-obl fire A-O be-pres tobacco-obl leaf
bɨ-nang ama ambia-kɨr
3sg-dat 1sg light-fut
Thus I set out (going my way) thinking: ‘Fire is at Peter’s garden. I will light a
cigarette there.’
2 The creation of knowledge

I did my first long-term research in Ambonwari between 1990 and 1992 and since
then spent extended periods in the village every few years. The last long-term peri-
od of fieldwork lasted for ten months between February and December 2011 when
I joined Daniela Vávrová who had arrived in the village in December 2010 and was
carrying out her own audio-visual anthropological research.1 One thing that became
1
Over twenty years of working with Ambonwari people an amount of recorded, transcribed and trans-
lated material has accumulated. This material includes a large number of myths, legends, and personal
life-histories, several thousands lines of all initiation and other celebratory songs which used to be sung
throughout the all-night singing and dancing events, and ‘modern’ songs, either Christian or those sung by
the village string-bands, when they were still fashionable. The first texts in Karawari were either recorded
on audio cassettes, and later transcribed and translated into Tok Pisin, or written straight into a note book.
Since 2005 when Daniela Vávrová began her own audio-visual research in Ambonwari, the number of video
recordings of daily events, casual discussions, public meetings, crosses, mourning periods, celebrations,
immediately obvious during our last stay was that people—including us—constantly
complained about the unreliability of information and untrustworthiness of its source.
This was far from the situation twenty years ago. Then, people listened to the older men
and women and did not jump to unsupported conclusions. The important knowledge
of the time was more embedded in and based on the internal knowledge of the past
than displaced and based on external hearsay and presumption about the present. The
pace of life was slower—less long distance travel and preoccupation with money—and
new technologies like wireless and mobile phones were not even dreamt of. In 2011,
the questions about who told you this or that, or who saw something, or how does
somebody know something, became weekly if not daily obsessions. Speculations were
constant. Some were short lasting, innocent, and not important.
When, for instance, children and some adults heard the sound of the outboard
motor they loudly called the names of those they believed were coming in a canoe.
Usually, they were wrong. Wireless and mobile phones provided another context. Sud-
denly long distance information became anticipated regardless of the fact that there
was no wireless or mobile phone connection in the vicinity of the village. There were
rumors that Digicel would build a tower—where else than straight in or near Ambon-
wari village—but these talks were never confirmed by any authority. The possibility
of having a Digicel tower produced an explosion of information that brought confu-
sion to some and amusement to others. Rustle and buzz in the phones without actual
connection provided space for imagination springing from the cultural environment,
which although Catholic in denomination continues to question the whereabouts and
nature of spirits of the dead, and the forest spirits.
One obvious consequence of new communication and information technology—
wireless and mobile phones in particular—is that it is not the source of information but
information per se that expands and swells. In other words, as people say, it receives its
‘extra’ from every intermediary messenger carrying the information (see earlier exam-
ples (11) and (12)). These messengers freely add their own assumptions, interpretations,
and conclusions to the actual facts. In this way they create their own position in the
whole event, their own voice, and build up their own importance if their postulations
subsequently prove right. However, if their assumptions prove wrong, they may well
and interviews accumulated. Just in 2011 Daniela recorded one hundred one-hour video cassettes and over
twenty hours of audio material, most of it in the Karawari language, all of which became a welcome sup-
plement to my own recordings. Transcriptions and translations into Tok Pisin, which were made together
with local elementary teacher Julius Sungulmari during this last fieldwork, amounted to over two thousand
pages. In regard to this chapter I would like to thank the Research Centre of the Slovenian Academy of Sci-
ences and Arts for continuous support, the Australian National University which assisted me not only when
I was a PhD student there but also later on during my many affiliations, the National Research Institute of
Papua New Guinea for providing me with a research visa, and the Firebird Foundation for Anthropological
Research for a grant which secured the proper recording equipment for our last long-term fieldwork in 2011.
Special thanks to Alexandra Aikhenvald and R. M. W. Dixon for inviting me to the workshop ‘The grammar
of knowledge’ at the Language and Culture Research Centre, James Cook University, and providing me with
comments on an earlier draft of this chapter.
270 Borut Telban
be labelled kambra mariawkusɨr-ar/-ma or kambra mariawk sur-ar/-ma ‘liar, deceiver,

man/woman who talks nonsense’. They can be looked down or despised as those who
are inclined to tell kambra mariawk ‘nothing talk, meaningless talk, empty talk, just
talk, lies’ and are not to be trusted. The term which is important here is not mariawk
‘talk, speech’ but kambra ‘nothing, just’. As I discussed the etymology and wider mean-
ings of kambra in Telban (1997: 315), let me here just note that kambra ‘nothing’ refers
to something or somebody without substance: kambra arɨm ‘dead body, old carving’
(Lit. ‘nothing skin’), kambra yam ‘empty house’, kambra yukum ‘empty basket’, kambra
karis ‘plain sago pudding without meat or fish’, kambra kapɨs ‘naked’ (Lit. ‘nothing
buttocks’), kambra kanapang ‘bald head’ (Lit. ‘nothing head’).
When recognizing that talk is still talk even when it is ‘nothing talk’, then a differ-
ent ‘logic’ becomes attached to it. For example, when people are calling the names of
those whom they don’t actually see but expect to come, they are actually making them
come. They are helping them to return. In such a case, kambra mariawk ‘nothing talk’
reflects people’s participation in the event of a return. From such a perspective, kambra
mariawk ‘nothing talk’ is simultaneously very much its own opposite, or at least it is on
its way—a possibility—to become: pan mariawk ‘very talk’. On the other hand, kambra
mariawk is also used in the sense of ‘no worries’, ‘it’s all right’, whenever one wants to
comfort another person and their deeds. Taking all these dimensions into account we
could say that in kambra mariawk ‘nothing talk’ we can detect morality, sociality, and
rituality in terms of both reflection and creation.
What is considered knowledge among the Karawari people? Twenty years ago
the answer would be: (a) names of people, spirits, and places to which legends and
other stories of importance from the past are attached and which organize people
in their social organization; (b) the steps and procedures of rituals and other cer-
emonies including all-night songs and ritual texts; and also (c) practical daily knowl-
edge available to but not mastered to the same degree by everyone (processing sago,
carving canoes, building houses, cooking, socialization that includes sharing and gift
giving). A common way of expressing practical knowledge is by nominalization and
then g ender-marking of any verb to characterize a person by his or her capability. For
example, if somebody carves a good hand-drum people say:
(19) mɨn wanggɨn-di pan karing-ar-ar
3sg hand.drum-pl very carve-nomz-masc
He is very much the carver of hand-drums (he knows how to make hand-drums).
In 2011, the general answer to the question about the most important knowledge which
concerned the Ambonwari included the complexities of rubber business and the ways
of finding money, completing school up to and beyond grade 10, mastering know-
how in church practices and in the Catholic charismatic movement, and bringing a
mobile phone connection to the village. However, what became apparent from people’s
accounts was that perspective on knowledge did not change very much: knowledge
continues to merge abstract with practical dimensions, and cognitive concepts with
bodily know-how. In other words: just talking is not enough; it has to materialize in
the visible world. People can say about anyone, child, woman, or adult man:
(20) mɨn pan wambung ngandɨkɨn
3sg very insideness with
She/he is a person with knowledge.
The above observation would be made also for a person who helps, who cares, who
gives, who mourns, who does things in culturally, socially, and morally expected ways.
In short, knowledge and affectivity are both expressed with wambung ‘insideness, cen-
tre of thoughts and feelings’. When noun phrases and clauses include wambung with
different verbs of movement (become, run away, jump, come out, arise, etc.) they relate
to different emotions, such as being angry or scared, happy or sad, excited or shocked,
and so on (see Telban 1998: 56–65).
3 Aykapіkan ‘remember, learn, reflect, know’

Ambonwari use two verbs to express knowing: sɨkan ‘feel, do, become’ and aykapɨkan
‘remember’. The first verb focuses more on feelings and emotions, and when with
wambung it captures the feelings of worry, care, sorrow, anxiety, and concern for
somebody or something. When used with other nouns it reflects obvious bodily
experiences such as pain, cold, fear, sadness, hunger, shame, and so on. Aykapɨkɨn
and aykapɨkɨnma refer to ‘knowledgeable or/and skilful man’ and ‘knowledgeable or/
and skilful woman’, respectively. The verb ayk-apɨkan (‘VIS-put.inside’) is an unusual
SVC with the first verb being the copula, a truly irregular verb (as in Yimas, Foley
1991: 226), and with the irrealis suffix -k, telling that remembering is not bound in
time. Therefore, the reference to visibility, to the relationship between visible things
and memory, shows that remembering does not come from inside but from outside.
This is also well reflected in practices when, for example, people burn and destroy
the belongings of their recently deceased sons and avoid the places which would
remind them of their boys. Negation of the intransitive verb form aykapɨkan refers
to not remembering, which is closer to not knowing (example (22)), and negation of
a transitive verb construction refers to forgetting something (example (23)). There is
no separate term for forgetting. It is external arɨm ‘skin’ that triggers associations to
the past and not wambung ‘insideness’. The verb aykapɨkan can be used either with
wambung ‘insideness’ (referring to understanding) or without it (referring to remem-
bering and knowing).
(21) wambung mɨn-an-a wara mɨn aykapɨ-kay-kan

insideness 3sg-poss-3sg not 3sg remember-stay-pres
He/she does not understand.
272 Borut Telban
(22) mɨn wara mɨn aykapɨ-r

3sg neg 3sg remember-immed.past
He/she does not remember (he doesn’t know).
(23) mɨn wara ya-n aykapɨ-r

3sg neg A-O remember-immed.past
He forgot it.
(24) ama kanggɨnɨng ya-ka say-aykapɨ-r

1sg something A-O hold-remember-immed.past
I know how to use something.
(25) mɨn kay i-n karing-aykapɨ-kan

3sg canoe A-O carve-remember-pres
He is learning to make a canoe.
Daniela and the village teacher, Julias, organized making a collage with pictures from
magazines and newspapers around the general theme of wambung. Julias explained
to elementary school children how to choose the pictures according to what they like:
(26) mi wambung minya-n-a aykapɨ-mbɨn bɨni

2sg insideness your-IIIsg-obl think-seq so
manda ama ma-ka may-aykapɨ -kan
this:IVsg 1sg A-O all-think-pres
Think first in your insideness so (that you can say): ‘I know about all this.’
(27) manda wara olsem ma-pora-mbuna piksa bɨni yaw

this:IVsg neg thus other-place-obl picture so hey
This is not a picture from some other place, so (that you would think): ‘Alas!’
(28) mba mi mba olsem mi-nyan bus-na o

enough 2sg enough thus 2sg-poss bush-obl or
mi-nyan imɨngga-n si-n sanggwa-ngor-a
2sg-poss village-obl A-O see-imper-3sg
That’s it. You (know) enough. Thus, it is from your bush or from your village,
you see it all the time.
(29) bɨni ama ya-ka may-aykapɨ-kan yanda samandɨkɨn

so 1sg A-O all-remember-pres this same
Thus (you think): ‘I remember all this. It is the same (like the one I know).’
(30) olsem ama imbian bɨni andɨ-ka sɨmɨng-gambi

thus 1sg pig so pot-1sg talk-pot
ama wambung amana-n-a andɨ-ka sɨmɨng-gambi
1sg insideness my-poss:IIIsg-obl pot-1sg talk-pot
sambɨs ama-na-kia-na ya-ka sanggwa-ndukun
eyes 1sg-poss-IIIpl-obl A-O see-rem.past
Thus, if I say ‘This is a pig,’ in my thoughts I should say: ‘I saw it with my eyes.’
As we can notice from the above examples knowing is equated with visual recognition.
4 Hearing and seeing

In Ambonwari, knowledge comes from seeing, listening, and touching. Taste is con-
ceptually related to the first sense, that is, seeing, and smelling to the second, that is,
listening. Touching is connected with practical skill, to know how, to sensory mastery
of activity, to bodily relationship with one another. Two verbs come to the fore in this
latter sense: sari-kan ‘grab, hold tight, have something between hands or legs, touch’
and say-kan ‘hold, keep, use, wear’. Evidential strategies in Karawari are also related to
the main three senses just mentioned.
What people consider as important knowledge is another matter; and to distinguish
between knowledge and knowing is yet another. Knowledge has to become knowing
(Keck 1998: 10), it has to become negotiable: a knowing process. This was the case in the
past, when knowledge was created through amicable or hostile relationships between
the living, the forest spirits, and the spirits of the dead; this is also the case in the present,
when knowledge is supposed to be created through relationships between the living,
God, and the Whites. However, the latter knowledge is not separated from the former
one. The former relationships actually continue to influence and compete with the lat-
ter ones and in people’s opinion do not allow them to get access to the real knowledge.
This, however, also means that people want and often do hear and see things ‘beyond’
and ‘behind’ the audible and visible, and they constantly speculate about them. It is in
these invisible and inaudible domains that the real knowledge is hidden.
Like the Urapmin, the Ambonwari too perceive that ‘hiddenness is a general condi-
tion of important things’ (Robbins 2004: 139). The Urapmin consider the eyes to be
the quintessential organs of revelation and vision ‘the primary knowledge-producing
sense’ (Robbins 2004: 141). Sight is prior to hearing as a means of obtaining knowledge
and ‘knowledge is in important respects equated with sight’ (Robbins 2004: 141). Rob-
bins says that ‘[i]t is impossible to overemphasize how strong the Urapmin concern for
vision is’ (Robbins 2004: 139). I can only confirm this in the case for the Ambonwari.
Under the influence of the Catholic charismatic movement, which entered the village
in December 1994, people became obsessed by saying that things, thoughts, and doings
should all come out to an open place (Telban 2009). It became a part of rejection of
274 Borut Telban
tradition, of the practices of their ancestors, who suddenly became accused of hiding
their knowledge. This pertained not only to seeing but also to hearing. Women and
uninitiated men were under strict taboos related to seeing the secret objects from the
men’s house, carved spirit-crocodiles and flutes in particular, and hearing of the names
of the spirits (Telban 2008). In 2007 and 2008, when some Ambonwari got the first
mobile phones, whenever they went to town, the phones had to be put on external
speakers so that everyone could hear what the person on the other side of the line said.
Seeing, on the other hand, is very much related to spatial perception and the present
tense sanggwan ‘see, look’ is homonym of sanggwan ‘where’. Interrogative sanggwandɨn
‘how, in which way, where from’ also finds its basis in visibility and emplacement.
(31) ama ya-ka sanggwa-ma-r
1sg A-O look-die-immed.past
I’ve watched him die.
(32) mi sanggwan mi kwasa-r-a
2sg where 2sg get.up-immed.past-inter
Where did you come from?
(33) yam sanggwandɨn ma awa-r-a
house how upriver burn-immed.past-inter
How did the house get burned?
The verb andɨkan ‘hear, listen’ is the verb often used in connection with wambung, that
is, with understanding:
(34) mɨn anduk apia-na-k kɨ-n andɨ-kan
3sg language 1pl-poss-Vsg A-O hear-pres
He understands our language.
Listening is also related to obedience and belief (also in God). About somebody who
doesn’t want to understand or obey, people say that he is kwandɨkas kanar ‘man with
no ears’. In a serial verb construction the verb andɨ- ‘hear’ often appears in redupli-
cated form andandɨ- (rarely andɨng-andɨng) where it refers to attentively listening to
somebody or something.
5 Mirativity
Mirative meanings, an expression of unexpected or surprising information or situa-
tion, or even an anticipation of a surprise, are marked by interjection kambay- ‘alas!’
inserted between two verbs, one of which or both are usually, though not always, either
andɨ- ‘hear’ or sanggwa- ‘see, look’:
(35) ama ya-ka andɨ-kambay-andɨ-r yanda mariawk
1sg A-O hear-interj-hear-immed.past this talk
I’ve been surprised to hear this talk.
(36) wakɨn ama ya-ka sanggwa-kambay-sanggwa-n

snake 1sg A-O see-interj-see-pres
I am in shock when I see a snake.
(37) ama kapuk andɨ-rarin kambra ama sanggwa-kambay-sa-n

1sg call hear-desid nothing 1sg look-interj-be/sit-pres
I just watch and listen as I want to hear their call (I think that they will call me).
In the above example sanggwa-kambay-sa-n can be replaced by andɨ-kambay-sa-n

‘hear-interj-be-pres’ or and-andɨ-kambay-sa-n ‘redup-hear-interj-be-press’ where
all the emphasis would be on listening. The last verb sa-n ‘sit-pres’ can be replaced by
kay-kan ‘stay-pres’; each of them can be translated by ‘be-pres’.
(38) mɨn kambra mɨn imɨng-kambay-imɨn-dɨ kapuk
3sg nothing 3sg stand-interj-stand-immed.past noise
bɨn-ang andɨ-mbɨn
3sg-dat hear-seq
He just stood in surprise after hearing the noise.
(39) mɨn wurumɨng-kambay-wurumɨn-dɨ wiya mɨn-ang sanggwa-mbɨn

3sg laugh-interj-laugh-immed.past dog 3sg-dat see-seq
He laughed in surprise after seeing the dog.
Kambay- with suffix -a, is often used as an independent exclamation kambaya ‘alas’,
which is one of the verbal expressions of shock or surprise.
6 Conclusion
Bateson (1958: 232) noticed that ‘Iatmul thought is characterised not only by its intel-
lectuality, but also by a tendency to insist that what is symbolically, sociologically, or
emotionally true, is also cognitively true’. All these aspects are also valid for the Kara-
wari people and are merged in their concept of wambung ‘insideness, seat of memory,
thinking and feeling’. Bateson (1958: 228) also writes: ‘The naming system is indeed a
theoretical image of the whole culture and in it every formulated aspect of the culture
is reflected’. Stimulated by Bateson’s work, additional research has been conducted
about the importance of ‘proper names’ in the Sepik (Harrison 1990; Wassmann 1991;
Telban 1998). Wassmann (1991: fn. 222), who correctly focuses on ‘proper names’ in
his analysis of Iatmul knowledge, totemic songs and ancestral past, also observes that
a theoretical discussion of two concepts, ‘proper name’ and ‘common name’, should
be dealt with on a linguistic level. So if we look at the production of knowledge, which
is simultaneously a production or re-production of a particular culture (including
changed or modified culture), we can say that it is based on visible ‘names’ and aural
‘discussions’, on wi and mariawk respectively.
276 Borut Telban
Let us make it clear that wi ‘name’ covers both proper and common names. In
the present chapter, which has been mainly about the ways of speaking, I alluded
to people’s hypersensitivity whenever their proper name was mentioned in con-
nection to some amoral, antisocial, or unpleasant situation. By calling the name
of a person, that name ‘swelled’, became visible, and the person suddenly found
himself or herself under the eyes of the people. All these dimensions are already
embedded in the term wi with three homonymic meanings: name, night, dream.
All of them represent border lines between usually visible and usually invisible
worlds, between the world of the living and the world of the dead. All of them
represent the surface, arɨm ‘skin, appearance’, covering wambung ‘insideness’, the
true centre of all intentions, thoughts, and feelings. Anggɨndar kwanar ‘personal
spirit, shadow’ is freer and is not bound to the body/skin (thus an image in water
or a mirror can be called either arɨm-iak ‘reflection’ or anggɨndar-kwa ‘spirit,
shadow’).
Moreover, beings can change the skin but retain their interiority: a man from the
Eagle Clan can in his dreams take up the body of an eagle and fly over the land and
tell in the morning that he saw a motor canoe coming. His knowledge is his capabil-
ity of seeing in dreams as an eagle. A dancer from the Cassowary Clan can decorate
his skin with cassowary feathers and during the ceremony become both the first
ancestor of the clan and the cassowary. Skin is clothed in perception: perceiving
the outside world in a particular way while being perceived according to the form
it has taken. Name, night, and dream can wi-kan ‘swell’ and while expanding they
begin to reveal the realities of beings which were not visible at first. Thus, invisible
things may become visible and vice versa. Once things and beings are called by
names, encountered at night (through different signs), or observed in dreams, they
are also expected to appear in their fully tangible form. They are expected to get
the skin, become visible, and take part in people’s lives. This may pertain to good
or bad things: money, travel, visitors, spirits, Whites, special powers, sickness, or
death. Just as men’s houses with initiation rituals created an environment where
spirits could get their visible skin, and just as the calling of the forest spirits’ names
can ‘wake them up’ and bring back—as the members of the Catholic charismatic
movement would say—mainly bad customary practices, so can communication
with mobile phones connect the Ambonwari to their spirits of the dead. Conse-
quently this should then provide them with all those advantages and goods which
in their view characterize the world of the Whites (where their spirits who are the
Whites dwell).
To conclude, the most resourceful knowledge is created between visible and invis-
ible domains of people’s lives, and the transition between two domains can be crea-
tively manipulated. Speaking, seen as creation, is just one way—along with many other
practices such as body decoration or dance, for instance—of bringing the desired
life-world into existence.
References
———. 2008a. The Manambu language of East Sepik, Papua New Guinea. Oxford: Oxford Uni-
versity Press.
———. 2008b. ‘Semi-direct speech: Manambu and beyond’, Language Sciences 30: 383–422.
Bateson, Gregory. 1958. Naven. London: Wildwood House.
Brison, Karen. 1992. Just talk: Gossip, meetings, and power in a Papua New Guinea village. Berke-
ley: University of California Press.
Foley, William A. 1991. The Yimas language of New Guinea. Stanford: Stanford University Press.
Harrison, Simon. 1990. Stealing people’s names: History and politics in a Sepik River cosmology.
Keck, Verena. 1998. ‘Reflections on knowledge in an Oceanic setting’, pp. 1–29 of Common
worlds and single lives: Constituting knowledge in Pacific societies, edited by Verena Keck.
Oxford: Berg.
Marchand, Trevor H. J. 2010. ‘Making knowledge: explorations of the indissoluble relation
between minds, bodies, and environment’, Journal of the Royal Anthropological Institute,
Special Issue: S1–S21.
Robbins, Joel. 2004. Becoming Sinners: Christianity and moral torment in a Papua New Guinea
society. Berkeley: University of California Press.
Rumsey, Alan. 2008. ‘Confession, anger and cross-cultural articulation in Papua New Guinea’,
Anthropological Quarterly 81(2): 455–72.
Schieffelin, Bambi. 2008. ‘Speaking only your own mind: Reflection on talk, gossip and inten-
tionality in Bosavi (PNG)’, Anthropological Quarterly 81(2): 431–41.
Telban, Borut. 1997. ‘Being and “non-being” in Ambonwari (Papua New Guinea) ritual’, Oceania
67(4): 308–25.
———. 1998. Dancing through time: A Sepik cosmology. Oxford: Clarendon Press.
———. 2008. ‘The poetics of the crocodile: Changing cultural perspectives in Ambonwari’, Oce-
ania 78(2): 217–35.
———. 2009. ‘A struggle with spirits: Hierarchy, rituals and charismatic movement in a Sepik
community’, pp. 133–58 of Religious and ritual change: Cosmologies and histories (Ritual Stud-
ies Monograph Series), edited by Pamela J. Stewart and Andrew Strathern. Durham, NC:
Carolina Academic Press.
Wassmann, Jürg. 1991. The song to the flying fox. Boroko: National Research Institute.
Young, Michael. 1971. Fighting with food: Leadership, values and social control in a Massim soci-
ety. Cambridge: Cambridge University Press.
Index of authors
A Coupe, Alexander R. 110–11, 239
Adelaar, Willem F. H. 242 Creissels, Denis 53, 74
Aikhenvald, Alexandra Y. 2–4, 6, 9–33, Croft, William 258
38–40, 44, 52–3, 64, 69, 71–6, 79–83, 92, Crouch, Isabel 29
99–100, 104, 115, 133–4, 142, 145, 162, 168, Crystal, David 78
195, 202, 225, 239, 242, 257, 261–2, 267–8 Culy, Christopher 255
Alamin, Suzan 246, 248, 251–2 Cyffer, Norbert 205
Alcazar, Asier 15
Alho, Irija 80, 84 D
Ameka, F. K. 256 Dahl, Östen 53
Dehé, N. 24
B Deibler, E. W. 27
Babel, Anna 12 DeLancey, Scott 4, 104, 110, 116, 133, 139,
Baranova, V. V. 148 231, 239
Bartee, Ellen 110, 239 Dendale, Patrick 20
Basso, Ellen B. 27, 34 Diewald, Gabriele 3
Bateson, Gregory 261, 275 Dimmendaal, Gerrit J. 19, 24, 26, 43, 245–6,
Beier, Christine 228 251–2, 255–6
Besnier, Nico 39 Dingemanse, Mark 257
Bickel, Balthasar 110 Dixon, R. M. W. 3, 10, 14–25, 31, 39, 42, 55,
Bläsing, Uwe 148–9, 164 70, 72, 133, 137, 188–9, 221, 242
Blass, Regina 15 Dubois, Betty 29
Boas, Frans 1–4, 19, 32, 44 Duranti, Alessandro 40–1
Boeschoten, Hendrik 86
Bokarev, Evgenij A. 67 E
Botne, Robert 15 Eberhard, David 8, 29–33, 36–7, 40–1
Boyeldieu, P. 15 Epps, Patience 233
Bradac, James J. 28 Evans, Nicholas 25
Bradley, David 109, 132
Brison, Karen 264 F
Brown, Michael 228 Fleck, David 6, 10–11
Bryan, Margaret A. 245 Floyd, Rick 10, 33
Buba, Malami 205 Floyd, Simeon 4
Burke, Peter 190 Foley, William A. 260–1, 271
Bybee, Joan 53, 249 Forker, Diana 5–6, 13–16, 25, 28, 30–6, 42,
53, 58, 63, 65
C Fortescue, Michael 39
Carlin, Eithne B. 16 Frajzyngier, Zygmunt 191, 195–8
Carol, Javier J. 17 Frawley, William 80
Casad, E. 100 Friedman, Victor A. 53
Chafe, Wallace L. 3
Chang, Anna Hsiou-chuan 89 G
Chirikba, Viacheslav 12, 21 Genetti, Carol 110–11, 239
Chirkova, Katia 132–3, 136, 139 Gerzenstein, Anna 17
Coly, Jules Jacques 13–14, 18, 24, 37, 41–2, 194 Giacalone Ramat, Anne 20, 22
Comrie, Bernard 15, 21, 65–7, 70, 232 Gilley, Leona G. 15, 250
280 Index of authors
Givón, Talmy 239 Koch, Grace 189

Goldberg, Adele E. 258 Koivisto, Vesa 80, 84
Gordon, L. 100 König, Christa 15
Greed, Teija 5–6, 13–17, 25, 28, 31–2, 35–6, Kopytoff, Igor 193
42, 70, 79 Korhonen, Riitta 80, 84
Greenberg, Joseph H. 245 Kracke, Waud H. 33
Grice, P. 38 Kroskrity, Paul V. 2, 26
Gronemeyer, C. 20 Kuipers, Joel C. 39
Grunow-Hårsta, Karen 35 Kuteva, Tania 25
Gueldemann, T. 101
Guillaume, Antoine 233 L
Lacan, Jacques 34
H LaPolla, Randy J. 133
de Haan, Ferdinand 3–4 Larson, Mildred L. 243
Haiman, John 144 Lazard, Gilbert 4, 19
Hakulinen, Auli 80, 84 Leger, Rudolf 196, 198, 202–3
Hale, Austin 139 Levinson, Stephen 251
Hardman, Martha J. 32 Li, Charles 143
Harner, Michael J. 228 Lidz, Liberty A. 8, 11, 35, 132–4, 139,
Harrison, Simon 43, 261, 275 145, 163
Haspelmath, Martin 100, 111 Lin, You-Jing 110, 239
Heine, Bernd 25, 43 Lindstrom, Lamont 39
Heinonen, Tarja Riitta 80, 84 Liu, Hui Qiang 132
Hewitt, B. George 20 Longacre, R. 101
Hieda, Osamu 20, 250 Loughnane, Robyn 12, 152
Hildebrandt, Kristine 110, 239 Lowe, Ivan 8, 18, 21
Hill, Jane H. 37, 69 Lüpke, Friederike 207
Hopper, Paul 240
Hsieh, Fuhui 100 M
Huang, Chenglong 132–3 MacDonald, Lorna 27, 100
Hyslop, Gwendolyn 10, 16, 18, 32, 37, 42, McLendon, Sally 37–8
109–11, 114–18, 124 Mann, Trischa 242
Marchand, Trevor H. J. 262
I Maslova, Elena 23, 160
Irvine, Judith T. 37, 69 Matisoff, James A. 110, 240
Isxakova, X. F. 73, 86 Meira, Sérgio 16
Michael, Lev 27, 209, 225, 231
J Migliazza, Ernest C. 15
Jacobsen, William H. Jr. 3–4 Mihas, Elena 8, 14–15, 21–7, 32–5, 39, 42–3
Jaggar, Philip 205 Miller, Cynthia 15, 250
Jakobson, Roman O. 3–4 Mithun, Marianne 217
Johanson, Lars 3, 5, 9, 15, 53, 81, 86 Monserrat, Ruth 10
Jungraithmayr, Hermann 191, 194 Moseley, Christopher 132
Mulac, Anthony 24, 28
K Mulaeva, N. M. 151
Kaltenböck, Gunther 25 Muysken, Pieter 242
Kany, Charles 12
Kavalova, Y. 24 N
Keck, Verena 273 Nasilov, D. M. 73, 86
Keenan, Edward L. 38 Neukom, L. 19
Khalilova, Zaira 53, 56, 62, 65, 67 Nevskaja I. A. 73, 86
Kibrik, Aleksandr E. 62 Newman, Paul 252–3
Kilian-Hatz, Christa 256 Nichols, Johanna 3, 65
Kirsner, R. S. 20 Nikitina, Tatiana 256
Index of authors 281
Noonan, Michael 110, 239–40 Sun, Hong Kai 132

Nordlinger, Rachel 19 Sun, Jackson T.-S. 133–4, 138–9
Nuckolls, Janis B. 31, 41
T
O Talmy, Leonard 251
Ochs, Elinor 38 Tatevosov, Sergei 53
Olbertz, Hella 12 Telban, Borut 24, 27, 32, 35, 40, 43, 270–1,
Overall, Simon E. 15, 20–4, 27, 33, 36, 39, 40, 273–5
43, 227, 233 Thompson, Sandra A. 20, 24, 28, 143, 232
Thurgood, Graham 100
P Timberlake, Alan 20
Pagliuca, William 53, 249 Topadze, Manana 20
Palmer F. R. 4 Tournadre, Nicolas 100, 114, 139
Palmer, Gary 32 Trask, R. L. 24
Pan, Chia-jung 5, 10, 12, 15, 17, 27–32, 36, Traugott, Elizabeth C. 28
42, 89 Travis, Catherine 12, 28
Payne, Doris L. 242 Tucker, A. N. 245
Payne, Thomas E. 71 Tung, Tung-ho 17
Perkins, Revere 53, 249
Pietrandrea, Paola 22 U
Plungian, Vladimir A. 3–4, 12 Urmson, James O. 24, 223
Polinsky, Maria 66–7 Utas, Bo 3, 5, 15
Pusch, Claus D. 15, 20
Pyurbeev, G. Z. 148, 152 V
Valenzuela, Pilar 12, 95
R Van Beek, Walter E. A. 202
Ramirez, Henri 29–30 Van Bogaert, Julie 20, 22
Reesink, G. P. 27 van der Auwera, Johan 3, 4
de Reuse, Willem J. 40 Vandamme, Marc 86
Robbins, Joel 4, 41, 273 Vilkuna, Maria 80, 84
Rubenstein, Steven 228 Voeltz, F. K. Erhard 256
Rumsey, Alan 41, 263
W
S Waag, C. 15
Sadler, Louise 19 Wälievä, Gölčäčäk 73
San Roque, Lila 12, 152 Wang, Dehe 132, 136
Sanžeev, G. D. 148, 168 Wassmann, Jürg 261, 275
Sapir, Edward 4 Watters, David E. 240–1
Sasse, Hans-Jürgen 254 Weber, David J. 4, 33–4, 37
Say, S. S. 148 Westermann, D. 15
Schadeberg, Thilo C. 245 Wiemer, Björn 20, 25
Schieffelin, Bambi 264 Wilkins, David 25
Schneider-Blum, Gertrud 251–3, 256 Willett, Thomas 3, 4, 10
Schuh, Russell G. 194, 197
Seesing, Olga 8–14, 20–1, 31–3, 36, 42, 151 Y
Šencova, I. V. 73, 86 Yang, Gloria Fan-pei 13, 34, 89
Serdobol’skaja, N. V. 148 Young, Michael 264
Sherzer, Joel 231
Skribnik, Elena 8–14, 20–1, 31–3, 36, 42, 151 Z
Smeets, Ineke 6 Zakiev, M. Z. 73–5, 77–80, 86
Smirnova, Elena 3 Zakirova, Rämzijä 81
Squartini, Mario 5, 20 Zhang, Sihong 6, 7, 10–11, 16, 20, 25, 28, 30,
Storch, Anne 13–15, 24–5, 37, 39, 41–2, 42, 138
207, 250 Zoch, Ulrike 194
Index of languages, language families,
and linguistic areas
A Bole-Angas branch of Chadic 195
Abkhaz 9, 12, 21 Bole-Tangale languages 191, 196, 202
Acholi 20 Bora 9
African languages 15 Bosavi 264
Afroasiatic languages 191, 252 Bura 191
Aguaruna 15, 20–4, 27, 36, 39, 43, 77, 227–44
Akha 10, 100 C
Akhvakh 53, 74 Campan languages 27, 209–26
Amazonian languages 7–8, 15, 23, 27–9, Carib languages 34
38–42, 209–26, 227–44 Catalan sign language 8
Ambonwari variety of Karawari, see Karawari Caucasian languages 5, 15, 20
Amdo Tibetan 133, 138–9 Cavineña 20
Andean languages 5–6, 9, 14, 42–3, 242 Central Khoisan languages 15
Andes, languages of, see Andean languages Chadic languages 42, 190–208, 252
Arawá languages 5 Chaghatay 86
Arawak languages 7, 15, 23, 27, 42, 209–26 Chantyal 240
Archi 62 Chorote 17
Arizona Tewa 2, 26 Classical Tibetan 109–10, 114
Asháninka Tambo-Ene 209 Colombian Spanish 28, see also South
Ashéninka Apurucayali 209 American Spanish, Spanish
Ashéninka Pajonal 209 Cora 100
Ashéninka Perené 8, 15, 23, 27, 32–5, 39, Cushitic languages 252
42–3, 209–26 Cuzco Quechua 29, see also Quechua
Ashéninka Pichis 209
Ashéninka Ucayali 209 D
Athabaskan languages 41 Dyirbal 17, 23, 34, 39, 42, 143, 155, 171–89
Australian languages 15, 17, 39, 42, 55, 143, Dzongkha 109
171–89
Austronesian languages 5, 13, 15–16, 39, E
89–107, 267 East Bodish 109
Avar 52, 61 East Tsezic 65–7
Aymara 32 East Tucanoan languages 7, 29, 33
Eastern Pomo 38
B English 12, 17, 20–5, 32, 40, 44, 109
Baima 133 Ersu 6–7, 10–11, 16, 20, 25, 28, 36, 132–47
Balkan languages 53 Estonian 12, 38
Balkans linguistic area 15, 53 Ethiopian Plateau, languages of 101
Bantu languages 15 European languages 4, 15, 22–5, 40, 63
Baré 15
Basque 15, 38 F
Benue-Congo languages 194, 245 Finno-Ugric languages 5
Bezhta 65 Formosan languages 13, 15, 17, 42, 89–107
Bhutan, languages of 108–31 French 19, 22–4
Bole 191, 207 Fur 15
Index of languages, language families, and linguistic areas 283
G Kurtöp 6, 10, 13–18, 32, 37, 42, 72, 108–31

Gahuku 27 Kwa languages 257
Gascony Occitan 20 Kwakiutl 4, 19
Georgian 20 Kwami 196, 202, 205
Gongola-Benue linguistic area 194 Kwoma 33
Gur languages 15
L
H Laal 15
Hausa 191, 193, 197, 207, 252 Lahu 240
Heiban 242, 246 Lega 15
Highlands of New Guinea, languages of 15, Lezgian 100
see also New Guinea, languages of Lhasa Tibetan 133, 139
Hindi 109 Língua Geral, see Nhêengatú
Hinuq 5–6, 9–10, 13–16, 25–36, 42, 52–68, 74, 76 Lithuanian 20, 25
Hone 31 Lower Sepik languages 43, 260–78
Huallaga Quechua 33–4, see also Quechua Luo 15
Hunzib 65–7
M
I Maaka 13–14, 18–19, 24, 37, 41–2, 190–208
Iatmul 265 Magar 35
Indo-European languages 44, 71 Maká 17
Ingush 65 Malagasy 38
Iranian 4, 15, 20 Mamaindê 8, 29, 30–3, 36, 40–1
Italian 20–4 Mamu dialect of Dyirbal 188
Manambu 23, 27, 33, 261, 267
J Mandarin Chinese 105, 139, 143
Jaqi 32 Mansi 20
Jarawara 5, 10, 14–17, 31–5 Mapudungun 5, 9, 31, 38
Jivaroan languages 227–44 Maricopa 100
Jukun languages 31 Mataco-Mataguayan languages 17
Matses 6, 10–11, 22, 35
K Meso-American languages 15
Kadugli 245 Mongolic languages 9, 42, 148–70
Kagwahiv 33 Munda languages 19
Kakinte 209 Mupun 195, 197
Kalapalo 37, 34 Mỹky 10
Kalmyk 8–12, 20–1, 31–3, 36, 42, 55, 78, 115,
148–70 N
Kampan languages, see Campan languages Nakh-Daghestanian languages 5, 15, 42,
Kanuri 191, 194, 197, 204–5 52–68, 74
Karawari 24, 27, 32–5, 40, 43, 260–78 Nambiquara languages 8, 29
Kathmandu Newari 139 Nanti 27, 225
Katla 245–6 Ndu languages 23
Katla-Rashad languages 245 Nenets 20
Kavalan 100 Nepali 109
Kham 240–1 New Guinea, languages of 15, 16, 27, 43
Khoisan languages 15 Nhêengatú 4
Khwarshi 62, 66–7 Niger-Congo languages 245, 252, 257
Kipchak-Bulgar (Western) group of Turkic Nilo-Saharan languages 15, 252
languages 70 Nilotic languages 15, 20, 250
Kordofanian languages 245–6 North American languages 4, 14
Ku Waru 263 North-east Caucasian languages 5, 15, 42,
Kupto 202 52–68, 74
284 Index of languages, language families, and linguistic areas
O Tibetan 100, 109–10, 114, 139

Occitan, see Gascony Occitan Tibeto-Burman languages 7–8, 10–11, 14,
Old Turkic Runic texts 86 35, 42, 108–31, 132–47, 239–40
Tima 19, 24, 26, 40, 43, 76, 245–59
P Tok Pisin 262, 265, 268–9
Panoan 6 Trio 16
Papuan languages 15, 260–78, see also New Tsafiki 9
Guinea, languages of Tsez 52, 66
Pastaza Quichua 31, 41, see also Quechua Tshangla 109
Piya 202 Tsou 13, 17, 34
Portuguese 23, 32 Tsouic languages 13, 17, 34
Purépecha 20 Tucano 9, 12, 20, 29–36, 41
Puxi dialect 133 Tucanoan languages 7, 29, 33
Tumtum (Kadugli) cluster, see Kadugli
Q Tupí-Guaraní languages 4, 33
Qiangic branch of Tibeto-Burman Turkic languages 5, 15, 20, 42, 53, 69–88
languages 132 Turkic languages 53
Quechua 9, 12, 22, 37, 41, see also Cuzco Tuvaluan 39
Quechua, Huallaga Quechua, Wanka Tzesic sub-branch of
Quechua, Pastaza Quichua Nakh-Daghestanian 52–68
R U
Rashad languages 245–6 Ubangian languages 252
Russian 60, 62, 70, 79 Uralic languages 15
Urapmin 273
S Usan 27
Saaroa 4–5, 10–12, 15–16, 26–32, 36, 42, Uto-Aztecan languages 15
89–107
Saharan languages 191 V
Santali 19 Vaupés River Basin linguistic area 15, 33, 40
Semitic languages 15
Sepik languages 23, 27, 43 W
Shilluk 9, 15 Wakashan languages 19
Shipibo-Konibo 12, 22, 33–5 Wanka Quechua 10, 33, see also Quechua
Sissala 15 Warekena 23
Siwu 257 West African languages 256
South American Spanish 12, see also Span- West Tsezic 66–7
ish, Colombian Spanish Western Apache 22, 40
South Arawak languages 209–26 Western Chadic 190–208
Southern Nambiquara dialect complex 8, Western Mongolic languages 148
18–19 Western Nilotic languages 20, 250
Spanish 22, 24, 32, see also Colombian Weyewa 39
Spanish, South American Spanish Widala 202
Wintun 32
T
Tacana languages 20 X
Taiwan, languages of 89–107 !Xun 15
Talodi 245–6
Tamangic languages 239 Y
Tanna 39 Yanomami 15
Tariana 7–12, 22–5, 29, 32–6, 42 Yimas 260–1, 271
Tatar 5, 6, 9–10, 13–14, 16–17, 25–6, 28, 31, 33, Yongning Na (Mosuo), see Yongning Na
35–6, 42, 55, 69–88 Yongning Na 8, 11, 35, 132–4, 139, 145
Tauya 27, 100 Yukaghir 9, 23
Index of subjects
A command 1, 2, 9, 10, 214–15, see also
adjective 3, 22–4, 52, 62–3, 70–1, 171–5, 186, imperative
188, 210, 256 common knowledge evidential 163–5,
adverb 3, 22–8, 43, 52–3, 63, 70–1, 79, 192–4, see also general knowledge
199, 214, 224, 249–51, 256–8 communication, new means of 36–7, 40,
agglutinating 70, 110, 194, 227 see also technology and the use of evi-
areal diffusion 242, see also language dentials; media and the use of evidentials
contact complement clause 20, 24–5, 28, 61–4, 67,
aspect 1, 2–6, 10, 18–20, 42, 44, 52–3, 71–2, 76–7, 86, 156–7, 231–2, 250, 254–6,
75, 92, 109–17, 120, 124, 127–8, 148–56, see also complementizer
196, 210, 230–3, 239, 247–50, 253, 257, complementation strategy 20, see also
see also tense complement clause
assumed evidential 12, 20, 23, 34, 72, 75, complementizer 13, 20, 26, 76–7, 86, 149,
79–81, 84, 152–3, 159–61 165–9, 192, 195, 198–200, see also com-
assumption 1–3, 9, 12, 20–5, 28, 31, 35, 37, plement clause
40, 42, 59, 61, 81, 84, 117, 122–3, 134–5, conditional 10, 19–21, 210, 217
210–11, 214–19, 226, see also assumed confess 263
evidential conjecture 7, 211–12, 216, see also inferred
assumptive marking, see assumption evidential
audibility 171, 179, 183–7, see also hearing conjunct person marking, see conjunct/
auditory information 19–20, 23, 220, disjunct person marking
see also hearing conjunct/disjunct person marking 139–40
auxiliary 25, 53, 70, 84, 149–52, 157, 161 contact language 15, 41, see also areal diffu-
avoidance style 39, 188 sion, language contact
control 29–30, 41, 93, 99–100, 104, 190–2,
B 200–2, 206–7
backgrounding 240–1, see also information converb 53, 63, 71, 75, 86, 148–51, 154–6, 165
structure co-occurrence of evidentials 11–2, 79, 95, 142
borrowing 110–11, see also language contact copula 10, 14, 16, 108, 111–16, 120–3, 128–9,
234–5
C counter-expectation 112, 124–8, 230–1
case 17–18, 52, 70, 90–1, 110, 148, 151, 156–61, cultural conventions 2, 3, 15, 23, 25, 29,
169, 172–85, 194, 210, 227, 232–4, 237–9 33–4, 37–8, 41–3, 65–9, 82–3, 90, 105–6,
certainty 10, 12–13, 20, 29–30, 44–5, 72, 79, 171, 188, 191, 194, 202, 206–7, 256, 265,
84, 113, 119–23, 127–8, 143–4, 153, 157–61, see also precision, requirement for
191–2, 195–9, 214, see also uncertainty, cultural postulates, see cultural conventions
epistemic modality
change in evidential use 2, 34–5, 41–3 D
clause-chaining 111, 120, 127, 227, 232 declarative clause 9, 19, 79, 81, 211, 214–16,
clitic 13–16, 32, 52, 59–67, 71–2, 84–5, 87, 90, 219, 223–5, 231, 234–7, 241
93–6, 100, 106, 109–13, 123–8, 210–17, deduction 5, see also inference
224–5, 246–54 default evidential 6, see also markedness
cognition 2, 22, 25, 28, see also cognition, demonstrative pronouns 19, 52, 71, 144,
verbs of; perception, verbs of 167–8, 177–80, 210, 214–15, 222
cognition, verbs of 22, 38, 76–7, 83, 165–7, dependent clause 233–4, 237–40, see also
169, 171, 195, 202–7, 225, 229, 256, 271–3 subordinate clause
286 Index of subjects
de-subordination 21 exclamatory sentences 1, 225, 231

direct evidential 9–10, 12, 20, 29, 33–6, 42, expectation of knowledge 108–28, 210–11,
114, 122, 133–5, 139–41, 145, 152–5, 217, 225–6, see also mirativity
see also firsthand evidential experiential 93, 103
direct quote 26–28, see also quotative, self- explicitness 211, 220–2, see also precision,
quote, reported speech requirement for, cultural conventions
direct speech 2, 26, 75–7, 149, 161–2, 211, eyewitness evidential 8, 55, 66, see also
222, 267–8, see also speech report firsthand evidential
direction 133, 142–6, 180–3, 186–8, 210, 250–6 eyewitness information source 192,
discourse genre, see genre 196–8, see also seeing, visually acquired
disjunct, see conjunct/disjunct person marking information
dizque as an evidentiality strategy in
Spanish 28 F
double marking of information finite verb 59–62, 70, 148–51, 165–9,
source 11–2, 79, 95, 142 230–40, see also non-finite verb
doubt 10, 61, 74, 84–6, 191–2, 197–9, 211, first person effect 29–31, 54–5, 73–4, 87,
see also dubitative, epistemic modality 98–100, 106, 138–9
dreams 33–4, 66, 82, 114, 155–6, 201–11, 219, firsthand evidential 5–9, 14, 17, 35, 56, 66,
228, 238, 276 71–3, 79, 83–4, see also direct evidential,
dubitative 12–13, 23–7, 44, 93, 104–5, 210–16, non-firsthand evidential
224–6, see also epistemic modality firsthand information 258, see also first-
hand evidential
E focus 93–5, 214, 249, 254, see also back-
egophoricity 108, 113–14, 123, 127, 139–42 grounding, emphasis, foregrounding,
emotions 27, 271 information structure, topic
emphasis 213–14, 217, 224 foregrounding 240–1
endangered language 15, 42, 132, 209 forgetting 271
epistemic extensions of evidentials, see epis- formal markedness, see markedness
temic modality frustrative 217, 231
epistemic modality 4, 10–20, 24–7, 40–4, functional markedness, see markedness
62–3, 72, 75–80, 84–7, 92–3, 97–8, 108, future 6, 10, 21, 31, 34–6, 42, 72, 76, 79–80,
113, 117–24, 127, 143, 153, 157–60, 169, 108, 119–20, 127–8, 135, 137, 149–51, 154,
190–4, 197, 223, 226, see also certainty, 157–62, 169, 174, 248, 252
probability, uncertainty
epistemicity, see epistemic modality G
epistemology 1, 37–44, 229–42, see also gender 1, 2, 4, 14, 44, 52, 174–80, 194, 210,
information source see also noun class
ergative 51, 62, 110, 172–3, 175–9, 182, general knowledge, inference based
199–200, 206–7, 246, 254 on 7–9, 20–1, 134, 210–11, 217
evidence 19–20, 24, 38, 44–5, 114–17, 121–3, generic knowledge, see general knowledge
126, 152–3, 157–9, 167–9, 250, 258 genre 13, 31, 35–6, 39, 41, 58, 65–7, 82–3, 92,
evidential extensions 10–14, see also evi- 105–6, 138–42, 241, 245, 261
dentiality strategy gossip 39, 264, 266
evidentiality neutral forms 2, 5, 14, 35, 42, grammaticalization 1–3, 7, 15, 21, 26–8,
53–8, 66–7, 71–3, 83, 86, 164, 235, 240–1 43–4, 52–3, 61–4, 75, 78, 87, 93, 101, 108,
evidentiality strategy 10, 19–22, 29, 32, 115–18, 149, 151, 162–3, 167, 172, 218, 232,
36, 39, 101–4, 142–3, 165–9, 197–202, 239, 242, 258
210–26, 228–42 Grice’s maxims 38
evidentials 1–19, 29–37, 40–3, 52–60, 63–7, gustatory perception 205, 220, see also tasting
72–83, 86–7, 89–101, 105–6, 108–28,
132–42, 145, 148, 152–67, 169, 191–2, H
195–7, 200–2, 206–7, 239, 261 habitual 149–57, 162–3, 167
evidentials in dreams 33–4, see also dreams hearing 7, 9, 20, 23, 134, 192, 200, 223, 273–5
2 Index of subjects 287
hearsay 4, 5, 13–14, 20–1, 32, 40, 60, 65, K

72, 75–8, 82–7, 200, 210, 223–4, 226, knowledge, creation of 269–3
242, see also reported evidential, knowledge, value of, see value of knowledge
quotative
hedging 26, 29, 222, 253–6 L
historical development of evidentials and language, long 262
information source markers 20–1, language, short 262
27, 53, 86, 90, 100–1, 210, 218, 223, see language contact 14–15, 41, 43, 110–11, 242
also grammaticalization, language language obsolescence 15, 42
contact lexical evidentiality 3, 44, see also lexical
hypothetical 7, 210–1, 262 expression of information source
lexical expression of information
I source 2–3, 8, 22–6, 39, 61–3, 83–6,
ideophone 24, 43, 256–8 104–5, 204–7, 229, 262–8
imperative 2, 211, 214–15, 218, 230–1, 252, linguistic area 14–15, see also language
see also command contact
imperfective 10, 108, 111, 117–19, 127–8, 149, location 250–6
155, 165, 196, 230, 238–40, 248–9 logophoricity 253–6
incorporation 32, 274–5 lying with evidentials 45
indicative 196, see also declarative clause
indirect speech 2, 26, 75, 86, 225, see also M
speech report markedness 4–7, 10, 14–18, 134–5, 139–42
indirectivity 155–7, see also non-firsthand maxims, see Grice’s maxims
evidential media and the use of evidentials 34–5,
inference 1–12, 20–2, 38, 43, 61, 134–6, 57–8, 66–7, 81–2, see also technology
138–9, 142–5, 152, 155–9, 169, 192, 195, and the use of evidentials
see also inferred evidential, inferential metalinguistic valuation of evidentials 40
inferential 7–8, 18, 20–1, 62, 93, 103–4, mirative meanings, see first person effect,
133–40, 142, 145, 210–14, 221, 226, mirativity
see also inferred evidential mirativity 31–2, 55, 62, 73–4, 85, 87, 93–4,
inferred evidential 11–12, 31, 35–6, 54, 97–128, 153, 168–9, 200, 212, 216, 226,
62, 134–6, 145, see also inferential, 231, 240, 274–5
inference modal verb 1–4, 22, 25, 28, 43–4, 80, 84, 190
information source 1–3, 37–42, 133–42, 145, modality 1–4, 10–12, 20–1, 32, 36, 43, 44,
195–207, 209–26, see also epistemology, 80, 87, 92, 191, 197, 210–11, 214, 217, 225,
evidentials 230–1, 235, 239–42, see also epistemic
information structure 75, 80, 93–5, see also modality
backgrounding, emphasis, foreground- mode 13, 33, 241, 394
ing, focus, pivot, topic mood 1–2, 4, 9, 19, 28, 43–4, 52–3, 70–1, 210,
intention 139, 267 230–5, 239–41
intentionality 27, 30, 40 multiple sources of information 12,
interjection 262, 274–5 see also co-occurrence of eviden-
internal state, verbs of 30, 139, see also cog- tials, double marking of information
nition, verbs of; perception, verbs of source
interrogative 71–2, 80–1, 85, 182–3, 211–12,
225, 231, see also questions N
intuition 8, 21–5, 190–2, 195–7, 206, 210, names 261, 270, 275–6
218–21, 226 narrative 114–16, 120, 132–9, 142, 145–6,
irony and evidentials 81 156–7, 164–5, 178, 187, 210–16, 220–5, 227,
irrealis 44, 93, 96, 101, 214, 271 231, 235–6, 240–2, 252, 255–8,
see also genre
J narrative converb 53, 63
jokes and evidentials 81 narrative modality 232, 235, 241–2
288 Index of subjects
negation 1, 10, 59, 71, 80, 96, 106, 110–12, person 1, 4, 26, 114–20, 127, 134, 138–42,
120–3, 128, 148–52, 155–68, 211–17, 221, 145, 172–3, 192–4, 210, 213, 219–25,
224–5 246–8, 255–6, see also conjunct/
nominalization 20–2, 36, 43, 110–11, 114, disjunct person marking, first person
117–18, 122, 127, 230–41, see also eviden- effect
tiality strategy pivot 172, 176, 179, 188, 240
nominative/accusative 70, 165–6, 172–3, 210 politeness 127
non-eyewitness evidential, see non- postposition 61
firsthand evidential, unwitnessed precision, requirement for 15, 23, 27, 32,
evidential 37–43, 171, 188, 211, 220–2, 225, see also
non-finite verb 53, 59, 71–2, 84, 149, 165, 227, cultural conventions
see also finite verb prediction 10, 12, 21, 33, see also prospective
non-firsthand evidential 5–9, 14–16, 20–2, preposition 23, 250–2, 258
31, 31–4, 42, 53, 55, 63, 71–4, 79–84, 87, probability 1–2, 12, 20, 23, 25, 29, 44,
see also firsthand evidential 199–200, 206, 210–12, 216–17, see also
in narratives 36–8 epistemic modality
non-firsthand information source 230–42 prohibitive 230–1, see also negation
non-propositional evidentiality, see noun prospective 9–12, 21, 36, 152–3, 160–1, 169
phrase, evidentials in
non-visual evidential 7–9, 23, 29–8, 257 Q
non-visual information source 171–88, 195, questions 28–9, 32, 124–8, 213–17, 225, 237,
see also smelling, hearing, touching see also interrogative
noun class 174–80, 246, 250, see also gender evidentials in 1, 5–6, 9–10
noun phrase, evidentials in 2, 4, 14, 16–21, quotation 2, 11, 26, 149–52, 161–9, 211, 215,
42, 171–89, see also scope 222, 225–6, see also direct quote, self-
number systems 1, 4, 14, 52, 246, 260 quote, reported speech
quotative 7–11, 14–16, 52, 59, 63–7, 72, 75–8,
O 84–7, 133–8, 142, 145
olfactory perception 205, 220, see also
smelling R
omission of evidentials 95, 133, 136, 145 realis 96, 210, 214, 232, see also irrealis
onomatopoeia 24, 183, see also ideophone reality status, see realis, irrealis
origin of evidentials, see historical develop- reanalysis 222–3
ment of evidentials and information reasoning 9, 23, 34, 134–6, 143, 211, 216, 221
source markers reliability 12–15, 23, 26–8, 34, 37, 40–3, 74,
79, 82, 85, 97, 106, 152–3, 165–6, 169,
P 190–1, 197, 223–4, 242, 268–9
parable 265–6 remembering 154, 158, 167–8, 171, 180–3,
parallelism 261 187–8, 262, 271–3
parenthetical 2, 22–5, 27–8, 33, 43, 61–2, 67, reportative, see reported evidential
113, 145, 210–11, 222–3, 225 reported evidential 4–17, 20, 26–7, 30–4,
participle 20, 25, 53, 71–5, 148–51, 154–5, 38–9, 42, 52, 63–7, 72, 78–89, 92–102,
162–8 105–6, 133–45, 152–3, 161–3, 190–2,
passive 19–21, 25 195–9, see also hearsay
past tense 5, 10, 14–16, 19, 53–9, 66–7, in narratives 35–6
see also tense reported speech 16, 21, 26–8, 35, 101–2,
perception, verbs of 2, 3, 20, 23, 25, 61–3, 67, 106, 112, 118, 124–5, 190, 195–7, 210–13,
76, 83, 87, 104–6, 142, 165–7, 169, 204–7, 219-26, 231–2, 253–6, 262–8, see also
210, 220–1, 256, 273–5, see also sensory direct speech, indirect speech, speech
perception, verbs of report
perfect 20–1, 72, 86 responsibility 20, 27, 30, 37–8, 41, 43–4, 93,
perfective 10, 108, 113–17, 127–8, 136, 149–51, 97–8, 103
230, 240 rumour 198, 264, 266, 269
2 Index of subjects 289
S tense 1–4, 10–21, 28, 31, 36, 42–4, 52–5, 59,

sarcasm 127 64–7, 70–3, 76, 79–87, 92, 96, 108–12,
scattered coding of evidentiality 14–16, 119–20, 124, 127–8, 133, 138, 174, 181, 210,
225, 258 230, 233–6, 239–41, 246–50, 253, 257
scope 2, 10, 13, 16–19, 21, 24, 28, 59, 96, 134, thirdhand evidential 8
143, 192, 195, 198 time 2, 10–12, 42–4
secondhand, see reported evidential time reference of evidential 10–12, 93, 96–7
seeing 7–9, 16, 23, 29, 35, 134, 200, 273–6, topic 192, 196, 213, 214, see also information
see also visually acquired information, structure
visual evidential touching 9, 23, 211, 220, 262, 264, 273
self-citation 268, see self-quote truth 18, 29, 36–9, 44–5, 85–6, 190–202,
self-quote 27, 210, 223–6, 267–8 207–8, 230, 266
semi-direct speech 26–7
sensory evidential 9, 20, 23–4, 29, 38, 134, U
141, 257 uncertainty 10, 12, 16, 30, 37, 44–5, 74, 80,
sensory perception, verbs of 142, see also 113, 117–20, 127–8, 135, 143–5, 153, 157–9,
perception, verbs of 160–1, 198–9, 216–17, see also certainty,
shamanic knowledge 29, 33–6, 40, 228–9 epistemic modality
smelling 2, 7–9, 19, 23, 44, 134, 210, 213, unintentional action, see intention,
220, 273 intentionality
sorcery 41, 38, 264 unmarked evidential, see markedness
speculation 21, 24, 35, 40, 80, 81, 210–12, unmarked for evidentiality, see evidentiality
216–17, 230–1 neutral forms
speech, figurative 265–6 unwitnessed evidential 53–60, 63–7,
speech act participant (SAP) 114, 134, see also non-firsthand evidential
138–42, 145–6, see also person
speech report 72, 75–9, 87, 197–8, 231–2, V
240–1, 267, see also reported speech value of knowledge 35, 40–1, 43, 61, 270–1
spirits 33–4, 171, 179, 186–8, 223, 228, 261, ventive 248, 251–3, 258
273–4 viewpoint 219, 226
stylistic effects of evidentials 35–6, see also vision 192, 195–6, 204–6, see also seeing
genre visual evidential 1, 6–9, 12, 17, 29–32, 35–8,
subordinate clause 53, 59, 71, 149, see also see also direct evidential, firsthand
dependent clause evidential,
surprise 213–16, 226, see also mirativity visually acquired information
switch-reference 234, 238–40 in demonstratives 17–18, 34
in dreams 32–3
T visually acquired information 152, 160, 166,
tactile perception 220–1, 266, see also 171–88, 210–11, 220–3, 257–8, 261, 271–6,
touching see also seeing
tasting 8, 19, 40, 210, 220, 273 volition 30
technology and the use of evidentials 34–5,
57–8, 66–7, 269, see also media and the Z
use of evidentials zero-marking of evidentiality, see marked-
temporal meanings, see tense ness, evidentiality neutral forms

Alexandra Y. Aikhenvald, R. M. W. Dixon (2014) The Grammar of Knowledge - A Cross-Linguistic Typology-Oxford University Press

Uploaded by

Copyright:

Available Formats

You might also like

Alexandra Y. Aikhenvald, R. M. W. Dixon (2014) The Grammar of Knowledge - A Cross-Linguistic Typology-Oxford University Press

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Alexandra Y. Aikhenvald, R. M. W. Dixon (2014) The Grammar of Knowledge - A Cross-Linguistic Typology-Oxford University Press

Uploaded by

Copyright:

Available Formats

The Grammar of Knowledge

published in association with the series

Language and Culture Research Centre

1 The grammar of knowledge: a cross-linguistic view of evidentials

7 Evidentiality in Kalmyk 148

8 The non-visible marker in Dyirbal 171

9 The grammar of knowledge in Maaka (Western Chadic, Nigeria) 190

10 Expression of information source meanings in Ashéninka Perené

Index of authors 279

Gerrit J. Dimmendaal is Professor of African Studies at the University of Cologne, Germany.

-DY past tense suffix in Tatar

POT, pot potential

Sa ‘active’ S, marked like A

The grammar of knowledge:

1 Evidentiality and information source

2 Evidentiality: a bird’s-eye view3

obligatory.4 We now summarize a few points relevant for an understanding of eviden-

2.1 Evidentiality systems: a snapshot

(1) u-sú u-piniatika paá

(2) jomee tiwa na-tafi-no awa?

(3) owa na-tafi-are-ka

(4) Kuyfi miyaw-ürke-y mawida mew

In (5), the same -rke- describes what one has inferred:

The ‘information-source-neutral’ form may refer to a variety of information sources.

(6) daëd ik-o-�h

(7) nënëchokid-n ak-ak

(8) nënëchokid-n ak-a�h

(9) nënëchokid-n-bi-en ak-chito-ak

(10) nɑ13 bu33 nɑ13 bu33 ʐi33 tu33-kuɔ 33 di31 tɑ13

2.1.2 Summing up: semantic parameters in grammatical evidentiality The semantic

I. Visual covers evidence acquired through seeing.

commands (see Aikhenvald 2010: 138–40, for details). An evidential can be

(13) �hëktenamë kuen-ak-onda-�h

(14) “ɕi33 gi13 ze33” pi33 phæ33 di33

(15) nehucma o-si nana [eainca to amo-su

(16) o-si eUsvUta a'o [nehucma tena cu la

The grammatical marking of information source in the Kurtöp verb is intertwined

2.1.6 Scattered coding of information source Evidential meanings can be expressed

2.2 The geography of evidentials

Evidentials are relatively poorly represented in familiar European languages (see

2.3 The scope of evidentials: information source of a noun phrase

(17) [[mee tabori botee]-mete-moneha]NP:OotaaA awa-hamaro ama-ke

(18) m-o mongsi co oko

3 Information source through other means

inference based on results or assumption hearsay

modalities, perfects, resultatives, passives, nominalizations

hearsay inference based on results or assumption

Scheme 1 The semantic range of evidentiality strategies.

draw. Nominalizations in Aguaruna are better considered evidentiality strategies

3.2 Information source in its further guises

(22) nu-hmeta-ka du-ñami-sita-sika

(23) nu-awada-ka du-ñami-sita-sika

evidentials, employs wuna mawulam ‘in my thinking, lit. in my inside’. Ashénin-

(24) a. She will, regrettably, have to sell her car

A verb or an adjective used in parentheticals may not have a corresponding adverb.

across the world (a further discussion of their grammatical properties in European

A link with information source is obvious—yet information source is an overtone of

(25) 'o-he: gi-ba na-t�