A grammar of Awa Pit (Cuaiquer):

An indigenous language of south-western Colombia

by

Timothy Jowan Curnow

A thesis submitted for the degree of

Doctor of Philosophy
of
The Australian National University

August 1997
This thesis is entirely my own work.
 © Copyright 1997
by
Timothy Jowan Curnow
 Acknowledgements
For something which condently states it is entirely my own work, this thesis
owes a great deal to many people, on two continents.
I am grateful for the assistance and friendship of many in Colombia. None
of this would have been possible without the help of the Awa, especially my in-
formants, the residents of Pialapí and Pueblo Viejo, and the Cabildo (in par-
ticular the Governors José María and Romel), who gave me their permission to
work on their traditional language. A great deal of assistance, both technical and
personal, was provided by the teachers in Pialapí and Pueblo Viejo, Demetrio,
Carmen and Lola, and their families.
Both the Alcaldía of Ricaurte and the Colombian Ocina de Asuntos In-
dígenas (through the director Aníbal Feuillet) allowed me ocial access to the
Awa; and Rocío Calvache put me in contact with the Awa in the rst place and
assisted me greatly by telling me what to write to whom, how and when, in which
level of bureaucracy.
Thanks also to Jon Landaburu and others at the Centro Colombiano de
Estudios en Lenguas Aborígines at the University of the Andes in Bogotá, without
whose help I would never have left Australia, would never have heard of the Awa,
and would not have had any linguistics for a year  they did all that, and many
of them (especially Miguel Ángel and Nydia) still had energy left over to have
lunch with me, chat, and make me feel like I belonged there.
I probably would not have survived the eldwork without the friendship
of a few other non-Awa. The sta and researchers at La Planada, especially Luz
Mary, Pilar and Guillermo, always made me feel welcome; whenever I escaped to
Pasto, I could rely on Gloria Narváez to pretend that I was a normal, sane human
being; and Germán and Yenny and their family in Ricaurte always shifted one
bed over to make room for me when I arrived unannounced on their doorstep,
which made me really feel like part of the family!
Back in Australia there are many others who have contributed to my thesis
in many ways, though only a few can be mentioned here.
I owe any linguistic knowledge I have to the Department of Linguistics,
Faculties, at the ANU; and I have a special debt to Bob Dixon, Cindy Allen and
Sasha Aikhenvald, who supervised this thesis. The nancial support provided by
the Australian government through an Australian Postgraduate Research Award
was essential.
Friends and fellow students are vital in coping with a thesis, and I would
like to thank them all, and in particular Angela, Dorothy, Helen, Michael, Patti,
Peita, Verna, and all the others who sit outside to indulge me.
And nally I must thank Tony, who lived through it all, but had the sense
to keep out except when invited.

v
vi
 Abstract
This thesis is a descriptive grammar of Awa Pit, previously known as Cuaiquer, an
indigenous language of the Barbacoan family spoken in south-western Colombia
and north-western Ecuador. The thesis concentrates on the variety of Awa Pit
spoken in the settlements of Pialapí and Pueblo Viejo, in the Municipality of
Ricaurte, Nariño in Colombia.
Chapter 1 gives a general introduction to the Awa and discusses previous
research on Awa Pit, as well as describing the eldwork for this thesis.
The phonetics and phonology of Awa Pit are described in chapter 2. Par-
ticular issues which have been problematic in earlier analyses of the sound system
of Awa Pit are examined closely: the fricatives, [R] as an allophone of /t/, the
status of voiceless vowels, and the phone [e].
Chapter 3 begins the description of the syntax of Awa Pit, looking at issues
which are denitionally important in the remainder of the thesis. After examining
constituent order, the contrasts between main and subordinate clauses, nite and
non-nite clauses, and complements and adjuncts are established. Following this
the syntactic functions and grammatical relations of Awa Pit are discussed, and
the various predicates types are illustrated.
After a survey of the word classes in chapter 4, together with a brief
discussion of loan words, chapter 5 looks at noun phrases, postpositional phrases
and Copula complements.
The following four chapters all deal mainly with verbs. Chapter 6 con-
centrates on verb stems and derivational processes, examining ambitransitivity,
non-productive derivation including compound verbs, and productive verbal de-
rivation, whether valency-increasing or valency-preserving. A survey of verb in-
ection is one of the major themes of chapter 7; however it also discusses number
marking in verbs, which appears to be derivational rather than inectional. In
chapter 9, the various tense, aspect and mood inections are discussed more fully.
Chapter 8 is an examination of one of the most interesting features of Awa
Pit  the person-marking system. There is a binary division of `person' in Awa
Pit verbs into Locutor (rst person in statements, second person in questions)
and Non-locutor (second and third person in statements, rst and second person
in questions), with the split being quite similar to the conjunct/disjunct division
found in some Tibeto-Burman languages. To complicate matters further, Awa
Pit relies partly on grammatical relations, partly on semantic roles, and on a
hierarchy (Locutor > Non-locutor) in determining which person marker to use
in any situation.
After discussing verbs, complex sentences are examined. Complement
clauses, adverbial clauses, relative clauses and clausal nominalizations  the four
types of subordinate clause  are discussed in chapter 10, followed by an exam-
ination of complex non-subordinate phenomena in chapter 11: mainauxiliary
structures, Serial Verbs, Conjoined Clauses, and juxtaposed clauses.
The interrogative and negative structures of Awa Pit, many of which are
interrelated, are looked at in depth in chapter 12.
vii
 Chapter 13 examines adjuncts and adverbs in Awa Pit: temporal, cir-
cumstantial and locational adjuncts; manner adverbials; degree adverbs; and the
structures used for comparison.
Finally the discourse particles are discussed in chapter 14. The majority
of the chapter is dedicated to the Topic marker, which is very frequently used in
Awa Pit, although the other particles are also examined.

viii
Contents
List of Figures xvii
List of Tables xix
List of Abbreviations xxi
1 Introduction 1
1.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.2 Physical environment and population . . . . . . . . . . . . . . . . 1
1.3 Previous anthropological and social research . . . . . . . . . . . . 4
1.4 Language classication . . . . . . . . . . . . . . . . . . . . . . . . 8
1.5 Previous linguistic research . . . . . . . . . . . . . . . . . . . . . . 13
1.6 Origins and areal typology . . . . . . . . . . . . . . . . . . . . . . 15
1.7 Language death? . . . . . . . . . . . . . . . . . . . . . . . . . . . 18
1.8 Fieldwork and data collection . . . . . . . . . . . . . . . . . . . . 20
2 Phonetics and phonology 23
2.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
2.2 Consonants . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
2.2.1 Stops . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
2.2.1.1 Geminate stops . . . . . . . . . . . . . . . . . . . 26
2.2.1.2 Stops and sandhi phenomena . . . . . . . . . . . 27
2.2.2 Fricatives (and aricates) . . . . . . . . . . . . . . . . . . 29
2.2.3 Voiceless lateral fricative . . . . . . . . . . . . . . . . . . . 32
2.2.4 Lateral approximant . . . . . . . . . . . . . . . . . . . . . 33
2.2.5 Nasals . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34
2.2.5.1 The phoneme /m/ . . . . . . . . . . . . . . . . . 34
2.2.5.2 The phoneme /n/ . . . . . . . . . . . . . . . . . 34
2.2.5.3 The phoneme /N/ . . . . . . . . . . . . . . . . . . 35
2.3 Glides or semivowels . . . . . . . . . . . . . . . . . . . . . . . . . 38
2.4 Vowels . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40
2.4.1 Voiced vowels . . . . . . . . . . . . . . . . . . . . . . . . . 40
2.4.1.1 The allophone [e] . . . . . . . . . . . . . . . . . . 42
2.4.1.2 Long vowels . . . . . . . . . . . . . . . . . . . . 42
2.4.2 Voiceless vowels . . . . . . . . . . . . . . . . . . . . . . . . 43
2.5 Phonotactics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 46
 2.5.1 Syllable structure . . . . . . . . . . . . . . . . . . . . . . . 46
2.5.2 Stress . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 46
2.6 Orthography . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 46
3 Basic clause structure 49
3.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49
3.2 Constituent order . . . . . . . . . . . . . . . . . . . . . . . . . . . 50
3.3 Clause types . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 52
3.3.1 Main and subordinate clauses . . . . . . . . . . . . . . . . 52
3.3.2 Finite and non-nite clauses . . . . . . . . . . . . . . . . . 53
3.3.3 Summary of clause types . . . . . . . . . . . . . . . . . . . 59
3.4 Complements and adjuncts . . . . . . . . . . . . . . . . . . . . . . 60
3.5 Syntactic functions and grammatical relations . . . . . . . . . . . 62
3.5.1 A, S and O syntactic functions . . . . . . . . . . . . . . . 62
3.5.2 Subject . . . . . . . . . . . . . . . . . . . . . . . . . . . . 63
3.5.2.1 Constituent order . . . . . . . . . . . . . . . . . . 64
3.5.2.2 Case marking . . . . . . . . . . . . . . . . . . . . 64
3.5.2.3 Number marking . . . . . . . . . . . . . . . . . . 66
3.5.2.4 Person marking . . . . . . . . . . . . . . . . . . . 66
3.5.2.5 The Conjoined Clause construction . . . . . . . . 67
3.5.2.6 The same-Subject purposive . . . . . . . . . . . . 68
3.5.2.7 Complements of intention . . . . . . . . . . . . . 69
3.5.2.8 Agentive adjectivizations . . . . . . . . . . . . . . 69
3.5.2.9 Agentive nominalizations . . . . . . . . . . . . . 70
3.5.2.10 Deverbal adjectives and resultatives . . . . . . . . 71
3.5.3 Object and Second Object . . . . . . . . . . . . . . . . . . 71
3.5.3.1 Case marking . . . . . . . . . . . . . . . . . . . . 72
3.5.3.2 Number marking . . . . . . . . . . . . . . . . . . 75
3.5.3.3 The First Person Object Imperative . . . . . . . 75
3.5.4 Copula Complement . . . . . . . . . . . . . . . . . . . . . 77
3.5.5 Core, oblique and external functions . . . . . . . . . . . . 78
3.6 Predicate types . . . . . . . . . . . . . . . . . . . . . . . . . . . . 79
3.6.1 Purely verbal predicates . . . . . . . . . . . . . . . . . . . 79
3.6.2 Copula Complement predicates . . . . . . . . . . . . . . . 79
4 Word classes 81
4.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 81
4.2 Phonological words and grammatical words . . . . . . . . . . . . . 81
4.3 Nominals . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 82
4.3.1 Nouns . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 82
4.3.2 Time nouns and place nouns . . . . . . . . . . . . . . . . . 83
4.3.3 Personal pronouns . . . . . . . . . . . . . . . . . . . . . . 85
4.3.4 Demonstrative pronouns . . . . . . . . . . . . . . . . . . . 87
4.3.5 Relational nouns . . . . . . . . . . . . . . . . . . . . . . . 87
4.3.6 Nominal postmodiers . . . . . . . . . . . . . . . . . . . . 88
4.4 Adjectives . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 90
x
 4.4.1 Distinguishing adjectives from nominals . . . . . . . . . . 90
4.4.2 Descriptive adjectives . . . . . . . . . . . . . . . . . . . . . 92
4.4.3 Quantiers . . . . . . . . . . . . . . . . . . . . . . . . . . . 93
4.4.4 Demonstrative adjectives . . . . . . . . . . . . . . . . . . . 94
4.4.5 Possessive adjectives . . . . . . . . . . . . . . . . . . . . . 94
4.5 Verbs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 94
4.5.1 Establishing subcategorization frames . . . . . . . . . . . . 96
4.5.2 Zero-argument (impersonal) verbs . . . . . . . . . . . . . . 98
4.5.3 One-argument (intransitive) verbs . . . . . . . . . . . . . . 98
4.5.3.1 Active intransitive verbs . . . . . . . . . . . . . . 98
4.5.3.2 Stative intransitive verbs . . . . . . . . . . . . . . 99
4.5.4 Two-argument verbs . . . . . . . . . . . . . . . . . . . . . 100
4.5.4.1 Transitive verbs . . . . . . . . . . . . . . . . . . . 100
4.5.4.2 Two-argument stative verbs . . . . . . . . . . . . 100
4.5.4.3 Two-argument active, non-transitive verbs . . . . 101
4.5.5 Three-argument verbs . . . . . . . . . . . . . . . . . . . . 101
4.5.5.1 Ditransitive verbs . . . . . . . . . . . . . . . . . . 101
4.5.5.2 Three-argument verbs with an oblique argument 102
4.5.5.3 Three-argument stative verbs . . . . . . . . . . . 103
4.5.6 Compound verbs . . . . . . . . . . . . . . . . . . . . . . . 103
4.6 Postpositions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 104
4.7 Time adverbs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 107
4.8 Place adverbs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 108
4.9 Manner adverbs . . . . . . . . . . . . . . . . . . . . . . . . . . . . 109
4.10 Other adverb-like words . . . . . . . . . . . . . . . . . . . . . . . 109
4.10.1 The negative marker . . . . . . . . . . . . . . . . . . . . . 110
4.10.2 Degree adverbs . . . . . . . . . . . . . . . . . . . . . . . . 110
4.10.3 The comparative marker . . . . . . . . . . . . . . . . . . . 110
4.11 Discourse particles . . . . . . . . . . . . . . . . . . . . . . . . . . 110
4.12 Interjections . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 111
4.13 Cross-classication of words . . . . . . . . . . . . . . . . . . . . . 111
4.14 Borrowed words . . . . . . . . . . . . . . . . . . . . . . . . . . . . 112
5 NPs, PPs and Copula Complements 115
5.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 115
5.2 Noun phrases . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 115
5.2.1 Lexical nominalizations . . . . . . . . . . . . . . . . . . . . 116
5.2.2 Reexives . . . . . . . . . . . . . . . . . . . . . . . . . . 116
5.2.3 Modication of nouns . . . . . . . . . . . . . . . . . . . . . 117
5.2.3.1 Deverbal adjectives . . . . . . . . . . . . . . . . . 119
5.2.3.2 Purposive adjectivizations . . . . . . . . . . . . . 122
5.2.4 Possessive constructions . . . . . . . . . . . . . . . . . . 122
5.2.4.1 Alienable possession and kinship relations . . . . 125
5.2.4.2 Wholepart . . . . . . . . . . . . . . . . . . . . . 125
5.2.4.3 Relational nouns . . . . . . . . . . . . . . . . . . 126
xi
 5.2.4.4 Locationobject . . . . . . . . . . . . . . . . . . . 128
5.2.4.5 Purposeobject . . . . . . . . . . . . . . . . . . . 129
5.2.4.6 Materialobject . . . . . . . . . . . . . . . . . . . 129
5.2.4.7 Specicgeneral . . . . . . . . . . . . . . . . . . . 129
5.2.5 Headless NPs . . . . . . . . . . . . . . . . . . . . . . . . . 130
5.2.6 Plural marking and nominalizations . . . . . . . . . . . . 131
5.3 Copula Complements . . . . . . . . . . . . . . . . . . . . . . . . . 132
5.4 Postpositional phrases . . . . . . . . . . . . . . . . . . . . . . . . 133
5.4.1 Postpositions and pronouns . . . . . . . . . . . . . . . . . 134
5.4.2 Ta: The Accusative marker . . . . . . . . . . . . . . . . . 134
5.4.3 Ta: Locative/allative . . . . . . . . . . . . . . . . . . . . . 135
5.4.4 Pa: Locative/allative . . . . . . . . . . . . . . . . . . . . . 137
5.4.5 Mal: Locative/allative . . . . . . . . . . . . . . . . . . . . 139
5.4.6 Ki: Locative . . . . . . . . . . . . . . . . . . . . . . . . . . 140
5.4.7 Kima: `Until' . . . . . . . . . . . . . . . . . . . . . . . . . 141
5.4.8 Pa: Possessive . . . . . . . . . . . . . . . . . . . . . . . . . 141
5.4.9 Kasa: `With' . . . . . . . . . . . . . . . . . . . . . . . . . 142
5.4.10 Akwa: `Because' . . . . . . . . . . . . . . . . . . . . . . . 143
5.4.11 Patsa: `Like' . . . . . . . . . . . . . . . . . . . . . . . . . . 143
5.4.12 Kana: Semblative . . . . . . . . . . . . . . . . . . . . . . . 144
5.5 NPs and PPs in apposition . . . . . . . . . . . . . . . . . . . . . . 145
6 Verb derivation and valency changes 147
6.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 147
6.2 Ambitransitive verbs . . . . . . . . . . . . . . . . . . . . . . . . . 148
6.2.1 S=A ambitransitive verb roots . . . . . . . . . . . . . . . . 149
6.2.2 S=O ambitransitive verb roots . . . . . . . . . . . . . . . . 149
6.3 Non-productive derivations . . . . . . . . . . . . . . . . . . . . . . 150
6.3.1 Non-productive a, ta and na suxes . . . . . . . . . . . . 150
6.3.2 One-word compounds . . . . . . . . . . . . . . . . . . . . . 151
6.3.3 Two-word compound verbs . . . . . . . . . . . . . . . . . . 152
6.3.3.1 Compound verbs are two words . . . . . . . . . . 153
6.3.3.2 Compound verbs are one word . . . . . . . . . . 153
6.3.3.3 Compounds as incorporation . . . . . . . . . . . 155
6.3.3.4 Loan verbs . . . . . . . . . . . . . . . . . . . . . 156
6.4 Productive derivations . . . . . . . . . . . . . . . . . . . . . . . . 157
6.4.1 Valency-changing productive derivations . . . . . . . . . . 159
6.4.1.1 Causative . . . . . . . . . . . . . . . . . . . . . . 164
6.4.1.2 Auxiliative . . . . . . . . . . . . . . . . . . . . . 165
6.4.2 Valency-preserving productive derivations . . . . . . . . . 166
6.4.2.1 Desiderative . . . . . . . . . . . . . . . . . . . . . 166
6.4.2.2 Prospective aspect . . . . . . . . . . . . . . . . . 168
6.4.2.3 Inceptive aspect . . . . . . . . . . . . . . . . . . 171
6.4.2.4 Learnative . . . . . . . . . . . . . . . . . . . . . . 173

xii
7 Verb inection and number 175
7.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 175
7.2 Inections . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 176
7.2.1 Tense inection . . . . . . . . . . . . . . . . . . . . . . . . 176
7.2.2 Aspect inection . . . . . . . . . . . . . . . . . . . . . . . 176
7.2.3 Mood inection . . . . . . . . . . . . . . . . . . . . . . . . 177
7.2.4 Negative and interrogative inection . . . . . . . . . . . . 177
7.2.5 Serial Verb inection . . . . . . . . . . . . . . . . . . . . . 177
7.2.6 Person inection . . . . . . . . . . . . . . . . . . . . . . . 178
7.2.7 Subordinating inection . . . . . . . . . . . . . . . . . . . 178
7.2.8 Other non-nite inections . . . . . . . . . . . . . . . . . . 178
7.2.8.1 Innitive . . . . . . . . . . . . . . . . . . . . . . 178
7.2.8.2 Imperfective Participle . . . . . . . . . . . . . . . 179
7.2.8.3 Perfective Participle . . . . . . . . . . . . . . . . 179
7.2.8.4 The adjectivizers . . . . . . . . . . . . . . . . . . 180
7.2.9 Order of inections . . . . . . . . . . . . . . . . . . . . . . 180
7.3 Plural marking . . . . . . . . . . . . . . . . . . . . . . . . . . . . 181
7.3.1 Plural marking in active verbs . . . . . . . . . . . . . . . . 182
7.3.2 Plural marking in stative verbs . . . . . . . . . . . . . . . 184
7.3.3 Plural marking as derivational . . . . . . . . . . . . . . . . 185
8 Person marking 187
8.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 187
8.2 Locutor and Non-locutor . . . . . . . . . . . . . . . . . . . . . . . 189
8.3 The Past paradigm . . . . . . . . . . . . . . . . . . . . . . . . . . 192
8.3.1 Locutor Subject marking . . . . . . . . . . . . . . . . . . . 193
8.3.2 Locutor Undergoer marking . . . . . . . . . . . . . . . . . 194
8.3.3 Non-locutor marking . . . . . . . . . . . . . . . . . . . . . 198
8.4 The non-Past paradigm . . . . . . . . . . . . . . . . . . . . . . . . 199
8.4.1 Person marking on non-verb constituents . . . . . . . . . . 202
8.5 Discussion and comparisons . . . . . . . . . . . . . . . . . . . . . 203
8.5.1 Person marking, but not cross-referencing . . . . . . . . . 203
8.5.2 Inverses and passives . . . . . . . . . . . . . . . . . . . . . 205
8.5.3 Anticipation and Sherpa . . . . . . . . . . . . . . . . . . 206
8.5.4 Conjunct/disjunct and Kathmandu Newari . . . . . . . . 206
8.5.5 Interior states in Japanese and Korean . . . . . . . . . . 208
8.5.6 Epistemic authority and Kathmandu Newari . . . . . . . 209
8.5.7 Evidentiality and Akha . . . . . . . . . . . . . . . . . . . 210
8.5.8 Mirativity and Lhasa Tibetan . . . . . . . . . . . . . . . 213
8.6 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 216
9 Simple tense, aspect and mood 219
9.1 Introduction . . . . . . . . . ... . . . . . . . . . . . . . . . . . . 219
9.2 Tense . . . . . . . . . . . . . ... . . . . . . . . . . . . . . . . . . 219
9.2.1 Past tense . . . . . . ... . . . . . . . . . . . . . . . . . . 220
9.2.2 Future tense . . . . . ... . . . . . . . . . . . . . . . . . . 222
xiii
 9.2.3 Present tense . . . . . . . . . . . . . . . . . . . . . . . . . 223
9.3 Aspect . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 224
9.3.1 Imperfective aspect . . . . . . . . . . . . . . . . . . . . . . 224
9.3.1.1 On-going activities and states . . . . . . . . . . . 225
9.3.1.2 Habitual activities and states . . . . . . . . . . . 226
9.3.1.3 Scheduled future . . . . . . . . . . . . . . . . . . 227
9.3.1.4 Generic statements . . . . . . . . . . . . . . . . . 228
9.3.2 Completive aspect . . . . . . . . . . . . . . . . . . . . . . 229
9.3.2.1 Completive aspect with telic clauses . . . . . . . 230
9.3.2.2 Completive aspect with atelic clauses . . . . . . . 231
9.3.2.3 Completive aspect with inchoative propositions . 232
9.3.2.4 Completive aspect with stative propositions . . . 232
9.3.2.5 Summary of the Completive aspect . . . . . . . . 233
9.3.3 Terminative aspect . . . . . . . . . . . . . . . . . . . . . . 234
9.3.3.1 Terminative aspect as a nal inection . . . . . . 234
9.3.3.2 Terminative aspect and Past tense . . . . . . . . 235
9.3.3.3 Terminative aspect and Future tense . . . . . . . 235
9.3.3.4 Terminative aspect and Imperfective aspect . . . 236
9.3.3.5 Terminative aspect and Completive aspect . . . . 236
9.3.3.6 Summary of the Terminative aspect . . . . . . . 238
9.4 Mood . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 238
9.4.1 Indicative . . . . . . . . . . . . . . . . . . . . . . . . . . . 239
9.4.2 Obligation and necessity . . . . . . . . . . . . . . . . . . . 239
9.4.2.1 Obligative . . . . . . . . . . . . . . . . . . . . . . 239
9.4.2.2 Necessitive . . . . . . . . . . . . . . . . . . . . . 240
9.4.3 Potential and Negative Potential moods . . . . . . . . . . 242
9.4.4 Counterfactual irrealis . . . . . . . . . . . . . . . . . . . . 243
9.4.5 Directives . . . . . . . . . . . . . . . . . . . . . . . . . . . 243
9.4.5.1 Imperatives . . . . . . . . . . . . . . . . . . . . . 243
9.4.5.2 Prohibitives . . . . . . . . . . . . . . . . . . . . . 247
9.4.5.3 Hortatives . . . . . . . . . . . . . . . . . . . . . . 248
10 Subordinate clauses 251
10.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 251
10.2 Complement clauses . . . . . . . . . . . . . . . . . . . . . . . . . 253
10.2.1 Complement clause 1: Sentence-like complement . . . . . . 253
10.2.2 Complement clause 2: Indirect questions . . . . . . . . . 256
10.2.3 Complement clause 3 . . . . . . . . . . . . . . . . . . . . . 259
10.2.3.1 Complement clause 3, complement of a verb . . . 259
10.2.3.2 Complement clause 3, complement of a postposition262
10.2.4 Complement clause 4 . . . . . . . . . . . . . . . . . . . . . 265
10.2.4.1 Complement clause 4 as Object: Intention . . . 265
10.2.4.2 Complement clause 4 as Subject: Evaluative pre-
dicates . . . . . . . . . . . . . . . . . . . . . . . 266
10.3 Adverbial clauses . . . . . . . . . . . . . . . . . . . . . . . . . . . 267
xiv
 10.3.1 Same-Subject purposive . . . . . . . . . . . . . . . . . . . 268
10.3.2 Dierent-Subject purposive . . . . . . . . . . . . . . . . . 269
10.3.3 After clauses . . . . . . . . . . . . . . . . . . . . . . . . . 271
10.3.4 Simultaneity clauses . . . . . . . . . . . . . . . . . . . . . 272
10.3.5 Concessive clauses . . . . . . . . . . . . . . . . . . . . . . 276
10.3.6 Counterfactual clauses . . . . . . . . . . . . . . . . . . . . 276
10.3.7 Absolute clauses . . . . . . . . . . . . . . . . . . . . . . . 278
10.3.7.1 Temporal, causal or conditional absolute clauses . 278
10.3.7.2 Accompanying circumstance absolute clauses . . 279
10.3.7.3 No-Subject purposive absolute clauses . . . . . . 280
10.4 Relative clauses . . . . . . . . . . . . . . . . . . . . . . . . . . . . 280
10.5 Nominalizations . . . . . . . . . . . . . . . . . . . . . . . . . . . . 284
11 Complex non-subordinate clauses 291
11.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 291
11.2 MainAuxiliary constructions . . . . . . . . . . . . . . . . . . . . 292
11.2.1 Imperfective Participle and auxiliaries . . . . . . . . . . . 293
11.2.2 Extended Perfective Participle and auxiliaries . . . . . . . 294
11.2.2.1 The Past Anterior . . . . . . . . . . . . . . . . . 295
11.2.2.2 The Resultative . . . . . . . . . . . . . . . . . . . 296
11.3 Negatives . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 299
11.4 The hearsay evidential . . . . . . . . . . . . . . . . . . . . . . . . 300
11.5 Serial Verbs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 302
11.6 Perfective aspect through Serial Verbs . . . . . . . . . . . . . . . 305
11.7 The Conjoined Clauses construction . . . . . . . . . . . . . . . . . 308
11.8 Juxtaposition of clauses . . . . . . . . . . . . . . . . . . . . . . . 310
12 Interrogatives and negatives 313
12.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 313
12.2 Interrogative/negative content words . . . . . . . . . . . . . . . . 314
12.2.1 M1n `who, no-one' . . . . . . . . . . . . . . . . . . . . . . . 316
12.2.2 Shi `what, nothing' . . . . . . . . . . . . . . . . . . . . . . 317
12.2.3 M1n= `where, nowhere' . . . . . . . . . . . . . . . . . . . . 318
12.2.4 M1n `which' . . . . . . . . . . . . . . . . . . . . . . . . . . 319
12.2.5 Yawa `how much, how many' . . . . . . . . . . . . . . . . 319
12.2.6 Mizha `how' . . . . . . . . . . . . . . . . . . . . . . . . . . 320
12.2.7 Mizha(pa)ka `when' and mizhuta `when, never' . . . . . . . 321
12.2.8 Shin `why' . . . . . . . . . . . . . . . . . . . . . . . . . . . 322
12.3 Polar questions . . . . . . . . . . . . . . . . . . . . . . . . . . . . 323
12.3.1 Polar questions marked by a sux . . . . . . . . . . . . . 323
12.3.2 Polar questions with ki or sa . . . . . . . . . . . . . . . . . 324
12.4 Tag questions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 328
12.4.1 Binary-choice tag questions . . . . . . . . . . . . . . . . . 328
12.4.2 Multiple-choice tag questions . . . . . . . . . . . . . . . . 330
12.5 Clausal negation . . . . . . . . . . . . . . . . . . . . . . . . . . . 331
12.5.1 Negatives with ma . . . . . . . . . . . . . . . . . . . . . . 332
xv
 12.5.2 The negative copula . . . . . . . . . . . . . . . . . . . . . 333
12.5.3 The negative auxiliary construction . . . . . . . . . . . . . 334
12.6 Non-clausal negation . . . . . . . . . . . . . . . . . . . . . . . . . 335
12.7 Interrogatives and negatives . . . . . . . . . . . . . . . . . . . . . 337
13 Adjuncts and adverbs 339
13.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 339
13.2 Temporal adjuncts . . . . . . . . . . . . . . . . . . . . . . . . . . 339
13.2.1 Time adverbs . . . . . . . . . . . . . . . . . . . . . . . . . 340
13.2.2 Nouns with temporal reference . . . . . . . . . . . . . . . . 343
13.2.3 Number of times . . . . . . . . . . . . . . . . . . . . . . . 345
13.2.4 Temporal postpositional phrases . . . . . . . . . . . . . . . 346
13.2.5 Temporal adverbial clauses . . . . . . . . . . . . . . . . . . 347
13.3 Circumstantial adjuncts . . . . . . . . . . . . . . . . . . . . . . . 347
13.3.1 Circumstantial postpositional phrases . . . . . . . . . . . . 347
13.3.2 Circumstantial adverbial clauses . . . . . . . . . . . . . . . 348
13.4 Locational adjuncts . . . . . . . . . . . . . . . . . . . . . . . . . . 348
13.5 Manner adverbials . . . . . . . . . . . . . . . . . . . . . . . . . . 349
13.6 Degree adverbs . . . . . . . . . . . . . . . . . . . . . . . . . . . . 352
13.7 Comparatives . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 353
14 Discourse clitics 357
14.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 357
14.2 The Topic marker . . . . . . . . . . . . . . . . . . . . . . . . . . . 358
14.2.1 The Topic marker on complements . . . . . . . . . . . . . 360
14.2.2 The Topic marker on adjuncts . . . . . . . . . . . . . . . . 364
14.2.3 External topics . . . . . . . . . . . . . . . . . . . . . . . . 367
14.2.4 The Topic marker on complete predications . . . . . . . . 368
14.3 The Restrictive marker . . . . . . . . . . . . . . . . . . . . . . . . 371
14.4 The Additive marker . . . . . . . . . . . . . . . . . . . . . . . . . 372
14.5 The Interrogative marker . . . . . . . . . . . . . . . . . . . . . . . 374
14.6 The Temporal marker . . . . . . . . . . . . . . . . . . . . . . . . 376
14.7 The Emphasis marker . . . . . . . . . . . . . . . . . . . . . . . . 378
Bibliography 381

xvi
List of Figures
1.1 Map of the Awa region . . . . . . . . . . . . . . . . . . . . . . . . 2
1.2 The short path and the Cabildo of Pialapí/Pueblo Viejo . . . . 3
1.3 Languages of southern Colombia and northern Ecuador . . . . . . 17
List of Tables
2.1 Consonant and glide phonemes of Awa Pit . . . . . . . . . . . . . 24
2.2 Environments of fricative and aricate allophones . . . . . . . . . 29
3.1 The order of basic clause constituents . . . . . . . . . . . . . . . . 50
3.2 The copula and verb suxes . . . . . . . . . . . . . . . . . . . . . 58
4.1 Awa kinship terms . . . . . . . . . . . . . . . . . . . . . . . . . . 84
4.2 Time and place nouns . . . . . . . . . . . . . . . . . . . . . . . . 85
4.3 Personal pronouns . . . . . . . . . . . . . . . . . . . . . . . . . . 86
4.4 Relational nouns . . . . . . . . . . . . . . . . . . . . . . . . . . . 89
4.5 Nominal postmodiers . . . . . . . . . . . . . . . . . . . . . . . . 89
4.6 Examples of descriptive adjectives in adjective types . . . . . . . . 92
4.7 Quantiers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 93
4.8 Postural/locational verbs in Awa Pit . . . . . . . . . . . . . . . . 99
4.9 Examples of compound verbs . . . . . . . . . . . . . . . . . . . . 104
4.10 Postpositions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 104
4.11 Time adverbs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 108
4.12 Manner adverbs . . . . . . . . . . . . . . . . . . . . . . . . . . . . 109
4.13 Discourse particles . . . . . . . . . . . . . . . . . . . . . . . . . . 110
5.1 Basic noun phrase structure . . . . . . . . . . . . . . . . . . . . . 117
6.1 Verb pairs related through the suxes a, ta or na . . . . . . . . . 151
6.2 Verb pairs related through the sux ap, pizh, pyan and kul . . . 152
7.1 Schema to cover possible inectional combinations . . . . . . . . . 181
7.2 Unspecied number and plural forms of the copula verb i . . . . . 184
8.1 The distribution of Locutor and Non-locutor . . . . . . . . . . . . 191
8.2 Allomorphs of the person markers in the Past system . . . . . . . 193
8.3 Allomorphs of the person markers in the non-Past system . . . . . 200
8.4 Conjunct and disjunct forms for Lhasa Tibetan imperfectives . . . 213
8.5 Conjunct and disjunct forms for Lhasa Tibetan perfectives . . . . 213
10.1 Subordinate clause features . . . . . . . . . . . . . . . . . . . . . 252
10.2 Features of the four dierent complement clause types . . . . . . . 253
12.1 Interrogative/negative content words and indenite words . . . . . 314
12.2 Translation equivalents of negative mainauxiliary constructions . 335
14.1 The discourse clitics . . . . . . . . . . . . . . . . . . . . . . . . . 357
14.2 Forms of the Restrictive marker . . . . . . . . . . . . . . . . . . . 371

xx
List of Abbreviations
The following abbreviations are used in glosses throughout this thesis.
. Separates two meanings in portmanteau morpheme
: Between words in a multi-word gloss
- Separates morphemes
= Separates clitics
() `Covert', unmarked meaning element
1 1st person
2 2nd person
3 3rd person
a Transitive Subject (approximately); see section 3.5.1
acc Accusative case
add Additive marker
adjzr Adjectivizer
after After marker
and Conjoined Clause marker
anter Past Anterior auxiliary
aug Augmentative
caus Causative
cntrfc Counterfactual marker
coll Collective action sux
Comp Comparative marker
comp Completive aspect
concess Concessive marker
CopComp Copula complement
cpltzr Complementizer
DescrAdj Descriptive adjective
DemAdj Demonstrative adjective
desid Desiderative
dim Diminutive
drop Perfective Serial Verb kway-
dspurp Dierent-Subject purposive marker
du Dual
dummy Dummy sux (with the Desiderative)
emph Emphasis marker
fut Future tense
give Perfective Serial Verb ta-
help Auxiliative
hort Hortative
imp Imperative
impf Imperfective aspect
impfpart Imperfective Participle
incep Inceptive aspect
inf Innitive
inter Interrogative marker
irr Irrealis mood
learn Learnative
loc Locative postposition
locut Locutor person marker
necess Necessitive mood
neg Negative
negadjzr Negative Adjectivizer
negpot Negative Potential mood
nmlzr Nominalizer
nom Nominative case
nonfut Non-future tense
nonlocut Non-locutor person marker
np Noun phrase
o (Transitive) Object (approximately); see section 3.5.1
Obj, obj Object
Obj2 Second Object
oblig Obligative mood
part The non-nite form wal, perhaps a participle
past Past tense
pfpart Perfective Participle
pl Plural
plt:imp Polite Imperative
poss Possessive postposition
PossAdj Possessive adjective
pot Potential mood
pp Postpositional phrase
prohib Prohibitive
prosp Prospective aspect
Q Question marker
Q:unsure Tentative question marker
Quant Quantier
rest Restrictive marker
s Intransitive Subject (approximately); see section 3.5.1
sg Singular
Subj, subj Subject
sv Serial Verb marker
temp Temporal marker
xxii
term Terminative aspect
throw Perfective Serial Verb kyan-
top Topic marker
under Undergoer
V Verb
when Simultaneity marker

xxiii
Chapter 1
Introduction
1.1 Introduction
Awa Pit is the traditional language of the Awa, an indigenous group who live
in the border regions of Colombia and Ecuador in South America, between the
Andes and the coast.
Throughout this thesis, the indigenous group will be referred to as the
Awa (literally `person' or `people') and the language as Awa Pit (`people mouth'
or `people language'), as these terms are becoming more popular for both outside
researchers and the Awa themselves. While the simple label Awa is generally
sucient, if those Awa who speak Awa Pit must distinguish themselves from other
groups, they use the label 1nkal awa `mountain people'. A more traditional
label (for outside research purposes) for both the group and the language is
Cuaiquer (Coaiquer, Cuayquer, Kwaiker, Kwayquer, etc.), and indeed this label
appears to be the ocial designation of the group for the Revista Colombiana de
Antropología (Colombian Journal of Anthropology), though they allow Awa in
parentheses (Fernández 198990).
The Awa themselves generally only consider Cuaiquer as part of a place-
name, the settlement of Cuaiquer Viejo. While it has sometimes been claimed
that this is the capital city of the Awa (Ortiz 1938:557), Nuestra Señora de
Cuaiquer (modern Cuaiquer Viejo, built on the River Güiza, formerly the River
Cuaiquer) was founded around the year 1600 by Garcia Tulcanaza, a cacique
(`chief') of the Pasto group who worked for the Spanish (Aragón 1974:69).

1.2 Physical environment and population

This section will give only a very general summary of the environment and ma-
terial conditions in which the Awa live  for many more details, see especially
Cerón Solarte (1986).
The Awa live in south-western Colombia and north-western Ecuador, on
the western slopes of the Andes, between about 500 metres and 1 500 metres
above sea level (Cerón Solarte 1986:13); to be more precise, in an area from
approximately 78°45 W to 77°45 W longitude, and 1°45 N to 1°00 N latitude
0 0 0 0
2 CHAPTER 1. INTRODUCTION

Figure 1.1: Map of the Awa region (based on Haug 1994:87)

(Martínez Santacruz 1992:89)  see Figure 1.1. The majority of the Awa live
in the department of Nariño in Colombia, with some in the provinces of Carchi,
Esmeraldas and Imbabura in Ecuador.
The eldwork on which this study is based was carried out in 1994, in the
settlements of Pialapí and Pueblo Viejo, in the area controlled by the Cabildo
Integrado (`integrated local indigenous government') of Pialapí/Pueblo Viejo, in
the municipality of Ricaurte, Nariño, Colombia. The Cabildo consists of a num-
ber of settlement areas around the settlements of Pialapí and Pueblo Viejo, which
are on what is known locally as the short path, a walking trail which connects
Chucunés to Altaquer, both of which are on the PiedranchaJunín stretch of the
PastoTumaco highway. The short path runs to the south-west of the high-
way (see Figure 1.2). The section of the trail from Chucunés to La Planada is
gravel, and can be negotiated by car. The Cabildo of Pialapí/Pueblo Viejo is the
most south-easterly modern settlement area of the Awa, and is relatively close to
non-indigenous settlements, only about 5 or 6 hours walk from the PastoTumaco
highway. The Cabildo was formed in 1990; some of the region is in the Resguardo
(`reservation') of Pialapí, ocially constituted in 1993.
The physical environment in which the Awa live is extremely mountainous,
1.2. PHYSICAL ENVIRONMENT AND POPULATION 3

Figure 1.2: A schematic diagram of the short path and the Cabildo Integrado
de Pialapí/Pueblo Viejo (not to scale, directions not precise)

and there is a very high rainfall, up to 7 or 8 metres annually. Access to the Awa
areas is necessarily by foot or donkey, with only dirt tracks crossing the region.
The Awa are traditionally slash-and-mulch agriculturalists; the main crops
are plantains (related to bananas), corn, sugarcane and beans. More recently,
some Awa have begun clearing areas of jungle to pasture, and farming cattle.
The Awa do not live in villages, but rather in a scattered settlement pat-
tern, with neighbouring houses often being 2 or 3 kilometres apart, and settle-
ment regions being 3 to 8 hours walk from one another (Cerón Solarte 1986:13).
In the region of Pialapí and Pueblo Viejo, the population density is higher than
average, with the two settlements being only about an hour apart, and houses
much closer together, although still usually 15 minutes or more apart. Few houses
are built on the main trails, with most being at least 5 minutes, sometimes up to
half an hour, o a main trail.
Often families will have a main house, but may own property at quite
some distance also, and have a secondary house there. At particular times of the
year, the entire family may move to the secondary house for a period of time in
order to work the land surrounding it. Additionally, many families have huts on
more distant parts of their properties, and various family members may live in
these huts during particular points of the agricultural cycle  for example while
clearing or planting an area, or during harvest.
The general inaccessibility of the region, the scattered settlement, the fre-
quent shifting of abode, and also the secrecy of the Awa (discussed below) have
all often made research with the Awa dicult, and it has been especially dif-
4 CHAPTER 1. INTRODUCTION

cult to obtain accurate population gures. Cerón Solarte (1986:1617) states

that there are 4 366 Awa in Colombia and 951 in Ecuador, with each settlement
being given a specic, listed population; Martínez Santacruz (1992) believes that
the population of Awa in Colombia is 6 549 (although she notes that only about
fty percent of her census data is recent); based on surveys carried out by a vari-
ety of Colombian governmental agencies, Henriksen & Obando Ordóñez (1985:xii)
suggest an estimate of 20 000 for Colombia;1 while based on an report of the Min-
isterio del Gobierno (Department of the Interior) from 1980, Telban (1988:337)
states that there are approximately 25 650 Awa.

1.3 Previous anthropological and social research

There have been three general discussion papers on the Awa produced for govern-
ments: Martínez et al. (1984) for the Colombian government; Carrasco Andrade,
Contreras Ponce, Espinoza V. & Moncayo Román (1984) for the Ecuadorian gov-
ernment; and more recently Martínez Santacruz (1992) for the Colombian Oce
of Indigenous Aairs. These studies have as their aim a description of the current
situation of the Awa, to provide information around which the various govern-
mental and non-governmental organizations can base their programmes for the
Awa.
Apart from Cerón Solarte (1986), which is a general but detailed overview
of the Awa, dealing with geographical and environmental issues as well as social
and cultural issues, there are three other ethnographic works on the Awa which
deal more specically with contact between the Awa and the mestizos (non-
indigenous Colombians), and the changes which have taken place in Awa culture
as a result of this. Aragón (1974) and Parra Rizo (1989) deal with the Colombian
Awa, in geographical and socio-cultural terms respectively. Ehrenreich (1989)
discusses the Awa in the Plan Grande of Ecuador, and is largely an ethnography,
or rather two ethnographies: by studying those Awa with lesser and greater
contact with mestizos, and around the time of the discovery of the Ecuadorian
Awa in 1974,2 he is able to compare pre- and post-discovery Awa culture, and
examine the impact of the presence of various institutions on this culture.
In addition, there is a body of work produced by the British social an-
thropologist Ann Osborn. Based on eldwork undertaken in the 1960s, Osborn
produced a general sketch of Awa culture (Osborn 1970), as well as studies
dealing with kinship and other aspects of social organization (Osborn 1969
72, Osborn 1974), including the relationship of compadrazgo (godparentship)
between the Awa and mestizos (Osborn 1991a). Osborn's later works were
based on eldwork carried out for the Instituto Colombiano de Bienestar Fa-
miliar (Colombian Institute for Family Welfare) and deal largely with health
and nutrition (for example Osborn & Melo 1991), although one of these studies
This is the gure used by Grimes (1996): 20 000 Awa in Colombia, 1 000 in Ecuador.
1
That is, the ocial recognition of the Awa by the Ecuadorian state, with the consequent
2
inux of government services and church and other non-governmental organizations.
1.3. PREVIOUS ANTHROPOLOGICAL AND SOCIAL RESEARCH 5

(Osborn 1991b), while ostensibly a study of Awa forms of child-rearing, is in fact

an ethnographic study, drawing on Osborn's earlier eldwork as well as the later.
The cultural practices of health among the Awa have also been studied by
Kempf (in, for example, Kempf 1982), together with the impact integration into
mainstream Ecuadorian society is having on these practices.
More recently health and indigenous medicine have also been the focus
of study of Gloria Narváez, who spent over ten years working as a nurse in and
with the Awa community, producing a number of studies of Awa culture and
traditional medicine with dierent communities (Narváez Reyes & Awa Indigen-
ous Community 1990, Narváez Reyes & Cabildo Cuchilla del Palmar 1992, Nar-
váez Reyes & Cabildo Ramos-Mongón-Manchuria-Mirador 1992), as well as Nar-
váez Reyes (1993), a more detailed study of traditional Awa medicine and how
it can be incorporated into the educational system of the Awa, which has been
reduced and published as Narváez Reyes (1994b); a briefer summary is given in
Narváez Reyes (1994a).
Most recently, a study examining the social construction of space, kinship
and power among the Awa has also been published (Haug 1994).
Perhaps the most important cultural aspect of the Awa for this study is
the phenomenon of secrecy or dissembling behaviour, discussed by Ehrenreich
(1989:252257). This phenomenon has been noted by all researchers, though the
terms in which it is phrased vary greatly: ignorance by the Awa of their own
culture; the Awa being ashamed of their culture;3 or hiding their culture from
outsiders. In her earliest work, Osborn described the phenomenon at a personal
level:
Quite apart from the demoralising eects of the mestizos and in-
uences of other people before them on the Kwaiker, they are a
dicult-to-get-on-with people, to say the least; they are reticent,
hostile, insecure and dour.
(Osborn 196972:215)
Ehrenreich on the other hand discusses the same aspect of Awa life and interac-
tions as a cultural attribute:
La esencia y núcleo de su cultura están en conicto directo con
las metas y métodos de la investigación etnográca. La cultura
estimula el silencio, el secreto y el ngimiento como la quinta esencia
de las respuestas a las indagaciones de los extraños y aún de otros
Coaiquer.4
(Ehrenreich 1989:2829)
3The Awa themselves often describe it in these terms, saying that other Awa tienen vergüenza
`have shame'.
4The essence and nucleus of their culture are in direct conict with the goals and methods
of ethnographic investigation. Their culture promotes silence, secrets and pretence as the
quintessence of replies to the enquiries of outsiders and even of other Awa. (This and following
translations are mine.)
6 CHAPTER 1. INTRODUCTION

Various authors believe that secrecy is a response to contact with the Spanish
and mestizos, that the reaction of these
ha creado una característica primordial en el Grupo [Awa] cual es
la permanente evasión a establecer comunicación con interlocutores
blancos y a ocultar sus manifestaciones de identidad cultural a tal
punto que dejan una sensación de un Grupo que carece de lengua
y de otros elementos de su cultura . . . 5
(Calvache Dueñas & Cerón Solarte 1988:14)
However Ehrenreich (1989:252257) believes that this phenomenon dates from
pre-contact times, and his comment, quoted earlier, that this phenomenon applies
to all strangers, whether white, mestizo or Awa, would appear to support this.
Equally relevant here is the non-sociability of the Awa. The pattern of
settlement described above does not lend itself to an active village social life,
and the cultural prohibition against travelling after dark further restricts this.
While some community activity is now found, it is largely run through the schools
(and normally only involves those Awa whose children are at school), and it would
appear that, traditionally, Awa society was very similar to that of the Epena
on the coast, where social contact between the various Epena family groups
of a community is limited to strictly business matters and community estas
(Harms 1994:3).
All of these cultural factors naturally lead one to question much of the
early ethnographic and linguistic material, based on short-term eldwork. As
Pérez comments after a fairly recent short eldwork experience with the Awa,
Si en estos tiempos, en que dichos autóctonos mantienen contacto
con los mestizos, no es posible obtener su conanza para una inves-
tigación, dudo de los resultados conseguidos por los autores citados
por Jijón en aquellos tiempos de total aislamiento . . . 6
(Pérez T. 1980:6)
It is unfortunate, but in early works it is often unclear what proportion of the
information was based on direct observation, and what proportion based on dis-
cussion with the Awa, who were acting under the cultural rules of secrecy and
evasion; nor indeed can we tell how much of the information was collected from
discussions with mestizos, who while they may have a great deal of contact with
the Awa, often have very incorrect ideas about the customs of the Awa, as Ehren-
reich (1989:46) notes.
5 has created a fundamental characteristic in the Awa, which is permanent ight from esta-
blishing communication with white interlocutors and hiding manifestations of cultural identity
to such a point that they leave a feeling of a group who lack a language and other elements of
their culture . . . 
6 If in these times, in which the said indigenous people maintain contact with the mestizos,
it is not possible to obtain their condence for a research project, I doubt the results obtained
by the authors cited by Jijón in those times of complete isolation . . . 
1.3. PREVIOUS ANTHROPOLOGICAL AND SOCIAL RESEARCH 7

The phenomenon of secrecy applies just as much to Awa Pit as to any

other cultural issue. As a result of eldwork in the mid-1940s, Lehmann notes
that in dealings with mestizo villagers, the Awa
parle espagnol, plus ou moins mal bien entendu. Il voudrait faire
croire que l'espagnol est sa langue maternelle, car le fait de parler
Kwaiker est pour lui une humiliation. Il le dit parfois lui-même:
J'en ai honte, c'est pourquoi je ne veux pas parler. Ou bien il
prétend qu'il est cholo [mestizo] . . . Quand ils disent qu'ils sont
cholos, les Kwaiker cherchent à faire croire qu'ils ne parlent que
l'espagnol.7
(Lehmann 1963:268269)
With regard to their language, Ehrenreich comments that among the Ecuadorian
Awa, pocos Coaiquer admitirán que ellos hablen otra cosa que no sea español 8
(Ehrenreich 1989:31). And Osborn reects on secrecy and Awa Pit as she sees
it:
Los Coaiquer son famosos entre las gentes curiosas y los cientícos
por su negación a hablar en otra lengua que no sea el español, y
por su reticencia a dar información sobre su lengua, las razones de
esto no son difíciles de entender, ellos no quieren ser conocidos co-
mo gentes extrañas, y ser considerados como distintos de los demás
colombianos, y el lenguaje es el primer elemento cultural que dis-
tingue cualquier sociedad de otra. En las palabras de ellos tiene
vergüenza, la implicación es que ellos han sido objeto de ridículo
y abuso por su lengua, pero la autora nunca ha observado mofa o
abuso en este contexto.9
(Osborn 1970:71)
This last comment, that Osborn never observed the mestizos ridiculing the
Awa for using Awa Pit, is interesting. On the few occasions when I observed Awa
Pit used in the presence of mestizos, the mestizos were fascinated, and rather
than mocking, wanted to hear more (though this could, of course, have been
7 speak Spanish, more or less badly, of course. They want it to be believed that Spanish is
their mother tongue, because for them speaking Awa Pit is a humiliation. They themselves say
from time to time: `I am ashamed, that's why I don't want to speak'. Or else they claim that
they are mestizos . . . When they say that they are mestizos, the Awa want people to believe
that they speak only Spanish.
8few Awa will admit that they speak anything other than Spanish.
9The Awa are famous among interested people and scientists for their refusal to speak in
anything other than Spanish, and for their reticence in giving information about their language,
the reasons for which are not hard to understand: they do not wish to be known as strange
people, and be considered as dierent from other Colombians, and language is the rst element
of culture which distinguishes any society from another. In their words, they are `ashamed', the
implication being that they have been ridiculed and abused for their language, but the author
has never observed mockery or abuse in this context.
8 CHAPTER 1. INTRODUCTION

related to my presence). What did occur is that any other Awa present, whether
they spoke Awa Pit or not, burst into embarrassed giggles. However the older
members of the community have assured me that when they were young, the Awa
would go to town, and attempt to buy products such as salt from the mestizos
using Awa Pit, whereupon the mestizos would call them stupid Indians, tell
them to stop speaking in that ugly way (hablar feo), and that they should speak
properly (hablar bien), that is in Spanish. These last two terms are still used
by a number of people in the community. Thus, if I was playing a recording of
someone speaking in Awa Pit, I would sometimes be asked (in Spanish) Who is
that who is speaking ugly?; and in recounting history, the Awa will often say
because in those times they couldn't speak properly. One of my informants, a
uent bilingual, occasionally used the expression conversar al derecho speak the
right way round, to refer to speaking in Spanish. Apparently one of the early
mestizo teachers to enter the region put it even more strongly when she told the
Awa that hablaban caca they spoke shit.

1.4 Language classication

Much of the energy of early examinations of Awa Pit was directed into trying to
discover the origins of the Awa through grouping their language together with
other languages of Colombia and Ecuador. Perhaps the most pertinent comment
about much of this work was made by Yolanda Martínez:
El origen de los Awá se desconoce hasta el momento. No se cuen-
ta con sucientes datos y estudios etnohistóricos, arqueológicos y
lingüísticos que permitan esclarecerlo, aunque se han planteado mu-
chas hipótesis al respeto.10
(Martínez Santacruz 1992:8)
The majority of the comparative studies undertaken have been based on
very small word lists collected by travellers through the region, and on studies of
place names and personal names. The word lists used are normally some subset of
those which appear in André (1884), Hidalgo (1913 [1894]), Gutiérrez (1920) and
Caldas (1946), as well as the more substantial list collected by Pankeri specically
for Jijón y Caamaño (1941).
One problem with much of the early comparative work (where comparat-
ive is used in a broad sense; all these works are based on inspection rather than
the comparative method) is the uncertainty of the existence of numerous groups
or languages discussed in them. Thus the various authors are in a great deal of
disagreement about whether, for example, groups such as the Telembís, Barba-
coans, Sindaguas, Pastos, Mallas or Muellamueses are distinct groups (and/or
languages), or simply alternate names for, or ancestors of, the Awa, known as
The origin of the Awa remains unknown to this point. There are insucient ethnohistori-
10
cal, archæological and linguistic facts and studies to clarify it, although many hypotheses have
been made.
1.4. LANGUAGE CLASSIFICATION 9

Cuaiquers (in various spellings) throughout these documents. Paz y Miño per-
haps takes the confusion further than most, claiming that the Cuaiquér were a
subgroup of the Pastos (Paz y Miño 1946:160); but one page later in the same
document that the Kuaikéres (note the spelling dierence) were a subgroup of
the Barbakóa group, one of the groups whose territory bordered on that of the
Pastos (Paz y Miño 1946:161). As Jaramillo & Acosta Pinzón (1990:29) note,
there is the further confusion that the term Cuaiquer could be used in historical
documents to name a cultural group, the Awa, or as a term to refer to any people
living along the Cuaiquer River (today known as the Güiza River).
Here I will only deal with the various classications of the Cuaiquer (in
whatever spelling), leaving aside whether these are identical with, a subgroup
of, or descendants of the various other groups, except where this directly aects
the classication of the Awa. The labels used to refer to indigenous languages
and peoples is somewhat complex, as the original references are followed here; to
simplify matters for the reader, the commonest label changes are the Cuaiquer
(for Awa), and their language Cuaiquer (Awa Pit), the Colorado (Tsachila) and
their language Colorado (Tsaqui), and the Cayapa (Chachi) and their language
Cayapa (Cha'palaachi).
Studies of place names have generally focussed on the ending -ker, which
occurs in the place names Cuaiquer Viejo, Altaquer and Mayasquer in the actual
territory of the Awa, as well as in many names of places in the Andes region, such
as Yacanquer (near Pasto). While García Ortiz (1949:280) claims to have found
an old Awa who, after oers of money and a great deal of guarapo (an alcoholic
drink) revealed to him that Coaiquer is formed from two Awa Pit words, coai
`people' and quer `force', although these words were no longer part of the current
Awa Pit vocabulary, I suspect he was being misled. It is possible that the Awa
in question had heard from an earlier researcher that quer meant `people', since
both Ortiz (1938) and Paz y Miño (1946) claim that it has this meaning in the
now extinct Pasto language.11 This has led some to link the Awa with the Pastos,
although Ortiz (1938:557) thinks that Cuaiquer and Altaquer are Pasto names
which emigrated to areas where Pasto was not spoken. Indeed as noted earlier,
Cuaiquer Viejo and also Mayasquer were established by a Pasto chief. Altaquer
was not, but oral tradition claims that Altaquer (which was founded at the end
of the 19th century) was given this name when it was still only a resting-stop for
travellers on the PastoBarbacoas road, after a huge tree at the location, of a
species known in Spanish as altaquer (Parra Rizo & Virsano Bellow 1992:1011).
Ortiz (1954) also analyzes the occurrence of many place names in pi, al
and a variety of other monosyllables, which occur in Awa territory and in the
Andes. It is quite unclear what meaning, if any, these monosyllables have in any
language  Awa Pit, Pasto or any other.12
Turning now to language classications, the earliest to include Awa Pit
appears to be that of Brinton (1891:198200), who includes the Cuaiquers as one
11 Alternatively, of course, the Awa in question may have been a speaker of Pasto.
12 Apart from pi, which is Awa Pit pii `river', and according to Ortiz has the same meaning
in a language in the Andes, presumably Guambiano.
10 CHAPTER 1. INTRODUCTION

of the eight groups in the Barbacoan linguistic stock, together with the Cayapas,
Colorados, Barbacoas, Iscuandeses, Manivis, Sacchas and Telembís. He notes
that André (1884) considers that the Cuaiquers speak Colorado (though he does
not mention that André also claims that the Telembís speak Cuaiquer).
The rst study to include this group as part of a wider classication is
Beuchat & Rivet (1910). They use Brinton's classication of the Barbacoan
group, and mix word lists of Awa Pit, Colorado and Cayapa to compare this group
with the Paniquitan and Coconuco groups. On the basis of this comparison by
inspection, they place Awa Pit in the Talamanca-Barbacoa subgroup of Chibchan.
This classication, with minor variations, has survived through the majority of
classications to the present day.
Ortiz (1937:75) considers Kuaiker as in the Barbakóa group of mixed lan-
guages, which also contains Sindagua, Moguex, Totoró, Kokonuko and Popayán.
Kuaiker is a language with vestiges of Mashakali. In his later classication
(Ortiz 1965:135), Kwaiker remains in the Barbakóa group of mixed languages, but
now accompanied by Kamsá (as well as Kayapa and Sáxchila/Colorado, though
since these are spoken only in Ecuador, he does not deal with them directly in
this work).
The most inuential classication of the Chibchan languages is probably
that of Jijón y Caamaño (1943). As far as Awa Pit is concerned, Coayquer turns
up together with Cayapa and Colorado, following Beuchat & Rivet's (1910) clas-
sication, although the group is now called the archaic or western group of
Chibchan, rather than Barbacoan. The inuence of Jijón y Caamaño's classica-
tion (that is, Brinton's classication), both in this work and his earlier grammar
of Coayquer (Jijón y Caamaño 1941), was such that authors such as Duque state
with full condence (although confusing languages and ethnic groups) that
La lingüística comparada demuestra nexos indiscutibles entre el
grupo Coaiquer de Nariño y el Cayapa-Colorado del Ecuador, am-
bos pertenecientes a la familia Macro-Chibcha.13
(Duque Gómez 1955:8)
Various recent works have also followed this classication. Thus Cost-
ales & Costales (1983:7586) group the language of the Chachis and Zatchila as
well as that of the Quijos together with Awa Pit in the Barbaco subgroup of
the Shillipanu family (which also includes Cofán). Stark (1985:159) notes, un-
fortunately without references, that glottochronological calculations show that
Cayapa-Colorado and Coaiquer separated around the year 50 BC. In his classi-
cation of American languages, Greenberg (1987:382) places Cuaiquer, Colorado,
Cayapa and Cara together to form the Barbacoa subgroup of Nuclear Paezan,
of Paezan, of Chibchan-Paezan, of Amerind. Obando Ordóñez (1992:3845), us-
ing a Greenbergian approach based on a Swadesh word-list compiled from old
sources, concludes that Awa Pit is related to Colorado and Cayapa, though the
Comparative linguistics shows indisputable links between the Awa of Nariño and the
13
Cayapa-Colorado of Ecuador, both belonging to the Macro-Chibchan family.
1.4. LANGUAGE CLASSIFICATION 11

latter two are more closely related. He notes that the aliation of this group to
Chibchan is uncertain.
However there have been occasional dissenting voices  usually not pro-
posing any other classication, but doubting the evidence on which conclusions
have been based. This doubt extends from the most general to the most specic
levels of classication. Thus various researchers have noted the deplorable lack of
comparative work carried out in proposing general Latin American classications,
and consider that
it is not an exaggeration to say that researchers working on native
South American languages have been embarrassed by the dilapid-
ated state of the classication schemes, at the same time as they
are chagrined at seeing those schemes taken up uncritically by oth-
ers [non-linguists].
(Urban & Sherzer 1988:294)
At the level of the inclusion of Barbacoan (Awa Pit, Tsaqui and Cha'palaachi)
in Macro-Chibchan, there are also doubts. Constenla, one of the few to have done
any strictly comparative work on any of the Chibchan languages, notes, after a
discussion of the literature referring to Macro-Chibchan, that
las clasicaciones incluidas en las subsecciones precedentes en su
mayor parte toman en cuenta lenguas o conjuntos de lenguas cuya
relación con las consideradas como propiamente chibchas no está
en absoluto comprobado.14
(Constenla Umaña 1985:164165)
In his reconstruction of Proto Chibchan, Wheeler (1972) made very occasional use
of South Barbacoan, reconstructed by Moore (1962) on the basis of Cayapa and
Colorado. He notes that the level of lexical resemblance between these languages
and Chibcha proper is very low, less than a ten percent correlation (Wheeler
1972:95). And for Adelaar, the comparative work of Constenla on Chibchan is
conclusive:
All these languages [Cayapa, Colorado and Cuaiquer] were formerly
classied as Chibchan, but recent research (Constenla 1981) has
shown the invalidity of that claim. A revised classication is
needed. For the moment Guambiano seems to share a common
origin with Cuaiquer and the Cayapa and Colorado languages of
Ecuador.
(Adelaar 1991:66)
14the classications included in the preceding subsections are largely taking into account
languages or groups of languages whose relation with those properly considered Chibchan is
totally unproven.
12 CHAPTER 1. INTRODUCTION

However even at the level of Barbacoan doubts have been expressed. After
discussing the classication schemes of Jijón y Caamaño and others, Carrasco et
al. consider that
si bien esta clasicación se la ha venido manejando por muchísimos
años, es conveniente tenerla más como marco de referencia mas
no como lo nal y correcto; sería conveniente estudiar comparati-
vamente estos tres lenguas (Chapalachi, Tsiqui [sic] y Kwaiker)
como primer paso para esclarecer, reformar o sustentar de una ma-
nera más cientíca, lo anterior.15
(Carrasco Andrade, Contreras Ponce, Espinoza V. &
Moncayo Román 1984:8)
Until very recently, all comparisons of Awa Pit with Cayapa and Colorado
have been based on inspection, and almost entirely on inspection of word-lists
collected in the rst half of the century by travellers and historians; at least in
the case of the Awa Pit word-lists, these are very limited and contain a great
number of errors. As Pérez says,
Con escaso material como el de los autores consultados por Jijón
y el mío, no es factible establecer conclusiones gramaticales ni lin-
güísticas . . . Sería un atrevimiento concluir, con escaso vocabulario,
que el Cuaiker pertenece a tal o cual familia lingüística.16
(Pérez T. 1980:67)
Recently the suggestion has been made that the Awa language and people
are a combination of a variety of languages and groups which previously existed
in the region (Cerón Solarte 1987). However the linguistic evidence on which
this is based is extremely weak. Citing evidence from West (1957), Cerón and
Calvache note that
algunos documentos de la audiencia de Quito del Siglo XVII ar-
man que los Nulpes, Panga, Guelmambí, Cuasminga, Chupa, Guapí
y Boya usan dialectos poco diferenciados de una lengua común de
la región.17
(Cerón Solarte & Calvache Dueñas 1990:18)
15 even though this classication has been used for many years, it is advisable to have it more
as a reference point but not as the nal and correct classication; it would be appropriate to
make a comparative study of these three languages (Cha'palaachi, Tsaqui and Awa Pit) as a
rst step in clarifying, reforming or supporting this classication, in a more scientic manner.
16With scarce material like that of the authors consulted by Jijón and my own, it is not
feasible to establish grammatical or linguistic conclusions . . . It would be rash to conclude, on
the basis of a small list of words, that Awa Pit belongs to this or that linguistic family.
17some documents of the courts in Quito in the 17th century state that the Nulpes, Panga,
Guelmambí, Cuasminga, Chupa, Guapí and Boya use little-dierentiated dialects of a common
language of the region.
1.5. PREVIOUS LINGUISTIC RESEARCH 13

As well as this conclusion being based on court documents, Calvache discussed the
same hypothesis in an earlier work, supporting it with synchronic observations:
ya se ha encontrado diferentes maneras de escribir y hablar el Kwai-
ker, por lo tanto, las palabras también varían en su signicado, lo
cual nos lleva a raticar una vez más el supuesto ya enunciado de
que aquellos proceden de varios Grupos indígenas precolombinos.18
(Calvache Dueñas 1987a:section 1 2)

But of course a variety of dialects is the natural situation of any language, es-
pecially in an area where communication between dierent communities is as
dicult as it is in the Awa region, and hence the existence of the dialects men-
tioned in the above quotes, both in the 17th century and in the present day, in
no way indicates that Awa Pit is a reduction of a variety of languages.
More recently there has been a strictly comparative study of Awa Pit and
other Colombian and Ecuadorian languages. Curnow & Liddicoat (forthcoming)
examine Awa Pit, Cha'palaachi (Cayapa), Tsaqui (Colorado), Guambiano, To-
toró and Paez, and show by reconstructing forty or fty words, together with
hypothesized sound changes, that the rst ve of these languages are apparently
genetically related, but that Paez is related very distantly, if at all, to the oth-
ers. Guambiano and Totoró are assigned to the North Barbacoan subfamily,
Cha'palaachi and Tsaqui to the South Barbacoan branch. Awa Pit is tentat-
ively placed with a GuambianoTotoró protolanguage in North Barbacoan, but
this is not clear and Awa Pit may, in fact, form a separate Central Barbacoan
subfamily.
This comparative work thus clearly places Awa Pit together with Guambi-
ano and Totoró (in Colombia) and Cha'palaachi and Tsaqui (in Ecuador) in the
Barbacoan family. However the relationship between this family and any other
remains to be established.

1.5 Previous linguistic research

Before the present investigation was undertaken, four main authors had written
on Awa Pit from a synchronic point of view, although there had also been a
number of meetings held to discuss the establishment of a ColombianEcuadorian
bilingualbicultural education programme, which necessarily involved discussion
of the phonemic system of Awa Pit, in order to establish a unied orthography
for the language.
Following an agreement between the Summer Institute of Linguistics (SIL;
Instituto Lingüístico de Verano) and the Colombian government in 1962, in the
late 1960s Lee and Lynne Henriksen arrived in Colombia to work with the Awa.
The SIL have since published a variety of works on and in Awa Pit, sometimes
18dierent manners of writing and speaking Awa Pit have been found, and words also vary
in their meanings, which brings support to the hypothesis mentioned earlier that they proceed
from various pre-Columbian indigenous groups.
14 CHAPTER 1. INTRODUCTION

jointly with the Colombian Ministry of Government. While these include a series
of readers in Awa Pit (Henriksen & Henriksen 1978) and a series of primers de-
signed to teach the Awa to read and write in their own language (Henriksen &
Henriksen 1986), they have included a number of more strictly linguistic publica-
tions: two papers on discourse (Henriksen & Levinsohn 1977, Henriksen 1978), a
preliminary phonology (Henriksen & Henriksen 1979), and a short paper on vari-
ous grammatical points (Henriksen 1985). More recently, a teaching grammar for
non-Awa Pit speakers has also appeared (Henriksen & Obando Ordóñez 1985).
In the past eight years or so, two theses on aspects of Awa Pit have been
produced, Calvache Dueñas (1989) and Obando Ordóñez (1992), the former at
the Universidad de los Andes in Bogotá, the latter at the University of Texas
at Austin. Both are essentially works on phonology, though both also contain
some aspects of morphology and syntax. The two works have many disagreements
between them, in some cases perhaps dialect dierences, and there are also a num-
ber of inconsistencies within each work  a discussion of this will be postponed
until the phonology chapter. A copy of the phonology and morphophonology
sections of Calvache Dueñas (1989) also appear in Centro Experimental Piloto
Nariño (1990) and Calvache Dueñas (1991).
The nal author is Luis Montaluisa, working on the Awa Pit spoken in
Ecuador, in contrast to the previous three authors who focus on Colombia. His
phonology/orthography (Montaluisa Chasiquiza 1991) will also be discussed in
the phonology chapter.
While none of the authors mentioned above refer directly to the phonology
proposals of the others (although Montaluisa notes that his work involves the
ideas of others as well as his own), Henriksen, Obando, Calvache and Montaluisa
were all involved in the rst meeting on the unication of the Awa Pit alphabet
held on 1721 May 1987 in La Planada, Colombia, and Montaluisa and Obando
also participated in the second meeting on 30 May3 June 1988 in Maldonado,
Ecuador.
In both these meetings a variety of alternate phonemic systems were dis-
cussed, and internal inconsistencies and conicts between the systems debated,
so the authors of works published after these dates were aware of the ideas of
the other participants. The major points of dispute, or areas where further cla-
rication was felt to be needed, after the rst meeting were: the vowel system
(the existence or otherwise of phonemically voiceless, nasal and long vowels); the
glottal stop; the existence of /ts/ and /r/; and possible consonant clusters /pj/,
/pw/, /kw/, /tw/, and so on (First Binational Meeting 1987). The majority of
these remained points of contention after the second meeting also (Second Bin-
ational Meeting 1988), although it was decided to adopt the suggested alphabet
as experimental, to be tested in practice for a year. Between the two meetings
a number of spectrograms were made to examine the existence and contexts of
voiceless vowels in the language (Calderón Rivera, Trillos Amaya, Reina & R. de
Montes 1987).
As far as I am aware, apart from general articles discussing the rst
(Calvache Dueñas 1987b) and second (Trillos Amaya 1988) meetings, the only
1.6. ORIGINS AND AREAL TYPOLOGY 15

published result of these meetings to appear has been the Ecuadorian Dirección
Nacional de Educación Indígena Intercultural Bilingüe's (1989) proposal for an
experimental unied Awa Pit alphabet, which seems to be largely based on Mont-
aluisa's work, though it takes into account a number of the issues raised during
the binational meetings.
Contreras Ponce (1984) also gives a proposal for an alphabet for Awa Pit,
however the phonological analysis on which this is based is highly dubious. For
example, /p/ sometimes is realized as [p], sometimes as [B], both in intervocalic
environments, with no indication of when it is one, when the other; and likewise
for a number of other phonemes.
Beyond these works on phonology/orthography, almost nothing has been
published on the morphology or syntax of Awa Pit, apart from the sketches of
Henriksen mentioned above, and the brief sections in Calvache Dueñas (1989) and
Obando Ordóñez (1992). These last two are both relatively short, and contain a
number of serious disagreements between the two.
Some of the dierences between analyses (although by no means all) could
be related to dialect dierences. It is clear that there are dierent dialects of Awa
Pit, at least on the basis of pronunciation. For example the Dirección Nacional
de Educación Indígena Intercultural Bilingüe (1989) suggests that orthographic
s should be pronounced like a Spanish s ([s]) in words such as sun `that'; in the
dialect on which this study is based, /sun/ is pronounced [tsun], and there is
some evidence that in some other regions it may be retroex [ù] in this position
(Obando Ordóñez 1992). Equally, there may be some syntactic and morphological
dierences between dialects, although lack of data clouds this issue  perhaps the
only clear example is the dierence between an apparently complex and obligatory
number cross-referencing system in some dialects (Lee Henriksen, p.c.) and the
much simpler optional system found in the data for this study (see section 7.3).

1.6 Origins and areal typology

One of the concerns which has perhaps plagued research on the Awa, and stopped
references to them in works dealing with wider issues, is the problem of how to
classify the Awa, culturally and geographically. Living as they do on the west-
ern slopes of the Andes mountains, down towards the Pacic coast, the Awa are
not an Andean group, nor are they a group of the Amazon basin; but equally
they are not a coastal group. The environment in which the Awa live is closest
to that in which lowland, that is Amazonian, groups inhabit, in terms of being
tropical and subtropical rainforest, with a very high rainfall; and indeed, Stark
(1985) classies the language of the Awa as a lowland language in her survey of
Indigenous languages of lowland Ecuador, although for most, `lowlands' implies
only the eastern side of the Andes (cf. Doris Payne 1990b:214). Discussing this
classicatory problem, Ehrenreich (1989:46) proposes the label of an inland cul-
ture (tierra adentro), to distinguish the environment and culture of the Awa,
Chachis and Tsachilas from Andean groups (tierra alta), coastal groups (costera)
and Amazonian groups (tierra baja).
16 CHAPTER 1. INTRODUCTION

The geographical dierentiation of inland groups from coastal and high-

land groups appears to be correlated with certain cultural dierences also. Un-
fortunately, while various theories have been proposed over the years seeking the
origin of the Awa in either the highlands or the coast (or even Central America or
Polynesia), no archæological work has been carried out in the Awa region. The
very little archæological research done near Awa territory appears to suggest that
the peoples living on the borders of the current Awa area had a culture distinct
from both the culture of the highlands and that of the coast. An analysis of
archæological nds made on the western slopes of the Cumbal volcano and in
Barbacoas shows that the ceramic types in this region were totally dierent from
those found in the highlands and the coast (Jaramillo & Acosta Pinzón 1990:24).
Of course, there is no evidence to link the Awa with nds in this area: they
may well have arrived in the region from the coast or the highlands after these
ceramics were produced.
The few aspects of Awa material culture which have been analyzed confuse
the highlandscoast debate further, since some aspects link the Awa with one,
some with the other. For example, traditional Awa music involves utes, appar-
ently originally a highlands instrument, and the marimba (a type of xylophone),
used on the coast and usually believed to have been brought there by African
slaves (Cerón Solarte 1986:168).
The entire living pattern of the Awa  housing style, dispersed settlement
pattern, low frequency of social interaction  appears much more like the coastal
Epena than highlands groups like the Quechua.
The current geographical distribution of the related Barbacoan languages
is also somewhat unrevealing (see Figure 1.3). Guambiano and Totoró are cur-
rently spoken in the highlands of Colombia to the north of Awa Pit; but Tsaqui
and Cha'palaachi are spoken by inland groups, like Awa Pit. However there are
further facts to be added to this geographical distribution.
The Guambianos and Totorós are currently found in the Andes highlands,
and appear to have been there for quite some time, as they are completely sur-
rounded by a traditionally war-like group, the Paez. This suggests that the
Guambianos and Totorós are a remnant group, who have been encircled by the
more recently arrived Paez.
While the Chachis (speakers of Cha'palaachi) and Tsachilas (speakers of
Tsaqui) currently live in inland Ecuador south of the Awa, both of these groups
have traditions claiming that they lived earlier in the highlands, and the Chachis
even date many techniques and artifacts in terms of pre- and post-migration
practices (Murra 1963:278). Thus the Tsachilas claim to have had the vol-
canos Cotopaxi and Chimborazo (near Quito) in their territory (Karsten 1988
[1925]:56), while the Chachis lived near Ibarra further north, near the Colombia
Ecuador border (Barrett 1925:31), perhaps having moved there from near Quito
(Vittadello 1988, vol.2:5).
Stark (1985:158160) gives a possible scenario of migration for the Awa,
Tsachila and Chachi peoples, as a result of invasion by another group, prob-
ably the Incas, and following post-Conquest contact with the Spanish. Her ac-
1.6. ORIGINS AND AREAL TYPOLOGY 17

Figure 1.3: Indigenous languages of southern Colombia and northern Ecuador

discussed in this work (based on Constenla Umaña 1991:70): Awa Pit, Tsaqui,
Cha'palaachi, Guambiano and Totoró (not shown, considered here a dialect of
Guambiano) form the Barbacoan family of languages
18 CHAPTER 1. INTRODUCTION

count needs to be altered in the light of the relationship of these three languages
with Guambiano and Totoró, perhaps pushing the boundary of the Barbacoan-
speaking peoples much further to the north.
Typologically, Awa Pit also appears to be much more Andean than coastal
or Amazonian. There are many typological similarities between all three groups,
but certain dierences. The closest Amazonian languages tend to have highly
complex classier systems (Doris Payne 1990b:220), completely absent from Awa
Pit, and a much more polysynthetic structure (Doris Payne 1990b:214). While
all languages in the region tend towards cliticized postpositions for case marking,
there are relatively few postpositions in many Amazonian languages. Coastal
languages such as Epena Pedee are ergativeabsolutive (Harms 1994:910); while
Awa Pit, Quechua and many languages on the Amazon side of the Andes are
nominativeaccusative and have similar case systems. Indeed, many of the case
forms given by Levinsohn (1976) for Inga Quechua appear similar or identical to
those of Awa Pit: an accusative and locative ta; a genitive pa; and a locative
meaning encoded by mal in Awa Pit versus ma in Inga Quechua. Unfortunately,
there is no evidence of lexical similarities, apart from a few words which are likely
to be loans (Inga Quechua forms rst): `father' (tayta, taytta), `cattle' (huagra,
wakata [waGaRa]), `gun' (illapa, iyaNpa), and so on.

1.7 Language death?

It is quite clear that not all members of the Awa indigenous group speak the tra-
ditional language  most are monolingual speakers of Spanish. Unsurprisingly,
it also seems that those regions (such as the eldwork site for this project) which
are closer to non-indigenous areas have a lower percentage of speakers.
Given the uncertainty surrounding the number of Awa, with suggestions
of between 5 317 and 25 650, together with the Awa cultural trait of secrecy, it is
not surprising that exact numbers of speakers of Awa Pit are simply unavailable.
The language is not accorded prestige in the community, and there are many
dentro de los propios indígenas quienes desean que sus hijos no
repitan la historia de explotación y marginamiento. Es decir, en su
perspectiva ideológica se olvidó de su lengua nativa y el Castellano
se presenta como clave del ascenso social, lo cual es comprensible
ya que empíricamente el indígena así lo comprueba.19
(Calvache Dueñas & Cerón Solarte 1988:1112)
There are some gures available giving percentages of speakers of Awa
Pit in dierent regions (Calvache Dueñas 1991:45), although it is unclear how
accurate they are. These gures are based on schoolchildren and their parents,
19among the indigenous people themselves who don't want their children to repeat the
history of exploitation and marginalization. That is to say, from their ideological perspective
they have forgotten their native language and Spanish becomes the key to social progress, which
is understandable given that the indigenous people have veried it themselves.
1.7. LANGUAGE DEATH? 19

living in ve dierent areas, and range from one area, where 0% of the children,
8% of fathers and 4% of mothers are bilingual, to the most strongly Awa Pit-
speaking areas where 44% of children, 57% of fathers and 55% of mothers are
bilingual. There is no evidence of any monolingual speakers of Awa Pit. The
gures in this survey for Pialapí have 0% of children, 5% of fathers and 5% of
mothers bilingual, which accords well with a survey carried out in 1994 by Lola
Caguasango, the teacher in neighbouring Pueblo Viejo, which found that about
5% of the adult population spoke Awa Pit.
Of course these gures must be taken with extreme caution. While there
appear to be no monolingual speakers of Awa Pit, it is clear that there are a few
semi-speakers of both languages, and the criteria on which someone was assigned
the status of `bilingual' in the above survey is unknown. Osborn points out that:
Los kwaiker que pueden hablar dos lenguas no son muy locuaces
en ninguna de ellas. Sin embargo, quieren que sus niños aprendan
español y se reservan su propio lenguaje (si es que lo conocen) para
conversaciones íntimas con otros adultos. Existen casos extremos
en este aspecto . . . en los cuales los adultos no se saben expresar en
ningún idioma en forma adecuada, ya sea por causa de matrimo-
nios mixtos de sus padres (por ejemplo, con alguien que no hable
kwaiker) o por represión cultural.20
(Osborn 1991b:193)
If non-uent speakers have been included in the data above, the actual gures
of speakers could be much fewer. Equally, it is not clear whether bilinguals are
those who actually speak Awa Pit in some aspect of their daily lives, or simply
those who feel that they can (but do not) speak Awa Pit.
Language loss within individual families appears to have been very sudden.
The major informants for this study are all former speakers: they state that they
were monolingual in Awa Pit until the age of about 10, when they started going
out to the non-indigenous town for purchases. The children of these speakers,
however, have never learnt any Awa Pit. While speakers claim not to have known
any Spanish before going to the town, this seems unlikely, as there are other
members of the community in whose families the language loss appears to have
occurred a generation or two earlier  while these Awa are of about the same age,
none ever learnt Awa Pit as children. However the Awa Pit speakers are from the
family which controls most land to the east of the Pialapí river, or the family with
territory around Pueblo Viejo;21 the family controlling the area between the two
20 The Awa who can speak both languages are not very talkative in either. Nevertheless,
they want their children to learn Spanish, and reserve their own language (if they know it)
for intimate conversations with other adults. In this respect there are extreme cases . . . in
which adults cannot express themselves adequately in either language, whether it is because of
a mixed marriage of their parents (for example, marrying someone who does not speak Awa
Pit) or because of cultural repression.
21There are only small phonetic dierences between the dialects of these dierent regions,
and these will be noted below where relevant.
20 CHAPTER 1. INTRODUCTION

has (to my knowledge) no living speakers of Awa Pit (there has, of course, been
intermarriage between the families), and perhaps this areal distribution accounts
for the lack of contact with Spanish at an early age for some speakers.

1.8 Fieldwork and data collection

As mentioned above, the eldwork for this study was carried out over about
10 months during 1994 in the settlements of Pialapí and Pueblo Viejo which
are, to all intents and purposes, monolingual Spanish-speaking areas. While
these conditions are not ideal for language study, various practical and political
considerations made this area the only possibility.
Working in an area with uent use of Awa Pit would obviously allow an
examination of Awa Pit in contexts which was impossible in the present study.
However there is evidence that the grammar of former speakers is in fact still vi-
able. Reporting on a former speaker of Salvadorean Lenca, Campbell & Muntzel
(1989:1834) note that the speaker's phonological system was intact; and this
is clearly the case with the Awa Pit informants, whose speech contained phon-
emes, allophones and combinations of phonemes not found in Spanish. The Lenca
speaker's syntactic system was likewise fully viable, although the speaker's actual
production consisted of simple constructions and phrases. This accords well with
Mithun's (1989) experience with an Oklahoma Cayuga speaker. The speaker con-
trolled the morphology, but tended not to use long combinations of morphemes,
even when she controlled all the morphemes individually. Indeed, the most strik-
ing feature of the Oklahoma Cayuga speakers was their complete retention of an
amazingly complex morphological and phonological system, under such limited
opportunities to use it (Mithun 1989:257). In contrast to these other areas of
grammar, discourse patterns seem to be lost relatively early by former speakers.
The present study involved work with ve dierent male speakers of Awa
Pit, with varying histories and uency in Awa Pit. All informants traced their
ancestry back, from two to four generations ago, to settlers from the region of
Cuaiquer Viejo, Altaquer and Vegas, areas much closer to the PastoTumaco
highway. It would appear that the migration to current areas occurred at the
time of increased settlement of the region around the highway by mestizo col-
onists, pushing the Awa out into the more inhospitable mountainous regions.
Informants stated that the families moved into areas occupied by indios bravos
`savage indians', who were less civilized  for example, they did not eat salt
 but did cultivate corn and were themselves speakers of Awa Pit.
The main informant for this study, E, was 54, and a resident of Pueblo
Viejo. He lived alone, his monolingual Spanish-speaking wife having died quite
some years before. It would seem that he never uses Awa Pit in his normal life.
A second informant, L, aged about 40, lived with two of his brothers in
his sister's house in Pialapí, as did the sister and her grandchildren. The three
men all admitted to being able to speak Awa Pit, although unfortunately the
two other brothers both had speech defects. While his sister claimed initially to
be unable to speak Awa Pit, and never ocially changed from this status, she
1.8. FIELDWORK AND DATA COLLECTION 21

was the only member of the community who ever greeted me (when meeting on
a path) in Awa Pit rather than Spanish, when alone or with her granddaughter.
The granddaughter, who was aged about 12, never spoke Awa Pit to me but
appeared able to follow the limited Awa Pit interactions. This suggests that the
family possibly speak some Awa Pit at home, although all are fully uent Spanish
speakers.
The third informant, T, nephew of L, was also probably in his early 40s,
and had grown up monolingually in Awa Pit, but had a monolingual Spanish-
speaking wife, and his children do not understand Awa Pit. While he does not
speak Awa Pit in the community in Pialapí, he spent about ve years living in
Ecuador in his late teens and early 20s, and often interacted with other Awa Pit
speakers there.
R, the fourth informant, was probably in his 70s or older, and was from
outside the region, having grown up in Arrayán, the next settlement to the west
from Pueblo Viejo. He had very little social contact with others, and lived with
his deaf-mute son. He had a great deal of diculty expressing himself in Spanish.
The nal informant, A, who was in his early 30s, was only consulted
with regard to a few points of phonology. Until the age of 5 he lived with his
grandmother and spoke only Awa Pit. After this, however, he spoke only Spanish
until his mid-20s, when he relearnt Awa Pit in a dierent area (some distance
west). His input was important for aspects of the phonology, as he is literate
in Spanish (though not in Awa Pit) and some of his spellings of Awa Pit words
provide important data.
It was almost impossible to acquire narrative texts from any of the in-
formants. The reasons for this are unclear, and could be related to any number
of phenomena: lack of a story-telling tradition; the informants not especially
good story-tellers; story-telling a social activity, and the lack of response on my
part leading to stories not being able to be told; and so on. Two stories were
recorded from L and one from E, however it was impossible to obtain more than
a free translation of these entire narratives; and speakers were unwilling to re-
peat unclear sentences, leading to only a very rough transcription of these texts.
Equally, informants refused to converse in Awa Pit with each other, at least in
my presence (two were in fact not on speaking terms, and one was ostracized by
everyone), even though they were aware that I was working with all individually.
The reasons for this are once again unknown, although they may be similar to
those of Mithun's (1989:245) Oklahoma Cayuga speaker, who enjoyed speaking
Cayuga but often answered Cayuga questions in English, as she felt the other
speaker was more correct, and worried that she might be making mistakes.
For these reasons, then, this thesis is essentially based on elicited material.
At times speakers gave spontaneous utterances, and these have been relied on
where possible; equally, while sentences in Awa Pit were sometimes oered to a
speaker to accept or reject as grammatical utterances (and, indeed, were often
rejected), in order to clarify some point, these have been relied on as little as
possible to minimize the eect of speakers accepting incorrect utterances to please
the researcher.
Chapter 2
Phonetics and phonology
2.1 Introduction
The phonetics and phonology of Awa Pit have been written about in a number
of previous studies, either studies of phonetics or phonology per se, or as prepar-
atory work for designing an orthography for the language. These studies include
Calvache Dueñas (1989), Henriksen & Henriksen (1979), Montaluisa Chasiquiza
(1991) and Obando Ordóñez (1992), as well as the discussion papers from the
two binational meetings held to discuss the formation of a unied orthography for
Awa Pit (First Binational Meeting 1987, Second Binational Meeting 1988). Ref-
erence will be made to the ideas from these works where appropriate throughout
this chapter.
This chapter will give an overview of the phonetics and phonology, but
clearly relies on the works listed above to some extent. Certain areas will be
focussed on in greater depth, where there has been disagreement among the
previous studies. These controversial areas are: whether there is a phoneme /r/
or whether it is an allophone of /t/; the number and distribution of fricatives
and their allophones; palatal and velar nasals, and nasalized vowels; the sound
[e]; and the phonological status of voiceless vowels.

2.2 Consonants
The suggested consonantal and glide phonemes of Awa Pit have been listed in
Table 2.1.

2.2.1 Stops
There are three stop phonemes in Awa Pit: /p/, /t/ and /k/. In general terms,
each of these phonemes has three allophones within a word (for sandhi phenom-
ena, see section 2.2.1.2). After a voiced consonant and before a voiced phoneme
(either a voiced vowel or a glide), the allophone is voiced; between voiced vowels
or glides the allophone is voiced and fricativized; and in other positions the al-
lophone used is a voiceless stop. However there is one major irregularity in this
24 CHAPTER 2. PHONETICS AND PHONOLOGY

Labio- Apico- Apico- Lamino- Lamino- Dorso-

labial dental alveolar alveopalatal palatal velar
Stop p t k
Voiceless fricative s S
Voiced fricative z Z
Lateral fricative ì
Lateral approximant l
Nasal m n N
Glide w j
Table 2.1: Consonant and glide phonemes of Awa Pit

pattern, and a number of additional allophones are in free variation with these
three under certain circumstances.
Thus the most common allophones are:
8 h i
>
>
>
> [b] = C voiced
>
> [voiced]
>
>
<
/p/ > [B] = V[voiced
!
or G
]
V or G
[voiced]
>
>
>
> [pw ] = 1
>
>
: [p] elsewhere
8 h i
>
> [d] = C voiced
>
>
>
< [voiced]

/t/ > [R] = V or G

!
V or G
>
> [voiced] [voiced]
>
>
>
: [t] elsewhere
2.2. CONSONANTS 25
8 h i
>
>
>
> [g] = C voiced
>
> [voiced]
<
/k/ > [G] = V or G
!
V or G
>
> [voiced] [voiced]
>
>
>
: [k] elsewhere
It should be noted especially that between two vowels, /t/ is produced as [R],
not as [D].1 In broad terms, the analysis of the stops presented so far agrees with
the four major previous works, except that Calvache Dueñas (1989) states that
/t/ and /r/ are separate phonemes, rather than the latter being an intervocalic
allophone of the former. Her reasons for claiming this, and arguments against it,
will be examined in sections 2.2.1.2 and 2.2.5.3. All four works also claim that
between a voiced consonant and any vowel we nd the voiced stop allophone;
in fact, this only occurs (at least in the dialect represented in my data) before
a voiced vowel. Before a voiceless vowel, we nd the voiceless stop allophone:
/kimpu/ is produced as [kImpu].
At the end of an utterance,
 there are two possible pronunciations for each
stop. Either they are unreleased ([p^], [t^] and [k^]), or else they have a nasal
release ([pm], [tn] and [kN]).2
The distribution of allophones is not quite so rigid as stated above. After a
voiced fricative, there is free variation between a voiced stop and a voiced fricat-
ive allophone (in this case, [D] for /t/)  [IZbUl] [IZBUl] `soul', [kazd1la] [kazD1la]
 

`day after tomorrow'. Similarly, after a voiceless fricative there is free variation
between a voiceless stop and a voiceless fricative allophone ([F], [T] and [x] re-
spectively)  [ISkamda] [ISxamda] `snake sp.', [maySti] [maySTi] `machete'.
 

In addition there is one situation in which [D] may appear rather than
[R] between two vowels. This occurs optionally if the speaker has produced an
alveolar stop or fricative earlier in the word. Thus the sequence /kutnja-ta/
`three-acc' is pronounced as either [kutñaRa] or [kutñaDa].3
The phoneme /p/ also has a labialized allophone [pw] which occurs before
the vowel /1/. This labialization also occurs with the bilabial nasal /m/.4
1 [R] as a phoneme of a dental stop is found in a variety of nearby Amazonian languages
(Aikhenvald forthcoming).
2Compare the use of (somewhat dierent) forms of nasalization to indicate pauses in other
South American languages (Aikhenvald 1996b).
3There is an additional complication with certain roots and suxes containing the sequences
/ata/ and /at1/. The word /kata-/ `bring' has a variant /kaa-/; and the root /pjat1s/ `sugar
cane' has a variant /pjas/. Other roots with the same sequence have not been found to vary.
Likewise, as will be discussed in section 9.2.1, the Past tense marker /t1/ is often elided after
the vowel /a/, whether this is part of the sux or the root: /kata-t1-zi/ or /kata-zi/ and
/ku-mtu-at1-zi/ or /ku-mtu-a-zi/.
4While phonetically this labialization cannot be distinguished from a phonetic [w] (that is,
[p 1] and [pwi] dier only in vowel quality), a superscript [w] will be used here to indicate it for
w
two reasons: rstly to mark its non-phonemic status, but also because there is some suggestion
that this labialization does not occur in all dialects of Awa Pit.
26 CHAPTER 2. PHONETICS AND PHONOLOGY

It should be noted that there is a great dierence in the frequency of

dierent intervocalic stops, and also an interesting morphophonemic rule. The
phoneme /t/ is found intervocalically (as [R]) in many words. The velar stop /k/
is found in intervocalic position in about ten words in the data, but it is not clear
for many of these words that this is a historically intervocalic /k/. At least one
of the words, wakata `cattle', is a loan from Quechua; another, piikam- `swim',
is historically a verb root plus a sux (pii-kam-); three other words (kuku kizh-
`bark (of dog)', mukul `tórtula (bird sp.)' and pwikuku `pabau (bird sp.)') are
probably onomatopæic; the origins of the other words with intervocalic /k/ are
unknown. The status of the intervocalic [B] allophone of /p/ is even less clear.
It occurs in only seven words in the data. One of these is an obvious loan from
Spanish  wipu `egg', Spanish huevo [ueBo]. Four others are clearly or probably
compounds in Awa Pit, with the rst element ending in a vowel, the second be-
ginning with /p/. This leaves two words with an intervocalic /p/  1lapa `old'
and tapayl `miserly'. In addition, all suxes which begin with /p/ undergo a
morphophonemic change to /w/ when following a vowel. This suggests that his-
torically /p/ had an allophone [w] between vowels, but word-internally this was
then synchronically analyzable as the phoneme /w/, leading to a stage where
Awa Pit had no intervocalic /p/, with the few words containing an intervocalic
/p/ having developed recently (either by the introduction of loans, through com-
pounding, or in the case of 1lapa and tapayl possibly from a disappearing nasal;
see section 2.2.5.3).
2.2.1.1 Geminate stops
It is important to note that in words of two or more syllables it is possible to
obtain a sequence of two stops in intervocalic position, with one stop being the
nal phoneme of the rst syllable, the second stop being the rst phoneme of the
second syllable. When these two stops are distinct, there is nothing unusual in
this sequence. However, when the two stops are identical, they can be pronounced
as a geminate (delayed release) stop, or as a single stop sound. Even when
pronounced as a single sound, the distinction between one phoneme and two is
still maintained, however, as a single stop between vowels is produced as a voiced
fricative (or [R]). Thus there are distinctions such as:
(1) /ittul/ [It:Ul] [itUl]
 `avocado'
/kwitu/ [kwiRu] `guinea-pig (loan, local Spanish)'
/ippa/ [Ip:a] [ipa]
 `lightning'
/kipa-/ [kiBa-] `swell'
/kukku/ [kUk:u] [kuku] `torch'


/mukul/ [muGUl] `tórtula (bird sp.)'

There are few words which contain geminate stops (about twelve in the data),
and some of these are clearly loans (shutta `hat' from Quechua), while others are
probably compounds (for example many fruits end in tul; compare ittul `avocado'
with tizhtul `lemon', where tizh is a word meaning `sour').
2.2. CONSONANTS 27

An alternative analysis would consider that the stops found in intervocalic

position in these words are allophones of a single stop with optional lengthening
(that is, /t/ becomes [t] or [t:]). However this solution has been rejected for a
number of reasons.
To begin with, it is clear that sequences of two identical stops can be
pronounced either with delayed release as a geminate or else as a single stop.
For example, the verb root pat- `wash (clothes)' can be followed by the imper-
ative sux ti, and the resulting word patti is produced with either [t:] or [t]
intervocalically.
Secondly, this analysis would require the introduction of three new phon-
emes to account for intervocalic [B], [R] and [G]. Additionally, two new morpho-
phonemic rules would be required to explain the allomorphy of all suxes begin-
ning with /t/ and /k/, which would have allomorphs beginning with /R/ and /G/
after vowels.5
These factors, plus the low frequency of these sounds, the clear origin of
some as two-stop sequences, and the necessity of two-stop sequences between vow-
els in any case (to account for non-homorganic stop clusters) lead to the analysis
given here, that these intervocalic [p] [p:], [t] [t:] and [k] [k:] are phonemically
  

/pp/, /tt/ and /kk/ respectively.6

2.2.1.2 Stops and sandhi phenomena
Previous works (Calvache Dueñas 1989, Obando Ordóñez 1992, Henriksen &
Henriksen 1979, Montaluisa Chasiquiza 1991) have only examined the phonology
at the level of words, without examination of inter-word phonology. But the
interactions which take place at word boundaries are of vital importance to the
understanding of Awa Pit phonology, particularly stop and fricative phonemes,
and can be used to resolve a number of issues, such as the existence or otherwise
of the phoneme /r/.
The biggest dierence between intra-word and inter-word allophony is that
the former (discussed above) is obligatory, while the latter is always optional 
speakers can pronounce words as they would in isolation, although in connected
speech they usually do not. It might be thought that the allophony associated
with connected speech would be similar to word-internal allophony, and in some
cases it is, but in others it is not. That is, there are signicant boundary phe-
nomena in Awa Pit or, in somewhat outdated structuralist terms, the word break
(`juncture') is an important phoneme in Awa Pit.
At the end of a phonological word, before a following vowel-initial word,
5Although, of course, such a morphophonemic rule is necessary to explain the change of /p/
to /w/ in similar circumstances.
6It is perhaps worth noting that single intervocalic stops never occur adjacent to voiceless
vowels. That is, while a word such as /attiS/ [at:IS][atIS] occurs, a word such as /atiS/ does
 voicing
not, in the analysis chosen here. It is clear that  and  fricativizing of stops does not
 occur
next to voiceless vowels in any case  /kimpu/ contains a [p] rather than the expected [b].
Thus [atiS] could be analyzed as /atiS/. However  the alternation between [t] and [t:] in words
 
such as these leads to their analysis as containing a sequence /tt/.
28 CHAPTER 2. PHONETICS AND PHONOLOGY

a voiceless stop always remains as such. Sometimes the stop is geminated, as

though the second word began with a homorganic stop:
(2) /ap ajmpiS/ [apajmpIS] [ap:ajmpIS] `my brother'


/wat ampu/ [watambu]  [wat:ambu]


 `good man'
At the beginning of a phonological word, after a word ending in a voiced
consonant, a voiceless stop can be voiced, as occurs in the same environment
word-internally:
(3) /akkwan pija/ [akwanbija] `much corn'
/sun t1m/ [tsUndIm] `that basket'
The most interesting changes are those which occur word-initially after a
voiced vowel. In this position a voiceless stop is voiced and fricativized, as occurs
word-internally  however in the case of /t/, the result is [D] rather than [R],
which occurs word-internally:
(4) /maza t1m/ [mazaDIm] `one basket'
/maza kuZu/ [mazaGuZu] `one pig'
All other occurrences of voiceless stops at phonological word boundaries
have the same form in isolation and in the same context within a phonological
word, and no sandhi phenomena occur.
As mentioned earlier, one of the most serious disagreements in previous
studies of Awa Pit phonology regards the status of [R], as either an allophone of
/t/, or as a separate phoneme /r/. Calvache Dueñas (1989) has taken the second
analysis; the other authors (Henriksen & Henriksen 1979, Obando Ordóñez 1992,
Montaluisa Chasiquiza 1991) have taken the former analysis. Calvache Dueñas
introduces /r/ because of the contrast she observes between intervocalic [D] (/t/
in her analysis) and [R] (/r/ in her analysis)  her examples of intervocalic [D]
are [kwaDam] `red' and [swaDaRe] `put over there'. The rst example will be dealt
with in section 2.2.5.3; however the second word is now explainable. The form
given by Calvache Dueñas for `put over there' is not, in fact, a single word, but
two separate words:
(5) su=wa ta-ti
there=in(approx) put-imp.sg
`Put [it] over there.'
The [D] is now expected, having been formed as a result of the application of
sandhi rules.
Given this analysis, plus the account of [kwaDam] in section 2.2.5.3, [R] can
be considered simply as the allophone of /t/ found in intervocalic position.7
There is additional evidence that /t/ does not have a (general) intervocalic allophone [D],
7
which comes from the Spanish spoken by Awa Pit speakers who are imperfect bilinguals. While
those who are uently bilingual have no trouble with the sound system of Spanish, second-
language speakers of Spanish have problems with a number of sounds in Spanish, and one of
these is intervocalic /d/, pronounced [D]. Words or names which contain an intervocalic /d/ in
Spanish, such as /la planada/ [laplanaDa], are produced by Awa Pit speakers as [lap1lanaRa]. If
[D] were an intervocalic allophone of /t/, as Calvache Dueñas claims, we might expect Awa Pit
speakers to produce this sound correctly.
2.2. CONSONANTS 29

[s] [ts] [S] [Ù] [z] [dz] [Z] [Ã]

initial - -p
- - - -
p

nal p
- - p
- p
- p

between vowels p p p
-
p p
- p

before voiceless C p
- - p
- p
- p

before voiced C - - - -
p p p p

after voiceless C - - - - - -
p p

after voiced C - - - - - - - p

Table 2.2: Environments of fricative and aricate allophones

2.2.2 Fricatives (and aricates)

There are a wide variety of fricative and aricate allophones in Awa Pit: [s] and
[ts], [ù] and [tù] (in some dialects), [S] and [Ù], [z] and (only in sandhi cases) [dz],
[Z] and [Ã]. These fricatives and aricates have been combined into phonemes in
a number of ways.
Calvache Dueñas (1989) considers that there are six phonemes  /s/,
/ts/, /z/, /S/, /Ù/ and /Z/8  with [Ã] an allophone of /Ù/, and all other al-
lophones in contrastive distribution. More commonly there are four phonemes
in the analysis  /s/, /z/, /S/ and /Ù/  as in Henriksen & Henriksen (1979),
Montaluisa Chasiquiza (1991) and Obando Ordóñez (1992), with /Ã/ an allo-
phone of /Ù/, and [Z] divided up by dierent authors into /S/, /Ù/ and /z/ in
dierent ways. The analysis suggested here is dierent again, although it re-
lies on the above allophones being divided between four phonemes. Some of the
variation in analyses may be dialect variation, but I believe that the majority of
variation at least can be accounted for in one analysis. The allophonic data on
which the analysis is based is given in Table 2.2.
To begin with, it is necessary to account for the distinction between [s] and
[ts]. In my data, [ts] is found initially, between vowels (apparently), and after
a voiceless consonant; [s] is found elsewhere, but also between vowels. While
both [s] and [ts] are found intervocalically, a sequence [VtsV] can, of course, be
treated as consisting of two syllables (the basic syllable structure of Awa Pit
being (C)(G)V(G)(C), see section 2.5.1), with a syllable break between the [t]
and the [s]. Then [ts] (realizing a single phoneme) is only used word-initially and
after voiceless consonants, [s] elsewhere.
This analysis is given additional support from informant A, who reads and
writes Spanish. At one point I wrote down matsuh for `face'. With no comment
from me, A stated that the word was [mat.tsu], separating out syllables, and
claimed that there were two ts in the word. 
This analysis of [s] and [ts] accords with Henriksen & Henriksen (1979:56),
although they state that [ts] and [s] are in free variation initially (which may be
8 In fact, in all the works in question, the symbols ², £, º and are used.
30 CHAPTER 2. PHONETICS AND PHONOLOGY

related to sandhi rules  see below). Obando Ordóñez (1992) makes no mention
of [ts], but does note a retroex allophone of /s/, [ù], in initial position, and this
would appear to be dialectal. The pattern of [s] and [ts] allophony is also found
in other South American languages  Wari' has [ts] and [s] (and [S] and [Ù])
as allophones of one phoneme (MacEachern, Kern & Ladefoged 1997:22), and
some dialects of Barasano have an aricate [ts] rather than a fricative [s] (Jones
& Jones 1991:10), for example.
However the analysis conicts with Calvache Dueñas (1989). She states
that there are two phonemes /s/ and /ts/ in contrastive distribution, giving ex-
amples /tsula/ [tsula] `tooth' and /sun/ [sun] `that', which begin with an identical
sound in the data used here. It is of course possible that there is a dialect dier-
ence here  this would make Awa Pit similar to the Yanomami language Sanuma,
where in Brazilian Sanuma /s/ and /ts/ are distinct phonemes, though hard to
distinguish and carrying little functional load, while in Venezuelan Sanuma there
is no distinction (Borgman 1990:220221). An alternative possibility is related
to sandhi phenomena discussed below, where after a word ending in a vowel, a
word beginning with /s/ is pronounced with [s] rather than the [ts] it would have
in isolation. Calvache Dueñas's data appears to be largely based on connected
speech, and as such those words beginning with /s/ pronounced after a vowel
would begin with [s], while others would begin with [ts]. Certainly in the dia-
lect studied here speakers make no distinction in initial consonant between those
words which Calvache Dueñas considers to begin with /ts/ and those which she
claims begin with /s/.
Thus an analysis of intervocalic [ts] as a sequence /ts/ rather than a unit
phoneme is supported by the data, and allows a collapsing of [s] and initial and
post-consonantal [ts] into one phoneme; it additionally removes the necessity for
very strong distributional restrictions on phonemes [s] and [ts].9
While all previous work has established a distinction between /S/ and /Ù/
(represented as /s/ and /c/), the parallelism between [s][ts] and [S][Ù] seen in
Table 2.2 suggests treating [Ù] as an allophone of [S], with the same distributional
conditions as [ts]. This analysis is supported once again by informant A, who
when asked the word for `white', wrote putcha and syllabied it as [put.Ùa].
As [pu.Ùa] would be an acceptable Spanish word, but [put.Ùa] is not, it would
seem that what has previously been considered a single intervocalic phoneme
/Ù/ should in fact be considered as two. Once this is done, [S] and [Ù] are in
complementary distribution.
Obando Ordóñez (1992) established /S/ and /Ù/ as separate phonemes, but
gave no examples of an opposition between them; he also notes that /S/ does not
occur initially, and elsewhere that /Ù/ does not occur nally. Calvache Dueñas
(1989) does give initial contrasts for these phonemes, but once again all words
While the phoneme in question will be considered to be the fricative [s], it would be possible
9
to consider that underlyingly it was an aricate /ts/, and likewise /Ù/ rather than /S/. This
would create interesting parallels and dierences between the stops (which fricativize and voice
between vowels but remain stops word-nally) and the aricates (which fricativize between
vowels and word-nally). It would, however, leave Awa Pit with at most one voiceless fricative
(/ì/) but two voiced fricatives (/z/ and /Z/) and will not be followed here.
2.2. CONSONANTS 31

given are pronounced with [Ù] in isolation and [S] after a previous vowel-nal
word by speakers consulted, as was the case for [s] and [ts].
From Table 2.2, it can be seen that two more phonemes must be estab-
lished, /z/ and /Z/, and these have limited distributions, never occurring initially
or after a consonant.10
This leaves one allophone, [Ã], unaccounted for  and this occurs in
a position unique to it, after a voiced consonant and before a voiced vowel.11
While it is perhaps most similar phonetically to /Z/, it will be treated here as
an allophone of /S/, for two reasons: the phoneme /S/ does occur after another
consonant (as [tS]), while there is little evidence of /z/ or /Z/ occurring in this
position; and the phoneme /S/ has an allophone [Ã] in any case, as a result of
sandhi.
Thus the word-internal distribution of the fricatives and aricates is:
8
>
>
< [ts] = # , C , C V
/s/ >
! [voiceless] [voiced] [voiceless]
>
: [s] elsewhere
8
>
>
>
> [Ù] = # , C , C V
>
> [voiceless] [voiced] [voiceless]
<
/S/ > [Ã] = C
!
V
>
> [voiced] [voiced]
>
>
>
: [S] elsewhere
/z/ [z]
!

/Z/ [Z]
!

As with the voiceless stops, the voiceless fricatives can demonstrate sandhi
when they occur word-initially. Following a word ending in a voiced stop, /s/ and
/S/ can be produced as [dz] and [Ã] respectively. Following a vowel-nal word,
there are three possibilities, aside from the usual word-initial aricates [ts] and
[Ù]: they may be simple voiceless fricatives [s] and [S]; voiced fricatives [z] and [Z];
or voiced aricates [dz] and [Ã]. While the voiced fricative realizations are those
of other phonemes, /z/ and /Z/, it is worth noting that there is no possibility of
ambiguity, at least in one sense, as no words begin with voiced fricatives.
Indeed the distribution of all of the fricatives is slightly unusual. The
system itself is clearly odd  a voicing distinction in fricatives would normally
be accompanied by a voicing distinction in the stops, which are all voiceless
in Awa Pit. Equally, Lass (1984:154) states that the number of fricatives in a
language is unlikely to be greater than the number of stops. In Awa Pit the
Except in the imperative verb sux zha.
10
It might be expected that there would be a corresponding allophone [dz] occurring in the
11
same environment. However this allophone was not found (except in sandhi cases).
32 CHAPTER 2. PHONETICS AND PHONOLOGY

number of fricative places is less than the number of stop places, but because of
the voicing contrast there are more fricatives in total.12
The phonemes /S/ and /Z/ have an anity for /i/, /i/ and /j/  they
 vowels or glides
are found in other environments, but there is often one of these
around, and indeed some words seem to have introduced (or possibly lost) a
glide adjacent to one of these fricatives; for example, [paSpa] and [pajSpa] are
both possible pronunciations for `child'. In the reverse situation, /z/ is never
found near high vowels, and /s/ appears next to /i/ in only a few suxes. This
suggests that /Z/, at least, may have developed from /z/, as there are only
about ten words with /Z/ not accompanied by a high vowel or palatal glide 
at least one of those words is a loan from Quechua (/kuZu/ `pig'), and, given
the lack of mirrors in traditional Awa society, /maZa/ `mirror' is also possibly
a loan. It is also interesting to note the informal term for a woman's sister,
/aZa/, compared with the formal term, [ajS-piS]: there is normally a fairly direct
relationship between the initial segment of the formal term and the informal
term.

2.2.3 Voiceless lateral fricative

The dialect of Awa Pit examined in this study has a voiceless lateral fricative
/ì/, with allophones [h] and [ì], the former being used before /u/ and /u/, and

also sometimes adjacent to other voiceless vowels, the latter used elsewhere:
8
>
>
< [h]
> = u, u

/ì/ > [ì] [h] = i, 1
!
, i, 1
>
>

   
: [ì] elsewhere
This phoneme corresponds to the phoneme /h/ or /x/ in previous works.
The allophones of this phoneme appear to be strongly variable across dialects:
Calvache Dueñas (1989:4546) lists [x], [ç] and [h] as allophones of /h/; while
Obando Ordóñez (1992:7778) has [ç] and [x] as allophones of /x/; Montalu-
isa Chasiquiza (1991:29) has [h] as the only allophone of /h/; and Henriksen
& Henriksen (1979:57) have [x] and [ì] as allophones of /x/, while a separate
phoneme /h/ has an allophone of [h]. In the dialect studied here it makes most
sense to consider the phoneme as /ì/, as the allophone with the least restricted
distribution, but it would appear that for most dialects of Awa Pit the phoneme
would be better considered as /x/, or perhaps /h/, as [ì] is not even an allophone
in those dialects. However if /ì/ is considered as the basic allophone, the dis-
tribution of the allophone [h] in this dialect is phonetically plausible  before a
back vowel, the fricative assimilates backness; associated with a voiceless vowel,
the fricative varies with one requiring the least articulatory eort.
12This is at least true in those dialects with an additional velar/glottal fricative, where there
are three stops versus ve fricatives. In the dialect studied here, the velar/glottal fricative is
better analyzed as a lateral fricative  and Lass (1984:156) considers this a stop. However
distributionally it appears to pattern with the lateral approximant or the fricatives.
2.2. CONSONANTS 33

The correct phonemic grouping of /ì/ is somewhat dicult. Distribution-

ally, /ì/ has exactly the same possibilities as the lateral /l/ and could perhaps
be considered as its voiceless counterpart: it does not occur initially, nor after a
consonant. This is also the distribution of the voiced fricatives. If /ì/ is grouped
with the (voiceless) fricatives instead, it is possible to predict a higher number
of occurrences of voiceless vowels (see below); however /ì/ does have a greatly
restricted distribution compared with /s/ and /S/.13
There also appears to be some sort of relationship (probably dialectal)
between [ì] and [t] after a voiceless vowel in a few words. There are three words
where one of my informants, R, has [ì] while the others have [t] (recalling that
[t:] is /tt/):
(6) [kIt:Us] [kIìtUs]
 `making (a bed)'

w 

[p It:Us] [p IìtUs] `plucking'
w
[ÙIt] [ÙIì] 
 `(noise made by a) grasshopper'
 
In the rst two words it appears that it would be possible to claim assimila-
tion of the /ì/ to a following /t/. That this is not the case can be seen by
the third word, and by using the rst verbs in dierent aspects, for example
[pw Itn1maro] [pwIìn1maro] `going to pluck', where there is no /t/ to trigger the


assimilation.  necessary to consider that for speakers who say the second
It seems
option, the verb root is /k1ì-/, while for speakers who use the rst it is /k1t-/,
and similarly for the otherwords. This result will become extremely important 
in the discussion of voiceless vowels in section 2.4.2.

2.2.4 Lateral approximant

The lateral approximant /l/ has a number of allophones, many in free variation.
After /i/ and /j/ the lateral is palatalized to [L], although for some speakers
this is in free variation with [dl]; before consonants and word-nally it is often
prestopped or even produced as the stop [d]; and after a non-front vowel it is
often produced as prestopped, occasionally as [d]. Thus:
8 ()
>
> i
>
> [L] [d
 l] =
j
<
/l/ >
[l] [dl] [d] = C, #
!
>
>
 

>
: [l] [dl] elsewhere
The precise frequency of the various allophones in free variation depends very
much upon the speaker  some of my informants almost never used [l], while
others used [l] almost exclusively in word-lists (although they also used [dl] in
connected speech).
13Unless of course the analysis separates /s/ and /ts/ as distinct phonemes, and also separates
/S/ and /Ù/, in which case the distributions of /s/, /S/ and /ì/ are identical.
34 CHAPTER 2. PHONETICS AND PHONOLOGY

2.2.5 Nasals
Certain aspects of the nasals of Awa Pit are relatively uncontroversial. However
the analysis presented here departs from previous studies in several important
ways.
Awa Pit has a variety of allophones involving nasality: [m], [mw], [n],
[ñ], [N], [], and nasalized vowels, as well as a series of disappearing nasals. In
this analysis, [m] and [mw] are allophones of /m/; [n], [ñ], [] and possibly some
examples of [N] are allophones of /n/; while nasalized vowels and disappearing
nasals are related to /N/.
It is important to point out that non-homorganic nasalstop clusters are
possible in Awa Pit, although in rapid speech the nasal tends to assimilate to
the place of the stop. Especially frequent is the sequence /mt/, found in several
roots and also in one of the allomorphs of the Imperfective sux.
2.2.5.1 The phoneme /m/
All studies have taken /m/ to be realized basically by [m], although in some
varieties of Awa Pit, such as the one examined here, both /m/ and /p/ have an
o-glide before /1/:
8
>
< [mw] =
/m/ 1
>
!
: [m] elsewhere
2.2.5.2 The phoneme /n/
The basic allophonic variation of /n/ is relatively uncontroversial:
8
>
>
>
> [N] = C
>
> [velar]
<
/n/ > [n] [ñ] = V
!
V
>
>


high [high]
>
>
[ ]
>
: [n] elsewhere
However /n/ is involved in a further allophonic variation, giving rise to [ñ]. This
allophonic rule is:
8
>
<
/nj/ > [ñ] [] = V
!
 a
: [ñ] = a
This is indeed controversial. The existence of [ñ] has been dealt with previously
in two separate ways. Obando Ordóñez (1992), Montaluisa Chasiquiza (1991)
and Henriksen & Henriksen (1979) all establish a phoneme /ñ/; in the last of
these it is noted that /ñ/ has free variation between [ñ] and [] in intervocalic
2.2. CONSONANTS 35

environments, with the latter copying nasalization onto the surrounding vowels.
Calvache Dueñas (1989) treats this dierently. She claims that a word such
as [tuña] [tua] `mouse' is phonemically /tuja/, with nasalization being copied


from the following vowel onto the glide, and then this nasalized glide [] sometimes
strengthening to a nasal [ñ].
I would claim that there is in fact no separate phoneme /ñ/, in part
because of distributional criteria. The [ñ] [] only occurs preceding /a/: it


thus occurs syllable-initially only, and only when the vowel is /a/. My reason for
rejecting Calvache Dueñas's (1989) analysis has to do with citation examples and
dialect variations. In citations, rather than in text, [ñ] is much more frequent
than [], suggesting that [ñ] may be the more basic form. Also, in the dialect
studied for this work, the phoneme /a/ (actually analyzed here as /aN/) has
been lost, being replaced by an unnasalized /a/. It is thus not possible in this
dialect to consider that phonologically the word for `mouse' is /tuja/, with the
nal nasalized vowel triggering nasalization of the glide, as /a/ has been lost
everywhere. Thus the nasalization must proceed from the glide. The sequence
/CjV/ has to be established for Awa Pit (for /pj/, /kj/ and /ìj/ sequences), and
hence it seems most sensible to use this already established sequence to explain
the phone [ñ]. Thus in my analysis,
/tunja/ [tuña] [tua] `mouse'


(7) /njan/ [ñEn] `other'

/pjan/ [pjEn] `bridge'
There is some supporting evidence (although limited) from local Spanish. Spanish
has a number of names which have the sequence /nja/, such as Lisenia /lisenja/.
In standard Spanish these should be pronounced [lisenja], however in local Span-
ish they are produced with an [ñ], [liseña]. It must be noted, however, that this
evidence is weak, as Spanish also has a phoneme /ñ/.
There is an additional concern with the pre-velar [N] allophone of /n/.
Within a root the sequence [ng] is never found, but this sequence has been found
in a sux. The Polite Imperative has two allomorphs, naka and nka, and the
latter is most commonly [nga] rather than [Nga] (although this does occur). While
this sux is hypothesized as originating as two separate suxes, n(a) and ka
(see section 9.4.5.1), this does not explain the presence of [n] rather than [N]
synchronically. Consequently while words such as [kwaNgwa] `grandmother' will
be analyzed here as /kwankwa/, it is possible that an analysis as /kwaNkwa/
would be more appropriate; the rules given below for loss of /N/ would become
more complex, however, as it is never lost in these words.
2.2.5.3 The phoneme /N/
Languages in the south of Colombia and nearby areas of Ecuador and Brazil
are known for their unusual nasalization phenomena. Probably best known are
the Tucanoan languages like Barasano in the Amazon to the east of Awa Pit,
which have no nasal consonants, only nasalized vowels, but where the nasaliza-
tion feature spreads across syllables aecting consonants also, until reaching a
36 CHAPTER 2. PHONETICS AND PHONOLOGY

consonant which blocks nasalization; thus [m] is an allophone of /b/, and so on

(Jones & Jones 1991:1316). An identical process occurs in Chocó languages on
the coast to the west of Awa Pit, in languages such as Epena Pedee (Harms 1985).
The other members of the Barbacoan family also have unusual nasalization
patterns, although little is known about them. Cha'palaachi has been analyzed
as containing no nasalized vowels, but containing a phoneme /N/, traditionally
only in syllable-nal position, which can be actualized as nasalization of the
preceding vowel in utterance-nal position (Lindskoog & Brend 1962:35-37). In
contrast, the closely related language Tsaqui is analyzed as having phonological
nasalization (Moore 1962:271). Guambiano is usually not considered to have
nasalized vowels (Vásquez de Ruíz 1988); and while Huber & Reed (1992) (using
information from Branks & Branks (1973)) give no indication of nasalization
in the introductory remarks about Guambiano, a footnote to an `blood' notes
that morpheme nal nasals are realized as nasal consonants homorganic with
the following obstruent; preceding non-obstruents, it is unrealized (Huber &
Reed 1992:33).
Awa Pit also has two instances of unusual nasalization  the nasalized
vowels and the disappearing nasals.
Nasalized vowels Nasalized vowels denitely exist, as phones, in Awa Pit (see,
for example, the spectrograms in Obando Ordóñez 1992:157158). Henriksen &
Henriksen (1979), Obando Ordóñez (1992) and Calvache Dueñas (1989) all recog-
nize the existence of phonological nasalized vowels, although all their examples
use the nasalized vowel word-nally.14 Montaluisa Chasiquiza (1991:910) also
considers that nasalized vowels exist phonologically, although he does note that
the frequency of occurrence of the nasalized vowels is very low, and that there
are a number of words where the vowel may be produced as nasal or as oral (his
examples are, once again, all word-nal).
As was mentioned above, in the dialect studied here all instances of [a]
have been lost.15 Words which other authors have transcribed containing [a],
such as [kup:a] `squirrel', are found with a simple oral vowel, [kup:a]. All other
words which can be produced with a word-nal nasalized vowel have occasionally
been found with a simple oral vowel, or more commonly with a nasalized vowel
followed by a nasal consonant [N].
The phonetically nasalized vowels have strong distributional restrictions
placed on them. They are normally found word-nally, although some examples
have been found in nal position of verb roots. Nasalized vowel plus glide se-
quences have also been found: [mu] [muN] `expensive, worth'. With the ex-


ception of these glides, no syllable-nal consonant is possible after a nasalized

vowel.
14With the exception of Calvache Dueñas (1989), who also has a number of nonword-nal
nasalized vowels, in contexts where for her the nasalized vowel is essential to cause nasalization
of a preceding /j/  in these contexts other analyses, including my analysis, consider that the
word contains a nasal consonant and following oral vowels; see the preceding section.
15If these are indeed nasalized vowels, this goes against the typological pattern discussed in
Ruhlen (1978:228229), that the rst vowels to be denasalized are high vowels.
2.2. CONSONANTS 37

This suggests that, in fact, the nasalized vowels are better treated as vowel
(plus glide) plus consonant sequences, as this would then explain the lack of a
nal consonant without recourse to extra rules  as there is already a consonant
present, /N/, no other consonant can occur here.
The analysis of nasalized vowels as vowels followed by /N/ gains force
from allophonic considerations. After a nasalized vowel, consonants take the
allophone appropriate for following a nasal. For example, the locative sux /ta/
has the form [da] after a nasalized vowel, whereas after an oral vowel it is [Ra]:
[da] [Nda] `in the ame'; the Imperfective aspect sux, /mtu/ after a vowel,


has the form /du/, normally only found after a voiced consonant: [dus] [Ndus]


`I am carrying'.
The nasalized vowels of other analyses thus correspond here either to an
oral vowel (in the case of /a/) or to a vowel followed by the velar nasal /N/.
Disappearing nasals There is one very odd phenomenon which occurs with
a few words in Awa Pit  they have disappearing nasals. This has not been
discussed in previous works, although Montaluisa Chasiquiza (1991:9) has noted
its existence, and it clearly operates also in the dialect studied in Calvache Dueñas
(1989), as while she does not mention it, Calvache Dueñas's examples contain
the word `woman' transcribed in dierent places as either [aSamba] or [aSaBa].
The phenomenon in question operates on a few words containing an /a/
(in one case an /u/), a nasal, and a (homorganic) stop. These are words such
as `woman', which in word-lists is usually pronounced with a nasal, as [aSamba].
However in connected speech this word can be pronounced as [aSamba], [aSabba]
or [aSaBa]. In fact, in one word-list recording session informant T self-corrected:
he began by saying [aSaBa], realized he had said it in a way which he con-
sidered incorrect, said aaah [aSabba], realized it still wasn't right and said aaah
[aSamba]. However, this phenomenon is word-specic, and so while the nasal in
[aSamba] `woman' can disappear, the nasal in [ambu] `man' is always present;
there does not appear to be a correlation with anything else, such as stress.
As well as being word-specic, the phenomenon appears to depend some-
what on dialects also. My informant from Arrayán always pronounced the nasal
in the word [aSamba], while informants from Pueblo Viejo and Pialapí often did
not. While the phenomenon clearly operates in other regions also, some dialects
appear to go the opposite way for some words, losing the stop rather than the
nasal  the reader Zhitzhu (Henriksen & Henriksen 1978) contains the word
kuanam `red', which in Pialapí is produced [kwandam] in word-lists.
With disappearing nasals before /p/ and /k/ there are relatively few
diculties. It would be possible to claim that a word such as `woman' has two
phonological representations, /aSampa/ and /aSapa/; while this would not cover
the pronunciation [aSabba], it would cover the two more common pronunciations.
However this analysis does not work for those clusters involving /t/, such as in
[kwandam]. If the /n/ were optionally deleted at the phonological level, we would
obtain:
38 CHAPTER 2. PHONETICS AND PHONOLOGY

(8) /kwantam/ [kwandam]

/kwatam/ [kwaRam]
In fact, much more common for all except one of my informants is the pronunci-
ation [kwaDam]. Thus this appears to be a phonetic, rather than a phonological,
process: the nasal causes the stop to be voiced; the nasal is converted to a stop;
the two identical stops become reduced to one; and the single stop is fricativ-
ized between vowels. For informant L, this process is still possible, and so he
sometimes produces [kwaDam] as well as [kwandam]; in addition, however, he
sometimes produces [kwaRam], so it would appear that he has variation in the
phonological representation of the word as well as the phonetic production. How-
ever all other speakers produce only [kwandam] and [kwaDam], never [kwaRam].
It is interesting to note that these words all contain /a/ followed by a nasal,
with the exception of [un-da] `there' which is possibly being analogized with [an-
da] `here', which does contain an /a/. After an /a/ is precisely the environment
in which /N/ has been lost in this dialect, leading to oral rather than nasalized
vowels. If words with disappearing nasals are analyzed as containing /N/, it
would then be possible to claim that before a stop within a root either /N/ is
obligatorily assimilated to the following stop, or else it is lost, but aects the
production of the following stop ([D] rather than [R]).
At this point it should be noted that Calvache Dueñas's (1989) second
example showing the necessity of a phoneme /r/ has now been discounted. She
had two examples of intervocalic [D], which she stated realized /t/, requiring a
phoneme /r/ to account for intervocalic [R]. Her rst example was dealt with
by stop sandhi patterns in section 2.2.1.2; the other example was [kwaDam],
which as shown above contains a phonological /N/ before the stop, leading to its
pronunciation as [D] rather than [R].
Throughout this work, then, a root-nal /N/ represents optional nasaliza-
tion of the preceding vowel, while a root-internal /N/ represents a disappearing
nasal. These are, perhaps, not the most appropriate analyses of these phenom-
ena, and more work needs to be done, particularly on the latter. However the
use of an /N/ to represent nasals which are sometimes there and sometimes not
at least signals those nasals to which this process applies.

2.3 Glides or semivowels

As in many languages, the relationship between the semivowels /w/ and /j/ and
the vowels /u/ and /i/ is complex in Awa Pit, and has been analyzed in a variety
of manners. The most appropriate analysis appears to be that there are indeed
glides separate from the vowels, but that under certain circumstances the vowels
can be pronounced as a glide in diphthongs. For the semivowels we obtain the
following:
/w/ [w]
!
2.3. GLIDES OR SEMIVOWELS 39
8
>
<
/j/ > [j] [é] = V
!
 V
: [j] elsewhere
First it is necessary to establish the existence of the semivowels as separate
from the vowels, that is, nd a contrast between the two. In the case of /w/ and
/u/ there are a number of words where this distinction can be seen. For example,
there are words:
(9) [kwaNgwa] `grandmother'
[kuwa] `sister'
[pwii] `owing'
Several problems arise if it is assumed that /w/ does not exist, and that in these
words there is a vowel /u/ which is produced as a glide. Firstly, because of
[kwaNgwa] the word for `sister' cannot be phonologically /kua/, since then there
is no explanation for /kuankua/ having the /u/ as [w], while /kua/ has the vowel
/u/ pronounced in full. Therefore, it is necessary to assume that there are four
segments in `sister'. One possibility is to assume that `sister' is in fact /kuua/
(and indeed this seems to have been the idea of Calvache Dueñas in a preliminary
work, Calvache Dueñas (1988)). However then there is a conict between `sister'
[kuua] and `owing' [puii], as it is unclear which vowel should be transformed into
a glide; in the rst word it is the second vowel, in the second word the rst vowel.
Therefore it seems that a phoneme /w/ must be established to account for this.
The phoneme /w/ is also needed to account for those [wV] which are
always [wV] versus those which alternate between [wV] and [uV]. For example,
there is a contrast:
(10) [kwaNgwa] `grandmother'
[kuamduaze] [kwamdwaze] `they were eating'


One way of distinguishing these involves morphology  the rst word is a single
morpheme, while the second consists of ve morphemes [ku-a-mdu-a-ze]. It would
be possible to say that where a morpheme boundary intervenes between a /u/
and an /a/, the two can be pronounced as separate vowels, while if there is no
morpheme boundary the two must be pronounced as a diphthong. But given
the necessity for a /w/ presented above, it seems better to consider that the
distinction is:
(11) /kwankwa/ `grandmother'
/kuamtuazi/ `they were eating'
and then the dierence can be shown purely on the level of the phonology, without
resort to morphology.
There is an additional contrast between /u/ and /w/, found in one word.
The sequence [aw] is found in many words, and it varies with [o]: [tsaw] [tso] 

`eld'. This can be analyzed as /aw/. This contrasts with the word [au] `we',
which is always pronounced as two separate vowels, and is analyzed as /au/.
40 CHAPTER 2. PHONETICS AND PHONOLOGY

Indeed, for some speakers such as L, this word often has an epenthetic [G]:
[au] [aGu]. This appears to be the same process as occurs in Inga Quechua


to the east, where [Gu] is analyzed as an allophone of /u/ found syllable-initially

after a syllable-nal vowel (Levinsohn 1976:26).
There is also evidence for the existence of a glide /j/, although less than
for /w/. This relies on the allomorphy of the person-marking suxes on stative
verbs. After a vowel /i/, the Locutor marking is /s/ and the Non-locutor is / /: ;

[i-s] `I am', [i] `she is'. However a verb root ending in [j] has dierent allomorphs,
those of consonant-nal roots: [waj-is] `I am lacking', [waj-i] `it is lacking'.
The phoneme /j/ has variation in intervocalic position in this dialect. In
previous works intervocalic /j/ (or /i/) is always [j]. However in intervocalic
position in this dialect it is sometimes found strengthened to a palatal stop, as
indicated above.
Having established the existence of phonemes /w/ and /j/ as distinct from
/u/ and /i/, we can then assume that in words containing [w] and [j] where there
is no possibility of contrast, we in fact have the phonemes /w/ and /j/ rather
than /u/ and /i/ used as glides. For example, we can assume that words such
as [wat] `good' and [jal] `house' contain the semivowels /w/ and /j/ in initial
position rather than /u/ and /i/.

2.4 Vowels
Awa Pit has both voiced and voiceless vowels, although the latter are controver-
sial. There are clearly three voiceless vowel phonemes, if there are any, /i/, /1/
and /u/; and the present analysis has four voiced vowels, /i/, /1/, /u/ and  /a/,

 a fth, /e/, has previously been considered phonemic. The nasalized
although
vowels have been considered as allophones of vowels with a following /N/ here 
see section 2.2.5.3. There are also long vowels, a series of two identical vowels.
In general terms, there is a distinction between allophones of vowels in
open and closed syllables: vowels in closed syllables tend to be more open or lax,
while vowels in open syllables tend to be more close or tense. Thus [I], [I], [U]
and [5] tend to appear in closed syllables, while [i], [1], [u] and [a] appear in open
syllables; but these are tendencies rather than absolute rules.

2.4.1 Voiced vowels

There are four voiced oral vowel phonemes in Awa Pit:
Front Central Back
High /i/ /1/ /u/
Low /a/
This analysis assumes only four voiced oral vowels, in contrast to Obando Ordóñez
(1992), Calvache Dueñas (1989) and Henriksen & Henriksen (1979), where ve
distinct voiced oral vowels are considered. Montaluisa Chasiquiza (1991:5) also
2.4. VOWELS 41

considers that there are ve voiced oral vowels, although he does note that /e/
parece hallarse en estado de conformación como fonema,16 thereby giving it
somewhat less status as a clearly distinct phoneme. Because of the contrasts
between analyses with ve voiced oral vowels and this one with four, a separate
subsection below is dedicated to an examination of the phone [e].
The phoneme /i/ has the following allophones:
8
>
<
/i/ > [i] [e] =
!
 #
: [I] [i]  elsewhere
For some speakers, the allophone [e] also appears (in free variation with [i]) in
words borrowed from Spanish which contain [e] in Spanish, such as /wipu/ [weBu]
 [wiBu] `egg'.
The phoneme /1/ has three allophones:
8
>
<
/1/ > [1] [e] =
!
 # (words of more than one syllable)
: [I] [1]  elsewhere
The phoneme /1/ also causes a preceding labial consonant (/p/ or /m/) to be
labialized ([pw1] or [mw1]).
There are three major allophones of /u/:
8
>
<
/u/ > [u] [o] =
!
 #, and a few other words
: [U] [u]  elsewhere
Of all the vowels in Awa Pit, /a/ has the widest allophonic variation, with
many of the allophones being in free variation.
8
>
>
>
> [O] = w C
>
> [velar]
>
>
>
< [æ] [E] [@] = [palatal
>   C , in a closed syllable
/a/ > !
]
>
> [aj] [a] = C 
> 
fricative
>
>
>
> alveopalatal
>
: [5] [a]  elsewhere
As well as these allophones, there is an optional reduction of [aj] to [e]
or [E], especially (though not only) before a nasal. This appears to be determined
by the speaker, with variation between informants: for example, R uses only the
form [aj]; E uses [E] before a syllable-nal nasal, but otherwise [aj]; while L nearly
always uses [E] or [e] before a nasal.
16 seems to be in the process of becoming a phoneme
42 CHAPTER 2. PHONETICS AND PHONOLOGY

2.4.1.1 The allophone [e]

From the preceding sections it can be seen that there are many possible sources
for the sounds [e] and [E]. Both /i/ and /1/ have [e] as a word-nal allophone;
some words with /i/ have free variation between [i] and [e]; /a/ has an allophone
[E]; and the combination /aj/ can, under certain circumstances, be pronounced
as [e] or [E].
There is also a dialect variation in the topic marker, with some dialects
using only [na], others using [na] and [ne] in free variation. However, once again,
it seems that phonologically this marker is /na/.
Thus while there are many dierent sources from which [e] and [E] may
arise, there does not seem to be a separate phoneme /e/, as no words have
been found containing [e] or [E] which cannot be accounted for in terms of other
phonemes.
The only word which possibly contradicts this statement is the Serial Per-
fective aspect marker, which is transcribed here as /kwaj-/. While it has been
heard on a very few occasions as [kwaj-], it is overwhelmingly pronounced [kwe-],
even when it is not followed by a nasal.
2.4.1.2 Long vowels
Awa Pit does not have phonologically long vowels; but it does have the possibility
of double vowels, or two identical vowels occurring together. When articulated
carefully, such as in word-lists, each of the vowels is given its own pronunciation,
and a word such as /pii/ `river' is produced as [pii], appearing to have two
vowel peaks, though no consonant intervenes. In rapid speech, the occurrence of
two identical vowels adjacent to one another is much more dicult to perceive,
although there are cases where there is a distinction in sound between a double
vowel and a short vowel. In the case of /ii/ and /uu/, these could be analyzed
as /iji/ and /uwu/. However this would leave /aa/ and /11/ unaccounted for.
These double vowels appear to have a very limited distribution  they
occur only in roots which would be monosyllabic if it were not for the presence
of the two identical vowels. However monosyllabic roots are also possible in Awa
Pit, so it is not the case that there is an automatic rule doubling the vowel of a
monosyllable. For example there is a distinction:
(12) /ki-/ `do'
/kii-/ `marry'
If suxes are added to a root with a double vowel, the two vowels are
maintained:
(13) /ki-mtu/ [kImdu] `doing'
/kii-mtu/ [kiImdu] `marrying'
It should be noted in the last example that the two vowels can take dierent
allophones, depending on the allophone appropriate for their position  here the
rst /i/ is in an open syllable, and therefore has the allophone [i], while the second
2.4. VOWELS 43

/i/ is in a closed syllable, and has the allophone [I]. In rapid speech there is only
one vowel peak for the two vowels, however it is the second vowel which is elided,
and the presence of two vowels can often be determined from the dierence in
vowel quality:
(14) /pastu/ [p5stu] `Pasto'
/paas/ [pa5s] [pas] `two'


2.4.2 Voiceless vowels

Listening to Awa Pit spoken, one of the more unusual phenomena which is noted
quite quickly is the existence of voiceless high vowels in the language, in words
such as:
(15) [ajmpIS] `male's brother'
[ÙIt:1] `hand'
[k1su] `sand'

[Ù1] `bone'
[tu] `higra (a shoulder bag)'

Voiceless vowels, while unusual cross-linguistically, are not unknown, being
found in languages such as Japanese (Hinds 1986:400401), Numic languages such
as Tümpisa Shoshone (Dayley 1989:400401),17 and Amazonian languages like
Cofán, spoken to the east of Awa Pit, where there is free variation word-nally
between voiced and voiceless vowels (Borman 1962:54), and Baniwa of Içana,
where they are allophones of (voiced) vowels next to voiceless liquids and glides,
and aspirated stops (Sasha Aikhenvald, p.c.). Voiceless vowels are also found
in the Andean Spanish of Tuquerres, quite close to the Awa region, where for
example pues is [pwis] and papitos is [papitos] (personal observation).
 a great deal of debate
There has been  over the phonological status of
voiceless vowels in another Numic language, Comanche. Canonge (1957) claimed
that these vowels are voiceless, while in Armagost (1986) and Armagost (1988)
it is maintained that a devoicing rule will account for at least the majority of
voiceless vowels as allophones of the voiced vowels. At a more theoretical level
Cho (1993) has claimed that there are no phonemic voiceless vowels in the world's
languages, although there are two dierent processes by which vowels are devoiced
(these processes will not be discussed here, as they are not strictly relevant to
the discussion).
The presence of voiceless vowel sounds is indisputable in Awa Pit, with
conrmation from spectrographic analysis (Calderón Rivera, Trillos Amaya, Re-
ina & R. de Montes 1987). However there is dispute over the phonological status
17 The association between voiceless vowels and preaspirated stops in Shoshoni, where these
stops developed from geminated consonants following voiceless vowels (Miller 1980:153) is inter-
esting; while Awa Pit appears to be the only Barbacoan language with voiceless vowels, based
on current descriptions, Tsaqui does have preaspirated stops (Moore 1966). Unfortunately,
almost no cognates have been found between the two languages which involve Awa Pit voiceless
vowels (Curnow & Liddicoat forthcoming).
44 CHAPTER 2. PHONETICS AND PHONOLOGY

of these vowels. Calvache Dueñas (1989) and Henriksen & Henriksen (1979) claim
they are allophones of the voiced vowels; Henriksen & Obando Ordóñez (1985)
and Obando Ordóñez (1992) claim they are separate phonemes; while Montalu-
isa Chasiquiza (1991) attempts to walk a middle line, claiming that they are not
phonemes, but rather they are caused by a following /h/ which has since been
lost (this is also the essence of the argument in Henriksen & Henriksen (1979)).
The analysis of voiceless vowels as caused by a following /h/ cannot be
maintained as a unitary explanation without destroying the simple syllable struc-
ture of Awa Pit. In the word for `male's brother' given above, there would be
a word-nal /hS/ sequence, but only one consonant is permitted word-nally in
Awa Pit.18
Turning to Calvache Dueñas's (1989) rules, she claims that /i/, /1/ and
/u/ are devoiced between two voiceless consonants, and also before a word-nal
voiceless consonant. These rules are obviously not correct, and easily falsied
using data provided by Calvache Dueñas herself. Words such as /pit/ `mouth'
and /up/ `your' would have to be produced with voiceless vowels according to
these rules, and Calvache Dueñas gives them with voiced vowels.
Calvache Dueñas (personal communication) has suggested a modication
of these rules: vowels are devoiced between two voiceless consonants, where at
least one of these is a voiceless fricative (or aricate). Leaving aside those words
which end in a voiceless vowel (which will be discussed shortly), almost all words
containing voiceless vowels conform to this pattern.
There are, however, a number of words, at least in the dialect studied here,
which break these rules. For example, there are words which should contain a
voiceless vowel according to the rules and do not:
(16) [tsIptu] `sewing'
[ÙUt:a] `hat'
[pIìt5m] `green/blue'
[tISnUl] `lemon'
[pIt:Us] `I am sleeping'
(The nal example here, and many others like it, contain the imperfective marker
/-(m)tu/ and the locutor marker /s/.) There are also a few words which con-
tain voiceless vowels without a fricative; these words are those discussed in sec-
tion 2.2.3, where there is dialect variation between a root in /ì/ and /t/:
(17) [kIt:Us] [kIìtUs]
 `making (a bed)'
[pw It:Us] [p

 w IìtUs] `plucking'
[ÙIt] [ÙIì] 
 `(noise made by a) grasshopper'
 
Thus the rules are starting to look doubtful, although they do cover the majority
of words with non-nal voiceless vowels.
When we look at words with nal voiceless vowels, the distinction between
voiced and voiceless vowels becomes even more apparent. Calvache Dueñas (1989)
18As Sasha Aikhenvald (p.c.) points out, this could be avoided by treating /h/ as a glide.
This solution does not, however, account for following issues.
2.4. VOWELS 45

claims that words do not end in voiceless vowels, that in fact there is a fricative
/h/, corresponding to /ì/ here, following the vowel. However this is clearly
not the case in the dialect being studied here  while there could be some
doubt as to whether there is a nal [h] after a voiceless vowel or not, in this
dialect the nal fricative would be produced as [ì], and it is quite easy to hear
its presence or absence. In fact, there are several words which do end with a
voiceless vowel followed by /ì/, such as [ÙIì] `grasshopper'; however there are
many more words which end with a voiceless  vowel, such as [Ùi] `bone'. There is
also spectrographic evidence that words such as [k1] `leaf' do not end in a nal
consonant (Calderón Rivera, Trillos Amaya, Reina & R. de Montes 1987).
Additional evidence that these words end in a nal vowel phonologically,
as well as phonetically, is available from the morphology. When the locative post-
position /ta/ is cliticized onto a noun, the /t/ is produced as [d] after a voiced
consonant, as [t] after a voiceless consonant, and as [R] after a vowel. When it
is added to the word [tu] `higra (shoulder bag)', the allophone appropriate for a
vowel-nal root occurs,[tu-Ra] `in the higra'. This means that Calvache Dueñas's
(1989) /h/ and Montaluisa Chasiquiza's (1991) underlying /h/ cannot be main-
tained.
Thus, faced with contrasts such as
(18) [ÙIt:1] `hand'

[ÙUt:a] `hat'
[tu] `higra (shoulder bag)'

[tu] `be in a place'
it must be concluded that Awa Pit has voiceless vowels in its phonological system,
as well as phonetically. Clearly this issue must be examined in greater depth,
given Cho's (1993) claims that voiceless vowels do not have phonemic status in
the world's languages.
Of course, it must be kept in mind that the distribution of voiceless vowels
is quite limited. They can only occur between two (phonologically) voiceless
consonants, or word-nally after a voiceless consonant. It appears likely that the
development of voiceless vowel phonemes is a relatively recent phenomenon, and
they have developed historically from an allophonic rule such as that suggested
by Calvache Dueñas  however with the apparent loss of some nal voiceless
fricative, and the development of a few words where the voicing of an internal
vowel is not predictable, the voiceless vowels have achieved phonological status.19
Thus the voiceless vowels are synchronically phonemes, with the allo-
phones:
/i/ [I] [i]
! 
  
/1/ [I] [1]
! 
  
/u/ [U] [u]
! 
  
19As noted before, however, there could be a connection between Awa Pit's voiceless vowels
and Tsaqui's preaspirated stops.
46 CHAPTER 2. PHONETICS AND PHONOLOGY

2.5 Phonotactics
2.5.1 Syllable structure
The syllable in Awa Pit has the following form:
 = (C)(G)V(G)(C)
where C is any consonant; G is one of the glides /w/ or /j/; and V is any of
the vowels. A single syllable can contain a segment in all ve positions: [kwajL]
`bad'.
There are a number of restrictions on the occurrence of phonemes, in terms
of position in the word. The phonemes /z/, /Z/, /l/ and /ì/ never occur at the
beginning of a word, nor do they occur after another consonant. The velar nasal
/N/ is only found in syllable codas (either word-nally or before a consonant).
The voiceless vowels only occur between two voiceless consonants or nally after
a voiceless consonant. The phoneme /1/ never occurs adjacent to the semi-vowels
/j/ or /w/.

2.5.2 Stress
Stress is not phonologically distinctive in Awa Pit. There are no words which
are distinguished by stress, and while speakers seem to have a preferred stress
for each word, for some words dierent speakers stress the word dierently (in
isolation); sometimes the same speaker will stress the word dierently on dier-
ent occasions. The morphology and syntax also interact with the stress: words
normally stressed on one syllable may be stressed on a dierent syllable depend-
ing on the morphemes or words surrounding it. It should be noted that while
the stress may shift, each phonological word does receive a stress somewhere, at
least in slow speech.
As stress is not distinctive (and indeed my informants would never correct
me if I repeated a word after them with a dierent stress pattern), it will not
be dealt with here any further. It is, clearly, an area which requires a great deal
of further work to establish regularities. In particular, the eects of clitics must
be examined, and the possibility of phenomena such as extraprosodic morpho-
logy, known to occur in South America (cf. Aikhenvald 1996b), where certain
morphological elements don't count for stress rules, while others do.

2.6 Orthography
In the remainder of this thesis, a practical orthography will be adopted to avoid
the necessity of unusual IPA symbols. The following symbols will be used (other
phonemes will be represented by their IPA symbol):
2.6. ORTHOGRAPHY 47

Phoneme Letter
/j/ y
/ì/ j
/S/ sh
/Z/ zh
/i/ ih

/1/ 1h

/u/ uh

These symbols have been chosen as being in common use, either in the various
orthographies devised for Awa Pit (in the case of the vowel symbols), or also
more generally. This is not, of course, a proposal for a practical alphabet, which
should be decided by speakers themselves.
Chapter 3
Basic clause structure
3.1 Introduction
This chapter begins the description of the syntax of Awa Pit, and provides an
overview and the theoretical preliminaries on which the remainder of the descrip-
tion is based.
With the exception of a few interjections, described in section 4.12, every
utterance in Awa Pit consists of a series of one or more clauses. While there are
a variety of dierent clause types, all have the same basic structure, although in
dierent clauses dierent elements may be present or absent. This basic clause
structure is described in section 3.2.
Following this there is a discussion of the various clause types (section 3.3).
There are two major overlapping divisions here: one divides clauses into main
clauses and subordinate clauses, while the other separates the nite clauses from
the non-nite clauses. While most main clauses are nite and most subordinate
clauses are non-nite, the divisions are not quite that simple, and there are small
classes of non-nite main and nite subordinate clauses.
One important theoretical distinction made in elements associated with
the predicate is the distinction between complements, those elements which are
(semantically) required by a predicate, and adjuncts, those elements which may
be added to a clause but are not required by the predicate. This distinction
is important in a range of circumstances in Awa Pit, much more so than the
distinction between core and oblique arguments, and is taken up in section 3.4.
Also important in the analysis of Awa Pit is the concept of syntactic
functions and their language-internal codication as grammatical relations, such
as Subject, Object and Second Object. These are introduced and justied in
section 3.5.
Finally the dierent classes of predicate are introduced and discussed in
section 3.6. Predicates are, of course, vital in the formation of clauses in Awa
Pit. Every clause must contain an overt predicate (either a verb or Copula Com-
plement), while all other elements may be absent, either ellipsed or unnecessary.
All the elements of the syntax of Awa Pit form a tightly integrated network,
and each part relies on the other parts. A discussion of clause types, for example,
50 CHAPTER 3. BASIC CLAUSE STRUCTURE

Subject Temporal Locational Circumstantial Non-Subject Manner Verb

Adjuncts Adjuncts Adjuncts Complements Adverbials
Table 3.1: The order of basic clause constituents

relies on knowledge of predicate types; but to explain predicate types, clause types
must be invoked. Thus throughout the following analysis of the syntax of Awa
Pit, and in this chapter especially, there is great deal of necessary reliance on
concepts which are not introduced until later. This has been kept to a minimum,
as far as possible, and many cross-references have been given to sections where the
concepts are exemplied and justied. Unfortunately, the analysis of a language's
syntactic system must start somewhere, and must be given a linear structure,
which does not always reect the network of the syntax itself.

3.2 Constituent order

Awa Pit is an AOV/SV language, with the basic order of constituents in all
clause types shown in Table 3.1. This constituent order is found quite commonly
in South America, especially in the region near where Awa Pit is spoken, although
in Awa Pit the order appears much more xed than in many languages. Unlike
languages such as Epena Pedee (Harms 1994:1012) and Retuarã (Strom 1992:3),
both AOV/SV languages where some peripheral elements follow the verb, the
normal position for adjuncts in Awa Pit is between the Subject and complements,
with no elements occurring after the verb at all, unless phonetically set o with
a pause.
The order given in Table 3.1 is not a completely exhaustive list of clause
possibilities. Various minor adverb-like words may appear, for example, between
the manner adverbial slot and the verb itself; the verb may be simple, a com-
pound verb, a mainauxiliary construction, or a Serial Verb construction. Equally
clearly, not every constituent is present in every clause: indeed, unsurprisingly,
no clause has been found which contains all constituents. Adjuncts are, by deni-
tion (see section 3.4), optional; and complements (section 3.4) may undergo both
denite ellipsis (ellipsis of a known participant) as in example (19), and indenite
ellipsis (ellipsis of an unknown or unimportant participant) as in example (20).1
(19) kwizha pay-t kway-ta-w.
dog buy-sv drop-past-locut:subj
[kwizha] corral=mal tu
[dog] yard=loc be:in:place.(impfpart)
`I bought a dog. [The dog] lived in the yard.'
For glosses, see the List of Abbreviations. Perhaps the major non-self-explanatory gloss
1
is Locut(or): essentially Locutor is rst person in statements, second person in questions (see
chapter 8 for details).
3.2. CONSTITUENT ORDER 51

(20) [X] na-wa pyan-t1-t ma-s

[X] 1sg-acc hit-term-pfpart anter-locut:under
`[X=they/someone] had hit me.'
In addition, clauses used in particular constructions may have certain obligatorily
missing elements. Consequently, the majority of clauses contain only one or two
elements, apart from the predicate.
The complement slots  both the slot for the Subject complement and
the slot for the non-Subject complements (lled with Object, Second Object
and various oblique complements, which appear in that order, and the Copula
Complement)  are usually lled by noun phrases, postpositional phrases or
adjectival phrases, although some verbs allow subordinate clauses in some of
these positions. Complements and their grammatical relations are discussed in
greater detail in section 3.5 below.
The various adjunct positions  temporal, locational and circumstantial
adjuncts and manner adverbials  may be lled by a variety of word and phrase
types, as will be discussed in chapter 13. In addition, some adverbial subordin-
ate clauses may ll these positions, although others necessarily either precede
or follow the matrix clause: same-Subject (section 10.3.1) and dierent-Subject
(section 10.3.2) purposives, and same-Subject absolute (section 10.3.7) subor-
dinate clauses may appear in the circumstantial adjunct slot; the two types of
purposive (10.3.1 and 10.3.2) may also occur after the matrix clause; and the
purposives (10.3.1 and 10.3.2), After (10.3.3), simultaneity (10.3.4), concessive
(10.3.5), counterfactual (10.3.6) and absolute (10.3.7) subordinate clauses can
appear before the matrix clause.
In addition to the above order of clause constituents, there are two other
positions which are, in a sense, outside the clause. These are the initial position,
which may or may not be set o from the remainder of the clause by a pause, and
the nal post-clausal position, which is obligatorily set o from the clause by a
pause, suggesting under Ziv's (1994) analysis that it is an afterthought position
rather than a right dislocation position. Any of the adjuncts or complements
may occur in the initial position, with the exception of the Copula Complements
and the manner adverbials, which are tightly linked to the verb. In addition,
an `external' topic may appear in the initial position (see section 14.2). The
nal position may be lled with any adjunct or complement; and one part of
two nps or pps in apposition may occur in the nal position (see section 5.5).
Any complement which occurs in nal position is necessarily accompanied by its
postposition, if it would have one appearing in the usual position.
It appears that at most one element may occur in the initial or nal posi-
tion in any clause. The discourse function of these positions is unclear, although
the initial position appears linked to some form of topicalization; further text
studies are required to establish the functions of these positions. It would be
interesting to then compare the functions of these positions with the functions
of clause-external elements in other languages, such as the clause-external initial
topic and nal antitopic in French discussed by Lambrecht (1984).
It also appears that the initial and nal positions, especially the latter,
52 CHAPTER 3. BASIC CLAUSE STRUCTURE

are used for `heavy', that is lengthy or complex, constituents. In particular,

while sentence-like Object complements to verbs such as speech verbs (see sec-
tion 10.2.1) and clausal Object complements to indirect question words (see sec-
tion 10.2.2) may occur in the standard Object position, they are more commonly
found clause-nally, separated from the matrix clause by a pause.

3.3 Clause types

The basic clause structure was discussed in the previous section. This struc-
ture is the foundation of all clauses in Awa Pit, although certain elements are
missing from certain clauses  for example, some subordinate clauses are obligat-
orily without Subjects; however the rest of the clause follows the usual structure
expressed in section 3.2, complete with the usual case marking, modication,
adverbs, and so on, although clause-external initial and nal elements may not
be permitted. There are a variety of clause types, in the sense meant here: nite
main clauses; non-nite main clauses; conjoined main clauses; nite subordinate
clauses; and non-nite subordinate clauses, including adjectivizations.

3.3.1 Main and subordinate clauses

The major division of clauses is into main clauses and subordinate clauses. Main
clauses are, unsurprisingly, those clauses which can stand alone and express a full
proposition; subordinate clauses, on the other hand, necessarily form only a part
of a proposition, and are always associated with or form part of a main clause,
on which they are dependent.
There are two types of main clause. The most common is the normal,
nite (see below) main clause, with a single verb stem (which may be complex
 see chapter 6) inected with some nite inection. The non-nite main clause
construction is much more interesting. It contains a predicate which, if it is
a verb, is not inected in the usual way, but rather in a fashion more usually
appropriate to a subordinate clause; and the predicate may be marked with the
Topic marker. The existence of this clause type will be justied in the following
section. Unlike nite clauses, which obligatorily contain a verb stem, non-nite
main clauses may be verbless.
There is also a Conjoined Clause structure. This construction allows clause
chaining of a series of clauses with the same Subject. One of the verb stems
(usually the nal one) is inected in the same way as any other verb stem in a
main clause; all other verb stems are given a special inection, indicating that
they are in a conjoined structure and equivalent to giving them all the same
inectional ending placed on the fully inected verb. See section 11.7 for a full
discussion of this construction; here it is sucient to note that the fully inected
verb stem has precisely the same inectional possibilities as any other verb stem
in a main clause, and may be nite or non-nite.
Just like the non-nite main clause, the fully nite subordinate clause is
something of an anomaly, and used in few constructions. A restricted number
3.3. CLAUSE TYPES 53

of nite subordinate clauses is common in areal terms: while some Amazonian

languages, such as Macushi (Abbott 1991:6768) have a few nite subordinate
clauses, many have none at all (Derbyshire & Pullum 1986b:19); and Imbabura
Quechua also lacks these clauses (Cole 1985:3334). As will be discussed in
section 10.2.1, these clauses are used in Awa Pit to show direct speech, and also
to encode the clausal complements of some verbs of cognition and all verbs of
perception. While a nite subordinate clause is, formally, identical to a (nite)
main clause, it is embedded within another clause, often even placed between
other clausal elements, and in some cases depends on that main clause for its
time reference (see section 10.2.1 for details).
Non-nite subordinate clauses are much more common. They are used
as clausal complements to a variety of verbs (see section 10.2), either with or
without a complementizer. Non-nite subordinate clauses are also used as clausal
complements to postpositions (section 10.2.3.2); in a wide variety of adverbial
clauses (section 10.3); they may form clausal adjectives (section 10.4); and are
used in a nominalization process (see section 10.5).

3.3.2 Finite and non-nite clauses

In the division of clause types above, much mention was made of nite and
non-nite clauses, however the distinction was not made clear, and no dening
features were given. This section deals with this distinction, and a few other
related issues.
Clearly, whether one wishes to consider niteness as a binary category, as
it is traditionally considered and as is done here, or as a scalar category as is done
by, for example, Givón (1990:852), in general terms most main clauses should be
(more) nite, while most subordinate clauses should be (more) non-nite. In
addition to this high-level distinction there is a low-level distinction: niteness
is associated with tense and person, while non-niteness is not, both in terms of
more traditional denitions and more scalar denitions.
The vast majority of subordinate clauses in Awa Pit have neither tense
nor person marking. The only exceptions are a small class of subordinate clauses
which are used for direct speech and for clausal complements of some verbs of
cognition and all verbs of perception; these clausal complements are all identical
to main clauses.2 As these clausal complements are identical to main clauses in
Awa Pit, they clearly cannot be distinguished in terms of niteness; rather these
subordinate clauses are considered fully nite.
With the exception of these nite complement clauses, all other subordin-
ate clauses lack tense and person suxes. It could then be suggested that for
Awa Pit the nite/non-nite distinction is shown by the presence versus absence
of tense and person marking.
2Cross-linguistically these clausal complements are precisely those which are most commonly
similar to main clauses (Noonan 1985), being the least bound variety of complement (Givón
1980).
54 CHAPTER 3. BASIC CLAUSE STRUCTURE

This distinction is not fully adequate, however. Consider the following

clauses:
(21) na=na kwa-t1-ta-w
1sg.(nom)=top eat-term-past-locut:subj
`I had eaten.'
(22) na=na kwa-t1-n1-s
1sg.(nom)=top eat-term-fut-locut
`I will have eaten.'
(23) na=na kwa-t1
1sg.(nom)=top eat-term
`I have eaten.'
The rst two clauses are nite by the above denition, having tense and person
marking suxes. The third clause has no (overt) tense marking, and no person
marking, suggesting that it is non-nite. However this third clause is directly
parallel to the preceding two, and can only be used as a main clause, never as
a subordinate clause, suggesting that it is nite. In fact in this case, the lack
of tense marking indicates Present tense.3 The lack of person marking is simply
a feature of the Terminative aspect sux  it can never be followed directly
by person marking. Thus despite the absence of (overt) tense marking and the
absence of person marking in the third example above, it is considered nite.
The Terminative and Completive aspect markers (see section 9.3) are
somewhat unusual, in that unlike most other Awa Pit verb morphology they may
be used in either nite or non-nite clauses.4 In nite clauses they are followed
by tense and person (except for the Terminative in the Present); in non-nite
clauses, they are followed by a subordinating inection (section 7.2.7) or one of
the other non-nite inections (section 7.2.8). The Imperfective is much more
complex, and is discussed below.
In the above case of the Terminative, of course, it is possible to consider
that the third clause does contain tense marking, it is simply that it is marked by
an absence, rather than a presence, which is quite usual with the Present tense
in Awa Pit. More complex cases arise with mood marking and evaluating the
niteness of clauses containing mood marking.
Consider the directives (see section 9.4.5), such as the various imperative
forms. Clearly we would wish to classify an Imperative such as kwa-t `eat!' as
nite, despite the lack of tense or person marking suxes. The notions usually
expressed through tense and person marking are, of course, inherent in directives:
the action is always in the future, and either second person (for imperatives)
or rst person (for hortatives). But subordinate, non-nite clauses often have
`inherent' tense or person, though it is usually acquired from the construction in
Through the contrast of this form with either the Past or the Future inections; see sec-
3
tion 7.2.1.
4 And consequently aspect has not been used in dening the concept of niteness.
3.3. CLAUSE TYPES 55

which the clause appears. Hence it is best not to consider `inherent or overt' tense
and person marking as the criterion for nite clauses. Nevertheless, directives,
which never appear in subordinate clauses, are considered nite.
Of the other mood inections (section 7.2.3), only the Necessitive can
combine with tense marking, and even then there is only a two-way contrast
between Past (marked) and Present (unmarked). The Necessitive is always ac-
companied by person marking, as is the Negative Potential satshi. These mark-
ers, with person, occur in nite clauses. A form identical to the Necessitive, with
identical allomorphy, does occur in subordinate clauses; however there (except
in the few nite subordinate clauses) it is never associated with tense or person.
This formally identical marker, in non-nite clauses, is historically related to the
Necessitive, but synchronically distinct in meaning and distribution.5
The two remaining mood markers, the Obligative and the Potential, can
never combine with overt tense marking, and they are seldom associated with per-
son marking. However this appears to be semantically determined rather than be-
ing a grammatical rule. Both inections are most commonly used with no Subject
(and are then unmarked for person), being `universal' in scope: `[one/someone]
must cook', `[one/someone] can see the school from here'. They are accompanied
by person marking when a particular Subject is present:
(24) t1lawa=na a-tpa-y
tomorrow=top come-oblig-nonlocut
`You must come tomorrow.'
(25) nyampi=kasa pishkatu ki-sina-y
hook=with sh(1) sh(2)-pot-nonlocut
`You (the addressee) can sh with a hook (since the river is up).'
Thus these other two mood markers occur in nite clauses, although they are
seldom used with person marking because of their function, and they cannot be
used with tense inections.
The Obligative tpa/tawa is distinguished from a formally identical subor-
dinating inection, just like the Necessitive. As a subordinating marker, tpa/tawa
means `after', quite distinct from its meaning as an Obligative; and while the sub-
ordinate verb inected with tpa/tawa has a Subject, this subordinate verb cannot
be inected for person, in contrast to its main clause behaviour.
5This is perhaps the most appropriate point to bring up a methodological issue in this thesis,
which applies largely to inections, although it is also relevant to other grammaticalmorphemes.
The dening criterion for considering two identical forms as separate morphemes has been
taken to be, apart from semantics, their possibilities of combination with dierent inections
and hence their use in dierent constructions. Thus, for example, the Necessitive morpheme
occurs in main clauses, where it combines with tense and person marking, and indicates that
someone had a personal need to do something or gives a predictive future reading, while the
dierent-Subject purposive morpheme occurs in subordinate clauses and cannot combine with
tense or person marking; hence these are considered to be separate morphemes, despite their
formal identity as npa/napa. Dierent theoretical approaches would wish to treat these as
separate morphemes or as the same morpheme in distinct constructions. These (and others)
have been treated as distinct here, on the criterion given above, although where morphemes
are clearly diachronically related, this is mentioned.
56 CHAPTER 3. BASIC CLAUSE STRUCTURE

The mood inections are thus all used in nite main clauses, although
many are often not accompanied by tense and person inection. While two of
the mood markers are identical in form to two subordinating markers, semantic
and morphological criteria show that they are distinct: the mood inections are
used in nite main clauses, the subordinating inections in non-nite subordinate
clauses.
In the preceding discussion, almost all the clausal inectional possibilities
have been divided into nite or non-nite. However three groups of inections
remain to be examined: the Serial Verb inection (section 7.2.5), the Clause
Conjoining morphology (section 11.7) and the other non-nite inections (sec-
tion 7.2.8).
The Serial Verb inection is outside the nite/non-nite distinction. It is
used to combine two verb stems into a single predicate, then this predicate itself
is either nite or non-nite, depending on the inections on the fully inected
verb stem.
The Clause Conjoiner is quite similar in behaviour. It combines two
clauses, the eect being that both share the inectional meaning of the fully
inected verb. The two clauses then share their niteness, both nite or both
non-nite. The two clauses are, however, always main clauses  they cannot be
subordinate.
This then leaves the other non-nite inections to be discussed, and the
verbless copula constructions. The Innitive inection is clearly non-nite, being
used only in subordinate clauses and never being associated with tense or person;
the other cases are more complex.
Awa Pit forms copula constructions in two ways: with or without the
copula verb i (see section 3.6.2); thus the structures for copula constructions
using adjective Copula Complements, for example, are:6
Noun Adjective i-. . .
Noun Adjective
If the copula i appears in a main clause, tense and person marking are obligatory;
if it is absent, tense and person cannot be expressed, and are semantically free,
being xed by context. It could be hypothesized that these two constructions
are identical, and it is simply the case that the copula is optional. The two
constructions do, however, have dierent distributions, even in main clauses.
The form with tense and person marking can, unsurprisingly, only be used in
cases where a nite clause would be expected, and it is clearly nite. However
there are also main copula clauses which do not have a copula verb.
Two of the other non-nite inections, the two adjectivizers, pattern in
precisely the same fashion. They can occur accompanied by the copula verb, or
without it:
Both of these constructions will be referred to as `copula clauses', despite the absence of
6
copula verbs from the latter construction, because of the parallelism in their semantics. `Copula
clause' is thus a cover term for equative clauses, identity clauses, ascriptive clauses, and so on,
all of these functions being carried out by clauses with or without i.
3.3. CLAUSE TYPES 57

(26) ampat1nkwa amta shaa-m (i)

Vieja at:night walk-adjzr (be.(nonlocut))
`La Vieja (Old Woman) walks at night (is at-night-walking).'
The distribution of these clauses is the same as that of the simple copula con-
struction with or without a copula verb. The adjectivizers essentially form an
adjective from a clause, and this can then be used as the predicate in a copula
construction, in the same fashion as any other adjective (or, indeed, noun).
The extended Perfective Participle (see section 7.2.8.3) is quite similar. In
the Resultative and Past Anterior constructions, the extended Perfective Parti-
ciple is used in what is almost a copula construction, although there are greater
restrictions placed on it so that, for example, the copula is necessarily non-plural.7
Despite these dierences from the copula construction, these two constructions
can appear with or without a copula verb.
(27) Demetrio a-ka=na, piya pak-ma-t
Demetrio come-when=top corn harvest-comp-pfpart
(a-t1-zi)
(be-past-nonlocut)
`When Demetrio arrived, the corn was (in a state of having been)
harvested.'
So far there is, perhaps, little need to establish a separate clause type
to deal with these cases. It could still be maintained that there is simply an
optional copula, with dierent distributions on the construction with and without
an (overt) copula. However the Imperfective Participle construction is distinct.
The Imperfective Participle (see section 7.2.8.2) can be used in a construc-
tion apparently parallel to that of the extended Perfective Participle, without a
copula verb:
(28) na=na ku-mtu
1sg.(nom)=top eat-impfpart
`I am/was/will be eating.'
This Imperfective Participle can also be used with other stative verbs acting as
an auxiliary (see section 11.2.1 for details):
(29) ku-mtu uz-is
eat-impfpart be:sitting-locut
`I am (sitting) eating.'
However it can never be used together with the copula verb i, unlike other non-
nite forms such as the extended Perfective Participle:
(30) *ku-mtu i-s
eat-impfpart be-locut
7 See sections 11.2.2.2 and 11.2.2.1 for details.
58 CHAPTER 3. BASIC CLAUSE STRUCTURE

Copula Active verb pana uz

Locutor Present i-s -(m)tu-s pana-s uz-is
Non-locutor Present i -(m)tu-y pana-y uz-i
Locutor Subject Past a-ta-w -(m)tu-ata-w pana-ta-w uz-ata-w
Locutor Undergoer Past a-t1-s -(m)tu-at1-s pana-t1-s uz-at1-s
Non-locutor Past a-t1-zi -(m)tu-at1-zi pana-t1-zi uz-at1-zi
Locutor Future a-n1-s -(m)tu-an1-s pana-n1-s uz-an1-s
Non-locutor Future a-n1-zi -(m)tu-an1-zi pana-n1-zi uz-an1-zi
Table 3.2: Tense and person forms of the copula, Imperfective active verbs, and
the stative verbs pana `be standing' and uz `be sitting'.

Instead, corresponding to the Imperfective Participle without a copula are the

fully nite Imperfective aspect forms:
(31) ku-mtu-s
eat-impf-locut
`I am eating.'
It is very clear that, historically the forms of the nite Imperfective have
developed from the Imperfective Participle followed by the copula verb. This can
be seen rstly from the very lack of this possibility: the Imperfective Participle
can be used with any of the other stative verbs as an auxiliary, including the
pseudo-copula ka, almost indistinguishable from the copula in meaning. This
relationship between the (non-occurring) Participle plus copula and the nite
Imperfective is also suggested by the use of the constructions: the fully nite
Imperfective forms correspond in range precisely to the copula construction with
an overt copula, and also to the Resultative and Past Anterior constructions with
a copula; while the Imperfective Participle forms have the range of the copula,
Resultative and Past Anterior constructions without a copula. But the corres-
pondence can be seen even more clearly in the forms of the nite Imperfective
of both active and stative verbs given in Table 3.2. The historical origin of the
nite Imperfective forms explains the epenthetic a or i which appears in many
of these forms. While originally the stem of the copula, it is now simply part of
the allomorph of the person or tense marking.
Imperfective forms developing from a participle plus a copula are not un-
known in the languages of the world. Imperfective constructions often appear to
have developed from earlier progressive constructions (Bybee & Pagliuca 1985),
and Bybee, Perkins & Pagliuca (1994:130131) list six cases of progressives con-
sisting of copula auxiliaries plus a non-nite verb form.
Synchronically the form (m)tu cannot be treated as only an Imperfective
Participle occurring in some sort of arrangement with a fused copula. It occurs,
for example, before the subordinating sux ka, where the (other) two inectional
aspect suxes are also possible. Thus there are two separate imperfective forms,
historically only one and having identical allomorphy. One is an aspect inection,
3.3. CLAUSE TYPES 59

and patterns with the other two inectional aspect markers; the other forms a
(non-nite) Imperfective Participle.
However this then creates a problem for clause types. A main clause, under
certain conditions, can consist simply of a clause with an Imperfective Participle,
a non-nite form, as in sentence (28) above. This main clause cannot be analyzed
as a copula construction with an ellipsed copula, as the corresponding clause
with an overt copula, sentence (30), does not exist, as noted above. It is necessary
to introduce a separate clause type, a non-nite main clause, which of course has
restrictions on when it may appear, as all clause types do.
Once a special non-nite main clause with particular restrictions on its
occurrence has been established, it seems counter-intuitive not to consider that
other main clauses, apparently non-nite and with exactly the same set of re-
strictions on them, also belong to this clause type. Rather than considering that
there are only nite copula clauses, nite clauses involving adjectivizers, and nite
clauses involving the extended Perfective Participle, and particular restrictions
on the ellipsis of the full copula or auxiliary copula, it seems far more logical to
analyze these forms into two clause types: a fully nite clause, where an explicit
copula or auxiliary copula is present, and which corresponds to a nite Imper-
fective clause; and a non-nite clause type, where there is no copula verb, which
corresponds to a non-nite Imperfective Participle clause.
The exact range of uses of non-nite main clauses is unclear at this stage,
and requires examination of contextualized uses. It is possible that these clauses,
while not syntactically dependent on others, are discourse dependent; this would
then be similar to the use of non-nite main clauses in Witoto to show the `setting'
(Petersen de Piñeros 1992:111).
It would also be instructive to compare the use and frequency of nite
and non-nite main clauses in Awa Pit with those in Damana, spoken in the
north of Colombia. While Trillos Amaya's (1989) analysis of Damana discusses
aspectual versus mood forms, it is clear that the aspectual forms are, in fact,
non-nite  they have no person marking, and may even be followed by suxes
identical to noun suxes, and they may also precede auxiliaries. In addition,
however, similarly to the Awa Pit Imperfective, the aspectual forms can be
followed by a sux ka and are then marked for person. It is possible that this
sux ka is, historically at least, a copula form.8

3.3.3 Summary of clause types

From the above discussion, four major clause types can be established, as men-
tioned earlier. The most common main clauses are the nite main clauses, in-
cluding nite conjoined clauses. There are also non-nite main clauses, including
non-nite conjoined clauses. While the non-nite main clauses are syntactic-
ally independent units, they may be highly dependent on preceding discourse.
Parallel to the two main clause types are the two subordinate clause types: the
8Compare the pseudo-copula ka in Awa Pit, and the widespread occurrence of an auxili-
ary/copula similar to ka in South America (David Payne 1990).
60 CHAPTER 3. BASIC CLAUSE STRUCTURE

highly restricted nite subordinate clause, and the much more common non-nite
subordinate clause.

3.4 Complements and adjuncts

With the exception of predicates and modiers of predicates (for example the
Manner adverbials), all clausal elements in Awa Pit fall into two classes: comple-
ments and adjuncts. The distinction between these two types was dened in the
following fashion by Andrews:
The distribution of complements is governed by potentially idio-
syncratic specications on verbs (or other predicators). Adjuncts
on the other hand appear whenever they would be semantically ap-
propriate. . . . it is reasonable to think of the complement/adjunct
distinction as overlapping the core/oblique distinction, with the
class of complements comprising all core nps and some obliques.
Adjuncts, on the other hand, always seem to be oblique.
(Andrews 1985:89)
Complements are, then, those arguments which are called for by the predicate;
adjuncts are those items which can be added to any sentence (if they would make
sense), but are not tied to the subcategorization frame of the predicate.
In Awa Pit, the vast majority of complements are core grammatical func-
tions (see section 3.5)  nearly all verbs are only subcategorized for some subset
of Subject, Object, Second Object and Copula Complement. However there are
a few verbs which require some oblique argument (see section 4.5.5.2). Some of
these verbs require a particular oblique argument, for example the verb mazh-
`change/trade' requires a Subject (the person doing the trading), an Object (the
thing traded) and an oblique marked with the postposition =kasa `with' (the
thing the Object was traded for).
(32) na=na [ maza atal=na ] [ paas
1sg.(nom)=top [ one chicken=top ] [ two
A O with.complement
pollo pashpa=kasa ] mazh-ta-w
chicken dim=with ] change-past-locut:subj
V
`I traded one chicken for two little chicks.'
Other verbs in this group require an oblique argument, but are not particular
about which oblique, as long as it is semantically coherent. Thus the verb win-
`put in a place' requires a Subject (the person doing the putting), an Object
(the thing put) and some oblique locational phrase (the place where the thing is
put). This locational phrase may be one of a number of locational postpositional
phrases; for example:
3.4. COMPLEMENTS AND ADJUNCTS 61

(33) na ; t1m=ta win-ta-w

1sg.(nom) ; basket=in put-past-locut:subj
A O location.complement V
`I put [the plantains] into the basket.'
(34) na=na ; su=wa win-ta-w
1sg.(nom)=top ; there=in(approx) put-past-locut:subj
A O location.complement V
`I put [the plate] over there.'
In morphological terms, these oblique complements are identical to oblique
adjuncts. Thus the kasa-complement in sentence (32) above appears morphologic-
ally identical to the kasa-adjunct in sentence (35) below; and the ta-complement
in sentence (33) above has the same form as the ta-adjunct in sentence (36) below:
(35) awa=na t1=kasa piya waa-mtu
person=top stick=with corn sow-impfpart
A with.adjunct O V
`The Awa sow corn with a stick.'
(36) ; esta=ta ashaNpa t1ta-n1-ma-tu
; festival=in woman search-prosp-comp-hort.1sg
A location.adjunct O V
`I'm going o to nd a wife at the festival.'
As morphologically there is no distinction between the complements and adjuncts,
and semantically they are almost identical, there seems to be no reason to claim
that there are dierent syntactic functions involved in the two pairs of sentences.
However there are syntactic dierences between the two usages.
There is a dierence in position between the complements and the ad-
juncts. As was seen in section 3.2, in the most basic clause structure the adjuncts
occur before the Object, while oblique complements occur after the Object; and
this can be seen in the sentences above, with the adjuncts in sentences (35)
and (36) occurring before the Objects, while in sentence (32) the oblique com-
plement occurs after the Object. In either case the postpositional phrase could
be moved to initial position in the clause; but the adjunct pps could not be
placed after the Objects, and the complement pps could not be placed between
the Subject and the Object.
A more subtle, but still syntactic, dierence also exists between morpho-
logically identical complements and adjuncts. The Serial Verb construction (see
section 11.5) combines two verbs into a single entity. The verbs in question, how-
ever, must have identical complement structures. Thus two intransitive verbs
can be combined, or two transitive verbs, but not an intransitive and a transitive
verb. While the verbs must have identical complements, one or the other verb
may have additional adjuncts, which need not be shared by the other verb; for
example, in the following sentence, the adjunct iyaNpa=kasa `with a shotgun'
belongs to the rst verb, pyaNta- `kill', but not to the second, ku- `eat':
62 CHAPTER 3. BASIC CLAUSE STRUCTURE

(37) Juan=na iyaNpa=kasa kukum pyaNta-t ku-mtu

Juan=top shotgun=with possum kill-sv eat-impfpart
A1/A2 adjunct O1/O2 V1 V2
`Juan killed the possum with a shotgun and ate it.'
The kasa-adjunct in the above sentence is outside the complement structure,
and can be used applying to only one of the verbs in a Serial Verb construction.
However a kasa-complement is not outside the complement structure, and con-
sequently a Serial Verb cannot be formed by, for example, combining pun- `ll'
 which takes a Subject (the person lling), an Object (the thing lled) and a
kasa-complement (the substance used to ll)  with kway- `drop'  which takes
a Subject and an Object  even if the two Subjects and the two Objects are the
same, as they would be in I lled the bowl with soup and dropped it. While
the Subjects and Objects of the two verbs here are identical, the rst verb also
has an oblique complement, and as this is not mirrored in the second verb, they
cannot be joined in the Serial Verb construction.
The complement/adjunct distinction thus plays an important role in the
syntax of Awa Pit, much more so than the distinction between core and oblique
constituents (see section 3.5.5). Verbs specify complements, whereas adjuncts
may appear wherever they are semantically appropriate; adjuncts precede com-
plements (except the Subject) in the basic clause structure; and the operation of
syntactic processes may depend on the number and type of complements, while
the presence or absence of adjuncts has no eect on these processes.

3.5 Syntactic functions and grammatical relations

This section establishes the basic grammatical relations and other syntactic func-
tions of Awa Pit. While the basic syntactic functions of a, s and o are universal,
and can be easily established, they are not necessarily the most useful tools for
language description. It is often easier to work in terms of grammatical relations
such as Subject and Object, where a grammatical relation is:
a [syntactic] function that is generally signicant for the workings
of the grammatical principles of [the] language, and which it would
therefore be reasonable (although not necessarily correct) to posit
as a primitive element in the sentence structures of the language.
(Andrews 1985:66)
In particular, the grammatical relations of Subject, Object, Second Object and
Copula Complement will be useful in describing many constructions of Awa Pit,
once they have been introduced and justied.

3.5.1 A, S and O syntactic functions

The basic syntactic functions of a, s and o used in this work are dened according
to Andrews (1985). If an np argument of a verb with two or more np arguments
3.5. SYNTACTIC FUNCTIONS AND GRAMMATICAL RELATIONS 63

is acting morphologically and syntactically identically to the Agent argument of

a Primary Transitive Verb (such as `kill'), it is considered to have the syntactic
function a. If it is acting morphologically and syntactically identically to the
Patient of such a verb, it is said to have the syntactic function o. If an np
in an intransitive sentence is being accorded the morphological and syntactic
treatment normally given to the single argument of a one-argument predicate, it
is considered to have the syntactic function s.
The somewhat more restrictive denitions of a, s and o given by Andrews
(1985) are being adopted, rather than those of Dixon (1994), to avoid some serious
theoretical concerns. Dixon considers that the single argument of any intransitive
verb is an s, and the two arguments of any transitive verb are a and o. However
there are languages which have a very small number of verbs requiring their
arguments to take unusual case marking. For example, in Modern Icelandic
there are a few intransitive verbs which have their single argument case-marked
in the same way as transitive objects are normally marked (Andrews 1985:102).
In Dixon's system, we would then be required to consider Modern Icelandic to
be a split-s language, as the single argument of an intransitive verb (always an
s following Dixon's denition) can be marked in more than one way; in fact, it
would be a four-way split-s language, as there are also a few verbs which require
their single argument to be in dative or genitive case. Clearly, however, we wish to
consider Icelandic a nominativeaccusative language, with a very few verbs with
unusual case-marking requirements, quite dierent from a split-s language such
as Dakota, where it is not possible to make a generalization over the entire system
of intransitive verbs, but rather there are two dierent systems operating. If we
establish denitions whereby a language such as Icelandic is a split-s language,
it then becomes necessary to consider further splits such as the `split-o' system
of German, where a few transitive verbs (such as helpen `help') have their non-a
argument marked in the dative rather than the accusative, and so on.
Equally, there are problems with considering a and o to be dened as the
two arguments of any transitive verb  it is unclear what exactly is a transitive
verb. Dixon (1994:6) simply says that clauses involving a verb and two or more
core nps is a transitive clause. If a core np is dened as a, s or o, we have a
circular denition. If a core np is not dened this way, it is unclear whether, across
languages, there is an appropriate denition of core nps. Following Andrews's
(1985) denition, it will be suggested that Copula Complements in Awa Pit are
core arguments, and in this case copula clauses contain two core arguments 
but we would not wish to consider these clauses as transitive in many languages,
including Awa Pit. To avoid these problems, Andrews's (1985) more restrictive
denitions of a, s and o will be used here.

3.5.2 Subject
Awa Pit is a nominativeaccusative language, both morphologically and syn-
tactically: as will be shown in this section, nps in a function and those in s
function act in the same way, both morphologically and syntactically, and dif-
64 CHAPTER 3. BASIC CLAUSE STRUCTURE

fer from nps in o function.9 It is thus possible to establish the grammatical

relation of Subject, which covers all nps in a and s function and a number of
others, since a and s have been dened restrictively. With the exception of
zero-argument verbs, every verb in Awa Pit is subcategorized for a Subject.
The morphological and syntactic tests given below are applicable to all nps
in a function, and all nps in s function which are arguments of active verbs.10
Unfortunately, while the morphological tests used to show the identity of s and
a all work for active verbs (though with minor dierences in the case of number
marking), the majority of syntactic tests are not applicable for clauses involving
stative verbs  the stative verbs simply cannot be used in these constructions.
The evidence for considering that arguments of stative verbs are Subjects
comes from three sources. The majority of stative verbs have, in fact, only one
argument, by denition s, and thus this argument is a Subject. This is not,
however, particularly convincing. Morphologically and syntactically, one of the
arguments of a stative verb acts in precisely the same fashion as an a or s of an
active verb  in terms of constituent order, case marking, controlling number
marking (with dierences) and person marking. There is one syntactic test for
Subject which can be used with stative verbs, the same-Subject purposive.
The grammatical relation of Subject covers a variety of semantic roles,
such as volitional agent (ayna- `cook'), non-volitional agent (kwata- `vomit'),
patient (kwaa- `drown', iiN- `ripen'), experiencer (izh- `see') and possessor (m1j
`have').
3.5.2.1 Constituent order
The rst piece of evidence for a Subject grammatical relation comes from con-
stituent order. In the most basic clause ordering (discussed in section 3.2), only
one constituent occurs before any adjuncts. This constituent is a or s, when one
of these is present in a clause, and hence any constituent which occurs within a
clause before adjuncts is the Subject.
3.5.2.2 Case marking
The system of case marking used in Awa Pit is relatively transparent. It is `ana-
lytic' rather than `synthetic' (Blake 1994:10), as the relationship between depend-
ent arguments and their verbal heads is expressed through cliticized postpositions
rather than case inections; however the same information is being oered as is
oered through case inection, and the use of particles or adpositions to express
the syntactic function of arguments is well known (Dixon 1994:4142).
In Awa Pit, the case-marking system is nominativeaccusative, with a
and s being marked in the same fashion, distinctly from o. nps which are in a
In the following discussion,  nps in o function will often be referred to. In fact, in the
9
analysis given here, many of these  nps are actually pps, as the Accusative marker ta is
considered to be a postposition. However the somewhat sloppy wording is used to avoid a
constant use of  nps or pps.
10See section 4.5 for the important active/stative distinction in Awa Pit verbs.
3.5. SYNTACTIC FUNCTIONS AND GRAMMATICAL RELATIONS 65

or s function are unmarked, while nps in o function are marked as Accusative

if they refer to referential humans, although they remain unmarked if referring
to entities which are not referential and human.11 The personal pronouns have
special Accusative suxes or clitics, while all other nps are made Accusative by
the addition of a clitic postposition ta (see section 5.4.2). The following examples
contain a variety of nps in a, s and o functions, and clearly show the unmarked
nature of s and a, and the division of o between referential human (Accusative)
and other (unmarked).12
(38) Demetrio=na t1lawa a-mtu-y
Demetrio=top tomorrow come-impf-nonlocut
S V
`Demetrio is coming tomorrow.'
(39) Demetrio na-wa pyan-t1-t1-s
Demetrio 1sg-acc hit-term-past-locut:under
A O (ref. human) V
`Demetrio hit me.'
(40) Demetrio kuzhu pay-t kway-zi
Demetrio pig buy-sv drop-nonlocut
A O (not ref. human) V
`Demetrio bought a pig.'
(41) na=na Demetrio=ta pyan-tu
1sg.(nom)=top Demetrio=acc hit-impfpart
A O (ref. human) V
`I hit Demetrio.'
(42) na=na nayN-ma-t1-s
1sg.(nom)=top fall-comp-past-locut:under
S V
`I fell.'
(43) nu=na Juan=ta pyan-t1-zi
2sg.(nom)=top Juan=acc hit-past-nonlocut
A O (ref. human) V
`You hit Juan.'
(44) nu=na pala ku-mtu-y
2sg.(nom)=top plantain eat-impf-nonlocut
A O (not ref. human) V
`You are eating plantains.'
11In fact, o can sometimes be marked by other cases, if the sentence is not highly transitive
 see discussion below, in section 3.5.3.1.
12It is perhaps worth noting that the clitic na, used in many of these examples, is not a
case-marking clitic, but rather a topic marker, and may occur on a wide variety of elements,
although it is most commonly found on Subjects (see section 14.2).
66 CHAPTER 3. BASIC CLAUSE STRUCTURE

Many more examples could be given, of course, of dierent types of nominals,

dierent verbs, dierent tenses, aspects and moods, and subordinate rather than
main clauses. However in all cases, a and s are treated in the same way, and o
is treated distinctly, when it refers to a referential human entity. Subjects, then,
are always unmarked nps, rather than case-marked pps.
3.5.2.3 Number marking
As will be discussed in section 7.3, speakers of Awa Pit seldom use any number
marking. However sometimes number marking is used, with special verbal suxes
used to indicate the plurality of one of the verbal arguments. For active verbs
there are two dierent suxes, a and na. If a plural s or a is to be indicated, a
is used; a plural o can be shown by na:
(45) kayl-a-n1-zi
return-pl:subj-fut-nonlocut
`They will come back.'
(46) uspa=na Santos=ta t1t-shi-a-mtu-y
3pl.(nom)=top Santos=acc cut-desid-pl:subj-impf-nonlocut
`They want to stab Santos.'
(47) na=na uspa=tuza t1t
1sg.(nom)=top 3pl=(3pl.)acc cut
kyan-na-ta-w
throw-pl:obj-past-locut:subj
`I stabbed them.'
Thus marking the plurality of an argument of an active verb by using the verb
suxes a and na does not distinguish between s and a, with both being marked
as plural through a, but does dierentiate these from o, which is indicated as
plural through the use of the sux na. The possibility of cross-referencing the
number of a human argument of an active verb with a can thus be used as a test
for Subject.
Number marking of stative verbs is done in a distinct fashion, through
the use of dierent, suppletive, axes (see section 7.3.2), and only occurs in the
Present tense. Only one of the arguments of a stative verb can trigger number
agreement, and for stative intransitives this is the s argument. Thus while number
marking is of a distinct form, it can be used to establish Subjects of stative verbs,
as well as active verbs.
3.5.2.4 Person marking
Awa Pit has a morphological marking system which corresponds in many ways
to a person cross-referencing system in other languages. It is not truly a cross-
referencing system, however, and only partially relies on grammatical relations,
3.5. SYNTACTIC FUNCTIONS AND GRAMMATICAL RELATIONS 67

with semantic roles and the involvement of participants in an action also being
important factors (see chapter 8 for details).
However this system of marking can contribute to identifying the gram-
matical relation of Subject. If a verb of any type, active or stative, is in the
Past tense, and has a nal w, then the s or a argument (if that verb has one)
is necessarily rst person in a statement, second person in a question.13 Thus if
a rst-person referent in a statement or a second-person referent in a question
causes a Past tense verb to appear with the sux w, the np referring to that
referent is in Subject relation in the clause.
3.5.2.5 The Conjoined Clause construction
The above tests have shown that, morphologically, s and a are identied in Awa
Pit, and the language can be established as morphologically accusative. Awa
Pit is also syntactically accusative, having an s/a pivot (to use the terminology
introduced in Dixon (1979)) for those syntactic constructions requiring a pivot.14
Given the morphological accusativity of the language, it is not surprising that
it should also be syntactically accusative  as has often been observed, any
language that is syntactically ergative will also have some ergative characteristics
at the morphological level (Dixon 1994:177).
The rst construction showing the s/a pivot of Awa Pit is the Conjoined
Clause construction, which is approximately equivalent to the conjunction of two
propositions.15 The important point here is that with this construction, which
can only be used with active verbs, the two propositions involved must share the
same Subject  and the two verbs may be both transitive, both intransitive, or
one transitive and one intransitive. Thus this construction treats s and a in the
same fashion, quite distinct from o. This can be seen clearly in the following
sentences:16
(48) Marcos=na a-t kit ; pala kwa-ma-t1
Marcos=top come-sv and ; plantain eat-comp-term
S1 V1 (A2=S1) O2 V2
`Marcos came and [Marcos] ate a plantain.'
(49) Santos=na Laureano=ta pyan kit ii-ma-t1
;

Santos=top Laureano=acc hit and die-comp-term

;

A1 O1 V1 (S2=A1) V2
`Santos hit Laureano and [Santos] died.'
13The reverse is not necessarily true, however: a rst person s in a statement does not always
imply a nal w in the Past tense.
14For the one minor exception, see section 3.5.2.10.
15See section 11.7 for full details of this construction.
16It is important to note, here and in the following few sections, that the use of a zero in these
sentences does not imply that there was an underlying sentence with a full np argument which
was later deleted. The zero is simply indicating that the following verb has a subcategorization
frame expecting an argument of this type, and assigns it as coreferential to another referent in
the sentence.
68 CHAPTER 3. BASIC CLAUSE STRUCTURE

(50) Tel
esforo=na Abelardo=ta pyan kit ;

Telésforo=top Abelardo=acc hit and ;

A1 O1 V1 (S2=A1)
shaa-mtu
walk-impfpart
V2
`Telésforo hit Abelardo and [Telésforo] walked o.'
In example (48), the rst clause contains an intransitive verb, and an s. The
second clause is (obligatorily) without an expressed a, and the construction re-
quires that the non-expressed a of the second verb be coreferential to the s of
the rst clause of the sentence. Similarly, in both the second and third example
sentences, (49) and (50), the second verb in each sentence has no expressed s
argument. There are two potential referents in each case, semantically speaking
 either the a or the o of the previous verb. However in both cases, because of
the syntax of this construction, the non-expressed s must be coreferential with
the a of the previous verb, not the o.
The Conjoined Clause construction thus obligatorily identies s and a,
and dierentiates them from o  it works on an s/a pivot: it is syntactically
accusative, and operates on the basis of the grammatical relation of Subject.
3.5.2.6 The same-Subject purposive
Awa Pit has a same-Subject purposive construction.17 Just as with the Conjoined
Clause construction above, the same-Subject purposive requires that a (non-
expressed) a or s of the purposive clause be coreferential with an a or s in the
matrix clause. Thus corresponding to sentence (51), where the (obligatorily non-
expressed) subordinate a corresponds to the matrix a, there is sentence (52),
where the (obligatorily unexpressed) subordinate a corresponds to the matrix s.
(51) Carmen piya k11-t kway-zi, ; atal pashpa
Carmen corn mill-sv drop-nonlocut ; chicken dim
A1 O1 V1 (A2=A1) O2
kwin-na
give-inf
V2
`Carmen ground corn to give to the baby chickens.'
(52) Demetrio=na t1lawa a-mtu-y,
Demetrio=top tomorrow come-impf-nonlocut
S1 V1
; s1 pyan-na
; rewood cut-inf
(A2=S1) O2 V2
`Demetrio is coming tomorrow to cut rewood.'
17 See section 10.3.1 for details.
3.5. SYNTACTIC FUNCTIONS AND GRAMMATICAL RELATIONS 69

The same-Subject purposive construction thus operates on an s/a pivot, identi-

fying s and a, and distinguishing them from o; the Subjects of the two clauses
must be coreferential.
While the subordinate verb must be active, the matrix verb in a same-
Subject purposive may be stative. This construction is the only one which shows
that the argument of stative verbs which is morphologically similar to s is, syn-
tactically, a Subject. Thus in the following sentence, the s of the rst, stative,
verb is necessarily coreferential with the non-expressed a of the second verb.
(53) na=na akki tansha-s,
1sg.(nom)=top here sit-locut
S1 V1
; Laureano=ta t1tizh-na
; Laureano=acc wait:for-inf
(A2=S1) O2 V2
`I am sitting here, to wait for Laureano.'
The same-Subject purposive can consequently be used to show the grammatical
relation of Subject, operating as it does across both active and stative verbs.
3.5.2.7 Complements of intention
There is a construction used in Awa Pit for complements of intention.18 This
construction involves an active subordinate clause which necessarily does not
have an expressed s or a. The s or a of the subordinate verb is understood to
be coreferential with the a of the matrix clause; for example:
(54) Santos=na [ ; a-n ] kizh-t1-zi
Santos=top [ ; come-inf ] say-past-nonlocut
A1 (S2=A1) V2 V1
`Santos said that he would come.'
(55) Ángel=na [ ; t1lawa shappi k11-na ] kizh-t1
Ángel=top [ ; tomorrow cane:juice mill-inf ] say-term
A1 (A2=A1) O2 V2 V1
`Ángel said that he would mill cane juice tomorrow.'
While the matrix clause necessarily has an a in this construction, this a is always
coreferential with the Subject of the subordinate clause; and s and a in the
subordinate clause are thus identied.
3.5.2.8 Agentive adjectivizations
As will be discussed in section 10.4, there is a construction in Awa Pit  the agen-
tive adjectivization construction  which involves the addition of the verb-sux
mu (or its allomorph m) to a clause to form an adjective. This adjective (which
18 For full details, see section 10.2.4.1.
70 CHAPTER 3. BASIC CLAUSE STRUCTURE

is actually a subordinate clause) states that its referent habitually performs the
action of the subordinate clause, and may be formed from intransitive, transit-
ive or ditransitive active clauses, retaining all the complements and adjuncts of a
full clause, with one exception. The subordinate clause cannot contain a Subject,
with the notional Subject being coreferential with the noun being modied by
the adjective. For example:
(56) na=na [
; kwisha attihsh=ta shaa-m ]
1sg.(nom)=top [ ; very far=in walk-adjzr ]
S1 (S2=head) V2
awa i-s
person be-locut
head (CopComp) V1
`I am a person who habitually walks to very far (I'm used to walking a
long way).'
(57) [ ; kal ki-m=ta ta-mu ]
[ ; work(1) work(2)-adjzr=acc pay-adjzr ]
(A2=head) O2 V2
awa pyal kaa-ma-t1-zi
person money lose-comp-term-nonlocut
head (A1 ) O1 V1
`The person who pays the workers [money] lost the money.'
The head of the np containing the agentive adjectivization is always coreferential
with either a notional a or s of the subordinate clause  once again, Awa Pit
has an s/a pivot, and is operating on the basis of a Subject.
3.5.2.9 Agentive nominalizations
The nal construction showing the existence of a clear s/a pivot in Awa Pit is the
agentive nominalization construction, discussed in section 10.5, which produces
a nominalized active clause with the meaning the one who/which carries/carried
out the action in the subordinate clause. The subordinate clause lacks either
an s or an a, depending on its transitivity, and the understood s or a of the
subordinate clause is coreferential to the referent of the nominalization itself:
(58) [ ; anshik a-t ]=mika=na wiya
[ ; yesterday come-pfpart ]=nmlzr.sg=top ght(1)
(S2=head) V2 head (S1 )
ki-ma-t ma-t1-zi
ght(2)-comp-pfpart anter-term-nonlocut
V1
`The one who came yesterday had fought.'
3.5. SYNTACTIC FUNCTIONS AND GRAMMATICAL RELATIONS 71

(59) [ ; Juan=ta pyan-ta ]=mika=na katsa

[ ; Juan=acc hit-pfpart ]=nmlzr.sg=top big
(A2=head) O2 V2 head (S1 )
`The one who hit Juan is big.
3.5.2.10 Deverbal adjectives and resultatives
There are two constructions in Awa Pit which operate on an s/o syntactic pivot,
rather than an s/a pivot. These are the deverbal adjectives (section 5.2.3.1) and
the Resultative construction (section 11.2.2.2). However, there are reasons for
discounting these constructions from the overall analysis of pivots in Awa Pit.
Both of these constructions are examples of resultative constructions,
showing that a certain entity has undergone a change of state. As Comrie
(1981:113) notes, languages will tend to show a bias towards ergativeabsolutive
syntax in resultative constructions regardless of their base type, because of prag-
matic factors associated with the resultative construction  with an intransitive
verb, any change of state is necessarily assigned to the s; with prototypically
transitive verbs, the entity which undergoes a change of state is the o. Thus any
change of state will most naturally be assigned to s for an intransitive verb, and
o for a transitive verb.
In addition, of course, the deverbal adjectives and the Resultative con-
struction do not apply across all verbs. They are restricted to a semantic grouping
of verbs, those which imply a change of state within their lexical meaning.

3.5.3 Object and Second Object

As well as the grammatical relation of Subject, two grammatical relations have
to be established to cover the syntactic function o and related functions: these
are the grammatical relations of Object and of Second Object. Once again,
to establish these grammatical relations it would be better to rely on syntactic
rather than morphological criteria. Indeed, of the two morphological processes
involving o, case marking gives very unclear results about a grammatical relation
of Object, while number marking is slightly better.
Unfortunately, there is relatively little syntactic evidence for Object and
Second Object in Awa Pit. Given the lack of a passive construction in Awa Pit,
there are only three syntactic constructions which clearly involve o: the Serial
Verb construction, the Resultative, and the use of zha, the First Person Object
Imperative form. The rst two of these cannot be used to establish which of the
non-Subject arguments of ditransitive verbs should be considered o, for dierent
reasons. The Serial Verb construction (see section 11.5) requires that both verbs
in the construction have the same referents for all grammatical relations, but
unfortunately also requires the verbs involved to have the same transitivity 
hence it is impossible to combine a transitive and a ditransitive verb to discover
which argument of the ditransitive verb corresponds to the o of the transitive
verb. The Resultative construction (section 11.2.2.2) with transitive verbs has
an s which corresponds to an o of a verb in a non-Resultative form; however the
72 CHAPTER 3. BASIC CLAUSE STRUCTURE

Resultative construction is only possible with verbs expressing a change of state,

and there are no ditransitive verbs which fall in this semantic class. Hence the
only syntactic construction which can be used to examine which of the two non-
Subject arguments of a ditransitive verb is in o function is the zha construction,
and this will be discussed following the morphological evidence.
3.5.3.1 Case marking
The case-marking system of Awa Pit is nominativeaccusative, as was shown in
section 3.5.2. The Subject of a sentence is always unmarked for case, while an
o may be marked. In general, as was shown in the examples in that section, a
referential human np is marked as Accusative in o function while other nps in
this function are unmarked.
Marking only referential human nps in o function with Accusative mark-
ing, while leaving others unmarked, is not uncommon in the languages of the
world.19 It could, in theory, lead to confusion about whether a particular un-
marked non-human np is in o function or not. In fact, it is usually clear from
context or semantics that a particular non-human referent is, or is not, in o
function. If ambiguity remains, it is always possible to use explicit nps for both
participants, and rely on the usual constituent order, with a preceding o, to
disambiguate the syntactic functions:
(60) ishu=na pitikku ku-m
tiger=top sloth eat-adjzr
A O (not ref. human) V
`Tigers eat sloths.'
It is not just important that the referent of the o be human; to be case-
marked Accusative it must also be referential, in Givón's (1982) terms  it
must not simply refer to some generic type, but rather the referent of the np
must have a specic identity. When an np with a human referent is used in o
function, but as a generic type np, rather than referring to a specic referent,
the np is unmarked:
(61) ashaNpa t1ta-mtu-s
woman search-impf-locut
O (human, not ref.) V
`I am looking for a woman [to marry].'
In fact, it is not quite correct to claim that referential human os have to
be followed by ta. In the vast majority of circumstances they must be, and in
all circumstances they may be, but in fact in certain restricted circumstances a
referential human o may be marked with the locative postpositions mal or pa (or
its allomorph wa). This distinct marking occurs only when (because of modiers)
a clause containing a transitive verb is very low in transitivity, in Hopper &
19Spanish and Hindi, for example, also employ this dierential marking; see Moravcsik
(1978:272281) for further discussion of this phenomenon.
3.5. SYNTACTIC FUNCTIONS AND GRAMMATICAL RELATIONS 73

Thompson's (1980) terms; for example, there is no transfer of action. Thus

beside sentence (62) using the usual marker of o, ta, despite the low transitivity
of the clause, there are sentences (63) and (64) using pa and mal respectively.
(62) Santos=ta=na m1za pyan-a-ma-t
Santos=acc=top almost hit-pl:subj-comp-pfpart
`They almost beat up Santos.'
(63) Demetrio=wa=na m1za pyan-ma-t
Demetrio=in(approx)=top almost hit-comp-pfpart
`[They] almost beat up Demetrio.'
(64) na=na pueblo=mal shaa-ta-w,
1sg.(nom)=top town=loc walk-past-locut:subj
Libardo=mal m1za pyan-ma-t
Libardo=loc almost hit-comp-pfpart
`When I was in town, [they] almost beat up Libardo.'
This usage of postpositions other than ta to mark the o is, however, very uncom-
mon, and any meaning dierence signalled by the choice of postposition remains
to be examined.
In general, then, an np in o function in a transitive clause is marked
with ta if its referent is human and referential, and is otherwise unmarked. In
developing the grammatical relation of Object, however, it is necessary to take
into account ditransitive clauses also, and here there are two nps which could,
potentially, be Objects. That is, given a verb such as kwin- `give', the donor is
clearly Subject, but either the gift or the recipient could be an Object. Because of
their semantics  related to the concepts of giving, taking away from, or asking
 the ditransitive verbs almost always have a human followed by a non-human
np. The rst is always marked with ta, the second always unmarked:
(65) Camilo=na Santos=ta pala kwin-t1-zi
Camilo=top Santos=acc plantain give-past-nonlocut
`Camilo gave Santos a plantain.'
From the point of view of case marking, it remains unclear which of these
two nps is an Object. If a non-human recipient is selected, this is always marked
with the Accusative ta:
(66) na=na kwizha=ta=na comida kwin-ta-w
1sg.(nom)=top dog=acc=top food give-past-locut:subj
`I gave food to the dog.'
This suggests, perhaps, that the rst of the two nps is not being treated like o,
as it is always marked with ta, even when it is not human. However, given the
semantics of the verbs involved, it is also possible that this rst np is an Object,
and it is simply humanized for the purposes of this sentence  as it is only
74 CHAPTER 3. BASIC CLAUSE STRUCTURE

possible in general to give something to a human, the dog in this sentence is

being treated like a human. This analysis gains support from the treatment of
non-humans as humans in possession under certain conditions (see section 5.2.4).
Equally, the second np in these constructions is always unmarked, even
when it is human:
(67) na=na Santos=ta pashu m1la-ta-w
1sg.(nom)=top Santos=acc daughter give-past-locut:subj
`I gave my daughter to Santos.'
While at rst glance this then suggests that the second np is not being treated like
o either, once again the unusual semantics of such a sentence could be overriding
natural assignments  in this sentence, the daughter is really not being treated
as human, but rather as a non-human,20 and consequently the lack of Accusative
marking could be explained by this.
Thus it is not clear from case marking in clauses involving inherently
ditransitive verbs which of the two non-Subject arguments should be considered
to be an Object, as the inherent semantics of ditransitive verbs assign a hu-
man status to the rst non-Subject argument, and a non-human status to the
second. However there is additional evidence from causative-like constructions
(see section 6.4.1) which can be used to examine this issue.
The causative of a transitive verb has three arguments  the causer is
assigned to Subject, and the agent and patient of the transitive verb are assigned
to the rst and second non-Subject positions respectively:
(68) na=na Marcos=ta=na anya=ta payn-nin
1sg.(nom)=top Marcos=acc=top brother=acc hit-caus
kway-ta-w
drop-past-locut:subj
`I made Marcos hit [my] brother.'
With a causative construction, which does not necessarily assign human status to
the rst non-Subject argument and non-human status to the second non-Subject
argument, both non-Subject arguments are followed by the postposition ta when
human and referential, as in the previous sentence, or are left unmarked if they
are non-human or non-referential:
(69) na=na kuzhu piya kwa-nin-ta-w
1sg.(nom)=top pig corn eat-caus-past-locut:subj
`I let the pig eat corn.'
Consequently, case marking cannot determine which of the two non-Subject ar-
guments of a ditransitive verb is the Object: both are marked identically to o,
with ta (normally) if the referent is a referential human, or unmarked otherwise.21
20Among the Awa, there is no tradition of children being married o to someone. A couple
decide to live together in consultation with their parents and siblings (Osborn 1991b:167168).
21An additional issue with establishing the Object relation on the basis of case marking
3.5. SYNTACTIC FUNCTIONS AND GRAMMATICAL RELATIONS 75

3.5.3.2 Number marking

The other morphological characteristic of os is the use of the verb sux na to
indicate plurality of the np in o function (as opposed to a, showing the plurality
of the np in a function). This plural marking is once again sensitive to the
status of the np referents  it can only be used if the np in question is human.22
For ditransitive verbs and transitive verbs in causative-like valency-increasing
constructions, the na plural marking has only been found referring to the rst of
the two non-Subject arguments:
(70) pashpa=ta pan kwin-na-ta-w
child=acc bread give-pl:obj-past-locut:subj
`I gave bread to the children.'
(71) Demetrio maza kuzhu pyaNta-wayn-na-zi
Demetrio one pig kill-help-pl:obj-nonlocut
`Demetrio helped them kill one pig.'
From this, it would appear that the rst non-Subject argument is more likely
to be treated like o, and consequently the Object. Just like case marking, how-
ever, number marking is morphological, and syntactic tests would be much more
convincing in the establishment of the Object grammatical relation.
3.5.3.3 The First Person Object Imperative
As was discussed in section 3.5.3, there are very few syntactic constructions in
Awa Pit which involve o; and of those which do, only one can be used as a test
for Object status: the First Person Object Imperative. This is a special form of
the verb used in an imperative construction when the inherent or derived o is
rst person:23
(72) ; ; t1tizh-zha!
; ; wait:for-imp.1obj
A=2nd O=1st
`Wait for me!'
arises from polysemy. As well as being used to mark os, there is a form ta which is used to
mark location or direction (see section 5.4.3). While these forms have presumably developed
historically from one source, they are synchronically two separate functions, as can be seen by
the dierences in constraints on their use  for os, ta only appears when the referent is human
and referential; the locational ta, on the other hand, appears on any type of np, provided that it
can be understood as a reference to location. This dierence can be used to clearly distinguish
the two functions, and consequently there is no confusion.
It is interesting to compare here the Spanish equivalents of the accusative and locat-
ive/direction markers. Spanish has a preposition a which is used for marking direction towards;
it also has a preposition (its exact status is somewhat controversial) a which occurs before
specic human Object nps. Both of these prepositions (which have a number of other uses)
have developed from the Latin preposition ad, used for marking direction towards.
22And it is not usually used even then; see section 7.3.
23See section 9.4.5.1 for details.
76 CHAPTER 3. BASIC CLAUSE STRUCTURE

(73) ; na-wa=na pit-nin-zha!

; 1sg-acc=top sleep-caus-imp.1obj
A=2nd O=1st
`Let me sleep!'
When used with a ditransitive verb, whether inherently or derived ditransitive,
it is the rst of the two non-Subject arguments (that is, the recipient in the case
of kwin- `give', the agent of the underived verb in the case of a causative) which
has to be rst person:
(74) ; ; an kwin-zha!
; ; more give-imp.1obj
A=2nd arg.1=1st arg.2
`Give me more!'
(75) ; ; pan pay-nin-zha!
; ; bread buy-caus-imp.1obj
A=2nd arg.1=1st arg.2
`Sell me bread!'
The First Person Object Imperative, then, singles out either the o of a transitive
verb or the rst of the two non-Subject arguments of a ditransitive verb.
The rst non-Subject argument of a ditransitive verb is thus morpho-
logically and syntactically equivalent to the o of a transitive verb; the second
non-Subject argument of a ditransitive verb is very similar to the o of a trans-
itive verb, in terms of its morphology, but there are some morphological and
syntactic dierences. The rst non-Subject argument of ditransitives will thus
be considered to form, together with the transitive o, the grammatical relation
of Object; the second non-Subject argument also requires a grammatical relation
rather than just a syntactic function, as it is required for the process of causativ-
ization of a transitive verb (see section 6.4.1)  as this relation is very similar
to the Object relation, it will be called Second Object.
Typologically speaking, it is not unusual that, for example with the verb
kwin- `give', the semantic role of recipient is coded by the Object, while the
gift is a Second Object. As Andrews (1985:125126) points out, it is a striking
typological generalization that if the [two object] nps behave dierently, it is the
one with the Recipient-like semantic role that behaves like a direct object [ie. in
the same way as o].
In summary, then, Awa Pit has two object-like grammatical relations:
Object, which has Accusative case marking if it is referential and human, controls
the use of the na number marking morphology on the verb, and can be referred to
with the special First Person Object Imperative form; and Second Object, which
has Accusative case marking if it is referential and human, but cannot control
number marking on the verb, nor be referred to through the First Person Object
Imperative form.
3.5. SYNTACTIC FUNCTIONS AND GRAMMATICAL RELATIONS 77

3.5.4 Copula Complement

In addition to the grammatical relations of Subject, Object and Second Object,
one further grammatical relation needs to be established for Awa Pit, the Copula
Complement. This relation is always unmarked, but has a number of features
which distinguish it from the Subject grammatical relation.
The majority of two-argument verbs (see section 4.5.4) clearly have a Sub-
ject and an Object argument, according to the tests developed in the previous
sections. However there are stative verbs such as the copula i and m1j `have'
and two active verbs, nam- `become (change into)' and paa- `become (develop
into)', which also take two arguments, but do not have an Object as their second
argument. The rst argument of these verbs is a Subject, being unmarked and
controlling number marking, but the second argument can be distinguished from
an Object, as it is never case marked, even when a referential human (see ex-
ample (76)), cannot control na number marking (in the case of the active verbs),
and cannot be used in the First Person Object Imperative construction;24 and
this second argument must also be distinguished from a Subject, as it cannot
control number marking on the verb (see sentence (77)).
(76) Ricardo=na ap aympihsh i
Ricardo=top my brother be.(nonlocut)
S CopComp V
`Ricardo is my brother.'
(77) ap tío=na paas paynkul m1j-i
my uncle=top two son have-nonlocut
S CopComp V
`My uncle has two sons.'
This second argument of these verbs is unusual in a number of other ways
also. Other verbal arguments are nps, pps, or subordinate clauses; this argument
may also consist of an adjective, with or without degree adverbs.25
(78) s1=na p1na pul i
rewood=top very dry be.(nonlocut)
S CopComp V
`The rewood is very dry.'
Other verbal arguments may, in particular sentences, not appear, even
when they are semantically present, being ellipsed either through denite or in-
denite ellipsis (see section 3.2). This second argument of these verbs, in contrast,
cannot be ellipsed. In addition, other verbal arguments can be moved to initial
position in the clause, while this argument necessarily appears in the position just
24 In fact no imperatives can be formed on stative verbs, so this is not, perhaps, clear evidence
for the statives; but it is for nam- and paa-.
25See section 5.3 for details of the internal structure of Copula Complements.
78 CHAPTER 3. BASIC CLAUSE STRUCTURE

before the verb.26 Indeed, it is possible to have a sentence with a Subject and a
Copula Complement but no verb at all, corresponding to a sentence with Subject,
Copula Complement and the verb i `be' (see section 3.6). These peculiarities of
this verbal argument show that, in fact, rather than being strictly an argument,
the words appearing in this position are forming a part of the predicate, which
is the only obligatorily expressed part of an Awa Pit sentence.
For these reasons, a grammatical relation of Copula Complement is es-
tablished. Elements in this grammatical relation can be recognized as they are
unmarked, and cannot be ellipsed. They express meanings such as equation,
identity and quality (with, for example, the verb i `be'), value (mwiN `be worth')
and possession (m1j `have').

3.5.5 Core, oblique and external functions

Andrews (1985:8082) establishes three distinct types of syntactic functions. He
rst distinguishes between external and internal functions, with the former not
actually forming a part of the basic clause structure and never bearing any specic
semantic role. The internal functions in turn can be divided into two types, core
and oblique. The distinction between core and oblique functions is a little less
distinct, being very much language-dependent. Core functions always include
a, s and o, and any other syntactic functions which are similar in behaviour
to these. The core functions commonly correspond to `syntactic cases' (that
is, those cases which express a wide variety of semantic roles), while oblique
functions correspond to `semantic cases' (those cases which tend to have a more
or less one-to-one correspondence with semantic roles).
In Awa Pit, there is a clearly identiable external function  the external
topic. Those nps which ll this position do not form a part of the basic clause
on phonological grounds, nor are they specically related to any semantic role in
the clause with which they appear.27
Subject, Object and Second Object quite clearly cover syntactic functions
which are core, and can be considered core grammatical relations. Subject and
Object are used to code a, s and o, and consequently are core by denition. The
Second Object grammatical relation is very similar to the Object grammatical
relation, suggesting that it is probably core, and there is supporting evidence
for this. The Second Object is involved in a minor way in syntactic processes,
being necessary for causativization, where the Second Object of a causativized
transitive or ditransitive corresponds to the Object of the non-causativized verb.
The Copula Complement should also be considered a core grammatical
relation. While there is less evidence for this, the Copula Complement does
have distinctive syntactic properties  it cannot be ellipsed and appears to form
part of the predicate, and may consist of an adjective phrase as well as noun
phrases, postpositional phrases and subordinate clauses  and it codes a variety
26Neither of these conditions actually holds for the second argument of the stative two-
argument verb pyan `know'.
27For further details on the external topic, see section 14.2.
3.6. PREDICATE TYPES 79

of semantic roles.
All other syntactic functions in Awa Pit appear to be oblique. They are
not involved in any grammatical processes, and correspond relatively closely with
semantic roles. The various oblique syntactic functions will not be discussed here,
because they are necessarily coded in Awa Pit through the use of postpositions,
and the close correspondence with semantic roles means that it is simplest to
discuss the oblique functions together with the various postpositions, which will
be done in section 5.4.
In Awa Pit, then, there is an external function, the external topic, dis-
cussed in section 14.2. There are four core grammatical relations, the Subject,
Object, Second Object and Copula Complement, which were discussed in the
previous sections. There are a variety of oblique syntactic functions, which cor-
respond to the semantic roles taken by these functions, and these will be discussed
in section 5.4 together with the postpositions which encode these functions.

3.6 Predicate types

As mentioned in the introduction to this chapter, all clauses in Awa Pit have
an explicit, overt predicate. However this predicate may be one of a number of
types. The major division is between those predicates which are purely verbal,
and those which involve a Copula Complement.

3.6.1 Purely verbal predicates

Most commonly predicates are purely verbal, and consist of single verb roots
or verb stems,28 or a construction with a non-nite verb stem followed by an
auxiliary verb (see section 11.2). In either case, the verb stem may in fact be
a Serial Verb, compounded of more than one verb stem (see section 11.5). The
verb stems involved may be active or stative, simple or compound, impersonal,
intransitive, transitive or ditransitive (see section 4.5 for exemplication of all
these distinctions). Purely verbal predicates may be inected as nite, non-
nite, or with the Conjoined Clause inection (see section 7.2 for inectional
possibilities), and may occur as main clauses or subordinate clauses.

3.6.2 Copula Complement predicates

Much more complex than the purely verbal predicates are those predicates which
involve a Copula Complement. There are a variety of stative verbs and two
active verbs which combine with a Copula Complement to form a predicate (see
section 4.5.4), and these may be nite or non-nite, depending on the inection
on the verb (see section 7.2).
In addition to these partially verbal predicates, a Copula Complement on
its own may form a predicate. The existence of this clause type was justied in
28 Where verb stems are verb roots plus derivational axes; see chapter 6.
80 CHAPTER 3. BASIC CLAUSE STRUCTURE

section 3.3.2 above; examples of adjective, noun phrase and subordinate clause
Copula Complements acting as predicates are:
(79) ap kwizha=na t1l
my dog=top black
`My dog was black.'
(80) na=na putsha awa
1sg.(nom)=top white person
`I am a white person.'
(81) awa pala ku-m
person plantain eat-adjzr
`The Awa are plantain-eaters.'
Unlike purely verbal predicates, which have a wide variety of subcategor-
ization options, predicates involving Copula Complements have only one com-
plement, a Subject. Of course, such predicates themselves involve a Copula
Complement as well, either as the entire predicate, or as a complement to the
verbal part of the predicate.
Chapter 4
Word classes
4.1 Introduction
As in all languages, there are a variety of dierent types of words  word classes
 in Awa Pit. While many of the word classes will be discussed in more detail in
other sections of this thesis, this chapter contains a summary of the word classes
which have been found, and the dierences between them.
There are three large open classes of words in Awa Pit  nominals, ad-
jectives and verbs  and a string of minor closed classes  time adverbs, place
adverbs, manner adverbs, other adverbs, postpositions, discourse particles and
interjections. These classes can be distinguished on the formal grounds of mor-
phology and distribution, both at the sentence level and at the level of discourse.
Nominals and adjectives, and adjectives and some adverb types are the most
dicult to distinguish, but there are grounds for claiming them to be separate
classes, as will be shown below.
After a brief examination of the concept of `word', the various word classes
of Awa Pit will be discussed, together with their subclasses. Where particular
word classes are discussed in more detail in later sections of the thesis, a short
discussion only will be given, with references to the fuller account.

4.2 Phonological words and grammatical words

As has often been shown in the literature, the term `word' is ambiguous, and
may refer either to a phonological or to a grammatical item (see, for example,
Anderson 1985, Dixon 1988). In any discussion of word classes, it is necessary to
focus on the concept of the grammatical word only, and in Awa Pit this does not
necessarily align with the phonological word.
A phonological word in Awa Pit can be recognized by a variety of means,
as was shown in chapter 2. Each phonological word carries a main stress; there
are possibilities of pausing before and after a phonological word; and there are
several phonemes whose pronunciation varies depending upon whether they begin
a phonological word or form part of the same phonological word as the preceding
phoneme (see especially section 2.2.1.2).
82 CHAPTER 4. WORD CLASSES

Grammatical words, on the other hand, can be found by examining dis-

tribution and cooccurrence phenomena. For example, the sequence of phonemes
/na/, the Topic marker, never occurs as a phonological word  it is always found
as the nal sequence on a larger phonological word. However, its status as a gram-
matical word, not an ax, can be easily recognized, as it may occur following
almost any (grammatical) word from the other word classes. As a grammatical
word which phonologically attaches to the previous word, it will be considered a
clitic. There are a number of clitics in Awa Pit  in particular, the postpositions
and the discourse particles all cliticize onto the previous element in a sentence.
The reasons for considering these as separate grammatical words, rather than as
suxes, will be discussed individually for the two word classes.
Thus grammatical word boundaries in Awa Pit do not necessarily coincide
with phonological word boundaries. One phonological word may consist of more
than one grammatical word. The boundary between two grammatical words in
one phonological word will be indicated with the equals sign (=), contrasting
with the hyphen (-) indicating a morpheme boundary within one grammatical
word.
It appears that one grammatical word in Awa Pit cannot be expressed
by more than one phonological word. The compound verbs would seem to be
a candidate for this, but the two parts of the one grammatical word, while
normally collocated, can be separated by the negative marker, and these verbs
will thus be treated as one lexical item represented by two (phonological and
grammatical) words  for further details, see section 4.5.6.
In the remainder of this chapter (and, indeed, this thesis) the term `word'
will be used to refer to grammatical words, with `phonological word' being stated
in full otherwise.

4.3 Nominals
The rst of the three open classes to be discussed is the class of nominals. The
major use of nominals is as the head of an np, although many of them may also
be used as modiers of an np head, or as predicates. Some of the subclasses may
be used as adjuncts.
There are six subclasses of nominals, and all but one of these, the class of
nouns, appear to be closed classes, although the cross-classication of words as
both nouns and relational nouns suggests that the subclass of relational nouns
may in fact also be open (see section 4.3.5).

4.3.1 Nouns
The largest subclass of nominals are the nouns. These form an open subclass, with
many nouns having entered from Spanish, particularly nouns referring to cultural
objects previously not found in the Awa region. There is no specic dening
characteristic of all nouns to distinguish them from other nominals  they are
simply those nominals not categorized in any other subclass. They are discussed
4.3. NOMINALS 83

at greater length in section 5.2. As in Imbabura Quechua (Cole 1985:129), but

unlike many South American languages, there are no noun classes and no nominal
classication system in Awa Pit; nor is there a mass/count noun distinction.
The semantic core of the nouns is, of course, concrete objects, whether hu-
man (such as ampu `man' and ashaNpa `woman'), animate (such as ulam `arma-
dillo' and kuzhu `pig') or inanimate (such as uk `stone' and uskulam `blowpipe').
Awa Pit, like Epena Pedee (Harms 1994:19), appears to have no abstract nouns.
Human nouns may be distinguished from the others by the presence of Accusative
marking with Object nps containing human nouns (see section 5.4.2), and their
ability to take the sux tuzpa as a marker of Collective action (see section 5.2.6).
Perhaps one section of the noun subclass which needs further discussion
is the area of kinship. Kinship relations between people are encoded via nouns in
Awa Pit, and the relation is shown in same way as possession (see section 5.2.4).
The traditional kinship system of the Awa, based largely on sibling groups
(Osborn 1974), appears to be breaking down in the region of the Resguardo de
Pialapí; and the traditional terminology would seem to be the rst thing to go.
Several speakers had trouble recalling the traditional terms for various kinship
relations, and many Spanish terms are used, either to cover kinship relations
which perhaps did not exist for the traditional Awa (such as p1timu `male cousin'
from Spanish primo), or as one term to cover a Spanish concept which was
previously divided into several dierent categories (kunyata `sister-in-law' from
Spanish cuñada to cover both mamsha `male's sister-in-law' and mayN `female's
sister-in-law').1
Several of the core kinship relations have two words, a formal and an
informal word. The informal word is always used to address one's own kin, and
also to talk about one's own kin or the kin of those closely related to oneself,
while the formal terms are usually used to refer to someone else's kin. The
kinship terms which were collected are listed in Table 4.1; underlined terms may
have been borrowed from Spanish. It is also interesting to note that one of the
words for father, taytta, is similar to words in Quechua, Kamsá and Paez, and
is used in the Spanish of southern Colombia to refer to traditional leaders of
various indigenous groups. The terms for cousin, p1timu and p1tima are also
used to mean `general relative', that is, to refer to someone who is known to be
a relative, but the precise relation is not known or does not have a cover term;
the politically aware Awa also use this term to refer to any Awa, or sometimes
even extend it to any indigenous person.

4.3.2 Time nouns and place nouns

There are many nouns which, through their semantics, indicate places or times,
such as yal `house', Pialap
 `the settlement of Pialapí' and anshik `yesterday'.
1The system of kinship among the Awa, and its breakdown, appears similar to that of
Imbabura Quechua. Like Awa Pit, Imbabura Quechua has separate terms for a male's brother,
a male's sister, a female's brother and a female's sister; and like Awa Pit it has also borrowed
terms for aunt, uncle, female cousin, male cousin, mother-in-law and father-in-law from Spanish
(Cole 1985:219).
84 CHAPTER 4. WORD CLASSES

Formal Informal Gloss Referent

taytta, pappihsh pappa father F
akkwihsh akkwa, mama mother M
aympihsh anya male's brother B
kwampihsh kuwa male's sister Z
nalpihsh, alpihsh ala female's brother B
ayshpihsh azha female's sister Z
paynkul son S
pashu daughter D
kwankwa grand-mother/-daughter FM, MM, SD, DD
pampa grand-father/-son FF, MF, SS, DS
tiyu uncle FB, MB
tiya aunt FZ, MZ
p1timu male cousin FBS, FZS, MBS, MZS
p1tima female cousin FBD, FZD, MBD, MZD
sup1tinu nephew BS, ZS
sup1tina niece BD, ZD
ampu husband (`man') H
ashaNpa wife (`woman') W
kunyatu male's brother-in-law ZH, WZH
mamshappihsh mamsha male's sister-in-law BW, WBW
mamappihsh mama female's brother-in-law ZH, HZH
mayNpihsh mayN female's sister-in-law BW, HBW
swik1tu father-in-law WF, HF
swik1ta mother-in-law WM, HM
kwanappihsh kwana son-in-law DH
nwita daughter-in-law SW
Table 4.1: Awa kinship terms
4.3. NOMINALS 85

aN= `here'
uN= `there (in contrast to aN, a physical place)'
su= `there (in that place, physical or metaphorical)'
m1n= `where, nowhere'
kwizh= `later'
Table 4.2: Time and place nouns

These may either be used in nps as arguments of a verb, or together with post-
positions in pps, in structures common to all nouns (see section 5.2). Nouns
which refer to time may also be used bare as adjuncts (see section 13.2.2). These
words are, formally, identical to nouns.
However there are a very small number of words expressing deictic place
and time which can be formally distinguished from nouns, as they cannot be used
bare, but are obligatorily followed by a cliticized postposition. These forms are
listed in Table 4.2. For further details on the rst four words, see section 13.4;
for the last word, see section 13.2.4.
It would be possible to consider that the combination of these words with
a postposition actually forms a single place or time adverb. However, just as
with any other noun, there is a choice of postposition for these time and place
nouns. Thus, for example, there is a contrast between aN=ta `here' and aN=pa
`around here', where the contrast is due to the dierence in postposition, and
parallel to the distinction between yal=ta `in the house' and yal=pa `in the
neighbourhood of the house'. Thus analyzing the time and place nouns and the
following postposition together as a unit  as time and place adverbs  it would
be necessary to enter each time/place noun plus postposition combination into
the lexicon as separate items, and lose the generalizations which hold across all
uses of the postpositions.

4.3.3 Personal pronouns

The non-interrogative personal pronouns, which necessarily refer to humans only,
are dierentiated for three persons and two numbers. There are also two inter-
rogative/negative personal pronouns and two indenite pronouns; these have no
number distinctions.2 (See Table 4.3 for the forms.) The existence of special
plural forms for the non-interrogative and non-indenite pronouns is one of the
factors which distinguish personal pronouns from other nominals, which have no
number marking.
While other referential human nps use the postposition ta to indicate that
they are in Object relation (see section 5.4.2), the personal pronouns have special
The indenites are clearly based on the interrogative/negative pronouns, plus a following
2
waza, which does not appear to occur elsewhere in the language.
86 CHAPTER 4. WORD CLASSES

Singular Plural
Nominative Accusative Nominative Accusative
1st person na na-wa au au . . . =m1za
2nd person nu nu-wa u u . . . =m1za
3rd person us us-a uspa uspa . . . =tuza
`who, no-one' m1n m1n-a
`someone' m1n-waza m1n-a-waza
`what, nothing' shi
`something' shi-waza
Table 4.3: Personal pronouns

Accusative forms. For the singular personal pronouns there are special Accusative
suxes, wa for rst and second person, a for third person and the interrogative
person pronoun. As these forms are obligatorily attached to the personal pro-
nouns, and no material may intervene between the pronoun and the Accusative
marker, they are considered suxes rather than clitics. Another possibility would
be to consider the combination of pronoun and Accusative marker as a suppletive
form, but the repetition of the same forms for more than one personal pronoun
(wa for two pronouns, a for two pronouns), and the fact that the form of the
Nominative personal pronouns occurs in the Accusative form, suggest that an
analysis of these forms as Accusative suxes is more appropriate.
The plural personal pronouns have special Accusative clitics, which attach
to the nal element of an Object np containing a plural personal pronoun, not
necessarily directly to the personal pronouns themselves. Normally the only
element in an np containing a personal pronoun is the personal pronoun itself,
as no modiers except numerals are permitted in such an np. However, when
numerals do cooccur with (plural) personal pronouns, they follow the pronoun
(as in (83)), whereas numerals precede nouns (as in (82)). Thus if a personal
pronoun with a numeral is used in Object relation, the Accusative clitic occurs
on the numeral rather than the pronoun itself (as can be seen in (84)).
(82) kutnya ampu
three man
`three men'
(83) au kutnya
we three
`we three'
(84) au kutnya=m1za
we three=(1/2pl.)acc
`us three'
4.3. NOMINALS 87

The nal dierence between personal pronouns and other nominals is that
other referential human nominals form a possessive using the postposition pa (see
section 5.4.8). The human interrogative/negative pronoun m1n `who' forms a pos-
sessive in this usual fashion; however the non-interrogative personal pronouns do
not have possessive forms in this sense, but rather uses the suppletive possessive
adjectives for the singular pronouns, and either the singular possessive adjective
or the bare plural personal pronoun itself for the plural case (see section 4.4.5).
More information on the non-interrogative personal pronouns is found in
section 5.2; the interrogative/negative personal pronouns are discussed in sec-
tions 12.2.1 and 12.2.2.

4.3.4 Demonstrative pronouns

There are two demonstrative pronouns in Awa Pit, corresponding to the two
demonstrative adjectives (see section 4.4.4): ana `this' and suna `that'. These
demonstrative pronouns cannot be modied.
The demonstrative pronouns are distinguished from other nominals in that
they cannot be followed by the Topic marker na. It would appear that, histor-
ically, these pronouns may have developed from the demonstrative adjectives
an `this' and sun `that' by the axation of the Topic marker, and this would
explain their inability to be used together with the Topic marker. However syn-
chronically they are unitary lexical items, as can be seen from two facts: (1)
they have only a single n, while a combination of the demonstrative adjective
and the Topic marker should have two adjacent ns  ana `this (demonstrative
pronoun)' versus an=na `this (demonstrative adjective)=Topic marker'; and (2)
the Accusative-marking postposition ta (and other postpositions) occurs after
the entire demonstrative pronoun, whereas when other Accusative nominals are
topicalized, the Topic marker occurs after the Accusative marking:
(85) ana=ta
this=acc
`this (Object)'
(86) kuzhu=ta=na
pig=acc=top
`pig (Object)'
The demonstrative pronouns are also unique in that they may refer either
to humans or non-humans. Other nominals are either inherently human or inher-
ently non-human. When in Object position, if the demonstrative pronouns are
referring to humans they will take Accusative marking; if referring to non-humans
they will appear without Accusative marking.

4.3.5 Relational nouns

Awa Pit has a series of relational nouns (or `relator nouns' in Blake's (1994:16)
terminology), used to indicate complex locational ideas such as `on top of'. These
88 CHAPTER 4. WORD CLASSES

ideas are expressed through the combination of a nominal and a relational noun,
followed by a locational postposition:
(87) mesa 1sal=ki=na libro tu-y
table top (relational noun)=at=top book be:in:place-nonlocut
`There is a book on top of the table.'
As in the case of English in the above translation, some of these relational
nouns are transparently related to nouns or adjectives. For example, the rela-
tional noun used in the above example, 1sal, is cross-classied as a noun meaning
`top', and can be used without a following locative postposition in the same
wholepart construction:
(88) mesa 1sal=na natam izh-ma-t=na
table top=top yellow see-comp-pfpart=top
`The top of the table is yellow (in appearance).'
If all relational nouns were transparently related to nouns or adjectives in this
fashion, it would be unnecessary to distinguish them as a separate nominal sub-
class. However, while all relational nouns presumably developed historically from
nouns or adjectives, they are no longer all transparently related to nouns or ad-
jectives. Some relational nouns have diverged in meaning from the associated
noun or adjective: kal, a noun, means `fence', while kal, a relational noun, means
`inside'. And some relational nouns cannot be used as nouns or adjectives: ma is
a relational noun meaning `beside', but there is no noun or adjective ma. Thus
it is necessary to set up a separate subclass of nominals, the relational nouns. In
addition, the relational nouns cannot be modied, unlike other nouns.
All relational nouns which have been found in Awa Pit are listed in
Table 4.4, together with the meaning of a homophonous noun or adjective for
those relational nouns which have such a correspondence. For more details on
relational nouns, see section 5.2.4.3.

4.3.6 Nominal postmodiers

The nal subclass of nominals, the nominal postmodiers, is a slightly anomalous
subclass. It contains only four elements, shown in Table 4.5. The four words
in this subclass are all cross-classied as nouns (and one also as a descriptive
adjective), but they may also be used in a dierent construction, placed after an
np or adjective to modify it. This contrasts with the standard modier position,
which is before the head. This construction can only be used with an np or
adjective which is lling the Copula Complement relation (see 5.3).
(89) katsa 1lapa
big (Adj) big (Nominal postmodier)
`huge'
4.3. NOMINALS 89

Form As relational noun As other noun/adjective

ayuk= `inside of'
izhkwak= `face to face with, in front of (person)'
1sal= `on top of, in the top of' `top' (N)
kal= `within, inside of (walls)' `fence' (N)
kut= `outside of'
kwaka= `on the (other) side of' `side (of river, road)' (N)
kwash= `above'
kwiya= `to the left of' `left' (Adj)
ma= `beside'
nil= `in front of' `correct, straight' (Adj)
nu= `to the right of' `right (e.g. hand)' (Adj)
pazhka= `in the middle of'
pula/puta= `below'
puns1h= `behind' `back' (N)
su= `under' `earth, land, ground' (N)
Table 4.4: Relational nouns

Word Noun Nominal postmodier

1lapa `old (adjective), old man (noun)' `big, augmentative'
pashpa `child' `small, diminutive'
ampu `man' `male'
ashaNpa `woman' `female'
Table 4.5: Nominal postmodiers
90 CHAPTER 4. WORD CLASSES

(90) pii 1lapa

river/stream big
`river'
(91) kwizha ashaNpa
dog female
`bitch'
While these nominal postmodiers are phonologically identical to nouns,
their meanings are dierent as nouns and nominal postmodiers. Two of them,
1lapa `big' and pashpa `small', are used almost as an augmentative and a dimin-
utive, contrasting with the descriptive adjectives katsa `big' and aynki `small' in
that the former often have a clear aective value or emotional tone lacking from
the latter. The second pair, ampu `male' and ashaNpa `female', are used when it
is necessary to distinguish the sex of animals, as human nouns are the only ones
which make an inherent distinction between sexes.

4.4 Adjectives
The second of the open word classes in Awa Pit is the class of adjectives. The
primary function of adjectives is as modiers, though with the exception of the
demonstrative adjectives they may also be used in nps with an ellipsed head (see
section 5.2.5), and as predicates (see section 3.6).
The major subclass of adjectives is that of descriptive adjectives, which is
the only open subclass. In addition there are three closed subclasses of adjectives:
the quantiers, the demonstrative adjectives and the possessive adjectives.
Adjectives used in nps come before the head, with a specic ordering of
the subclasses, allowing their separation into these dierent subclasses on formal
grounds (see section 5.2). The only exception are numerals used with plural
personal pronouns, where the numeral follows the pronoun (see section 4.3.3).

4.4.1 Distinguishing adjectives from nominals

As in many other languages, such as Imbabura Quechua (Cole 1985:73), nom-
inals and adjectives are morphologically similar in Awa Pit. In addition, both
nominals and adjectives can appear as the only element in an np (with the excep-
tion of demonstrative adjectives), as modiers of an np head (with the exception
of demonstrative pronouns, time and place nouns, relational nouns and some
personal pronouns), and as predicates (with the exception of demonstrative pro-
nouns, demonstrative adjectives and some personal pronouns). Despite these
similarities in distribution, it is possible to distinguish nominals from adjectives.
To begin with, there are two series of demonstratives. Demonstrative
pronouns may only occur as the heads of nps, while demonstrative adjectives
may only modify np heads. Additionally, there are dierent interrogatives for
questioning nominals and adjectives: m1n `who' or shi `what' for nominals, versus
m1n `which' for adjectives.
4.4. ADJECTIVES 91

While there is little nominal or adjectival morphology, there is one sux,

the Collective action sux, used when a group of animate referents are acting
together in a task. This sux has two forms, one for nouns (tuzpa), and a
separate form (tuz) which occurs on adjectives when these are the only element
in an np.3
Both adjectives and most nominals may be used as modiers of a head in
an np, but the distribution is slightly dierent. When an explicit head is present,
either may be used as a modier:
(92) katsa yal
big house
Adj N
`a large house'
(93) uk yal
stone house
N N
`a stone house'
However once a referent has been introduced into a conversation, it may be
referred back to by a modier only (without an explicit head) if the modier is
an adjective, but not by a modier nominal:
(94) katsa
big
Adj
`the large [house]'
(95) *uk
stone
N
(acceptable as `the stone', but not as `the stone [house]')
This dierence is, perhaps, unsurprising, as uk `stone' can refer simply to `the
stone', and using it with no explicit head could lead to ambiguity.
Equally, although both nominals and adjectives can be used as the only
word in an np, there is evidence that they have dierent statuses. A nominal
used alone in an np is truly the head of that np (which is why a nominal modier
cannot be used without an explicit head); while an adjective used alone in an np
is still modifying a head, it is simply that this head has been ellipsed. Within
a connected discourse it is possible to refer to an introduced referent by an np
containing simply an adjective, such as
(96) katsa a-t1-zi
big come-past-nonlocut
`The large [person] came.'
3 The latter form is also used on nominalizations.
92 CHAPTER 4. WORD CLASSES

Semantic type Word Meaning

Dimension katsa `big'
nas `thin'
Physical property ii `hot'
sam `hard (solid)'
Colour kwaNtam `red'
t1l `black'
Human propensity ak `shy'
t1mpa `stupid'
Age 1lapa `old'
maas `new'
Value wat `good'
kwayl `bad'
Difficulty kal `dicult'
Qualification nil `correct'
Table 4.6: Examples of descriptive adjectives in adjective types

However a sentence containing an np with only an adjective cannot be used to

introduce a new participant, while an np with a head noun can be used as the
rst reference to a participant; thus while the previous sentence could be said
after a sentence such as `Two people said they were coming today', it could not
be used without some prior sentence. Thus rather than the adjective being the
head of the np, the adjective is modifying an ellipsed head, coreferential to a
previously mentioned nominal; Imbabura Quechua allows precisely the same use
of an adjective with an absent head (Cole 1985:76).
Nominals and adjectives, then, are similar in many ways in Awa Pit, but
they can be distinguished  there are two sets of demonstratives and interrog-
atives, depending on whether a nominal or adjective is required; nominals can
only be used as modiers with an explicit head, while adjectives can be used as
modiers with or without an explicit head; and nominals can be used to intro-
duce new participants, while adjectives can only be used alone in an np (with an
ellipsed head) if the referent is already established.

4.4.2 Descriptive adjectives

The only open subclass of adjectives are the descriptive adjectives, of which there
are a wide variety in Awa Pit. In a string of adjectives, the descriptive adjectives
follow all other adjective subclasses. Of Dixon's (1991:78) ten semantic types of
adjectives, Awa Pit has adjectives belonging to eight types  dimension, phys-
ical property, colour, human propensity, age, value, difficulty and
qualification (see Table 4.6 for examples). While there are examples of these
eight types, there are few examples of the difficulty or qualification types.
4.4. ADJECTIVES 93

maza `one'
paas `two'
kutnya `three'
ampata `four'
akkwan `much/many'
pitshiN `a little/few'
mamaz `other'
maza maza `some (distinct objects)'
yawa `how much/many?'
wan `all'
Table 4.7: Quantiers

The two missing types are encoded in other word classes  speed concepts
through manner adverbs (see section 4.9), and similarity through the use of
the postposition kana `like' (see section 5.4.12).
There is one interrogative descriptive adjective, m1n `which'. This is used
to ask for a dierentiation of items through their description, and is discussed
further in section 12.2.4. There is also an indenite descriptive adjective m1nat
`some (don't know which)'.4

4.4.3 Quantiers
In Awa Pit there are four traditional numerals (see Table 4.7), with numerals
greater than four being generally borrowed directly from Spanish.5 Aside from
the numerals, there are six other quantiers, including an interrogative quantier
(see Table 4.7). Quantiers precede any descriptive adjectives in a string of
adjectives, and are incompatible with each other, as well as being incompatible
with the comparative adverb an `more'.
The numeral maza `one' has occasionally been found in contexts in which
a numeral was not appropriate, but an indenite article would be appropriate:
(97) maza awa a-t1-zi
one person come-past-nonlocut
`Someone came (while I was away).'
4This indenite clearly consists at least historically of the interrogative m1n `which' plus a
sux at; but this sux is not found elsewhere.
5In those cases where a Spanish numeral contains within its expression another numeral less
than or equal to four, speakers will sometimes use the Awa Pit word for that numeral within a
Spanish frame, but more often simply use the entire Spanish numeral. For example, as with the
corresponding English expression four hundred, the Spanish equivalent cuatrocientos contains
within it the numeral cuatro `four', and some speakers will sometimes say ampata cientos; but
more commonly all numerals above four are fully expressed in Spanish.
94 CHAPTER 4. WORD CLASSES

Because of the common shift from the numeral `one' to an indenite article, it is
unclear whether this is a result of language interference (Spanish has an indenite
article un(a) identical to its quantier `one'); whether Awa Pit is developing
an indenite article under the inuence of Spanish, as has occurred in Pipil
(Campbell 1987:272); or whether it is an entirely separate development.

4.4.4 Demonstrative adjectives

The subclass of demonstrative adjectives is very small, containing only two words,
an `this' and sun `that'. In a string of adjectives, these words may only appear
before all other adjectives (see section 5.2).
Positionally, the demonstrative adjectives and the possessive adjectives
(see section 4.4.5) form a single subclass, and it is impossible to combine words
from the two subclasses. However there are some distributional dierences. The
possessive adjectives may be used as Copula Complements or predicates, while
the demonstrative adjectives cannot; and the possessive adjectives can also be
used in an np with an ellipsed head, while under these circumstances the demon-
strative pronouns rather than the demonstrative adjectives are used, although it
is possible that historically the demonstrative pronouns have developed from the
demonstrative adjectives.6

4.4.5 Possessive adjectives

The subclass of possessive adjectives is quite closely associated with the demon-
strative adjectives, although there are dierences (see section 4.4.4), with only
one word from the two subclasses permitted in any string of adjectives.
There are only three possessive adjectives in Awa Pit, corresponding to
the three singular personal pronouns (see section 4.3.3): ap `my', up `your' and
paynya `his/her'. There are no special plural possessive adjectives  to indicate
something which belongs to more than one person, either the singular possessive
adjective of the appropriate person may be used, or else the plural personal
pronoun is placed before the noun which is possessed.7

4.5 Verbs
Like many South American languages, Awa Pit is a strongly verb-marking lan-
guage, with a great deal of information being packed into the verb. Verbs are
easily recognizable, being the only words which can be marked with many axes,
showing such things as aspect, tense and mood (see chapter 9) as well as a num-
ber of derivational axes (see section 6.4). With the exception of a few equative,
identity and description sentences (see section 3.6), all sentences in Awa Pit must
contain a verb; even minimal responses in answer to polar or content questions
contain a verb  in response to the equivalent of Who went? a speaker must
6 See section 4.3.4 for details on the demonstrative pronouns.
7 See section 5.2.4 for many more details on possession and related phenomena.
4.5. VERBS 95

reply Remigio went not just Remigio, or replying to Did you go? a speaker
will say the equivalent of either I went or I did not go.
There are three cross-cutting classications of verb roots which must be
taken into account in establishing the verbal subclasses. These are the distinc-
tions between active and stative verb roots, between simple and compound verb
roots, and between verb roots of various argument frames (valence and transit-
ivity). In addition there are a small number of irregular verb roots. Aside from
the copula, which has a stem i in the Present and a otherwise, the irregular verbs
all consist of roots with an alternation in their ending between ku, which occurs
before a sux beginning with a bilabial or velar consonant or a vowel, and kwa,
which is the form they take otherwise. The most common of these is ku/kwa-
`eat'; others are ishku- `scare', kutku- `tell a lie' and walku- `steal'.
The major classication of Awa Pit verb roots is into two groups, the
stative verbs and the active verbs. This distinction is a semantic distinction
between those verbs describing predicates which are inherently imperfective 
such as i `be', m1j `have' and pana `be standing'  versus those which are not
inherently imperfective  such as 1 `go', pak `harvest' and pyan `hit'. This
semantic distinction corresponds to a morphological distinction: active verbs may
be suxed by the Imperfective aspect marker, while stative verbs cannot take this
sux, being inherently imperfective. This distinction is important in a number
of cases; for example, the present tense is unmarked, with just the stem form for
stative verbs, but the stem plus Imperfective for active verbs (followed potentially
by person marking). Only two verbs have been found which apparently can be
either active or stative, but in fact there are meaning distinctions between the
two formally identical verb roots: shaa `walk' is an active verb, while shaa `be
around' is a stative verb; tu `lie down' is an active verb, but tu `be permanently,
be lying down' is a stative verb. As will be discussed below (section 6.4.2.1), verb
roots suxed with the Desiderative marker may also be either active or stative,
although there is no discernable meaning dierence in this case.
The vast majority of verbs in Awa Pit are simple verbs  the lexical
meaning of the verb is carried by a single phonological and grammatical word.
However there are a number of compound verbs in the language, where the lex-
ical meaning is expressed in two phonological words. Sometimes the words in
a compound verb are also individually meaningful words in the language, and
the combined meaning is related to the meanings of the two individual words,
although the precise meaning may be unpredictable. In other cases, one of the
two phonological words in a compound verb is not a separate lexical item in the
language. Compound verbs are always active, and have a variety of argument
structures. As compound verbs can be understood by a slight modication of the
framework required for simple verbs, they will be discussed following the simple
verbs, in section 4.5.6.
Verbs in Awa Pit may have zero, one, two or three `obligatory' arguments
(complements). The use of the term `obligatory' is slightly odd here, as these
arguments may not necessarily appear in any given sentence, due to the use of
both denite and indenite argument ellipsis in Awa Pit. However there are some
96 CHAPTER 4. WORD CLASSES

tests which can be used to establish the number of complements of a verb  see
section 4.5.1.
Dierent verbs require their complements to have dierent sets of gram-
matical relations.8 Many verbs require some subset of core grammatical relations
 Subject, Object, Second Object, Copula Complement , but a few verbs
require some oblique grammatical relations, such as a locational element.
In the following discussion of verbal subclasses, verbs will be divided up on
the basis of their status as simple or compound verbs (with all compound verbs
being discussed last, in section 4.5.6), on the basis of their subcategorization
frames (both number of arguments and the grammatical relations of those argu-
ments), and on the basis of the active/stative distinction. Prior to the discussion
of the individual subclasses, the method for establishing verbal subcategorization
frames will be discussed.

4.5.1 Establishing subcategorization frames

As Awa Pit uses both denite ellipsis (ellipsis of an argument if its referent is
retrievable from context) and indenite ellipsis (ellipsis of an argument if its
referent is considered irrelevant for the purposes of communication), from any
given sentence it is sometimes dicult or impossible to tell the valence of the verb.
However some tests can be used to ascertain verb valences and subcategorization
frames.
Subject, Object, Second Object and Copula Complement grammatical
relations are always complements, never adjuncts. Consequently, if a verb can
take an argument in one of these relations, but appears sometimes without it,
either the argument is ellipsed in that particular instance, or the verb has more
than one subcategorization frame.
The only real diculty occurs in establishing whether a verb which appears
in sentences with only one argument (a Subject) or with two (a Subject and an
Object) has two subcategorization frames, or is simply occurring with an ellipsed
argument in the rst of these cases. For many verbs, however, there is a simple
way of testing this, through the Resultative construction (see section 11.2.2.2);
unfortunately, only those verbs which involve some sort of transformation can be
used in this construction.
Given sentences such as
(98) na=na 1-ta-w
1sg.(nom)=top go-past-locut:subj
`I went.'
(99) na=na kwa-ta-w
1sg.(nom)=top eat-past-locut:subj
`I ate.'
8 See section 3.5 for an explanation and justication of the grammatical relations in Awa Pit.
4.5. VERBS 97

(100) na=na pyan-ta-w

1sg.(nom)=top hit-past-locut:subj
`I hit [someone].'
it appears at rst that all three verbs 1- `go', ku- `eat' and pyan- `hit' might have
only one argument, a Subject. However ku- `eat' and pyan- `hit' can also occur
with both a Subject and an Object, while 1- `go' cannot:9
(101) na=na pala kwa-ta-w
1sg.(nom)=top plantain eat-past-locut:subj
`I ate a plantain.'
(102) na=na Juan=ta pyan-ta-w
1sg.(nom)=top Juan=acc hit-past-locut:subj
`I hit Juan.'
With this evidence, it appears that 1- `go' is intransitive, while ku- `eat' and
pyan- `hit' are transitive, but have ellipsed Objects in sentences (99) and (100).
However applying the Resultative construction to the above three verbs gives
dierent results. The intransitive 1- `go' can be put in the Resultative, with the
Subject of the Resultative corresponding to the Subject of the non-Resultative:
(103) na=na 1-ma-t i-s
1sg.(nom)=top go-comp-pfpart be-locut
`I am in a state of having gone.'
The transitive pyan- `hit' in the Resultative has a Subject which necessarily
corresponds to the Object in the non-Resultative:
(104) Juan=na pyan-ma-t i
Juan=top hit-comp-pfpart be.(nonlocut)
`Juan is in a state of having been hit.'
(105) *na=na pyan-ma-t i-s
1sg.(nom)=top hit-comp-pfpart be-locut
(Can't mean `I am in a state of having hit'; ne as `I am in a state of
having been hit'.)
With ku- `eat', two sentences are possible, with the Subject of the Resultative
corresponding to either the Subject of the non-Resultative in (99) or the Object
of the non-Resultative in (101):
(106) na=na ku-ma-t i-s
1sg.(nom)=top eat-comp-pfpart be-locut
`I am in a state of having eaten.'
9The dierence between the marking of the Objects in the following sentences relates to
the rst being non-human, and consequently unmarked, while the second is human, and con-
sequently marked as Accusative.
98 CHAPTER 4. WORD CLASSES

(107) pala ku-ma-t i

plantain eat-comp-pfpart be.(nonlocut)
`The plantain is in a state of having been eaten.'
Thus the use of the Resultative construction shows that, while 1- `go' is clearly
intransitive and pyan- `hit' is clearly transitive, ku- `eat' has two frames, and may
be intransitive or transitive, with the two Resultative sentences (106) and (107)
corresponding to the intransitive sentence (99) and the transitive sentence (101)
respectively.
Thus while ellipsis can cause some problems for resolving the number of
arguments of verbs, the Resultative construction can be used to establish the
valence of many of these verbs, and whether they have a single subcategorization
frame or multiple frames. Verbs with multiple frames, ambitransitive verbs, will
be discussed in section 6.2.

4.5.2 Zero-argument (impersonal) verbs

There is a very small closed subclass of impersonal verbs in Awa Pit. These verbs
refer to time, and have no arguments at all. The verbs are k1n- `dawn' and nash-
`afternoon, get late', and are both active.
(108) nash-m1z-i
be:afternoon-incep-nonlocut
V
`It is getting late.'

4.5.3 One-argument (intransitive) verbs

There are a wide variety of single-argument verbs in Awa Pit. All of these verbs
behave identically, in that their one argument is the Subject of the verb. The
verbs may be either active or stative.
4.5.3.1 Active intransitive verbs
The active intransitive verbs form an open subclass. These verbs cover meanings
such as movement (1- `go', p1l- `y'), bodily processes (az- `cry', kwata- `vomit'),
physical processes (iiN- `ripen', p1t- `rot'), mental or emotional processes (kazhpa-
`dream'), social activities (sal- `play', put- `sh'), and ways of speaking (kwiyan-
`yell', kutku- `tell a lie').
(109) iiN=na kwi-t1-zi
ame=top go:out-past-nonlocut
Subj V
`The ame went out.'
4.5. VERBS 99

kun be standing (cf. kun- `stand up')

pana be standing
p1la be hanging
sum be crouching/sitting (cf. sum- `crouch')
shaa be around (cf. shaa- `walk')
tala be lying
tansha be sitting
tu be somewhere (cf. tu- `lie down')
uz be sitting, be in a place permanently, live (cf. uz- `sit down')
Table 4.8: Postural/locational verbs in Awa Pit

4.5.3.2 Stative intransitive verbs

There are a small number of stative intransitive verbs in Awa Pit, the majority
of them postural/locational. There are a number of dierent postural/locational
verbs (see Table 4.8), a few of which are related, formally and semantically, to
active intransitive verbs. It is unclear what the meaning dierence between some
of these verbs is (for example, between kun `be standing' and pana `be standing').
The postural/locational verbs can be used with only a Subject argument,
in which case they indicate the posture of someone or something:
(110) Santos=na pana-y.
Santos=top be:standing-nonlocut
na=na tala-s
1sg.(nom)=top be:lying-locut
`Santos is standing. I am lying down.'
In addition to their use as postural verbs, these stative intransitives can
also be used to indicate the location of something, with the item in question being
the Subject, and the location indicated through the use of a locational adjunct
phrase:
(111) Demetrio=na banca=ta tansha-y
Demetrio=top bench=in be:sitting-nonlocut
Subj Loc V
`Demetrio is (sitting) on the bench.'
The postural verbs may also be used as auxiliary verbs, as in many other
languages, and this will be discussed in section 11.2.
It is worth pointing out that while nearly all of the postural verbs clearly
have only one obligatory argument, the status of tu `be somewhere' is not so clear.
It will be treated here as a two-argument stative verb, as while it appears to
syntactically parallel the other postural/locational verbs, it never occurs without
100 CHAPTER 4. WORD CLASSES

some locational phrase. That is, it is perhaps best treated as having not one
obligatory argument but two, a Subject and an oblique locational phrase.10
In addition to the postural/locational verbs, there are three other intrans-
itive stative verbs: wat `be good', waa `there is' and way `be lacking'.
(112) pala way-i
plantain lack-nonlocut
Subj V
`There are no plantains (plantains are lacking).'

4.5.4 Two-argument verbs

Verbs which have two obligatory arguments may be active and transitive, active
but not transitive, or stative and not transitive.
4.5.4.1 Transitive verbs
Awa Pit has a large open subclass of transitive verbs. These verbs are all active,
and have a Subject and an Object. They cover meanings such as physical impact
(pyan- `hit'), transformation (il- `roast', kit- `wash (body)'), construction (sa-
`make, construct', put- `plait, weave'), transport (iN- `carry', kwaa- `carry, wear'),
movement with respect to something (kwak- `cross', namna- `follow, catch up to'),
acts of perception or emotion (izh- `see', isha- `hate'), and social acts directed at
someone (mun- `baptize', t1tizh- `wait for').
(113) na=na Santos=ta namna-mtu-s
1sg.(nom)=top Santos=acc follow/catch:up-impf-locut
Subj Obj V
`I am following Santos.'
(114) uspa=na s1 pak-tu
3pl.(nom)=top bean harvest-impfpart
Subj Obj V
`They are harvesting beans.'
4.5.4.2 Two-argument stative verbs
There is a small closed subclass of stative verbs which take two arguments, a
Subject and a Copula Complement, distinguishing them from the transitive verbs,
which have an Object rather than a Copula Complement. These verbs are i `be',
ki `(not) be', ka `be permanently', m1j `have', waj `have (body part)', pyan `know
(someone or something)' and mwiN `be worth'.
(115) tunya nya=na kwashmayN i
rat meat=top tasty be.(nonlocut)
Subj CopComp V
`Rat meat is tasty.'
10 See section 4.5.5.2 for verbs which clearly have obligatory oblique arguments.
4.5. VERBS 101

(116) us=na ashaNpa m1j-i

3sg.(nom)=top woman have-nonlocut
Subj CopComp V
`He has a wife.'
(117) na=na Gloria pyan-is
1sg.(nom)=top Gloria know-locut
Subj CopComp V
`I know Gloria.'
Many of these verbs can also be used in a number of auxiliary constructions
(see section 11.2). MwiN `be worth' can also be used with no Copula Complement
present, in which case it means something closer to `be expensive'.
4.5.4.3 Two-argument active, non-transitive verbs
There are two very unusual two-argument verbs  nam- `become (change into)'
and paa- `become (develop into)'. These verbs are active, as they are used with
the Imperfective aspect marker. However all other two-argument active verbs
have a Subject and an Object, while these have a Subject and a Copula Com-
plement, similar to the two-argument stative verbs. This presumably occurs as
while the two verbs are, conceptually, active (in that they are not inherently im-
perfective, with the possibility of an instantaneous change), they are not at all
transitive (in the sense of Hopper & Thompson (1980)), with no transfer of ac-
tion from one participant to another; indeed, there is really only one participant
involved in a sentence involving nam- `become (change into)' or paa- `become
(develop into)'.
(118) Juan=na gobernador nam-t1-zi
Juan=top governor become-past-nonlocut
Subj CopComp V
`(When his father died,) Juan became governor.'

4.5.5 Three-argument verbs

There are three subclasses of verbs which take three arguments: the ditransit-
ive verbs, the three-argument verbs with an oblique argument, and the three-
argument stative verbs.
4.5.5.1 Ditransitive verbs
There is a fairly small, but probably open, subclass of ditransitive verbs in Awa
Pit, all of which are active. These verbs have a Subject, an Object, and a Second
Object. They cover three areas of meaning: they may be related to the concept
of giving (for example, kwin- `give', an- `pass'), taking (naka- `take away', sap-
`receive'), or asking (m1ma- `ask'). As well as a Subject, coding the agent (donor,
102 CHAPTER 4. WORD CLASSES

taker, or asker), these verbs subcategorize for an Object, coding the recipient of
the gift or query, or the source, and a Second Object, the theme.
(119) Camilo=na na-wa pala kwin-t1-s
Camilo=top 1sg-acc plantain give-past-locut:under
Subj Obj Obj2 V
`Camilo gave me a plantain.'
(120) na=na pashpa=ta t1 naka-t
1sg.(nom)=top child=acc stick take:away-sv
Subj Obj Obj2 V
kway-ta-w
drop-past-locut:subj
V
`I took the stick away from the child.'
4.5.5.2 Three-argument verbs with an oblique argument
There are a small number of active verbs which take a Subject, an Object, and
an oblique argument. The status of the oblique argument as an obligatory verb
complement rather than an adjunct can be established positionally  adjuncts
occur before Objects, while oblique complements are placed after a verb's Object
(see section 3.2).
Two verbs have been found which take an obligatory oblique locational
phrase: win- `put (in a place)' and nuk- `put (into something)'. It is important
to note that, like put and similar verbs in English (Andrews 1985:91), while these
two verbs require some locational phrase, the precise phrase-type is not determ-
ined by the verb  thus the location may be indicated using any semantically
appropriate postpositional phrase.
(121) na=na pala t1m=ki
1sg.(nom)=top plantain basket=at
Subj Obj location.complement
win-ta-w
put-past-locut:subj
V
`I put the plantains in the basket.'
The two other verbs in this subclass require an oblique phrase marked
with the postposition kasa `with'. These verbs are mazh- `change/trade [Object]
for/with [kasa-Oblique]' and pun- `ll [Object] with [kasa-Oblique]'.
(122) Santos=na wakata paas kuzhu=kasa mazh-t1-zi
Santos=top cow two pig=with change-past-nonlocut
Subj Obj oblique V
`Santos traded a cow for two pigs.'
4.5. VERBS 103

4.5.5.3 Three-argument stative verbs

Finally there is one three-argument stative verb, pwii `owe'. This verb clearly has
a Subject, the person who owes, but the grammatical relations of the other two
arguments are unclear. As this verb is stative, and involves no transformation,
it cannot be used in the Resultative construction, and there is no syntactic test
for which of its other two arguments, the person to whom something is owed
and what is owed, is an Object (if either of them are). Morphology is also not a
reliable guide in this case: what is owed is always unmarked, but as it is never
human this could either be a Copula Complement or an Object or Second Object.
Thus there appear to be two possibilities for the assignment of semantic roles to
grammatical relations: Subject (ower), Object (owee) and Second Object (thing
owed); or Subject (ower), Object (owee) and Copula Complement (thing owed).
While there appears to be no method of deciding between these two options, the
rst is perhaps most likely, as this assigment of roles is identical to the assignment
of the ditransitive (active) verbs, while no verbs have been found with a Copula
Complement in addition to an Object relation.
(123) na=na Santos=ta dos mil pwii
1sg.(nom)=top Santos=acc two thousand owe.(impfpart)
Subj Obj Obj2 V
ka-s
be:permanently-locut
Aux
`I owe Santos 2000 pesos.'

4.5.6 Compound verbs

As well as the simple, mono-lexemic, verbs discussed above, Awa Pit also has
compound verbs, clearly consisting of two phonological words. The second of
these is always formally identical to a simple verb, most commonly ki- `do, hap-
pen', while the rst may be either formally identical to another word in the
language, or not. The two-word status of these compound verbs and many other
aspects of their use will be discussed in section 6.3.3.
With the exception of a handful of the compound verbs, they are all
formed using ki, which as a simple verb means `do, happen'. While ki- `do' is a
transitive verb and ki- `happen' is an intransitive verb, the compound verb may
be impersonal (alu ki- `rain'), intransitive (kal ki- `work'), or transitive (wiya ki-
`ght'). The rst element in the compound may be identical to a noun or adjective
in the language, or not a separate lexical item at all, and the relationship between
the meaning of the noun or adjective and the compound verb may be clear or
more obscure, as can be seen from the examples of compound verbs given in
Table 4.9.
In the remainder of this thesis compound verbs will be treated as a single
lexical item, consisting of two phonological words and two grammatical words.
The second word is a verb, and acts in precisely the same way as any other active
104 CHAPTER 4. WORD CLASSES

alu N `rain' ki- `do' alu ki- `rain'

ii Adj `hot' ki- `do' ii ki- `be hot'
kal Adj `dicult' ki- `do' kal ki- `work'
yak N `envuelto (foodstu)' ki- `do' yak ki- `be hungry'
 ki- `do' ku ki- `dance'
 ki- `do' t1hj ki- `sow'
sula N `tooth' ku- `eat' sula ku- `bite'
p1nk1h N `paper' paa- `strike' p1nk1h paa- `write'
(kwinta V from Spanish `tell') kizh- `say' kwinta kizh- `tell a story'
Table 4.9: Examples of compound verbs
=akwa `because'
=kana `like'
=kasa `with'
=ki `at'
=kima `until'
=mal `loc'
=pa `poss'
=pa `in(approx)'
=patsa `like'
=ta `in'
=ta `acc'
Table 4.10: Postpositions

verb. The rst word is an unvarying item, occurring in the position between any
manner adverbial and the negative marker. Compound verbs will be glossed with
two glosses, but only one meaning, as: alu ki- [rain(1) rain(2)] `rain'.

4.6 Postpositions
Postpositions play a very important role in Awa Pit. As a language without a
case-sux system (except for the pronouns), Awa Pit relies very strongly on its
postpositions to indicate the relations which noun phrases may play in a sentence.
Many postpositions are also used following non-nite clauses, essentially creating
constructions similar to English subordinating conjunctions.
The Awa Pit postpositions are listed in Table 4.10, together with an ap-
proximate English gloss. These postpositions are never independent words,11 but
rather cliticize onto the preceding word, which is usually, but not always, the nal
11For a possible exception to this statement, see example (133) and associated discussion
below.
4.6. POSTPOSITIONS 105

word of an np or another postposition. Cliticized postpositions are quite common

in Amazonian languages to the east of Awa Pit (Doris Payne 1990b:220); and to
the west, Epena Pedee has a system of case marking through postpositions very
similar to Awa Pit, with the postpositions attaching to clauses as well as nps
(Harms 1994:6579). A similar situation holds in Ika, spoken in northern Colom-
bia, where locative postpositions can cliticize onto clauses (Frank 1990:102103).
The initial phoneme of the postpositions undergo the usual within-word
phonetic realizations, with the exception of those beginning with /p/, which
becomes /w/ after a vowel  as discussed in section 2.2.1, this is the case for all
morpheme-initial /p/ after a vowel. This is clear evidence that the postpositions
form one phonological word with the preceding element.
Grammatically, however, there is a dierent situation. While previous
analyses of Awa Pit have considered the postpositions to be nominal suxes
(Calvache Dueñas 1989, Obando Ordóñez 1992), there are clear reasons for ana-
lyzing them as separate grammatical words (clitics) rather than parts of the
preceding word (suxes), although not all types of evidence are found with all
postpositions.
First, many of the postpositions (all those which have some relation to
location) have been found following other postpositions. This is not so unusual
in those cases where the rst postposition is the genitive postposition pa (as
in example (124)), as sequences of two noun suxes have been found in other
languages when the rst is a genitive; but many also occur with one locational
postposition following another, as in example (125):
(124) Santos=pa=mal nya waa-y
Santos=poss=loc meat there:is-nonlocut
`There is meat at Santos's [place].'
(125) m1n=ta=kima=ma 1-mtu-s?
where=in=until=inter go-impf-locut
`Up to where (how far) are you going?'
This suggests that locational postpositions, at least, may not be noun suxes,
although analyses have been given which deal with case-stacking similar to this
in Australian languages (see Dench & Evans (1988) for a survey; Andrews (1996)
for an analysis of semantic case-stacking in an LFG framework).
Secondly, postpositions are found following an entire (non-nite) clause,
as well as following an np:12
(126) us ii-ta=ki=na, cruz
3sg.(nom) die-pfpart=at=top cross
pana=na
be:standing.(impfpart)=top
`At (the place where) he died stands a cross.'
12 See section 10.2.3.2 for many more examples.
106 CHAPTER 4. WORD CLASSES

(127) Demetrio kayl-na=kima, kal ki-n1-s

Demetrio return-inf=until, work(1) work(2)-fut-locut
`Until Demetrio returns, I will work.'
Postpositions may follow the Imperfective Participle, the Perfective Participle
and the Innitive. Clearly, as non-nite verbal forms, these are not nouns, and
hence the postpositions cannot be considered to be noun suxes.
There is also evidence that the postpositions go on the nal element of
an np, regardless of whether this is a noun or not. This evidence is perhaps less
strong, however it is required for the possessive marker pa and kasa `with', as these
have not been found following another postposition, nor following a clause. The
evidence for pa being a postposition rather than a sux is not strong, consisting
simply of sentences such as
(128) Carmen=na paynya nalpihsh=pa=ta
Carmen=top her brother=poss=in
ayna-ma-t1-zi
cook-comp-past-nonlocut
`Carmen cooked at her brother's [place].'
where the postposition pa has scope over the entire preceding np, paynya nalpihsh
`her brother'. The case for kasa `with' being a clitic postposition rather than a
nominal sux is much stronger. When speakers of Awa Pit include Spanish words
in an Awa Pit sentence, the word order of the phrase containing one or more
Spanish words may either follow Spanish or Awa Pit word order. In particular,
using a Spanish noun or adjective or both within an np will sometimes cause
the np word order to be nounadjective rather than adjectivenoun. In this case
the postposition follows the adjective (that is, the entire np) rather than being
attached to the noun:
(129) na=na cuchillo mocho=kasa=yN kuzhu
1sg.(nom)=top knife blunt=with=rest pig
nak-ma-ta-w
skin-comp-past-locut:subj
`I skinned the pig with a blunt knife.'
Here it is clear that kasa follows the entire np, and is thus clearly not a nominal
sux but rather a clitic to the entire np.
As a noun phrase may have an omitted head in Awa Pit, these post-
positions may also, of course, follow such things as descriptive adjectives and
possessive adjectives:
(130) katsa=wa kwizha
big=poss dog
`the big [person]'s dog'
(131) na=na ap=ta 1-mtu-s
1sg.(nom)=top my=in go-impf-locut
`I am going to my [home].'
4.7. TIME ADVERBS 107

(132) paynya=mal 1-t1-zi

his=loc go-past-nonlocut
`He went to his [house].'
I have no evidence of the postpositions being used except cliticized to a
preceding element. However Obando Ordóñez (1992:112) has a class of locative
deictics, ta and pa, the proximal and distal respectively.13 He does not associate
these in any way with the locative suxes ta and pa; and gives only one example
of use:14
(133) ta 1-mtu-s
over:there go-impf-locut
`I am going over there.'
Similar examples using ta and pa are found in Henriksen & Obando Ordóñez
(1985:91):
(134) pa shaa-n1-tu
over:there walk-prosp-hort.1sg
`I'm going to walk over there.'
It would seem best to treat these locative deictics simply as uses of the locative
postpositions ta and pa, with an ellipsed np. This analysis would explain the
parallelism between the locative deictics and the locative suxes. Given this
analysis, it seems clear that ta and pa, at least, cannot be treated as noun suxes,
but must be treated as separate words, as they can occur with the np ellipsed.
The evidence, then, suggests strongly that at least many of the elements
in Table 4.10 are separate words rather than noun suxes. As all words appear
to be alike in their behaviour (except for any semantic anomalies), all are treated
as separate words. Whether they are classed as `postpositions' or `case markers'
depends, to some extent, on one's theory of case. Here they will be treated as
postpositions  given the possibility of using one followed by another, and the
range of meanings expressed, this seems the most sensible analysis.

4.7 Time adverbs

Eleven time adverbs have been found in Awa Pit (see Table 4.11). Time adverbs
are only used as temporal adjuncts, which occur in the rst adjunct position in
a clause, after the Subject (see section 3.2). This distinguishes the time adverbs
from place adverbs (see section 4.8), manner adverbs (section 4.9) and other
adverb-like elements (section 4.10), which occur in other positions in the sentence;
and also from nouns referring to time, which may occur in the temporal adjunct
position, but may also be used as complements to a verb (see section 13.2.2).
13This could relate to a dialect dierence.
14The orthography, morpheme breaks and glosses in this example and the next have been
changed to t my analysis; the Awa Pit words and the translation are from the original sources.
108 CHAPTER 4. WORD CLASSES

Durative
k1ns1h `until dawn'
mansuh `all day'
Iterative
shil `every day'
Point time
amta `at night'
t1lawayN `in the early morning'
Relative time
kayas `early'
nashka `late'
anya `rst, before, earlier'
Continuative
mama `still'
Interrogative/negative
mizha(pa)ka `when?'
mizhuta `when?, never'
Table 4.11: Time adverbs

The two interrogative/negative time adverbs are discussed in sections 12.2.7; all
other time adverbs are discussed and exemplied in section 13.2.1.
There is one word which is analyzed as a time noun rather than a time
adverb, even though it may only occur in the temporal adjunct slot: kwizh=
`after, later'. See section 13.2.4 for a discussion and justication of this analysis.

4.8 Place adverbs

Locational adjuncts (see section 13.4) in Awa Pit are nearly always postpositional
phrases. There are four words which could potentially be analyzed as place
adverbs, but have been analyzed here as place nouns  aN= `here', uN= `there',
su= `there' and m1n= `where, nowhere'. See section 4.3.2 for discussion and
justication of this analysis, which parallels the analysis of kwizh= `after, later'
as a time noun rather than a time adverb.
However there is one remaining word which can be considered a place
adverb: akki `here'. This word is problematic in a number of ways.
The most likely source for akki `here' is the Spanish word aquí `here',
pronounced in the same way. However akki is not simply a clear loan word, as it
is integrated into the structure of Awa Pit. For example, the elative sux s (see
section 5.4.3) may be added to akki to form akki-s `from here'.
This use of akki with the elative sux is unusual, as other than with
this word, s is only used following postpositions. This suggests that akki should
4.9. MANNER ADVERBS 109

aynsuh `quickly'
aza `quickly'
impata `slowly'
ka `like this'
watsha `for sure, certainly'
manaz `again'
m1za `almost'
mizha `how'
shin `why'
wan `completely'
Table 4.12: Manner adverbs

be analyzed, not as a unit, but as a place noun ak= followed by the locative
postposition ki. Unfortunately there are two problems with this idea. First,
semantically, akki is used in cases where the postposition ki would not be appro-
priate, but rather one of the locative postpositions ta or pa should be used. But
more tellingly, if ak= were a place noun, it would be possible to combine it with
other postpositions to form, for example, ak=ta or ak=pa (see section 4.3.2).
However these words/postpositional phrases simply do not exist.
It seems then that akki `here' must be analyzed as a place adverb in Awa
Pit. While it was presumably borrowed from Spanish, it has fully integrated
itself into Awa Pit, probably because of the formal similarity of the nal segment
to the locational postposition ki, and also perhaps because of the similarity of
the initial segment to the place noun aN= (see section 4.3.2), which has the same
meaning.

4.9 Manner adverbs

There is a small class of manner adverbs in Awa Pit. These words, listed in
Table 4.12, may only occur in the manner adverbial position, between verbal
complements and the verb (see section 3.2). These words will be discussed in
greater detail in section 13.5; the only comment which will be made here is to
note that wan `completely' is formally identical to the adjective wan `all', but
semantically slightly dierent, as will be discussed in section 13.5.

4.10 Other adverb-like words

There are a number of other words which are in some ways similar to the manner
adverbs. Many of these are often found in almost the same position as the manner
adverbs, but there are dierences.
110 CHAPTER 4. WORD CLASSES

=na `Topic marker'

=miN `Restrictive marker'
=kas `Additive marker'
=ma `Interrogative marker'
=ma `Temporal marker'
=ka `Emphasis marker'
Table 4.13: Discourse particles

4.10.1 The negative marker

The negative marker shi `not' is one of the few elements which may appear
between the manner adverbs and the verb (the others are p1na `very' and the
rst part of a compound verb). In addition to this position, however, the negat-
ive marker may appear after verbs, adjectives, noun phrases and postpositional
phrases. When used sentence-nally, the negative marker may be inected for
person. See chapter 12 for further details.

4.10.2 Degree adverbs

There are two degree adverbs in Awa Pit, kwisha `very' and p1na `very'. Kwisha
may only occur before adjectives to modify them; p1na occurs before manner
and time adverbs, verbs, nouns and adjectives as a modier. See section 13.6 for
further details.

4.10.3 The comparative marker

The comparative marker an `more' may occur before manner and time adverbs,
adjectives and nouns to establish a comparative relationship between two things.
This is discussed fully in section 13.7.

4.11 Discourse particles

There are six discourse particles in Awa Pit (shown in Table 4.13), which cliti-
cize onto the preceding element of a clause, regardless of its syntactic status,
and without changing its syntactic possibilities. These discourse particles are
discussed in chapter 14. Note that this group of words is semantically diverse 
they are anlayzed together because of their formally similar behaviour.
4.12. INTERJECTIONS 111

4.12 Interjections
There are presumably a variety of interjections in Awa Pit, as there are in any
language. Unfortunately, only two are found in the data for this study  ashtash
`thank you' and watsha `clearly, obviously, that is true'. The latter is formally
identical to one of the manner adverbs, watsha `for sure, certainly'. The inter-
jections are the only words in Awa Pit which fall outside the level of sentence-
grammar, and so form complete utterances on their own, with no requirement
for a predicate.
It is perhaps interesting to note at this point that there are no words for
`yes' or `no' in Awa Pit. A positive answer is given by repeating the verb, in an
appropriate form; a negative answer by giving the verb negated in an appropriate
form.

4.13 Cross-classication of words

While the vast majority of words in Awa Pit belong to one word class or another,
there are a few words which are cross-classied: while formally identical and
semantically similar, there are actually two words involved, of dierent word
classes. It is important to point out that these are individual words which are
involved in cross-classication, not entire classes. For example, the temporal
adjunct slot can be lled by an unmarked time adverb or by an unmarked noun
with a temporal reference. As the latter group is based on a semantic feature,
and any noun dealing with time can be used in this slot, these words are treated
simply as nouns rather than being cross-classied as nouns and time adverbs 
it is simply a feature of Awa Pit that any noun with time reference can be used
in this way. In contrast, the word 1lapa can be used both as a noun meaning
`old man' or as an adjective meaning `old'. This contrasts with other nouns in
the same semantic eld, for example pashpa `child', which cannot be used as
adjectives as well. As this use in two word classes is particular to this one word,
it is treated as two separate words, one a noun and one an adjective.
As well as the various words discussed below which are cross-classied,
it is possible to change the word-class assignment of roots through derivational
processes. All productive derivational processes which have been found change
words into verbs (from nouns, adjectives and verbs), and will be discussed in
sections 6.3 and 6.4.
There are a number of nominals which are cross-classied within the sub-
classes of nominals. All nominal postmodiers are also nouns, and these were
discussed in section 4.3.6. Many of the relational nouns are cross-classied as
nouns (see section 4.3.5).
Verbs are also cross-classied between subclasses. There are two verbs
which can be both active and stative: shaa- `walk' and tu- `lie down' are active,
while shaa `be around' and tu `be somewhere' are stative. There are a variety
of ambitransitive verbs, those verbs which are both transitive and intransitive 
see section 6.2 for a discussion of these verbs.
112 CHAPTER 4. WORD CLASSES

A few words in Awa Pit are truly cross-classied between major classes.
The noun 1lapa `old man' is formally identical to the adjective 1lapa `old'; and
some relational nouns are formally indistinguishable from adjectives (see sec-
tion 4.3.5). There is one adjective wan `all' which is clearly related to the manner
adverb wan `completely'. And another manner adverb watsha `for sure, certainly'
is formally identical and semantically similar to the interjection watsha `clearly,
obviously, that is true'.

4.14 Borrowed words

Awa Pit is spoken in Colombia, which has Spanish as its national language. The
Awa region is not fully self-sucient, relying on interactions with the outside
world: for example, in the region where the data for this thesis were collected,
the majority of families have one or more members who go to town every week or
two. Equally, even within the Awa region, there are a large number of monolingual
speakers of Spanish, and in the data-collection region the everyday language of
interaction within the community is Spanish. Thus all speakers of Awa Pit are
also speakers of Spanish, to a greater or lesser degree. This has, naturally, led to
some borrowing from Spanish into Awa Pit.
There are a wide range of borrowed words, and these fall on a continuum
from permanent to temporary borrowings. While in part the integration of
a borrowed word depends on a speaker's Spanish competence, dierent words
said by the same speaker appear to be dierentially integrated into Awa Pit,
which accords with Mougeon & Beniak's (1989:307) comment that variation in
the integration of loans depends on individual words, as well as other factors.
There are a few words which appear to have been borrowed from Spanish,
but have fully and completely integrated in Awa Pit, both syntactically and
semantically. For example, all speakers use the words mayshti `machete', pishkatu
`sh' and kwinta kizh- `tell a story', although these are quite probably borrowed
from the Spanish words machete `machete', pescado `sh' and cuenta/contar
`tell' (kizh- is the Awa Pit word for `say'). What distinguishes these borrowings,
if they are borrowings, from others is that speakers appear to have no realization
that these are borrowings, and always pronounce them with Awa Pit phonology
and phonetics, not Spanish.
In contrast to these permanent borrowings, there are semi-permanent bor-
rowings. These words have clearly been borrowed, but have indeed acquired the
status of Awa Pit words, in that there is, in general, no traditional word to
describe them, and speakers will consistently use the same Spanish word for
the concept, even if Spanish has a variety of words available. The important
distinction between these borrowings and the previous set is that speakers are
aware that they have been borrowed, and while speakers who are not highly u-
ent in Spanish will tend to say them in a way which is allowable according to
the phonology of Awa Pit, speakers who are more uent in Spanish vary, saying
these words more according to the phonology of Awa Pit when speaking quickly,
and more according to Spanish phonology when speaking more slowly. For ex-
4.14. BORROWED WORDS 113

ample, for the concept of `plate', speakers will use a range of pronunciations from
the Spanish [plato], through [p1lato] or [platu], to [p1latu] and nally to [pw1latu],
which accords completely with Awa Pit phonetics for a phonologically well-formed
word p1lattu. Some of these semi-permanent borrowings have even shifted mean-
ing slightly within Awa Pit, while retaining their link to Spanish: for example,
the word [tumiNgu] [domiNgo] has clearly been borrowed from Spanish domingo


`Sunday', but in Spanish this can only refer to the day of the week, while in Awa
Pit, whether pronounced [tumiNgu] or [domiNgo], it may mean either `Sunday'
or `week'. Particularly frequent semi-permanent borrowings are the numerals
greater than four: the traditional Awa Pit counting system only reached ampata
`four', and all higher numerals have been borrowed from Spanish, but all speakers
are fully aware of this, as numerals are essential in their trade interactions with
non-Awa.
Finally there are temporary loans from Spanish in the speech of native
speakers of Awa Pit. As nouns and adjectives in Awa Pit are almost always
unmarked, it is very easy to simply borrow a Spanish noun or adjective and
place it in the appropriate position for an equivalent Awa Pit noun or adjective.
With its complex verb morphology, borrowing verbs into Awa Pit should be more
dicult  but speakers simply borrow the third-person, present, indicative form
of the Spanish verb, then use it as the rst, invariant, part of a compound verb,
and use the verb ki- `do' as the second part, which carries all the morphology.15
These loans appear to be simply one-o loans; a speaker may commonly use
a traditional Awa Pit word, but on one particular day borrow a word from
Spanish for the same concept. This process is presumably aided by the fact that
speakers in the region where this data was collected use Spanish as their daily
language of interaction, and consequently are constantly using Spanish words for
concepts, rather than Awa Pit words; hence the Spanish words spring to mind
more readily.
In the example sentences throughout this thesis, loans from Spanish are
indicated by a non-italic font.

15 See section 4.5.6 for further details on compound verbs.

Chapter 5
Noun phrases, postpositional
phrases and Copula Complements
5.1 Introduction
This chapter is concerned with various elements associated with the predicate
in Awa Pit, either complements, which are obligatory in a semantic sense, or
adjuncts, which are optional. These elements are nps, pps or adjectives, and
these constituents sometimes have complex internal structures, with nps being
embedded within pps within nps, for example.
The rst section deals with nps and their internal structure. Following
this, Copula Complements are examined, as these have slightly dierent possibil-
ities from other grammatical relations in terms of their constituency. The various
postpositions found in Awa Pit are then discussed, before a short description of
nps and pps in apposition.

5.2 Noun phrases

The head of an np is, naturally, a noun or pronoun, although under certain con-
ditions this head can be ellipsed (see 5.2.5). With the exception of the nominal
postmodiers (see 4.3.6), all other nominals discussed in section 4.3 can be used
as the head of an np: nouns, time and place nouns, personal pronouns, demon-
strative pronouns, and relational nouns (although there are strong restrictions on
all of these except nouns).
As well as a noun or pronoun head, an np has many other possibilit-
ies. Awa Pit has a productive clausal nominalization strategy, discussed in sec-
tion 10.5; however certain nominalizations have diverged from this pattern, and
apparently been lexicalized (section 5.2.1). While there is no true reexive con-
struction in Awa Pit, the techniques used by Awa Pit to reect the same meanings
are largely based on the selection of appropriate nps (section 5.2.2). Nominal
heads of nps can, of course, be modied, and the strategies for modication are
discussed in section 5.2.3; one particular modication strategy, used for posses-
sion, has a wide range of functions, and these are examined in section 5.2.4.
116 CHAPTER 5. NPS, PPS AND COPULA COMPLEMENTS

After a discussion of headless nps (5.2.5), the peculiarities of plural marking

are looked at, and the relationship between this and clausal nominalizations.

5.2.1 Lexical nominalizations

The verb sux mu (and its allomorph m, which occurs after a vowel) has a variety
of functions (section 7.2.8.4). In addition to these productive uses, however, there
are a number of words which appear to be lexicalized forms from a verb root plus
mu. These are all formed from intransitive verbs, and often an unpredictable
semantic element is present in the meaning of the noun. To use the terms of
Comrie & Thompson (1985), the nominalization involved may be an agentive
nominalization or an instrumental nominalization.
The agentive nominalization use of mu, where the resulting noun is a thing
which performs the action of the verb, is similar to the agentive relative clause
function of mu discussed in section 10.4. An example of this is the noun p1l-mu
`bat', related to the verb p1l- `y'.
The instrumental nominalization use of mu is quite distinct from all pro-
ductive functions of this morpheme which are either agentive or purposive. The
object denoted by the resulting noun is an item which is used in carrying out the
action denoted by the verb involved. Examples of this are pihshka-m `broom'
from pihshka- `sweep', and uz-mu `seat' from uz- `sit'.
There is one slightly problematic example of an instrumental nominal-
ization. While most speakers used the word mazha for `mirror', one speaker,
T, consistently produced matsuh izh-mu `face see-mu'. This is clearly an instru-
mental nominalization  `thing used for seeing [one's own] face'. Unlike all other
instrumental nominalizations, this is based on a transitive verb with its Object,
parallel to the productive agentive nominalizations and relative clauses using mu.
As no other examples of a productive use of instrumental nominalization have
been found, it is possible that matsuh izhmu `mirror' is in fact a xed expres-
sion, a relic from a time when mu was used to form productive instrumental
nominalizations.

5.2.2 Reexives
Awa Pit does not have a specic reexive construction; however the techniques
used to convey the notion of reexivity are more closely related to a nominal
reexive system than a verbal one such as is found in Imbabura Quechua (Cole
1985:134135).
Reexive situations are reported in a variety of ways in Awa Pit, with
two being the most common. With non-third person, there is always the option
of simply including separate Nominative and Accusative pronouns referring to
the same person, and the same structure is sometimes used for third person as
well, although then there is ambiguity between reexive and non-reexive; this
same sort of system, with the ambiguity present, is used in a few other South
American languages such as Pirahã (Everett 1986:215217). In this case, and
5.2. NOUN PHRASES 117

08>
91
PossAdj >
NP B@ DemAdj =CA (Quant) (DescrAdj)* Noun
<
!
>
: PP > ;
Where: NP = Noun phrase
PossAdj = Possessive adjective
DemAdj = Demonstrative adjective
PP = Postpositional phrase
Quant = Quantier
DescrAdj = Descriptive adjective

Table 5.1: Basic noun phrase structure

in the following one, where the verb inection would distinctly mark Locutor
Subject or Locutor Undergoer (see chapter 8), the Locutor Undergoer marking
is chosen.
More frequently, reexives are stated as sentences with an ellipsed Subject,
which can be understood to refer to the same participant as the Object (which
may also be ellipsed). In this case, an adjunct postpositional phrase ap miNpayN
`through my own idea', paynya miNpayN `through his/her own idea', and so forth
(section 5.4.4) is normally present:
(135) ap miN=pa=yN t1t kway-t1-s
my thought=in(approx)=rest cut drop-past-locut:under
`I got cut and it was my idea; I cut myself.'
(136) paynya miN=pa=yN payn-t1-t1-zi
his thought=in(approx)=rest hit-term-past-nonlocut
`He got hit, it was his idea; he hit himself.'
Note that the sentences are not strictly reexive, this is simply a possible in-
terpretation: the second sentence, for example, would also be appropriate in a
context in which he got into a ght with the intent of having someone hit him.
That is, the pp involving miN can also be used in non-reexive contexts. The
use of an adjunct to suggest a reexive context in sentences like these is also
found in Retuarã, where reexive clauses usually contain an adverbial phrase like
`knowingly' (Strom 1992:133134).

5.2.3 Modication of nouns

Head nouns in an np can often be modied, with the basic schema presented in
Table 5.1. However there are many restrictions and variations on this schema.
118 CHAPTER 5. NPS, PPS AND COPULA COMPLEMENTS

It does not include deverbal adjectives or purposive adjectives, modication by

nouns, or modication of adjectives. For negation of elements within an np, see
section 12.6.
Of the nominals which can be heads of nps, only the nouns can be modied
in this way. As was noted in section 4.3.3, personal pronouns can only be modied
by quantiers, and in this case the quantier appears after the pronoun rather
than before: au kutnya `we three'. Other nominals  time and place nouns,
demonstrative pronouns and relational nouns  cannot be modied at all.
Where it is semantically plausible, quantiers and descriptive adjectives
themselves can be modied by the degree adverbs kwisha and p1na or by the
comparative marker an, as can the head noun. These constructions are discussed
and exemplied in sections 13.6 and 13.7 respectively.
Not all postpositional phrases can occur within an np. Those which have
been found are possessive (discussed at length below), locational and similarity
pps.
Examples of a variety of nps with basic modication are:
(137) alizh kwizha
angry dog
DescrAdj N
`a erce dog'
(138) an k1h
this leaf
DemAdj N
`this leaf'
(139) ampata katsa t1l kuzhu
four big black pig
Quant DescrAdj DescrAdj Noun
`four big black pigs'
(140) ap kutnya kwizha
my three dog
PossAdj Quant Noun
`my three dogs'
(141) sun paas awa
that two person
DemAdj Quant Noun
`those two people'
(142) [ mesa=ta libro ] kwin-zha
[ table=in book ] give-imp.1obj
LocPP Noun
`Give me the book on the table.'
5.2. NOUN PHRASES 119

(143) [ Cumbal=ta-s awa ] a-mtu-y

[ Cumbal=in-from person ] come-impf-nonlocut
LocPP Noun
`The person from Cumbal is coming.'
(144) [ ap=kana kwizha ] sula kwa-t1-s
[ my=like dog ] bite(1) bite(2)-past-locut:under
SimilPP Noun
`The dog like mine bit me.'
The basic structure given above allows for indenitely long strings of ad-
jectives, in particular long strings of descriptive adjectives. In actual speech, a
maximum of two adjectives have been observed in any np. However in elicitation
sessions speakers were quite willing to accept longer strings, with strings of three
or four adjectives being perfectly allowable. This restriction on the use of adject-
ive strings does not appear to be a case of ungrammaticality, but rather that
speakers prefer simpler structures, and will reorganize sentences involving longer
strings into two or three shorter sentences. Because of the lack of spontaneous
use of strings of descriptive adjectives, it has not been possible to investigate the
relative order of dierent classes of such adjectives.
The only exception to adjectivenoun word order occurs occasionally when
Spanish adjectives are used, presumably as a result of interference from Spanish.
Within Spanish, adjectives follow the nouns they modify, and speakers sometimes
use this feature in Awa Pit, especially when the noun in question is also a loan:1
(145) na=na shihtt1 izquierdo=mal k1hshpizh
1sg.(nom)=top arm left=loc scratch
kway-ta-w
drop-past-locut:subj
`I scratched my left arm.'
(146) na=na cuchillo mocho=kasa=yN kuzhu
1sg.(nom)=top knife blunt=with=rest pig
nak-ma-ta-w
skin-comp-past-locut:subj
`I skinned the pig with a blunt knife.'
This phenomenon of nounadjective ordering with Spanish loan adjectives has
also been noted for the Salvadorian language Pipil, with a usual adjectivenoun
order (Campbell 1985:108).
5.2.3.1 Deverbal adjectives
In addition to lexical adjectives, Awa Pit has two classes of derived adjectives, the
deverbal adjectives (discussed here) and the purposive adjectivizations (the fol-
lowing section). These adjectives, which are derived from verbs, retain some
1 As elsewhere, loans are indicated here in non-italic.
120 CHAPTER 5. NPS, PPS AND COPULA COMPLEMENTS

verbal features  in particular, rather than occurring before the noun they
modify, they follow it. The two constructions have not been found after a noun
which has modiers before it also.
There are very strict semantic restrictions on the verbs which can be de-
verbalized: a deverbal adjective can only be formed from a verb which implies
a change of state and indicate the result of that change. Thus while a deverbal
adjective ii-ta `dead' can be formed on the verb ii- `die', there is no deverbal
adjective corresponding to az- `cry', as this verb does not signal a change of
state.
Deverbal adjectives can be formed on the basis of intransitive or transitive
verbs (there are no ditransitive verbs with appropriate meanings). The deverbal
adjective is formed through the use of the Perfective Participle sux, and gives
the meaning that the noun with which it is associated has, at some time in the
past, undergone the change of state indicated by the verb.
In the case of intransitive verbs, the deverbal adjective refers to the Subject
argument of the corresponding verb:
(147) wakata ii-t1-zi
cattle die-past-nonlocut
`The cow died.'
(148) wakata ii-ta
cattle die-pfpart
`a/the dead cow'
With transitive verbs, the deverbal adjective is associated with the noun which
has undergone the change of state marked by the verb  and for transitive verbs,
the Object is the argument which undergoes any change of state:
(149) na=na camisa pat-ta-w
1sg.(nom)=top shirt wash-past-locut:subj
`I washed the shirt.'
(150) camisa pat-ta
shirt wash-pfpart
`a/the clean shirt'
As can be seen from the above examples, there is a major dierence
between the distribution of (lexical) adjectives and deverbal adjectives. Within
the np, adjectives precede the noun which they modify; deverbal adjectives fol-
low the noun. In English it is reasonable to consider that passive participles used
as adjectives are indeed adjectives, derived at the level of the lexicon, since they
exhibit all the properties of adjectives (see Wasow 1977); however in Awa Pit the
deverbal adjectives are clearly quite separate from the lexical adjectives.
Not only are deverbal adjectives dierent from adjectives in terms of their
position in the noun phrase, there is another dierence between the two. Ad-
jectives in Awa Pit do not have complements; deverbal adjectives can bring
5.2. NOUN PHRASES 121

with them any optional adjuncts (pps) which the corresponding verb may have.
For example, the verb kutil- `skin' is a transitive verb, and can optionally be
accompanied by a pp indicating the method of skinning used:
(151) na=na iiN=kasa kuzhu kutil-ta-w
1sg.(nom)=top ame=with pig skin-past-locut:subj
`I skinned a pig with re.'
The corresponding deverbal adjective, kutilta `skinned', may likewise be accom-
panied by this pp:2
(152) kuzhu [ iiN=kasa kutil-ta ]=na kwashmayN
pig [ ame=with skin-pfpart ]=top tasty
i
be.(nonlocut)
`Pig skinned with re is tasty.'
Deverbal adjectives (and any associated pps) clearly form a constitutent
together with the noun they modify, as a single noun phrase. Any marking
associated with the noun phrase  either the topic marker or a postposition
indicating case  occurs not on the noun itself, but after the deverbal adjective,
as in the preceding example and the following ones:
(153) [ shap ayna-ta ]=na kwashmayN i
[ ripe:plantain cook-pfpart ]=top tasty be.(nonlocut)
`Cooked ripe plantain is tasty.'
(154) na=na [ awa ii-ta ]=ta
1sg.(nom)=top [ person die-pfpart ]=acc
izh-ta-w
see-past-locut:subj
`I saw the dead man.'
The noun itself cannot be marked by the topic marker or case marking, if it is
not the nal element in the noun phrase:
(155) *wakata=na ii-ta=na
cattle=top die-pfpart=top
This clearly indicates that the two words, the noun and the deverbal adjective,
are in the same np, rather than being associated in an appositional structure.
2Note that the English translation of this sentence is almost identical to the Awa Pit con-
struction, allowing for the dierence in positioning in the two languages of pps with respect to
the verb.
122 CHAPTER 5. NPS, PPS AND COPULA COMPLEMENTS

5.2.3.2 Purposive adjectivizations

The adjectivizer sux mu has a number of functions, and one of these is to
form a nominal modier from a transitive verb, indicating that the noun being
modied is (good) for Verb-ing or is to be Verb-ed. The noun is necessarily
something which could ll the Object slot of the mu-marked verb. When used
attributively rather than predicatively, the mu-marked verb follows the noun it
modies as deverbal adjectives do, rather than preceding it like lexical adject-
ives. The only examples of an attributive use of purposive adjectivizations which
have been found are together with the verb waa `there is', which is intransit-
ive. Consequently it is not clear where case marking postpositions would occur
with a noun modied by a purposive adjectivization. However, given the paral-
lels with the deverbal adjectives, it appears likely that the noun and a following
mu-marked verb form one noun phrase.
(156) [ chicha ku-m ] waa-y
[ chicha eat-adjzr ] there:is-nonlocut
`There's chicha (an alcoholic drink) to drink.'
(157) Santos=pa=mal [ wakata pay-nin-mu ]
Santos=poss=loc [ cattle buy-caus-adjzr ]
waa-y
there:is-nonlocut
`There are cattle for sale at Santos'.'

5.2.4 Possessive constructions

While the semantics of the dedicated modiers within the np  such as the
descriptive adjectives and the quantiers  is relatively straightforward, the
semantics of the possessive construction needs closer examination.
The possessive construction can be used to encode a wide range of func-
tions: locationobject, wholepart, materialobject, and meanings similar to
English complex prepositions, which are encoded through the use of relational
nouns. Many of these functions have been found cross-linguistically to be coded
similarly to possession,3 and indeed in Awa Pit some of the above functions can
also be marked in the same way as possession, either through the use of the
possessive postposition pa (see 5.4.8), or through the possessive adjectives.
In some cases, however, there is an alternation between a pp with pa, a
possessive adjective, and a bare noun modier, where the bare noun acting as a
modifer must occur directly before the noun it modies. The alternation between
these constructions can be seen to depend on humanness and referentiality.4 If
3See, for example, Ultan (1978b:2835).
4`Referential' here is not used to imply the logical idea of `existence in some world', but rather
is used in the sense Givón (1982:84) gives it, where a referential noun is one whose  specic
identity is important rather than its generic type membership  (emphasis in original). Hopper
& Thompson (1984:711) prefer the label `manipulable' to `referential'.
5.2. NOUN PHRASES 123

the modifying noun is human and referential, the postpositional construction is

used (or a possessive adjective, if the non-head entity is pronominal); otherwise
a bare noun modier is used:
(158) Santos=pa pimpul
Santos=poss leg
`Santos's leg'
(159) ap pimpul
my leg
`my leg'
(160) kwizha pimpul
dog leg
`the leg of (a/the) dog'
(161) awa=wa pit
person=poss language/tongue
`(the) person's tongue [most likely]; (the) person's way of speaking
[unlikely]'
(162) awa pit
person language/tongue
`the language of the people (Awa Pit) [most likely]; human tongues
[unlikely]'
Once this division between human, referential entities (expressed through
the postpositional or possessive adjective constructions) and other entities (ex-
pressed through a bare noun modier) has been drawn for those functions where
the constructions are in alternation, it is possible to see that this distinction un-
derlies all uses of the constructions, and in fact explains why some functions are
necessarily expressed by one construction or the other.
For example, the ideas of alienable possession and kinship relations neces-
sarily involve the use of the postpositional or possessive adjective constructions
(with the exception of plural possessive adjectives, see below), rather than a
bare noun modier. However this follows from the semantics: alienable posses-
sion and kinship are inherently characteristics of referential humans. In those
few cases where the idea of possession or relation are expanded to include higher
animates, Awa Pit personies the entity in question, treating it like a human:
(163) kwizha=wa pelota
dog=poss ball
`the dog's ball'
Equally, in those constructions which encode the relations of locationobject,
unitmass, materialobject and purposeobject, the rst entity is necessarily
either non-human or non-referential, and consequently the postpositional and
possessive adjective constructions cannot be used.
124 CHAPTER 5. NPS, PPS AND COPULA COMPLEMENTS

The plural possessive adjectives provide a slightly peculiar exception to

the above regularity. Singular possessive adjectives occur in the expected slot,
and descriptive adjectives may occur between the possessive adjective and the
head noun:
(164) ap katsa t1l kuzhu
my big black pig
`my big black pig'
However as noted in section 4.4.5 there are no special plural possessive adjective
forms. In order to translate a phrase such as our house, speakers of Awa Pit have
two options. The more common option is to simply use the singular possessive
adjective:
(165) ap yal
my house
`my house, our house'
The alternative, with no apparent dierence of meaning, is to use the plural
Subject pronoun, unchanged, in the bare noun modier slot:
(166) au yal
we house
`our house'
In this case no adjectives can intervene between the pronoun and the following
noun.
It could perhaps be maintained that the dierence between singular and
plural possessive adjective constructions is related to `individuation', in the
sense of Hopper & Thompson (1980:253). Pronominal referents which are human,
singular and referential  the more individuated nominals  take one position,
while those which are non-singular occur in a dierent position. In this case, how-
ever, a distinction would need to be drawn between the way the language treats
`individuation' in possession with pronouns (where singular and plural are dis-
tinguished by position) and nouns (where unmarked nouns with plural reference
are treated identically to unmarked nouns with singular reference). A distinction
would also need to be drawn between the treatment of possessive adjectives
(where entities which are singular, human and referential are distinguished from
others) and Accusative marking (where entities which are human and referential,
whether singular or plural, are distinguished from others).
On the whole it seems more sensible to simply consider that there is a
gap in the language, a lack of plural possessive adjectives, and that the speakers
make up for this gap either by using a singular possessive adjective or a plural
Subject pronoun in the bare noun modier position, rather than attempt to nd
a theoretical explanation.
In the following sections the various functions of the pa-postpositional
construction, the possessive adjective construction and the bare noun modier
5.2. NOUN PHRASES 125

construction will be examined. It is important to remember here that the con-

structions used to encode all the following functions are identical  the following
are simply a family of related meanings that are expressed in the same way.
5.2.4.1 Alienable possession and kinship relations
Probably the most obvious relationship which can hold between two entities is
that of alienable possession. In this relation the non-head nominal is necessarily
referential and human, as noted above, and consequently the construction neces-
sarily involves the postposition pa or a possessive adjective. Kinship relations
(see Table 4.1) are treated identically to alienable possession in Awa Pit; indeed,
within an np the possession of body parts is treated in the same way, but has
been treated separately here as a partwhole relation (see below), as in body part
possession the non-head noun is not necessarily referential or human.5 That is,
there is no distinction in Awa Pit between alienable and inalienable possession
within an np; dierent constructions exist for referential human possession and
others, regardless of the nature of the possession.
(167) Santos=pa kuzhu
Santos=poss pig
`Santos's pig'
(168) Carmen=pa ayshpihsh
Carmen=poss sister
`Carmen's sister'
(169) paynya cuchillo
his knife
`his knife'
(170) ap akkwa
my mother
`my mother'
5.2.4.2 Wholepart
Partwhole relations are encoded in the same way as possession also, with the
whole as the modier, and the part as the head noun. The modier whole may
be a referential human (in which case the construction with pa or a possessive
adjective is used), or a non-referential human or a non-human (encoded through
a bare noun modier):6
5 It should be noted that while alienable possession, kinship relations and possession of
body parts are treated identically within an np, there is a dierence in the coding of these
relations through verbs: alienable possession and kinship relations use the verb m1l `have',
while the possession of body parts is described using the verb waj `have (body part)'.
6It is interesting to note here that while Imbabura Quechua similarly encodes partwhole
relationships with either a genitive marker or a bare noun, the distinction in that language is
between animates and inanimates, rather than referential human and others (Cole 1985:117).
126 CHAPTER 5. NPS, PPS AND COPULA COMPLEMENTS

(171) Santos=pa sayl

Santos=poss arm
`Santos's arm'
(172) ap pimpul
my leg
`my leg'
(173) awa k1zpu
person head
`a human head'
(174) t1 aya
tree skin
`(tree) bark'
5.2.4.3 Relational nouns
As discussed in section 4.3.5, there is a special class of nouns in Awa Pit, the rela-
tional nouns, which are largely used to cover meanings for which many languages
have adpositions or complex adpositions. While the meaning of these is perhaps
not strictly that of a relation between two nominals, formally they are construc-
ted in exactly the same way as possessive constructions, with the relational noun
as head, and the other nominal modifying this  referential human nouns take
an enclitic pa, or a possessive adjective if pronominalized; other nouns are simply
placed before the relational noun. This distinguishes this Awa Pit construction
from a similar construction found in Imbabura Quechua where, except for chawpi
`between, among', the complex postpositions are simply placed after the noun
to which they refer, with no marking indicating the relationship even for animate
nouns (Cole 1985:120124).7
(175) Carmen=pa ma=mal
Carmen=poss side=loc
`beside Carmen'
(176) ap puns1h=mal
my back=loc
`behind me'
(177) yal kwash=pa
house above=in(approx)
`above the house'
Just as with other nps, the modier is not essential, and can be ellipsed
if it can be understood from context:
7Compare this with the English compound prepositions where, as Ultan (1978b:30) notes
in his discussion of locative-governed genitives, in my front is not equivalent to in front of me.
5.2. NOUN PHRASES 127

(178) kwizha kal=mal tu-y

dog inside=loc be:in:place-nonlocut
`The dog is inside [the house].'
This sentence was used while standing on the verandah of the house, as an ap-
propriate answer to `Where is the dog?'.
A list of the relational nouns which have been found in Awa Pit was given
in Table 4.4. However the precise uses of a few of these relational nouns require
more discussion: in particular, the dierences between pula/puta= `below' and
su= `under'; 1sal= `on top' and kwash= `above'; and ayuk= `inside' and kal=
`within'.
Pula/puta= has been glossed as `below' and su= as `under', and these
have clearly dierent uses. Pula/puta=8 is used to refer to any point which is
geographically lower than another, for example:
(179) yal pula=mal pii waa
house below=loc river there:is.(impfpart)
`There is a river below the house.'
(180) pijtam libro=na kwaNtam libro pula=mal
green/blue book=top red book below=loc
tu-y
be:in:place-nonlocut
`The green book is under the red book.'
On the other hand, su=, with its relation to the noun su `earth', means something
closer to `on the ground under', and could not be used in example (180), though
it could be used in example (179).
While 1sal= has been glossed `on top of, in the top of' and kwash= has
been glossed `above', it must be noted that they have overlapping uses. The latter
word, kwash=, can be used when one entity is strictly `above' another, with space
between them (sentence (181)), or when one entity is resting on another (sentence
(182)):
(181) yal kwash=mal p1l-tu
house above=loc y-impfpart
`[The bird] is ying above the house.'
(182) yal kwash=mal sum=na
house above=loc sit.(impfpart)=top
`[The bird] is sitting on top of the house.'
On the other hand 1sal= begins its range where kwash= stops, and can be used
anywhere from one entity resting on another surface (sentence (183)) down to a
situation where one entity is within the other, though still near the top (sentence
(184)):
8 The two forms pula= and puta= are from dierent dialects.
128 CHAPTER 5. NPS, PPS AND COPULA COMPLEMENTS

(183) mesa 1sal=ki=na libro tu-y

table top=at=top book be:in:place-nonlocut
`On the table there is a book.'
(184) shitshu=na t1 1sal=mal uz-i
bird=top tree top=loc sit-nonlocut
`The bird is sitting in the top of the tree.'
The exact distinction between kal= `within' and ayuk= `inside' is not
clear. Kal= is used for being inside a building or in a eld; on the other hand
ayuk= is the more normal word for being in something which is conceived of as
all around, but is also used to state that something is under a table.
5.2.4.4 Locationobject
The expression within one np of an object and its inherent location is expressed
using a bare noun modier  the location is the modier, and as a location can
never be a referential human, the postpositional construction and the possessive
adjective construction are never used. This construction is appropriate, for ex-
ample, when the Awa (`people') are required to specify themselves as a group of
people separate from other groups. Then they refer to themselves as:
(185) 1nkal awa
mountain people
`mountain people'
It is also extensively used for labelling various body parts. The word ash `hair/fur'
has a wide range of referents, and these can be made more precise with this
construction:
(186) k1zpu ash
head hair
`(head) hair'
(187) kasu ash
eye hair
`eyelash'
Similarly, some body-part terms also refer to other objects, and the use of a
locationobject construction causes them to refer specically to a body part:
(188) pu nul
male:genitals small:ball
`testicle'
5.2. NOUN PHRASES 129

5.2.4.5 Purposeobject
The bare noun modier construction can be used to give the meaning of an
object with a specic purpose. For example, a saw is a cultivated eld. In order
to indicate what the purpose of the eld is  that is, what is growing in it, or
what is to be grown in it  this construction is used:
(189) pala saw
plantain eld
`plantain eld'
Naturally the postpositional construction cannot be used, as the purpose is always
non-human.
5.2.4.6 Materialobject
It is often necessary to state what material an object is made from. In this case
the object is the head noun, and the material is a modier. The material is, of
course, non-human, and hence a bare noun modier is used. It does not matter
whether the object has been constructed by people or is naturally occurring:
(190) kwalt1 yal
gualte house
`house made from gualte (local wood)'
(191) uk 1za
stone ridge
`stone ridge'
5.2.4.7 Specicgeneral
There is one further construction which appears to use a bare noun modier,
which is the specicgeneric construction. Unfortunately, little data is available
on this construction.
While there do not appear to be any cases in Awa Pit where it is necessary
to use a specicgeneral construction, these constructions are sometimes used
with plants and animals. For example, kalputut `mouse sp.' is a particular variety
of tunya `mouse', and simply saying kalputut is sucient to identify this animal
 but often it will be referred to as kalputut tunya. Likewise rather than wisha
`white person', speakers will often say wisha awa `white:person person'.
These constructions could perhaps be considered as a form of noun clas-
sier, a classication device which accompanies the noun regardless of the pres-
ence of modiers. However noun classiers are commonly used anaphorically
(Craig forthcoming), and there is no evidence of this in Awa Pit, nor is there
evidence of the system extending beyond animals and plants.
In addition to the specicgeneral constructions, there are also similar
constructions involving types of trees. However with these constructions the
modier has fused with the word t1 `tree' to form one word. Thus for example
130 CHAPTER 5. NPS, PPS AND COPULA COMPLEMENTS

there is a particular variety of local tree called kwalt1, and bamboo is called
minant1.

5.2.5 Headless NPs

In section 4.4.1 it was noted that the head of an np can sometimes be ellipsed,
leaving only modiers in the np. Of course, this occurs only under restricted
discourse conditions  essentially, an entity must have been introduced into the
discourse rst, before it can be referred to by a headless np.
It is important to point out that a headless np is quite distinct from an
adjective used as a Copula Complement (section 5.3), and the two have quite
dierent properties. A Copula Complement can only consist of a single adjective
(plus a degree adverb or comparative marker), while a headless np can contain
a string of adjectives.
As well as appropriate discourse conditions, an np must have the right
sort of modier before its head can be ellipsed. The head can be ellipsed if
the modier is a possessive adjective, quantier, descriptive adjective or deverbal
adjective, or a possessive pp:
(192) na=na ap=ta 1-mtu-s
1sg.(nom)=top my=in go-impf-locut
`I am going to my [house].'
(193) na=na paas=ta izh-ta-w
1sg.(nom)=top two=acc see-past-locut:subj
`I saw two [people].'
(194) na=na katsa=ta=na payn-ta-w
1sg.(nom)=top big=acc=top hit-past-locut:subj
`I hit the big [person].'
(195) pat-ta=na watsal i
wash-pfpart=top beautiful be.(nonlocut)
`Washed [shirts] are beautiful.'
(196) Carmen=na paynya nalpihsh=pa=ta
Carmen=top her brother=poss=in
ayna-ma-t1-zi
cook-comp-past-nonlocut
`Carmen cooked at her brother's [house].'
As noted in section 4.4.1, the head of an np can never be ellipsed if the np has
a bare noun modier, presumably to avoid ambiguity in the status of this noun.
Purposive adjectives and non-possessive pps have also never been found without
a head (although they are used as Copula Complements). Finally, demonstrative
adjectives are never found without heads; the related demonstrative pronoun is
used instead.
5.2. NOUN PHRASES 131

5.2.6 Plural marking and nominalizations

With the exception of personal pronouns, nominals in Awa Pit do not have num-
ber as a grammatical category, and thus awa can indicate one person, a number
of people, or people as a group.9 However there is one sux, the Collective ac-
tion sux tuzpa, which can be attached to nouns to show something similar to
number. Rather than simply marking multiple referents, the sux marks that
the referents form a coherent group, similar to the collective plural in Tlingit
(Anderson 1985:174); additionally, the group has to have been acting together to
perform some process:
(197) ampu-tuzpa kal ki-n1-ma-t1
man-coll work(1) work(2)-prosp-comp-term
`Together the men went o to work.'
(198) kutnya ampu-tuzpa kal ki-mtu-y
three man-coll work(1) work(2)-impf-nonlocut
`The three men are working together.'
While the group marked by tuzpa must have been acting together in one
sense, the use of this sux as an equivalent of a reciprocal indicates that the
group need not have been acting cooperatively:
(199) uspa-tuzpa=yN wiya ki-a-ma-t1
3pl.(nom)-coll=rest ght(1) ght(2)-pl:subj-comp-term
`They were ghting among themselves.'
As this sux is used to indicate that a group were acting together, it
is restricted to appearing on Subjects, and can in fact only occur with verbs
indicating an action.10
The form tuzpa does not occur on headless nps. There is, however, another
form tuz with the same semantics used on adjectives in headless nps:
(200) kutnya-tuz kal ki-mtu-y
three-nmlzr.pl work(1) work(2)-impf-nonlocut
`The three [men] are working together.'
(201) yawa-tuz=ma kal ki-mtu-y?
how:many-nmlzr.pl=inter work(1) work(2)-impf-nonlocut
`How many are working together?'
The form tuz also occurs in clausal nominalizations in Awa Pit, where it
means `the ones who . . . ' and contrasts with the singular mika `the one who . . . '
9Number of human participants can be shown on the verb; see section 7.3.
10It would appear that Inga Quechua has a noun sux with a similar use, pura. Levinsohn
(1976:95) lists this form as one of several plural markers, translating it as `among', the usual
translation the Awa use for tuzpa. Unfortunately the details of the use of Inga pura remain
unclear.
132 CHAPTER 5. NPS, PPS AND COPULA COMPLEMENTS

(see section 10.5). The two uses, on headless nps and to nominalize a clause,
are presumably related historically at least, both being used to nominalize a
non-nominal element.
In fact, mika can also be used on adjectives as a nominalizer, although
this is rare:
(202) 1lapa-mika
old-nmlzr.sg
`the older one (of two brothers)'
However while tuz and mika are being treated as adjective suxes here, it would
be possible to consider that the adjective is, in fact, an entire non-nite copula
clause. This has not been done as tuz (but not mika) on adjectives appears to
be restricted to Subject position, unlike clausal nominalizations with tuz.

5.3 Copula Complements

As was mentioned in section 3.5.4, a Copula Complement may be an np or
a pp, which both occur as other grammatical relations also, or it may be an
adjective with or without a comparative marker or a degree adverb, and it may
also be a relative clause (see section 10.4). With the exception of a demonstrative
adjective (which is replaced by a demonstrative pronoun), any type of adjective
can be used as a Copula Complement, including deverbal adjectives and purposive
adjectivizations:
(203) an yal=na [ ap ]
this house=top [ my ]
Subj CopComp=PossAdj
ka-y
be:permanently-nonlocut
V
`This house is mine.'
(204) Laureano=wa pashu [ ampata=na ]
Laureano=poss daughter [ four=top ]
Subj CopComp=Quant
`Laureano has four daughters (Laureano's daughters are four).'
(205) [ kwayl ] i
[ bad ] be.(nonlocut)
CopComp=DescrAdj V
`It's bad.'
(206) ap yal=na [ an katsa ]
my house=top [ more big ]
Subj CopComp=Comp DescrAdj
`My house is bigger.'
5.4. POSTPOSITIONAL PHRASES 133

(207) an pyat1s [ k11-m ] ka-y

this sugar:cane [ mill-adjzr ] be:permanently-nonlocut
`This (type of) sugar cane is good for milling (it has lots of cane juice).'
(208) sun nya [ il-mu ] ka-y
that meat [ fry-adjzr ] be:permanently-nonlocut
`That meat is good for frying.'
There is an important dierence between nps as Copula Complements and
other nps: an np as a Copula Complement necessarily has a head. While this
may seem trivial, given that an adjective can be used as a Copula Complement
anyway, this rules out strings of adjectives in a Copula Complement.
An np or adjective being used as a Copula Complement can be followed
by one of the nominal postmodiers discussed in section 4.3.6: the diminutive
pashpa, the augmentative 1lapa, the male marker ampu or the female marker
ashaNpa.

5.4 Postpositional phrases

Postpositional phrases are formed in Awa Pit using a variety of postpositions (for
a full list, see Table 4.10). Postpositions have an important role in the language,
as they are often used to mark the relations which particular nps bear to the
predicate. In addition to pps being used as complements and adjuncts to verbs,
however, they may also be used within nps and other pps. Postpositions follow an
np or another pp, and cliticize onto the word which precedes them. In addition,
postpositions may be used following a clause (see section 10.2.3.2 for details).
At a theoretical level, the postpositions and postpositional phrases of Awa
Pit may be divided into a number of types. For example, in many theories of
grammar the Accusative marker would be treated as a case marker, a grammatical
item with no meaning, while other postpositions would be considered to be true
adpositions, full and complete words with their own semantic structure. In some
more recent work, however, this explicit dichotomy has been questioned (see,
for example, Kumashiro (1994), who discusses the case marker versus adposition
problem in Japanese).
It is clear that in Awa Pit there is some dierence between the Accusative
marker, which can show the grammatical relations of Object and Second Object
and marks a variety of semantic roles, and other postpositions, which do not
mark grammatical relations but rather explicitly mark semantic roles. However
at a formal morphosyntactic level all postpositions behave in the same fashion,
with the exception of suppletive forms being used for Object pronouns and pos-
sessive pronouns, and consequently after a brief discussion of the combination of
pronouns and postpositions, all postpositions will be treated in the same fashion.
134 CHAPTER 5. NPS, PPS AND COPULA COMPLEMENTS

5.4.1 Postpositions and pronouns

The expected combination of postposition plus pronoun does not always occur in
Awa Pit. Some postpositions  for example, all the locational postpositions 
are not used together with the personal pronouns for obvious semantic reasons,
however others are. With most postpositions, there is no distinction in the use
of the postpositions from their use with nouns:
(209) us=kasa 1-mtu-s
3sg=with go-impf-locut
`I am going with him/her.'
However expected combinations of the Accusative postposition or the possessive
postposition with pronouns do not occur.
The possessive marker pa is not used together with the rst, second and
third person personal pronouns (it is used together with the interrogative per-
sonal pronoun m1n `who'). Rather, there are special possessive adjectives instead
(see 4.4.5 for details). However, these forms are not merely suppletive forms,
but rather a quite separate item, as the possessive adjectives may be used pre-
dicatively, while a postpositional phrase of possession may not be used in this
way:
(210) an yal=na ap ka-y
this house=top my be:permanently-nonlocut
`This house is mine.'
(211) *an shutta Santos=pa i
this hat Santos=poss be.(nonlocut)
When a singular personal pronoun is being used as an Object or Second
Object, there are special suppletive Accusative pronoun forms, rather than simply
using the Nominative singular pronoun followed by the Accusative postposition.
In the case of plural personal pronouns, the Nominative form is used, together
with a special Accusative clitic, distinct from the Accusative postposition. For
further details and a full listing of these Accusative pronominal forms, see sec-
tion 4.3.3.

5.4.2 Ta: The Accusative marker

The postposition ta is used to indicate that the referential human noun phrase
it follows is acting in the grammatical relation of Object or Second Object. As
these grammatical relations are extensively discussed in section 3.5.3, the types
of verbs which take Object and Second Objects are discussed in sections 4.5.4
and 4.5.5, and examples are given in section 3.6, no further comments will be
made here.11
11It is interesting to note that a case-marker ta is found in Inga Quechua (Levinsohn 1976:96),
marking both accusative and a locative case.
5.4. POSTPOSITIONAL PHRASES 135

5.4.3 Ta: Locative/allative

While Obando Ordóñez (1992:107) lists ta as an allative, as with pa and mal
it is better considered as basically locative; it can be used to indicate direction
towards, but also position at, and when suxed with s, it is used for motion
away from. As a locative, its primary use appears to be to indicate location in a
container:
(212) shi=ta=ma comida kwin-ta-w?
what=in=inter food give-past-locut:subj
`In what [container] did you give [it] food?'
(213) wisha payl=ta=yN ayna-t ku-m
white:person pot=in=rest cook-sv eat-adjzr
`[We] cooked and ate in clay pots.'
However it is also used for location in a settlement or town; and is used with
an interrogative as the standard method of asking the location of something or
someone:
(214) profesor=na m1n=ta tu-y?
teacher=top where=in be:in:place-nonlocut
`Where is the teacher?'
(215) Nulpe Medio=ta tu-y, profesor=na
Nulpe Medio=in be:in:place-nonlocut teacher=top
`The teacher is in Nulpe Medio.'
(216) pueblo=ta=na esta waa-zi
town=in=top party there:is-nonlocut
`There was a party in the town.'
The postposition ta diers from the other locative/allative postpositions
in that it is used with a specic location  the preceding np must refer, at least
conceptually, to a quite specic, bounded area, such as a container or a town, in
which the action or participant is (or to which or from which someone is going). It
thus contrasts specically with pa, which refers to a general area around a point,
and this distinction appears to parallel the `specic locative' and the `general
locative' in Guambiano (Vásquez de Ruíz 1988:7273); ta also contrasts with
mal, which refers to a specic part of a larger entity (see examples in sections
5.4.4 and 5.4.5 respectively).
As well as its use as a locative, ta is used as an allative, with no change
in form. In this use as an allative, ta has the same constraint as it does when
functioning as a locative  the np referred to must be a specic, bound space:
(217) Chucun
es=ta Pasto=ta-s tazh-ta-w
Chucunés=in Pasto=in-from go:down-past-locut:subj
`I went (down) from Pasto to Chucunés.'
136 CHAPTER 5. NPS, PPS AND COPULA COMPLEMENTS

(218) mes-ayzhpa Ricaurte=ta puz-tu-s

month-each Ricaurte=in go:out-impf-locut
`Each month I go out to Ricaurte.'
While synchronically the Accusative marker ta and the locative ta are
distinct postpositions, it seems likely that diachronically they are related. The
development of markers of some sort of location into markers of accusative (espe-
cially with human nouns) is relatively common in languages of the world. It has
occurred in a fully documented way in the change from the Latin allative ad `to'
to the modern Spanish a, which is used both as an allative, an indirect object
marker, and a marker of human direct objects. Markers for human direct objects
and some form of locative are also identical in two of the languages related to Awa
Pit, Tsaqui (Moore 1966:98) and Cha'palaachi (Vittadello 1988:23), although
the forms in these languages, ka and nu respectively, are quite dierent.
Both the postposition ta and the postposition pa (see following section)
may be suxed by s. This sux converts these postpositions from locative or
allative to elative, showing motion away from the indicated point. The distinction
between ta-s and pa-s is as expected, with the former having a more precise
reading, the latter being more approximate.
(219) na=na Pasto=ta-s
1sg.(nom)=top Pasto=in-from
a-ma-ta-w=ma
come-comp-past-locut:subj=temp
`I came recently from Pasto.'
(220) na=na anshik=na Pueblo Viejo=ta-s
1sg.(nom)=top yesterday=top Pueblo Viejo=in-from
a-ta-w
come-past-locut:subj
`Yesterday I came from Pueblo Viejo.'
(221) na=na uN=pa-s kayl kway-ka=na,
1sg.(nom)=top there=in(approx)-from return drop-when=top
kal ki-mtu-an1-zi=ma
work(1) work(2)-impf-fut-nonlocut=temp
`When I return from there, [my friends] will already be working.'
As well as showing motion away from a place, the elative sux s can also
be used to indicate someone's origins, apparently only with ta and not with pa,
presumably because one is from a particular place rather than from somewhere
around a place:
(222) Cumbal=ta-s ka-y
Cumbal=in-from be:permanently-nonlocut
`They are from Cumbal.'
5.4. POSTPOSITIONAL PHRASES 137

(223) na=na 1nkal awa shi ki-s,

1sg.(nom)=top mountain person neg be.neg-locut
pueblo=ta-s i-s
town=in-from be-locut
`I'm not a mountain person (ie. Awa), I'm from town.'
Obando Ordóñez (1992:108) does not consider s to be a separate sux,
but rather considers ta and pa to be allative suxes, and separate suxes tas
and pas to be ablative. However there is evidence that s is in fact a separate
sux (apart from the repetition the alternative analysis requires). This comes
from the word akki, discussed in section 4.8. As well as following ta and pa, s
may appear on the word akki `here' (although not on the postposition ki `at' nor
on any other words):
(224) akki-s puz-tu-s
here-from go:out-impf-locut
`I am leaving (from) here.'

5.4.4 Pa: Locative/allative

As mentioned in the previous section, pa (or its allomorph wa after a vowel) can
be used as a locative or allative, where it has a less precise meaning than ta;
while ta refers to a specic, bounded location, pa refers to the region around a
point (and will thus be glossed `in(approx)'). Thus while aN=ta and uN=ta refer
to specic `here' and `there' points, with pa these are much more general:
(225) aN=pa awa su paa-ma-t1.
here=in(approx) person earth become-comp-term
uN=pa 1nkal i
there=in(approx) mountain be.(nonlocut)
`Around here (this side of the river) it's become cultivated. Over there
(the other side of the river) it's bush.'
Similarly the use of pa in the following example implies that the meat is not
being sold from Santos's house, but from somewhere in the vicinity.
(226) Santos=pa=wa nya waa-y
Santos=poss=in(approx) meat there:is-nonlocut
`There is meat (for sale) near Santos's.'
As was discussed and exemplied at the end of the previous section, pa may be
followed by the sux s, which converts it into an elative, and indicates motion
away from an approximate place.
As well as its use as a strict locative or allative, pa is used to indicate the
route along which something went:
138 CHAPTER 5. NPS, PPS AND COPULA COMPLEMENTS

(227) na=na Santos=ta izh-ta-w,

1sg.(nom)=top Santos=acc see-past-locut:subj
mii=wa shaa-zi
path=in(approx) walk-nonlocut
`I saw Santos, he was walking along the path.'
The postposition pa is also used to indicate the means of transportation
by which someone has travelled:
(228) na=na Pasto=ta-s carro=wa=yN
1sg.(nom)=top Pasto=in-from vehicle=in(approx)=rest
a-ta-w
come-past-locut:subj
`I came down from Pasto by bus.'
(229) pueblo=ta mitt1=wa=yN 1-ma-ta-w
town=in foot=in(approx)=rest go-comp-past-locut:subj
`I went to town by foot.'
There also seems to be a somewhat metaphorical extension of the previous
use. This is the use of pa in forming a pp with miN `thought' and a possessive
adjective, indicating that someone did something intentionally. This pp occurs
frequently in sentences with reexive-like meanings (see section 5.2.2 for other
examples):
(230) ap miN=pa=yN t1t-ma-t
my thought=in(approx)=rest cut-comp-pfpart
`I got cut, it was my idea; I cut myself.'
This pp is not restricted to reexives, however:
(231) ap miN=pa=yN shihtt1 kit-ta-w
my thought=in(approx)=rest hand wash-past-locut:subj
`I washed my hands because I wanted to.'
Pa is also used to indicate on what date, or in which month or season
something happens. These words are all borrowed concepts; traditional `seasons'
are clausal  `when it rains', `when it dries out'  and pa is not used in this
construction. Days of the week (also borrowed) are used with no postposition.
(232) siete de julio=wa=na=ma akki
the seventh of July=in(approx)=top=temp here
tu-an1-s
be:in:place-fut-locut
`I will be here on the 7th of July.'
5.4. POSTPOSITIONAL PHRASES 139

(233) diciembre=wa=na Pueblo Viejo mii=na pul=miN

December=in(approx)=top Pueblo Viejo path=top dry=rest
a-zi
be-nonlocut
`In December, the Pueblo Viejo path was dry.'
(234) verano=wa=na ii ki-mtu-s
summer=in(approx)=top be:hot(1) be:hot(2)-impf-locut
`In summer it is hot (to us).'
The postposition pa has also very occasionally be found marking an Ob-
ject. For examples of this, and discussion, see section 3.5.3.1.

5.4.5 Mal: Locative/allative

The nal locative/allative marker in Awa Pit is mal. Obando Ordóñez (1992:108)
considers that this is an allative marker, with a meaning of the nal terminus a
process reaches. However, like ta and pa, in my data this marker may be used
as a locative also. While ta indicates a specic location, and pa is more general
in location, mal appears to mark a location within a larger location. Thus it is
used in sentences such as:
(235) Dolores Pueblo Viejo=mal tu=ma
Dolores Pueblo Viejo=loc be:in:place.(impfpart)=inter
ka ki?
be:permanently.(impfpart) Q.(nonlocut)
`Is Dolores in Pueblo Viejo?'
(236) mitt1=mal aympi tuk-tu
foot=loc blood suck-impfpart
`[Mosquitos] sucked blood on my feet.'
In the rst example here, the question is really `Is Dolores at the school?' (she
is the teacher, and lives and works there), that is, is she at a particular point in
the (very diuse) settlement of Pueblo Viejo. The second example likewise states
that mosquitos have bitten parts of my feet, not my entire feet.
Similarly mal is used as an allative under the same conditions, where the
goal is in fact a part of the np. It is commonly used for large areas, such as
countries, where one presumably does not go to the entire country, but an area
within it:
(237) Ecuador=mal 1-ta-w=ma kutnya año
Ecuador=loc go-past-locut:subj=temp three year
paa-ma-t1
become-comp-term
`I went to Ecuador three years ago.'
140 CHAPTER 5. NPS, PPS AND COPULA COMPLEMENTS

(238) ap su=mal 1-shi-mtu-s=ma

my earth=loc go-desid-impf-locut=temp
`I want to go to my country.'

5.4.6 Ki: Locative

Unlike the preceding three postpositions, ki is used only as a locative, never as
an allative nor as a temporal marker. It corresponds quite closely to the English
preposition `at' (and is glossed as such in examples), and describes a location in
terms of position with respect to another point, rather than in terms of being in
another point.12 Thus for example:
(239) paas pala kwal=ki pana-y
two plantain trunk=at be:standing-nonlocut
`The two [people] are standing at (beside) the trunk of the plantain
tree.'
(240) Isabel=na profesora nueva Pialap
=ki tu-y
Isabel=top teacher new Pialapí=at be:in:place-nonlocut
`Isabel is the new teacher at Pialapí.'
Because it describes location in terms of another point, it is frequently used with
clauses to indicate `at the place where':13
(241) us ii-ta=ki=na, cruz
3sg.(nom) die-pfpart=at=top cross
pana=na
be:standing.(impfpart)=top
`At (the place where) he died stands a cross.'
In addition to the locative use of ki, it is also used with monetary amounts
to state the price of goods:
(242) yawa=ki=ma pay-ta-w panela=na?
how:much=at=inter buy-past-locut:subj panela=top
`How much did you buy the panela for?'
(243) doce mil=ki pay-ta-w
twelve thousand=at buy-past-locut:subj
`I bought it for twelve thousand [pesos].'
At rst glance the postposition ki also seems to appear in the Awa Pit word akki `here'.
12
However, as explained in section 4.8, this word is better treated as an unanalyzable whole.
13See section 10.2.3.2 for further discussion.
5.4. POSTPOSITIONAL PHRASES 141

5.4.7 Kima: `Until'

The nal local-related postposition in Awa Pit is kima `until'.14 As a local post-
position, it indicates a limit on a distance, and follows the postpositions ta or
pa:
(244) m1n=ta=kima=ma 1-mtu-s?
where=in=until=inter go-impf-locut
`On to where are you going?'
(245) pueblo=ta=kima 1-ma-ta-w
town=in=until go-comp-past-locut:subj
`I am going as far as the town.'
As well as being used as a locational postposition, kima is also used in
relation to time. When it is used for time, kima directly follows a measurement
of time, and indicates that this is a limit of time, the point up to when something
happened or will happen, as in the rst two examples below. This construction
is very common, as it occurs in the normal parting expression, given in sentence
(248). It is also used following non-nite clauses (see section 10.2.3.2), where the
time of the action in the non-nite clause gives a time limit for the action in the
matrix clause, as in sentence (249).
(246) ma=kima kwinta kizh-tu-s
now=until tell(1) tell(2)-impf-locut
`I am talking until now.'
(247) noviembre=kima ma-mtu-s mama
November=until stay-impf-locut still
`I am still going to be here until November.'
(248) t1lawa=kima
tomorrow=until
`Goodbye (until tomorrow).'
(249) Demetrio kayl-na=kima kal ki-n1-s
Demetrio return-inf=until work(1) work(2)-fut-locut
`Until Demetrio returns, I will work.'

5.4.8 Pa: Possessive

The possessive postposition pa (wa after a vowel) occurs following referential
human noun phrases. For precise details about when it occurs and its func-
tions, which include possession, partwhole relationships, and relational nouns,
see section 5.2.4.
14The Awa Pit form kima is quite similar to the Inga Quechua postposition kama `until' given
in Levinsohn (1976:96).
142 CHAPTER 5. NPS, PPS AND COPULA COMPLEMENTS

(250) na=na Santos=kasa Demetrio=wa yal

1sg.(nom)=top Santos=with Demetrio=poss house
1-ta-w
go-past-locut:subj
`I went with Santos to Demetrio's house.'
(251) m1n=pa shutta=ma ka-y?
who=poss hat=inter be:permanently-nonlocut
`Whose hat is this?'

5.4.9 Kasa: `With'

Like many languages, such as Imbabura Quechua (Cole 1985:114) and Ika (Frank
1990:3738), but unlike three-quarters of the languages of the Americas in Stolz's
(1996) sample, Awa Pit has a single marker indicating both comitative and in-
strumental functions, kasa.
With animate nps, kasa marks a person who accompanies the Subject:
(252) m1n=kasa=ma 1-mtu-s? anya=kasa 1-mtu-s
who=with=inter go-impf-locut brother=with go-impf-locut
`Who are you going with? I'm going with my brother.'
(253) na=na Santos=kasa Demetrio=kasa miimal
1sg.(nom)=top Santos=with Demetrio=with Chucunés
1-ta-w
go-past-locut:subj
`I went to Chucunés with Santos and Demetrio.'
As can be seen from the last example, Awa Pit often uses kasa where other
languages might use a word similar to English and. However it is clear that kasa
means `with' rather than `and', from sentences such as:
(254) Jos
e=na Pedro=kasa shi 1-ma-y
José=top Pedro=with neg go-neg-nonlocut
`José didn't go with Pedro (Pedro went).'
If kasa were conjoining two noun phrases, we would expect the previous sentence
to mean `José and Pedro did not go'. However the negated verb phrase applies
only to José, and hence it is clear that the postpositional phrase with kasa is
comitative rather than conjoined.
While with animate nps kasa is used as a comitative marker, with non-
animates, it expresses the idea of an instrument:
(255) na=na mantel=kasa mesa kihja-ta-w
1sg.(nom)=top cloth=with table wipe-past-locut:subj
`I cleaned the table with a cloth.'
5.4. POSTPOSITIONAL PHRASES 143

(256) anya=na awa=na shi=kasa=ma payna

before=top person=top what=with=inter deer
nak-tu?
skin-impfpart
`What did people skin deers with before?'
(257) kuzhu iiN=kasa kutil-ta=na kwashmayN
pig ame=with skin-pfpart=top tasty
`Pig skinned using re is tasty.'

5.4.10 Akwa: `Because'

The postposition akwa expresses a reason for something having happened. It is
most commonly found used with a clausal complement (see section 10.2.3.2):
(258) mama ish-tu=akwa 1-ta-w
mother sick-impfpart=because go-past-locut:subj
`I went because [my] mother was sick.'
Akwa is slightly dierent, semantically, from other postpositions in that its com-
plement must semantically be a proposition, and cannot be an entity. However
structurally it is identical, in that the concept of an action may be expressed in
an np:
(259) ap kwankwa ii-ma-t1.
my grandmother die-comp-term
suna=akwa yal mazh-ma-t i-s
that=because house change-comp-pfpart be-locut
`My grandmother died. We moved because of that.'

5.4.11 Patsa: `Like'

The exact status of the postposition patsa (and its variant watsa which follows
vowels) is unclear. It does not have a high frequency, and in all cases where
patsa was found it could be replaced by kana, though the reverse is not true.15
It expresses an equative construction, with the precise similarity being expressed
through the predicate or following adjective. Unlike kana, which is much more
general in use, patsa seems to be used to indicate ideas like `as big as', `as much
as', `as bad as', and so on.
(260) an pashpa ampu=watsa ku-mtu
this child man=like eat-impfpart
`This child eats like (as much as) a man.'
15 See the following section for details of this latter word.
144 CHAPTER 5. NPS, PPS AND COPULA COMPLEMENTS

(261) suna p11=watsa pya-m

that shit=like smell-adjzr
`That smells like shit.'
(262) na=watsa=yN katsa i
1sg=like=rest big be.(nonlocut)
`He is just as big as me.'

5.4.12 Kana: Semblative

Awa Pit has a semblative postposition, kana. This postposition describes an
object as being like something else:
(263) shitshu=kana kwiyan-tu=na
bird=like cry-impfpart=top
`She cries like a bird.'
(264) na=na an=kana ki-mtu-ata-w
1sg.(nom)=top this=like do-impf-past-locut:subj
`I did (like) this [demonstrating with hands].'
(265) na=na sun=kana=yN kizh-ta-w
1sg.(nom)=top that=like=rest say-past-locut:subj
`I said the same as him.'
Kana can also be used after a locative, or after a non-nite clause:
(266) na=na Pasto=ta=kana=yN tu=na
1sg.(nom)=top Pasto=in=like=rest be.(impfpart)=top
`I live (in a place) like Pasto.'
(267) shaa-t=kana=yN mitt1 ish-tu-s
walk-pfpart=like=rest foot hurt-impf-locut
`My feet hurt as though I had walked.'
As in many languages, the exact status of the semblative marker is dicult
to determine. In the above cases it appears with the same distribution as other
postpositions. However it can also be used in a slightly dierent construction,
although few examples have been found. Here it is used with a non-nite clause,
as before, but kana itself is marked for person:
(268) paas akkwihsh m1j kana-s
two mother have.(impfpart) like-locut
`It's like I have two mothers.'
The precise analysis of this sentence (and others like it) is unclear. One possibility
is that kana is both a postposition and a verb. However it is also possible, given
the hypothesized origin of the person markers and their occurrence on other non-
verbal elements (see 8.4.1), that kana is simply a postposition, but can be marked
if it is sentence-nal, acting as a Copula Complement.
5.5. NPS AND PPS IN APPOSITION 145

5.5 NPs and PPs in apposition

As was discussed in section 3.2, there are two positions in Awa Pit which are, in
some senses, outside the clause, the sentence-initial and sentence-nal positions.
While these two positions can be lled by a variety of elements which appear
there rather than in their expected position in the clause for a variety of reasons,
or with an external topic (see section 14.2), it is also possible to ll these positions
with what appears to be a headed or headless np or a pp referring to the same
entity as another np or pp within the clause, as occurs with the extraposed katsa
and Australiatas in the following sentences:
(269) ampu pyan-ta-w, katsa
man hit-past-locut:subj big
`I hit the man, the big one.'
(270) Australia=ta-s, uN=ta-s i-s
Australia=in-from there=in-from be-locut
`I am from there, from Australia.'
There are two potential analyses of sentences such as these. The Awa Pit
sentences could be similar to the English translations or phrase juxtaposition in
Macushi (Abbott 1991:3033) or Apalai (Koehn & Koehn 1986:3739), where two
noun phrases or postpositional phrases are juxtaposed, with one placed before
or after the remainder of the clause; in this case there are indeed two separate
nps or pps referring to the same entity. Alternatively, these constructions could
be similar to those in languages like Fox, where elements such as nps may be
discontinous, with some elements of an np appearing in the usual position, other
elements of the same np appearing after the verb (Dahlstrom 1987).
However in Awa Pit, unlike in Fox, there are reasons for considering that
these sentences do, in fact, contain two separate nps or pps. In Awa Pit, unlike
in Fox, nps are only marked once for items such as locative markers:
(271) an(*=ta) yal=ta
this(*=in) house=in
`to this house'
Consequently the elements Australia and uN= in sentence (270) above must be-
long to dierent nps (and hence pps) as the locative marking occurs on each of
them.
Similarly, the Topic marker can occur only once on each np in Awa Pit,
not separately on each element:
(272) katsa(*=na) ampu=na
big(*=top) man=top
`the big man'
However when the two words are separated, as in example (269) above, both
elements can occur with the Topic marker, suggesting two separate nps:
146 CHAPTER 5. NPS, PPS AND COPULA COMPLEMENTS

(273) ampu=na pyan-ta-w, katsa=na

man=top hit-past-locut:subj big=top
`I hit the man, the big one.'
There is also a suggestion of two nps from the translations into Spanish
which speakers give for examples such as (269): yo pegué al hombre, al grande.
The words are separated, with the post-verbal section being placed sentence-
nally after a pause, and, if it is an adjective, it is `nominalized' through the use of
the denite article. This construction is odd in Spanish, except in `afterthought'
constructions.
Thus it is clear that in Awa Pit sentences of the type given in (269) and
(270) have two nps or pps in apposition, rather than one discontinous np or pp.
In many cases the alternative analyses of these constructions is unimportant, in
that the same result would be obtained either way, however the distinction does
become important in section 10.5.
Chapter 6
Verb derivation and valency
changes
6.1 Introduction
As has been noted several times in the preceding chapters, Awa Pit is a strongly
verb-oriented language, with a great deal of information being expressed in the
verb, and the majority of morphology being verbal morphology. The following
four chapters concentrate exclusively on verbs and verbal morphology.
In chapter 4, and more specically section 4.5, the verb roots of Awa Pit
were classied into a variety of classes, depending upon the distinctions between
active and stative verb roots, between simple and compound verb roots, and
between verb roots of various valencies. The last distinction is of particular
importance for this chapter  while verb roots may have particular argument
frames, there are some derivational morphological processes which can alter the
argument structure of verbs, deriving a verb stem of dierent valency from the
verb root. Inectional axes may then be added to the derived verb stem.
This chapter deals with two areas of the grammar of Awa Pit which are
theoretically separate; however they are dealt with jointly, as they are deeply
intertwined. One is the various derivational processes which verb roots (and
sometimes other elements) may undergo to produce verb stems; the other are the
various valency relationships which pairs of verbs in Awa Pit may bear to each
other.
There are three sections in this chapter, each dealing with a slightly dier-
ent aspect of valency or verb derivation: (1) ambitransitivity, (2) non-productive
derivation (including compounding) and (3) productive derivation. The rst two
areas contrast with the third, in that the former are extremely non-productive,
and in fact the pairs of verbs discussed in these sections must both be in the lex-
icon; the latter, in contrast, consists of highly productive derivations. The rst
of the three sections, that of ambitransitivity, deals with pairs of verbs which
are formally identical, while one of the pairs of verbs discussed in the other two
sections has more phonological material than the other.
Before continuing onto these three interrelated areas, it is necessary to
148 CHAPTER 6. VERB DERIVATION AND VALENCY CHANGES

point out some processes which could be mistaken for valency changes, but in fact
are not. The rst of these is the phenomenon of ellipsis, discussed in section 3.2.
As both indenite and denite ellipsis are possible in Awa Pit, a verb often
appears with fewer arguments than its subcategorization frame states it to have.
As pointed out in section 4.5.1, this is purely a process of suface ellipsis, and
does not change the valency of the verb in question.
The second process which could be considered as a form of valency chan-
ging is the apparent addition of an Object argument to an intransitive or cop-
ula sentence through cross-referencing on the verb. In fact, however, as will
be shown in chapter 8, the verbal suxes indicating person are not necessarily
cross-referencing any grammatical relation in a sentence, and although a verb in
an intransitive sentence may have a sux indicating that an entity other than
the Subject is involved in the action, this other entity cannot actually appear
as an argument of the verb  that is, although reference is made to this extra
entity within the verb, the verb still only has the usual number of arguments.
In addition to these apparent valency-changing processes, there are several
phenomena related to valency changing and derivation which will be discussed
elsewhere: the Resultative construction, adjectivizations, and nominalizations.
In none of the cases is a verb stem formed, and hence they will only be briey
outlined here.
The Resultative construction, discussed in section 11.2.2.2, is a syntactic
construction which can be valency reducing, causing a transitive verb to obligat-
orily appear with only one argument. It is similar in syntactic eect to a passive,
although like many South American languages (Derbyshire & Pullum 1986b:19)
Awa Pit does not have a passive. Related to this are the deverbal adjectives,
where a (slightly unusual) adjective may be formed from a verb indicating a
change of state; as discussed in section 5.2.3.1, the formation of these adjectives
may involve the loss of a (conceptual) argument. The sux mu is involved in
a variety of adjectivizations and nominalizations, some productive and others
lexicalized, as discussed in sections 5.2.1, 5.2.3.2 and 10.4. And nally there are
agentive nominalizations, formed with the clitic mika, where an entire clause is
nominalized, as discussed in section 10.5. In many constructions with mu, and
all with mika, one of the arguments of the verb root is obligatorily missing.
In addition to the derivations discussed below, verbal number marking is
perhaps best considered as a derivational process in Awa Pit, although this is not
entirely clear. It is discussed in section 7.3, however, together with arguments
for treating it as either derivational or inectional.

6.2 Ambitransitive verbs

There are a small number of ambitransitive verb roots, or verb roots with two
argument frames, in Awa Pit. These verb roots are cross-classied as both trans-
itive and intransitive, and are of two types, depending on whether the Subject in
the intransitive root corresponds to the Subject or the Object of the transitive
root.
6.2. AMBITRANSITIVE VERBS 149

These pairs of verbs, all of which are active, are identical in form; that
is, there is no morphological marking associated with the change in valency. It
has unfortunately been impossible to determine for these pairs whether native
speakers consider one valency or the other to be more basic (cf. Dixon 1991:286-
293 for ambitransitive verbs in English).

6.2.1 S=A ambitransitive verb roots

Only one ambitransitive verb root has been found where the Subject in its in-
transitive use (s) bears the same semantic role as the Subject in its transitive use
(a): ku- `eat'. In Awa Pit the following sentences are both possible:1
(274) na=na pala kwa-ta-w
1sg.(nom)=top plantain eat-past-locut:subj
`I ate a plantain.'
(275) na=na kwa-ta-w
1sg.(nom)=top eat-past-locut:subj
`I ate.'
Given the common use of ellipsis, it could be suggested that ku- is a transitive verb
only, and that in the second sentence it is simply used with an ellipsed, indenite
Object. However, as was shown in section 4.5.1, this is clearly a case of the verb
root having two dierent subcategorization frames, being either intransitive or
transitive, and the arguments establishing this through the use of the Resultative
construction will not be repeated here.
While ku- `eat' is the only s=a ambitransitive verb which has been found
in Awa Pit, it is possible that there are others. Because of the formal identity
(except in the Resultative) between the use of a transitive verb with an ellipsed
Object, and the same verb used intransitively, it is very dicult to identify which
verbs may in fact be s=a ambitransitive, rather than simply transitive.

6.2.2 S=O ambitransitive verb roots

Awa Pit contains a number of other ambitransitive verb roots, in which the
Subject of the intransitive verb bears the same semantic role to the verb as the
Object of the transitive verb. The following verbs have been found in the data
(where the corresponding English verb is ambitransitive, only one translation is
given; otherwise the translation of the intransitive verb root is given before the
translation of the transitive root): alizh kul- `get annoyed, annoy', ii kul- `keep
warm', ishkwin-/ishkwa- `be startled, scare', kaa- `be born, bear', kaa- `get lost,
lose', ki- `happen, do', kil- `dry', k1z- `break',2 payl- `nish', pya- `smell (bad),
smell'.
1Recall that ku- `eat' has two forms, ku- and kwa-, depending on the following sound.
2It should be noted that k1z- as a transitive verb has only been found in the Perfective Serial
construction with kway- and kyan- (see section 11.6).
150 CHAPTER 6. VERB DERIVATION AND VALENCY CHANGES

(276) Roberto anya kaa-t kway-zi

Roberto before be:born-sv drop-nonlocut
`Roberto was born rst.'
(277) ashaNpa pashpa=ta kaa-t kway-zi
woman child=acc bear-sv drop-nonlocut
`The woman gave birth to a child.'
(278) na=na ii kul-m1z-is
1sg.(nom)=top keep:warm(1) keep:warm(2)-incep-locut
`I am going to get (begin to keep) warm.'
(279) ashaNpa=na Santos=ta ii
woman=top Santos=acc keep:warm(1)
kul-t1-zi
keep:warm(2)-past-nonlocut
`The woman kept Santos warm.'

6.3 Non-productive derivations

While the ambitransitive verbs discussed in the previous section consist of form-
ally identical pairs of verbs with dierent valencies, there are verbs which are
clearly related to other words, but dier from them by containing additional
phonological material. Similar to the ambitransitive verbs, but unlike the pro-
ductive derivations discussed in the following section, the verb derivation dis-
cussed in this section is non-productive  as it is unpredictable in both form
and function, the derived verbs must be included separately in the lexicon.
Three types of non-productive derivational morphology will be discussed
in this section. Firstly, there are a number of pairs of verbs which dier through
the addition of a, ta or na. Many of these, though not all, are pairs of unmarked
intransitive and marked transitive, with the transitive being a causative of the
intransitive. The second and third types of non-productive derivation are similar
to compounding in many ways, with many, though once again not all, of the
roots and the suxes being otherwise meaningful elements in Awa Pit, not
necessarily verbs. The distinction between the two categories is the type of
verb which results: it may be simple, consisting of one phonological word, or
compound, consisting of two phonological words.

6.3.1 Non-productive a, ta and na suxes

There are a number of pairs of verbs in Awa Pit where one member of the pair
is identical to the other, with the addition of a sux a, ta or na. A list of
the pairs which have been found is given in Table 6.1. The pairs are usually
related in some causative sense, though not always, with the unmarked verb being
intransitive while the marked is transitive, or the unmarked being transitive with
6.3. NON-PRODUCTIVE DERIVATIONS 151

sul- intr `boil' sul-a- tr `boil'

kawi- intr `grow up' kawi-ta- tr `bring up'
pak- intr `burn' pak-ta- tr `burn'
ay- intr `cook' ay-na- tr `cook'
kwi- intr `go out (ame)' kwi-na- tr `put out (ame)'
pii- intr `bathe' pii-na- tr `bathe'
m1l- tr `take' m1l-a- ditr `give'
kwaa- tr `carry' kwaa-na- ditr `load'
ap- tr `squash' ap-ta- tr `squash'
m1- tr `hear' m1-na- tr `listen'
Table 6.1: Verb pairs related through the suxes a, ta or na

the marked ditransitive. In these cases the Subject of the intransitive corresponds
to the Object of the transitive in semantic role, or the Subject and Object of
the transitive correspond to the Object and Second Object respectively of the
ditransitive, with the Subject of the marked verb indicating a causer in both
cases.3
One pair related through ta and one pair related through na do not have
this change in valency, with both verbs of the pairs being transitive. In one case
there is a dierence in meaning (m1- `hear', m1na- `listen'), but in the other case
(ap-, apta- `squash') no meaning dierence has been found.

6.3.2 One-word compounds

The non-productive suxes discussed in the previous section appear to have been
in origin a morphological causative marker. They are only found on verbs, the
result is a verb, and the usual additional causative meaning is accompanied by a
change in valency.
There is a second group of non-productive derivations, in some ways sim-
ilar to the previous group, but distinct from it. These derivations involve the
suxes ap, pizh, pyan and kul. In fact, this second group is a very heterogen-
eous one, covering those pairs of words which are single words (unlike the forms
discussed in the following section), but not related through a causative meaning
and the suxes a, ta or na. The suxes ap and kul have only been found in
these words, while pizh and pyan are also found as verb roots, meaning `grasp'
and `hit' respectively, and these meanings appear related to their use as deriv-
ational morphemes. All these suxes can be attached to phonetic sequences
3This correspondence of grammatical relations is identical to that of non-causativecausative
pairs using the productive causative; see section 6.4.1 for further details.
152 CHAPTER 6. VERB DERIVATION AND VALENCY CHANGES

uz stat `be, live' uz-ap stat `be'

ishuz- intr `sit down' ishuz-ap- intr `sit down'
kan- tr `tie up, hang up' kan-ap- tr `hang, put ropes on'
kuz- tr `hit with st' kuz-ap- tr `hit with st'
us- tr `blow' us-ap- tr `blow'
pwitayl N `untruth' pwitayl-ap- intr `tell a lie'
- - - azh-pizh- tr `open'
k11- tr `mill, grind' k11-pizh- tr `massage'
k1hsh- tr `slice' k1hsh-pizh- tr `scratch'
tul- tr `squeeze' tul-pizh- tr `squeeze'
- - - azh-pyan- tr `pull apart'
alizh Adj `angry' alizh-kul- intr, tr
`get annoyed,
annoy'
ii Adj `hot' ii-kul- intr, tr `keep warm'
kwayzh- intr `become tired' kwayzh-kul- intr `rest'
p1n- intr `pass through' p1n-kul- intr `pass through'
pya- intr, tr `smell' pya-kul- tr `smell
(place, of s.thing)'
Table 6.2: Verb pairs related through the sux ap, pizh, pyan and kul

which are otherwise verbs, adjectives or nouns; and the change in meaning which
occurs may be minimal or quite great.
As these words are all, synchronically, single items, and all four suxes
are non-productive, it is perhaps simplest just to list the various forms, which
has been done in Table 6.2. It is worth noting that azh, found in azhpyan- and
azhpizh-, has not been found as a separate word, although it seems likely that
diachronically it was a verb meaning something similar to `open' or `apart'.

6.3.3 Two-word compound verbs

The two-word compound verbs were introduced briey in section 4.5.6, where a
number of points were made about them. These are verbs like sula ku- `bite'
and kal ki- `work' (see section 4.5.6 for a longer list). These verbs are clearly
phonologically two words, but in other ways act as a single lexical item. The
second phonological word is always identical to another verb in the language,
often ki- `do, happen', while the rst phonological word may be identical to
a noun or an adjective, otherwise unknown in the language, or a verb from
Spanish. The majority of these compound verbs are xed and non-productive;
the exceptions are those compounds involving Spanish words  apparently any
Spanish verb (at least action verbs) can be used in Awa Pit by placing it in the
basic Spanish form and compounding it with ki-.
While the issue of these compound verbs will not be discussed exhaustively
here, a number of points must be made about them, and in particular about their
6.3. NON-PRODUCTIVE DERIVATIONS 153

Janus-faced nature, to borrow a term from Kiefer's (199091) discussion of an

apparently similar phenomenon in Hungarian.
6.3.3.1 Compound verbs are two words
Phonologically, as was noted in section 4.5.6, these compound verbs are clearly
two words. While there is some evidence from strictly phonological criteria,
syntactic criteria also make this clear, and additionally show that the compound
verb must be treated as two grammatical words.
In terms of phonology, in slow speech phonemes at the word boundary
may not undergo certain allophonic variation which is obligatory within a word.
Thus, for example, between vowels within a phonological word, /k/ is obligatorily
[G]; phonological word-initially, but in an intervocalic environment, this allophony
is optional, with /k/ appearing as either [k] or [G] in slow speech. Compound
verbs exhibit the latter process  alu ki- `rain' may be produced as [aluki] or
[aluGi] in slow speech.
However overwhelming evidence for the two word status of compound
verbs comes from their syntactic behaviour under negation. In simple negation
in Awa Pit, a negative word shi appears before the verb, and a special sux or
auxiliary occurs on or after the verb (see section 12.5). When a compound verb
is negated, shi occurs between the elements of the compound verb:
(280) alu shi ki-ma-y
rain(1) neg rain(2)-neg-nonlocut
`It didn't rain.'
The separation of the two halves of a compound verb by an intervening word
clearly shows that they are composed of two words: a word cannot occur between
two parts of another word, whether the words in question are phonological or
grammatical.4
6.3.3.2 Compound verbs are one word
While it is easy to show that compound verbs consist of two words, there is also
evidence that in another sense they form one word.
Compound verbs are treated as a single unit for the purposes of modica-
tion (except negation). For example, intensication of a verb is done by placing
the word p1na `very' before the verb: pyan- `hit', p1na pyan- `hit hard, hit a
lot' (see section 13.6). With compound verbs, modiers appear before the rst
element, rather than before the second:
(281) *alu p1na ki-ma-t1-zi
rain(1) very rain(2)-comp-past-nonlocut
4This use of negation to clearly show the two-word status of compound verbs is similar to
the use of the emphatic particle t an in Tamil, which appears between elements of compound
verbs  see Steever (1979:2856).

154 CHAPTER 6. VERB DERIVATION AND VALENCY CHANGES

(282) p1na alu ki-ma-t1-zi

very rain(1) rain(2)-comp-past-nonlocut
`It rained very hard.'
If the modier were modifying the noun alu `rain', akkwan `much' would be used
rather than p1na `very', however this is not possible. Thus, with the exception of
negation, modiers treat the compound verbs as single units.
Some compound verbs have presumably undergone a shift in meaning after
their original formation (cf. Mithun 1984:88990). Thus while kal `dicult' and
ki- `do' are presumably the origin of the compound verb kal ki-, the compound
means `work' rather than `do [something] dicult'.
Two other phenomena are associated with this lexicalization of compound
verbs. There are a number of compounds where the rst element of the compound
is not a (synchronic) lexical item in the language; for example ku ki- `dance', with
no word ku. It could be hypothesized that ku was an Awa Pit word at an earlier
stage of the language, having since disappeared, only surviving in this compound.
This would be identical to what Matiso (1981:309) believes has happened with
a number of compounds in Lahu.5
Secondly, the lexicalization of these compounds (followed by meaning shift
of words) has sometimes resulted in the loss of the hyponymic status of words.
For example, sula ku- `bite' has clearly been formed from sula `tooth' and ku- `eat,
drink'. However synchronically ku- on its own cannot cover the meaning `bite'.
Thus following a statement that a dog bit (sula ku-) me, a following sentence:
(283) p1na kwa-t1-s
very eat-past-locut:under
can only have the (somewhat questionable) meaning `it ate me a lot', not `it bit
me hard', which could only be indicated through:
(284) p1na sula kwa-t1-s
very bite(1) bite(2)-past-locut:under
As well as the above issues, there is the problem of the status of the
rst element, if it is a separate word. While this is not so much of an issue
when it is identical to an adjective (where it could, theoretically, be acting as a
manner adverbial; see section 13.5), it causes a problem for those words which
are identical to nouns. As with similar instances in, for example, Hungarian
(Kiefer 199091) and Tamil (Steever 1979), this noun is not an argument of the
verb. Syntactically, there is often no argument slot available  for example ku-
`eat, drink' takes two arguments (Subject=agent and Object=patient) with all
other nps necessarily obliquely-marked adjuncts; sula ku- `bite' has exactly the
same argument frame, and hence sula cannot be an argument of ku-. This noun
also cannot be modied or quantied, unlike arguments. At a more discourse-
oriented level, this noun is not a participant in the discourse, either: it cannot
It should be noted, however, that as Awa Pit uses this process synchronically for adopting
5
words from Spanish, ku and other words like it may in fact be borrowings from another language.
6.3. NON-PRODUCTIVE DERIVATIONS 155

be topicalized with na, preposed or postposed in the sentence, and it cannot be

referred back to in a following sentence  it is inherently non-referential.
Thus while these compounds are phonologically two words, they are lex-
ically one word, and must be entered explicitly in the lexicon.6 Grammatically
they are dual-natured  for nearly all processes they are treated as a single unit,
however negation treats the second item as a verb, and the rst as some sort of
modier. This dual treatment of compound verbs under syntax  two words
for negation, one unit for most other processes  is actually remarkably similar
to the treatment of compound verbs in American Sign Language (ASL) with re-
spect to the inchoative morpheme. In ASL compounds, there are clear reasons
for treating compounds of two verbs as a single lexical item (for example, think
+ marry `believe'); but the inchoative derivation, which normally aects the
beginning of a verb aects the beginning of the second half of the verb in these
compounds (Liddell & Johnson 1985). The conclusion, both for ASL and for
Awa Pit, is that a lexical compound . . . is a morphologically simple word (i.e., a
single morpheme) with two phonological parts (Liddell & Johnson 1985:95); in
the terms used here, a compound verb is a single lexical item consisting of two
phonological words and two grammatical words, although the two grammatical
words almost always act together in the same way as a single grammatical word.
6.3.3.3 Compounds as incorporation
Setting aside the use of compound verbs involving a Spanish verb, the other com-
pound verbs show many similarities to what is often considered as incorporation.
Unfortunately, the majority of work has been done on noun incorporation  the
incorporation of a noun into a verb. Examining those cases where a synchronic-
ally existing noun in Awa Pit is identical to the rst half of a compound verb, we
can see that Awa Pit probably has what Mithun (1984) has termed Type I and
Type II noun incorporation.
In Mithun's (1984) Type I incorporation, an argument of a verb has been
incorporated into the verb, with the complex having one less argument than the
original verb. It is not clear that this occurs in Awa Pit, although it may be the
case with alu ki- `rain'. If we assume that this verb is formed by incorporating
the Subject alu `rain' into the verb ki- `happen' then it is true that the resulting
compound has one less argument than the original verb  none rather than
one. However in her discussion of Type I noun incorporation Mithun (1984)
only discusses transitive verbs becoming intransitive through the incorporation
of an object. Equally, her Type I incorporated nouns are necessarily in particular
semantic relations to the verb  patient, location or instrument  and alu `rain'
is probably none of these in alu ki- `rain' (though it does depend on how the term
`patient' is interpreted).
There is also only one example of what may be Type II noun incorporation
in the data, the compound sula ku- `bite' from sula `tooth' and ku- `eat, drink'.
This compound is, in fact, identical to the results of the system of instrumental
6 Except for some examples of Spanish verbs; see below.
156 CHAPTER 6. VERB DERIVATION AND VALENCY CHANGES

prexes found in Uto-Aztecan languages, where a body-part term . . . signals

the instrument by which the action of the verb is realized, as in Pipil tan-kwa-
`bit' from tan- `tooth' and kwa- `eat' (Campbell 1985:96). This appears to be an
example of Mithun's (1984) Type II incorporation; however unlike her Type II,
this incorporation does not aect the arguments (Subject=agent, Object=patient
in both cases), but rather incorporates an instrumental adjunct.
While these few compound verbs with the rst part identical to a noun
may be treated as noun incorporation, it is clear that the phenomenon extends
well beyond this in Awa Pit, with incorporation of nouns, adjectives and various
unknown elements. As the process is non-productive except for the use of Spanish
verbs, which are clearly not noun incorporation, these verbs have simply been
called compound verbs, ignoring the various processes by which the compounds
may have been formed.
6.3.3.4 Loan verbs
As noted a number of times in the previous sections, there is one case in which
compound verbs are formed in a fully productive fashion: when Spanish verbs
are used in Awa Pit. Spanish words from most word classes can be used in Awa
Pit with no change (beyond phonetic/phonological alignment), as Awa Pit has
almost no morphology, except for verbs. However when a speaker wishes to use
a verb from Spanish, the verb is put into a compound structure with ki- `do,
happen', and this second element can then carry the morphology. The Spanish
verb itself is placed in its most basic form, the third person singular present
indicative form. Thus from the Spanish trabajar `work', an Awa Pit verb trabaja
ki- `work' may be formed, or from contar `count' comes cuenta ki- `count'.
This process of forming compound verbs from non-native words through
the use of an auxiliary verb meaning `do' can be compared in an interesting way
with similar constructions in a number of other languages.7 Japanese appears
to have a similar construction using the verb suru `do'  thus doraibu-suru is
formed from English drive (Hinds 1986:371-2). It must be noted, however, that
this construction is only used with foreign loan nouns, not verbs; and the hyphen
presumably indicates that these two elements can never be separated.
Clearer parallels are with Indian languages, whether Dravidian (such as
Tamil) or Indo-European (such as Panjabi). In the majority of these languages
there are a series of complex verbs, formed by combining an element such as a
noun, verb or adjective with some sort of auxiliary, most commonly `do'. Bi-
lingual PanjabiEnglish speakers in Birmingham have taken advantage of this
construction (Romaine 1985), and often use lexical items (usually verbs) from
English, followed by the auxiliary k@rna `do'. While it is not clear from Ro-
maine's discussion of Panjabi whether the resulting compound is a single word or
7 While only examples of languages using an auxiliary verb `do' will be looked at here, there
are other ways of achieving the same ends. For example, Warlpiri has a productive system of
compound verbs, and English loan verbs are entered into the language using this pattern (Bavin
& Shopen 1985:82); however as cases such as these do not involve a verb meaning `do', they
will not be discussed here.
6.4. PRODUCTIVE DERIVATIONS 157

two words, the equivalent construction in Tamil discussed in Annamalai (1978),

using English verbs and the auxiliary pannu `do', clearly does form two separate
phonological words  as mentioned earlier,

in Tamil an emphatic particle can
occur between the rst and second parts of the compound.
This use of a verb meaning `do' to form verbs from non-native words
is thus found in a variety of languages, although its exact status is, perhaps,
a little unclear. Annamalai's (1978) discussion of Tamil is in the context of
the anglicised Indian languages: a case of code mixing, and Romaine's (1985)
article likewise considers the Panjabi construction as some form of mixed system;
however in both languages there is a relatively high number of compound verbs
which do not involve non-native elements. In his discussion of Japanese, on the
other hand, Hinds (1986) discusses this structure being used with English loans
and Chinese-based words, but as a loan structure, not as code mixing  this is
presumably established on the basis that the majority of Japanese speakers are
not bilingual speakers of Chinese.
In Awa Pit it would appear that these uses of Spanish, as other uses of
Spanish words, range on a continuum from true loans through to tag-switching,
the use by bilinguals of a single lexical item from Spanish in the midst of Awa Pit.
A few of the compound verbs involving Spanish words, such as trabaja ki- `work',
appear to be well xed in the language, and understood and used by all speakers,
and indeed the only way of expressing the concept in question (the native word
kal ki- `work' is not used for paid work); while others appear to be once o uses
when a speaker temporarily cannot recall an Awa Pit word  indeed, sometimes
a speaker will use one of these Spanish-based compound constructions, then this
will be replaced in the following sentence by a native verb. Thus a relatively
small number of these compounds involving Spanish verbs have to be entered
into the lexicon of Awa Pit, while the majority do not.

6.4 Productive derivations

There are six productive derivational suxes which apply to verbs in Awa Pit
(excluding number marking, discussed in section 7.3 below), and these cover a
variety of meanings, from Causative and Desiderative markers, suxes that mean
`help' and `learn to', through to Inceptive and Prospective aspect suxes. Only
one of these suxes, the Desiderative shi, can be used with stative verbs; how-
ever all are fully productive with active verbs, being used wherever semantically
appropriate.
While the productive derivations discussed here have a wide variety of
meanings, and diering eects on the syntax of a sentence, they nevertheless form
a unied group on positional and distributional grounds. All occur directly after
the verb root (which may itself contain non-productive derivational morphology)
and before any other suxes.
The productive derivational suxes do not have a xed order among them-
selves. The ordering of the suxes, if there is more than one, depends on the
meaning which is to be expressed; compare the order of Causative nin and De-
158 CHAPTER 6. VERB DERIVATION AND VALENCY CHANGES

siderative shi and the eect of this on meaning in the following two sentences:
(285) na=na Juan=ta wakata pay-nin-shi-mtu-s
1sg.(nom)=top Juan=acc cattle buy-caus-desid-impf-locut
`I want to sell cattle to Juan.'
(286) na=na Juan=ta wakata pay-shi-nin-mtu-s
1sg.(nom)=top Juan=acc cattle buy-desid-caus-impf-locut
`I am making Juan want to buy cattle.'
It must be noted that some orderings are not possible, however these have clear
semantic problems (?`I helped Juan want to buy cattle') rather than an issue of
xed morpheme order.
In addition to a lack of xed morpheme order, these six derivational suf-
xes (and number marking) may appear in circumstances where no inectional
suxes may appear. For example a verb root with a productive derivational
sux may be followed by the various imperatives, which do not allow any other
inections to combine with them:
(287) na-wa pat-payn-zha!
1sg-acc wash-help-imp.1obj
`Help me wash [clothes]!'
This combination of a verb root plus derivation may also occur in subordinate,
non-nite structures where otherwise only lexical stems are possible:
(288) na=na pay-nin-mu i-s
1sg.(nom)=top buy-caus-adjzr be-locut
`I am a seller.'
It is clear from the above that these six suxes form a unied group, but it
has not yet been clearly stated why they are considered derivational rather than
inectional. In the theoretical literature, there is a great deal of dispute over the
precise nature of the dierence between inectional and derivational morphology,
if this distinction indeed exists in any real sense.8 However there are a number of
criteria on which the derivational status of this group can be established, given
the common assumptions about the distinction.
To begin with, while it does not establish the group as derivational in
nature, it is perhaps worth noting that these suxes occur directly after the verb
root. This means that any other (inectional) suxes are outside them, and
they are thus potentially derivational (Anderson 1992:126).
It was noted above that there is no xed order among the six suxes of this
group, but rather dierent orderings correspond to dierent meanings. Axes
which allow this sort of variation are usually considered derivational; inectional
axes have xed order.
8 For general discussion of the dierence between derivational and inectional morphology
and further references, see Anderson (1992:7485, 125128), Matthews (1991:4254) and Spen-
cer (1991:912, 193197).
6.4. PRODUCTIVE DERIVATIONS 159

In establishing this set of six axes as a unied group, it was also observed
that a verb root plus one of these suxes may occur in environments where
no other suxes are possible, for example before imperatives or in non-nite
structures. That is, a verb root plus one of these six suxes is treated as a
completely productive lexical verb. This is a property of derivations rather than
inections  only derivations derive new lexical items.
Finally, it can be clearly shown that at least four of the suxes are ne-
cessarily derivational, because of changes between verbs with and without them.
The Causative nin and the Auxiliative payn change the valency of the verb root
to which they attach, adding an additional argument to intransitive and trans-
itive verbs, and alter the correspondence of grammatical and semantic roles for
ditransitive verbs. The Desiderative sux shi and the Inceptive aspect sux m1z
optionally change the subclass of a verb, converting an active verb into a stative
verb. These sort of changes are only possible with derivational markers.
Having established this group of six suxes as a unied group of deriva-
tional suxes, the remainder of this section discusses the individual suxes and
their semantics. For the purposes of discussion they are divided into two groups,
valency-changing and valency-preserving, as this avoids some repetition.

6.4.1 Valency-changing productive derivations

Awa Pit has two productive derivations which change the valency of the verb root
to which they are suxed: the Causative nin and the Auxiliative payn (with its
post-vocalic allomorph wayn). These two suxes add an additional argument to
intransitive and transitive verb roots, and cause a rearrangement in the argument
structure of ditransitive verb roots. For an examination of the semantics of these
two suxes, see the following two subsections  here only the syntactic eect of
these derivations is examined. To simplify the discussion, the terms `causer' and
`causee' will be used here, and all examples will use the derivational sux nin.
It must be kept in mind that the discussion is equally valid for the Auxiliative
payn (for examples of sentences involving this sux, see below), and `causer'
represents either the person causing an activity or helping with an activity, while
`causee' is either the person caused to carry out an action or the person helped
to carry out an action.
Addition of nin or payn to an intransitive verb converts it to a transitive
verb, with the Subject (s) of the underived verb corresponding in semantic role
to the Object (o) of the derived verb, and the causer being shown by the Subject
(a) of the derived verb:
(289) ap aympihsh kal ki-mtu-at1-zi
my brother work(1) work(2)-impf-past-nonlocut
Subj V V
`My brother was working.'
160 CHAPTER 6. VERB DERIVATION AND VALENCY CHANGES

(290) na=na ap aympihsh=ta kal

1sg.(nom)=top my brother=acc work(1)
Subj Obj V
ki-nin-tu-ata-w
work(2)-caus-impf-past-locut:subj
V
`I was causing my brother to work.'
Thus the correspondences of roles for an intransitive verb before and after sux-
ation by the Causative or Auxiliative marker may be represented as:
(291) Underived Derived
Subject (causer)
Subject () Object
When the underived verb is transitive, the Causative or Auxiliative deriv-
ations change it to a ditransitive verb, with the Subject of the underived verb (a)
corresponding to the Object of the derived verb (o); the Object of the underived
verb (o) corresponds to the Second Object of the derived verb (obj2); and the
causer enters the sentence as a Subject (a):
(292) Jaime maza atal pay-t1-zi
Jaime one chicken buy-past-nonlocut
Subj Obj V
`Jaime bought a chicken.'
(293) Carmen=na Jaime=ta maza atal
Carmen=top Jaime=acc one chicken
Subj Obj Obj2
pay-nin-t1-zi
buy-caus-past-nonlocut
V
`Carmen sold a chicken to Jaime (caused Jaime to buy a chicken).'
The Object of the derived verb (corresponding to the Subject of the underived
verb) in a Causative or Auxiliative construction is often human, from the se-
mantics of the Causative and Auxiliative markers. Hence if one of the Object
and Second Object is non-human (as in the previous example), the two objects
may occur in either order, as it is usually clear which of the two is the Object
and hence the causee. If both agent and patient of the basic action are human,
then the order of the two objects of the derived verb is xed  Object followed
by Second Object  to enable comprehension:
6.4. PRODUCTIVE DERIVATIONS 161

(294) na=na Marcos=ta=na anya=ta

1sg.(nom)=top Marcos=acc=top brother=acc
Subj (causer) Obj (causee) Obj2 (aected party)
pyan-nin-ta-w
hit-caus-past-locut:subj
V
`I got Marcos to hit [my] brother.'
(295) na=na anya=ta Marcos=ta=na
1sg.(nom)=top brother=acc Marcos=acc=top
Subj (causer) Obj (causee) Obj2 (aected party)
pyan-nin-ta-w
hit-caus-past-locut:subj
V
`I got [my] brother to hit Marcos.'
The correspondences for underived and derived transitive verbs are thus:
(296) Underived Derived
Subject (causer)
Subject () Object
Object () Second Object
In earlier work on causativization (for example, Comrie 1975, Comrie
1981), the focus was on the paradigm case of morphological causativization,
whereby the Subject of the underived verb corresponds to the Indirect Object of
the derived verb in the causativization of a transitive verb. Cases such as that of
Awa Pit were merely considered as doubling on direct object (Comrie 1975:14
16). In fact, however, a process much more complex than doubling is occur-
ring: rather than simply a correspondence of the Subject of the underived verb
to another grammatical relation, both of the semantic roles lling the grammat-
ical roles of Subject and Object of the underived verb are required to change
grammatical relations. As pointed out by Baker (1988:164165), while this cor-
respondence of Subject to Object and Object to Second Object is rare in the
languages of the world, it has been found in a number of languages, including
Cebuano (Gibson 1980) and Choctaw (Davies 1981); it also occurs in Imbabura
Quechua, where a subject becomes a direct object, while the original object re-
mains marked with accusative (rather than dative, used for ditransitives) but
shows no other object properties (Cole 1985:135141).9
The Object status of the causee can be seen from sentences involving both
the Causative or Auxiliative marker and the First Person Object Imperative zha
(see section 3.5.3). Thus compare the following two sentences:
9A similar but not identical change in grammatical relations of two arguments happens in
other languages, where the Subject of the underived verb form corresponds to the Object of
the derived form, and the Object of the underived form corresponds to an oblique relation; this
occurs, for example, in Tolai (Mosel 1984:155).
162 CHAPTER 6. VERB DERIVATION AND VALENCY CHANGES

(297) nu=na na-wa shitshu sa-nin-tu-s

2sg.(nom)=top 1sg-acc breast touch-caus-impf-locut
`You let me touch [your] breasts.'
(298) shitshu sa-nin-zha
breast touch-caus-imp.1obj
`Let me touch [your] breasts.'
From the second sentence, it is clear that the entity who will carry out the action
of the verb roots in these sentences, `I', is expressed through the grammatical
relation of Object, as it is only possible to use zha when the Object is rst
person. A rst person Second Object does not act in this way.
Causatives and Auxiliatives of ditransitive verbs cause a problem for Awa
Pit. A simple clause has a maximum of three core arguments  Subject, Object
and Second Object. A non-causativized ditransitive clause already has these three
positions lled. It is possible to causativize a ditransitive verb, however only two
of the original three arguments may be expressed. Thus from a ditransitive verb
such as kwin- `give'
(299) Demetrio=na Carmen=ta pala kwin-t1-zi
Demetrio=top Carmen=acc plantain give-past-nonlocut
Subj Obj Obj2 V
`Demetrio gave Carmen a plantain.'
it is possible to derive a causative verb kwin-nin- `make [someone] give', but only
two of the core arguments correspond to the arguments of the underived verb, in
dierent ways:
(300) na=na Demetrio=ta pala
1sg.(nom)=top Demetrio=acc plantain
Subj Obj Obj2
kwin-nin-ta-w
give-caus-past-locut:subj
V
`I made Demetrio give a plantain.'
(301) na=na Carmen=ta pala
1sg.(nom)=top Carmen=acc plantain
Subj Obj Obj2
kwin-nin-ta-w
give-caus-past-locut:subj
V
`I made a plantain be given to Carmen.'
The two sentences above with the derived verbs are syntactically identical; but the
correspondence of grammatical and semantic roles is distinct. The interpretation
of the Object as corresponding to either the Subject of the underived verb (as
6.4. PRODUCTIVE DERIVATIONS 163

in (300)) or to the Object of the underived verb (as in (301)) depends purely
on contextual information, and out of context both sentences are ambiguous.
Thus for ditransitive verbs there are two patterns of correspondences between
underived and derived forms:
(302) Underived Derived
Subject (causer)
Subject () Object
Object () (unexpressed)
Second Object () Second Object
(303) Underived Derived
Subject (causer)
Subject () (unexpressed)
Object () Object
Second Object () Second Object
The notional argument which is unexpressed by a core argument of the derived
form (Object in the rst case, Subject in the second) simply cannot be expressed,
not even as a non-core argument.10
While a priori there would appear to be a third possibility, with under-
ived Subject corresponding to derived Object, underived Object corresponding
to derived Second Object, and an unexpressed underived Second Object, this is
not possible.
Comrie (1975:911) describes the idea of a xed number of syntactic slots
as causative blocking, and gives examples from Songhai, discussed by Shopen
& Konaré (1970). The latter point out that in Songhai there are only a limited
number of syntactic nodes available to verbs and if there are too many semantic
functions, one of them has to be left out (Shopen & Konaré 1970:215), and note
that the Songhai sentence
(304) Garba neere-ndi bari di Musa se
Garba sold-caus horse the Musa IO
is ambiguous, as Musa se may be the original Subject, `Garba had Musa sell the
horse', or the original Indirect Object, `Garba had the horse sold to Musa'. While
Awa Pit has dierent syntactic relations (Object and Second Object rather than
Indirect Object and Direct Object), the process is the same: when a ditransitive
verb is causativized, only one of the original Subject and the original Object may
be expressed, and the sentence is ambiguous.
It should be noted that any non-core arguments (whether complements
or adjuncts) may be retained by a verb when it is causativized; and new ad-
juncts may be added if semantically appropriate. The interpretation is normally
a matter of context; given the sentence
10Other arguments may, of course, be unexpressed in any given sentence, with sentences
involving derived verbs having the same possibilities as any other sentence for denite and
indenite ellipsis.
164 CHAPTER 6. VERB DERIVATION AND VALENCY CHANGES

(305) Juan=na pyalmiN=kasa Marcos=ta kuzhu

Juan=top axe=with Marcos=acc pig
pyaNta-nin-t1-zi
kill-caus-past-nonlocut
`Juan made Marcos kill a pig with an axe.'
in Awa Pit, as with the English translation, it is left to context as to whether
Juan used the axe to force Marcos to kill the pig, or whether Juan caused Marcos
to kill the pig, and Marcos used the axe to do so. Placing the adjunct phrase
initially in the sentence creates a strong suggestion that Juan used the axe, but
even this can be overridden by an appropriate context.
6.4.1.1 Causative
The rst of the two valency-changing derivational morphemes is the Causative
nin. This sux is fully productive with active verbs, but cannot be used with
stative verbs.11 It causes the addition of an extra argument to intransitive and
transitive roots and a rearrangement of the valency of ditransitive roots, in the
way explained above.
The Causative in Awa Pit has a wide range of use, simply indicating that
someone did something (accidentally or on purpose), and because of this (directly
or indirectly, with or without force) the event happened (with participants being
willing or unwilling). Thus the Causative in Awa Pit is not sensitive to any of the
usual causative parameters, such as volition, intention, directness or animacy. It
may correspond to all of the English causative-like verbs: cause, make, get, have,
force, let, and so on. Further context is needed to be more specic:
(306) Marcos=ta=na kal ki-nin-ta-w
Marcos=acc=top work(1) work(2)-caus-past-locut:subj
`(Marcos didn't want to work.) I made Marcos work.'
(307) Doris=ta=na pala kwa-nin-ta-w
Doris=acc=top plantain eat-caus-past-locut:subj
`(Doris really wanted to eat plantains.) I let Doris eat plantains.'
Awa Pit uses its Causative construction quite widely, even where many
other languages would choose to use a separate lexical item, or have an am-
bitransitive verb:
(308) uspa=na wakata nya pay-nin-a-mtu-y
3pl.(nom)=top cattle meat buy-caus-pl:subj-impf-nonlocut
`They are selling (causing [people] to buy) beef.'
Stative verbs can be causativized in a sense, by using a dierent verb root. Rather than
11
`make someone be standing (stative)', one can `make someone stand up (active)'; rather than
`make something be (stative) red', one can `make something become (active) red'.
6.4. PRODUCTIVE DERIVATIONS 165

(309) alcalde=ta p1nk1h paa-t 1-nin-t1-zi

mayor=acc paper write-sv go-caus-past-nonlocut
`She wrote and sent a letter (caused a letter to go) to the mayor.'
(310) t1nta awa uk man ki-nin-tu
strong person stone move(1) move(2)-caus-impfpart
`The strong man moved the stone (caused the stone to move).'
The semantics of the Causative marker, then, are simply that something
happens, and it would not happen if the causer had not done what he or she did.
6.4.1.2 Auxiliative
Like the Causative nin (see previous section), the Auxiliative payn (with an allo-
morph wayn after vowels) adds an extra argument to the verb. The Auxiliative
derivational sux is syntactically identical to the Causative sux, with an in-
transitive or transitive verb root increasing valency by one, while a ditransitive
verb root rearranges its argument when suxed by payn. Like the majority of
other derivational suxes, payn is only used with active verbs.
The semantics of the Auxiliative are very simple. When a verb root is
suxed by the Auxiliative (glossed as `help'), the clause indicates that the
Subject of the new verb stem is intentionally, volitionally helping the major
agent of the verbal action (coded as an Object) to carry out the action:
(311) na=na Demetrio=ta=na s1
1sg.(nom)=top Demetrio=acc=top rewood
pyan-payn-ta-w
hit-help-past-locut:subj
`I helped Demetrio cut rewood.'
(312) Demetrio maza kuzhu pyaNta-wayn-na-zi
Demetrio one pig kill-help-pl:obj-nonlocut
`Demetrio helped them kill a pig.'
(313) ap aympihsh=ta kal ki-wayn-tu-ata-w
my brother=acc work(1) work(2)-help-impf-past-locut:subj
`I was helping my brother work.'
It should be noted that the semantics of the Auxiliative necessarily cause
both the `helper' (Subject) and the `helped' (Object) to be human, or at least
personied. Only humans have the volition to intentionally help others; and
only humans (or highly personied animals) would be assisted in carrying out a
desired action.
166 CHAPTER 6. VERB DERIVATION AND VALENCY CHANGES

6.4.2 Valency-preserving productive derivations

In addition to the two valency-changing productive derivations, the Causative
and the Auxiliative discussed in the previous sections, Awa Pit has four valency-
preserving productive derivations: the Desiderative sux, the Prospective aspect
sux, the Inceptive aspect sux and the Learnative sux.
As valency-preserving suxes, these derivations have no syntactic eect in
a sentence  any verb stem formed from a verb root plus one of these derivations
has the same valency and syntactic possibilities as the verb root itself.
While all four derivations are fully productive with active verbs, only the
Desiderative may be used with stative verbs. In addition to this, the Desiderative
sux and the Inceptive aspect sux can change an active verb root into a stative
verb stem.
6.4.2.1 Desiderative
The Desiderative sux shi may be added to all verbs in Awa Pit, whether active
or stative. Its most common usage is, unsurprisingly, to indicate that a Subject
wants to carry out an action, or be in a state:
(314) an izh-shi-mtu-s
more see-desid-impf-locut
`I want to see more.'
(315) na=na 1-shi-mtu-ata-w
1sg.(nom)=top go-desid-impf-past-locut:subj
`I wanted to go.'
As it is not a separate verb, the Desiderative can only be used to express same-
subject desires; when a situation is to be described where one person wishes
another to carry out an action, a periphrastic way of stating this is used, usually
something like it would be good (for me) if you . . . .
As well as its use to indicate strict wanting, the Desiderative may also be
used for ideas closer to necessity than desire:
(316) pit-shi-ma-ta-w
sleep-desid-comp-past-locut:subj
`I got tired (I suddenly wanted to sleep).'
It can also be used with inanimate Subjects, as a sort of prediction of what will
happen soon:
(317) k1h p1t-m1z-shi-mtu-a-zi
leaf rot-incep-desid-impf-past-nonlocut
`The leaf was about to start rotting.'
(318) alu ki-shi-mtu
rain(1) rain(2)-desid-impfpart
`It is about to rain.'
6.4. PRODUCTIVE DERIVATIONS 167

Both of these extensions of meaning are common developments for desiderative

suxes and verbs (cf. Harkins 1996, ch. 6).
In the examples given so far, the verb stem, consisting of a verb root plus
the Desiderative sux, has always been active. But a verb stem consisting of
an (active) verb root and the Desiderative derivation can also be treated by the
grammar as stative. For example, in the Present tense (not formally marked)
active verbs are necessarily followed by the Imperfective aspect inection, while
stative verbs are followed directly by person marking. Desiderative-marked verbs
may be used with the Imperfective, as in sentence (314) above, showing that they
can be active; but they may also be followed directly by person marking, or in
questions followed directly by the negative copula, another property of stative
verbs only:
(319) tuk-shi-s
suck-desid-locut
`I want to smoke.'
(320) shi izh-shi ki-s
neg see-desid be.neg-locut
`I don't want to see.'
This ambivalent nature of verbs marked with the Desiderative is under-
standable. Given the strong distinction between active and stative verbs in Awa
Pit, it is perhaps not surprising that when a derived form is produced, this de-
rived form has a strong tendency to remain in the same verbal subclass as the
non-derived form. But balancing this tendency is the pull of semantics. Want-
ing something is normally a state rather than an action, and states are strongly
associated in Awa Pit with stative verbs, actions with active verbs. Thus the
ambivalent nature of Desiderative-marked verbs is unsurprising.
There is an additional morpheme ta (glossed as `dummy'), with very lim-
ited distribution, which is associated with the Desiderative sux. It only occurs
when a Desiderative-marked verb is being treated as a stative verb, and it is
perhaps possible that, historically, it was used whenever a stative reading was
desired, although this is clearly not synchronically true (see the preceding two
sentences, for example).
(321) maza=yN kwa-shi-ta-s
one=rest eat-desid-dummy-locut
`I want to eat just one.'
(322) uN=ta pana-shi-ta-s
there=in be:standing-desid-dummy-locut
`I want to be standing over there.'
(323) mizha izh-ta=ma pay-shi-ta-s?
how see-pfpart=inter buy-desid-dummy-locut
`Which colour [shirt] do you want to buy?'
168 CHAPTER 6. VERB DERIVATION AND VALENCY CHANGES

(324) na=na pyan-kam-shi-ta-ta-w, . . .

1sg.(nom)=top know-learn-desid-dummy-past-locut:subj
`I was wondering (I wanted to learn to know) . . . '
The synchronic function of this ta morpheme is unknown; with stative treatment
of the Desiderative-marked verb, shi and shi-ta appear to be in free variation.
One nal point which must be made about the Desiderative sux is its
unusual behaviour with the question marker ma. Normally to query a current
action or state, ki is used (see section 12.3.2), and this is possible with the
Desiderative marker also:
(325) tuk-shi-mtu ki-s?
suck-desid-impfpart Q-locut
`Do you want to smoke?'
However ma may also be used together with the Desiderative marker to query a
present action or state:
(326) nul=na kwa-shi-ma-s?
chontaduro=top eat-desid-Q.past-locut
`Do you want to eat chontaduro?'
This use of ma to query present occurrences is unusual; in all other contexts
it can only be used to question past activities or states (see section 12.3.1).
The distinction between these two question forms and their uses is not fully
understood and requires further study. It may perhaps parallel the distinction
between English do you want this? and did you want this?  in the latter the
speaker has some reason for believing that, in the past at least, the interlocutor
did want it, and is checking on the current desires of the interlocutor, using a
past tense.
6.4.2.2 Prospective aspect
There are a variety of ways of expressing aspect in Awa Pit, in a broad sense of
the term. The most common aspectual relations are shown through inection,
with the Imperfective, Completive and Terminative aspect inections (see sec-
tion 9.3). Perfective aspect can be shown through a Serial Verb construction (see
section 11.6). And two aspects, the Prospective and the Inceptive, are shown
through derivational suxes.
The Prospective aspect marker n1 indicates the intention of the Subject to
carry out the action expressed, and can only be used with animate Subjects, who
could be expected to have such intentions. It also conveys the meaning that at
the reference time, the Subject was (metaphorically) on his or her way to carrying
out the action  the action is not merely in an indenite future, but rather the
earliest steps have been taken to carry out the action. For this reason, it tends
to be used as an immediate future, and is especially common with hortatives.
Any combination of aspect, tense and person marking may follow n1, which
is only used with active verbs. The tense and aspect marking focus on the path
towards carrying out the action, not on the action itself  in saying
6.4. PRODUCTIVE DERIVATIONS 169

(327) us=na piikam-n1-ma-t1

3sg.(nom)=top swim-prosp-comp-term
`He is just going to swim.'
the speaker is not claiming that he (the Subject) has already entered the water,
but rather that the process leading to this has begun. The action itself does not
need to be completed:
(328) Gloria=ta izh-n1-ma-ta-w
Gloria=acc see-prosp-comp-past-locut:subj
`I went to see Gloria (but she wasn't there).'
In this case, while the intention was to see Gloria, and I was (at the time referred
to by the tense) on a path towards fullling that goal, the intended activity
(seeing Gloria) was not carried out.
Further examples of this construction are:
(329) na=na p1hj ki-n1-ma-tu
1sg.(nom)=top weed(1) weed(2)-prosp-comp-hort.1sg
`I am o to weed [a eld].'
(330) kwa-n1-shayN!
eat-prosp-hort.pl
`Lets eat!'
(331) na=na yaz
1sg.(nom)=top clear:land(1)
ki-n1-ta-w
clear:land(2)-prosp-past-locut:subj
`I went to clear (a eld).'
(332) ampu-tuzpa=na kal ki-n1-ma-t
man-coll=top work(1) work(2)-prosp-comp-pfpart
`Among men they went to work.'
The origin of this sux is very clear  historically, it is the Innitive n(a)
used as a same-Subject purposive marker (see section 10.3.1), together with the
verb 1- `go'. For those verbs where the Innitive is na (rather than just n), the a
has been elided before 1, which is a regular phenomenon in rapid speech.
However, while the verb 1- `go' plus the same-Subject purposive use of the
Innitive n(a) is the historical source of this construction, it is clear that the con-
struction has grammaticalized, with a distinct meaning and distinct construction.
A subordinate clause marked with n(a) is usually found after the main clause,
while in the Prospective aspect construction it is always found before the main
verb 1- `go': contrast sentence (328) above, containing the Prospective aspect
marker, with a sentence containing a same-Subject purposive:
170 CHAPTER 6. VERB DERIVATION AND VALENCY CHANGES

(333) 1-ma-ta-w, Gloria=ta izh-na

go-comp-past-locut:subj Gloria=acc see-inf
`I went, (in order) to see Gloria.'
In addition to this syntactic change, the physical movement component of
1- `go' is no longer a necessary part of the meaning of the Prospective, as can be
seen in example (330) above, said at the table in front of food; and indeed the
Prospective aspect sux can be found on the verb 1- `go' itself:
(334) 1-n1-pay!
go-prosp-hort.du
`Lets go!'
It is interesting to note that the Prospective aspect, with its immediate
future meaning, derives from a movement verb together with any tense or (inec-
tional) aspect, plus a purposive construction. While the development of future
tenses from movement verbs is well known in the languages of the world, Bybee,
Pagliuca & Perkins (1991) (and later Bybee, Perkins & Pagliuca (1994:268269))
hypothesize that
the semantics of movement is not sucient in itself to give rise to
the future sense. Rather, movement constructions that are sources
for future grams actually signal that the subject is in the process of
moving towards a goal. That is, along with movement as a compon-
ent of meaning, the source of such futures includes an imperfective
(or progressive) component and an allative component.
(Bybee, Pagliuca & Perkins 1991:30)
The Awa Pit Prospective marker clearly contains neither an imperfective compon-
ent nor an allative component. While Bybee, Pagliuca & Perkins (1991:30) note
that these components need not be overtly marked, it seems clear that in Awa Pit
not only are they not overtly marked, they are neither present nor necessary for
the development of future meaning  the idea of movement in order to do some-
thing is all that is necessary. Given that an allative construction is a common
source for purposive constructions in the world's languages (Haspelmath 1989),
it would be interesting to reexamine those languages used in Bybee, Pagliuca &
Perkins (1991) and Bybee, Perkins & Pagliuca (1994), to see how many of the
movement-derived futures are derived from movement plus purposive construc-
tions  especially given that of the ten languages having `go'-futures in the later
work, only one is explicitly noted as having an imperfective component, and only
two as containing an allative component. It seems possible that the notion of
future arises directly from the movement verb plus a purposive, and that imper-
fectives and allatives are associated with the purposive rather than the future
meaning, as it is clear that a future notion can develop from movement without
these two correlates, as has happened in Awa Pit.
6.4. PRODUCTIVE DERIVATIONS 171

6.4.2.3 Inceptive aspect

Just as the Desiderative can form either an active or a stative verb stem from an
active verb root, so too can the Inceptive aspect sux m1z, although this deriv-
ational marker can only be used with active verb roots. It cannot be combined
with the Completive aspect marker, presumably for semantic reasons, but may
occur with all other inectional morphology.
The Inceptive aspect sux has one very odd feature, shared by no other
inectional or derivational morphology. While other suxes always retain their
positional and structural integrity (although changing allomorphs depending on
surrounding axes), the Inceptive m1z combines with the Plural Subject marker
a to form a new portmanteau morpheme m1taz:
(335) uspa=na 1-m1taz-tu-y
3pl.(nom)=top go-incep.pl:subj-impf-nonlocut
`They are about to go.'
Aside from this, the Inceptive has no other allomorphs, always retaining the form
m1z.12
Inceptive aspect focusses on the beginning of an activity. The way in
which this focus on the beginning of an activity corresponds to the real world
depends on whether the Inceptive is used to create a stative verb or an active
verb.
When used to form a stative verb stem, the Inceptive indicates that at
the reference time indicated by the tense, the beginning of the activity described
had just happened: the reference time was in the beginning of the activity. The
Inceptive in this use necessarily indicates that the activity has begun  it was
sometimes translated by speakers as the Spanish equivalents of has begun to
Verb (if Present) or kept on Verb-ing (if Past). Examples are:
(336) pak-m1z-is
harvest-incep-locut
`We are harvesting/have begun harvesting.'
(337) pa mal-m1z-i
sun shine-incep-nonlocut
`The sun is shining, has started shining (said at 7am).'
(338) pashpa az-m1z-i
child cry-incep-nonlocut
`The child has begun to cry/is crying.'
12This unusual portmanteau behaviour of the Inceptive sux suggests that in origin it may
be a combination of two distinct morphemes, perhaps m1(t) and z. However there is no other
synchronic evidence of either of these suggested morphemes.
172 CHAPTER 6. VERB DERIVATION AND VALENCY CHANGES

(339) na=na a-t kway-ka=na, kuzhu

1sg.(nom)=top come-sv drop-when=top pig
kutil-m1z-at1-zi
skin:with:re-incep-past-nonlocut
`When I arrived, they had already begun to skin/kept on skinning a
pig.'
(340) na=na a-t kway-ka=na, uspa=na
1sg.(nom)=top come-sv drop-when=top 3pl.(nom)=top
shappi k11-m1z-an1-zi
cane:juice mill-incep-fut-nonlocut
`When I arrive, they will have begun milling the cane.'
Thus the stative verb formed by the Inceptive aspect marker describes the be-
ginning of an activity as a state which is current at the reference time indicated
by the tense of the verb.
An active verb formed by the Inceptive aspect has a slightly dierent
focus. As an active verb it naturally has a stronger focus on change, rather than
on a state, and this is often emphasized by using the Imperfective aspect marker
with it even when it is not (grammatically) required, for example with Past and
Future tenses. An active verb with the Inceptive focusses on the very beginning
of an activity, rather than the initial state: compare sentence (341) below with
sentence (338) above.
(341) pashpa az-m1z-tu-y
child cry-incep-impf-nonlocut
`The child is beginning to cry.'
The tense marking gives the time reference. Present tense (no tense marking, but
Imperfective aspect) indicates that the very beginning of the activity is happen-
ing; Past shows that at some reference time in the past, either shown by adjuncts
or from context, the action commenced; and Future states that at a determined
point the event will commence:
(342) na=na piya waa-m1z-tu-s
1sg.(nom)=top corn sow-incep-impf-locut
`I am beginning to sow corn (at this time of year).'
(343) k1n-ka=na, Marcos=na yaz
dawn-when=top Marcos=top clear:land(1)
ki-m1z-tu-a-zi
clear:land(2)-incep-impf-past-nonlocut
`Marcos began clearing land at dawn.'
(344) shappi k11-m1z-tu-an1-s
cane:juice mill-incep-impf-fut-locut
`I will begin milling cane.'
6.4. PRODUCTIVE DERIVATIONS 173

The active Inceptive-marked verb is rarely used without the Imperfect-

ive aspect. It is occasionally marked simply by the Past tense, where it can be
distinguished as active rather than stative by the Past allomorphs t1/ta, appro-
priate for active verbs as in sentence (345), rather than the allomorphs at1/ata,
appropriate for stative verbs as in sentence (339) above.
(345) na=na kayas yal pihshka-m1z-ta-w
1sg.(nom)=top early house sweep-incep-past-locut:subj
`I began sweeping the house early.'
The active Inceptive-marked verb is also occasionally used with the Ter-
minative aspect. In this case the verb combines the meanings of the two suxes,
indicating that the very beginning of the activity (active Inceptive) has just oc-
curred (Terminative):
(346) az-m1z-t1
cry-incep-term
`She has just begun to cry.'
This combination of the derivational Inceptive and the inectional Terminative
is very seldom used, however.
6.4.2.4 Learnative
Awa Pit has a transitive active verb kam- `learn'. As well as this, there is a
homophonous derivational ax kam, used for `learn to Verb'. This can be used
together with any semantically appropriate active verb, with the usual full range
of inectional possibilities:
(347) Awa Pit pata-kam-tu-s
person language speak-learn-impf-locut
`I am learning to speak Awa Pit.'
(348) na=na kuzhu kutil-kam-ma-t=ma
1sg.(nom)=top pig skin:with:re-learn-comp-pfpart=temp
`I have learnt how to skin a pig over the re.'
Together with the word pyan- `know', the derivational sux kam is used
to indicate entering into knowledge, coming to know a fact about the world not
previously known:
(349) pyan-kam-shi-mtu-s
know-learn-desid-impf-locut
`I want to know.'
In addition to its productive use, the derivational ax kam has lexied
together with the word pii- `bathe' to form the word piikam- `swim':
174 CHAPTER 6. VERB DERIVATION AND VALENCY CHANGES

(350) us=na piikam-n1-ma-t1

3sg.(nom)=top swim-prosp-comp-term
`He is just going to swim.'
The lexied status of piikam- `swim' is shown in its ability to be followed once
again by the derivational sux kam, in its productive usage:
(351) Gloria=na piikam-kam-ma-mtu-a-zi
Gloria=top swim-learn-comp-impf-past-nonlocut
`Gloria was learning how to swim.'
The sux kam could be considered as simply being the verb kam- `learn' in
some form of verb compounding. This has not been done here, as synchronically
this is the only verb which can undergo such compounding as a productive process
(but compare the non-productive uses of pizh- `grasp' and pyan- `hit' as suxes
in compound verbs, section 6.3.2).
Chapter 7
Verb inection and number
7.1 Introduction
In the previous chapter, a variety of verb derivations were discussed, both pro-
ductive and non-productive ones. These derivations produce verb stems, which
can then be modied through inections, which can also be added to verb stems
which are verb roots. That is, the productive derivations in the previous chapter
are all morphologically and syntactically optional, although clearly they have
semantic eects.
While there are a variety of inections which can be added (singly or
multiply) to verb stems, the inections are clearly not all of the same type, but
form dierent groups of inections, with dierent syntactic and morphological
eects. Some inections must be the only inection on a verb stem, others
combine with dierent inections. Some inections can only be used in main
clauses, others must only be used in subordinate clauses, while some can be used
in both.
This chapter begins with an overview of the dierent inectional pos-
sibilities which exist in Awa Pit. `Inectional' possibilities is, in fact, perhaps
a slightly incorrect shorthand  two of the `inections', the Imperfective Parti-
ciple for stative verbs and the Serial Verb `inection' for consonant-nal stems are
zero inections. That is, the morphological form of the Imperfective Participle
of stative verbs and the morphology of the Serial Verb form of consonant-nal
stems are identical to the forms of the verb stems themselves.
With the exception of the two zero inections mentioned in the previous
paragraph, all inections in Awa Pit are suxes  there are no prexes or
inxes, and inection is not shown through reduplication or a change in form of
a root; and suppletion is only used to show plural marking in the copula verb,
which appears to be derivational rather than inectional, as will be shown in
section 7.3. South American languages generally have only a few prexes but
many suxes (Doris Payne 1990b:215), and there are some, such as Epena Pedee
(Harms 1994:13) and Barasano (Jones & Jones 1991:5) which like Awa Pit are
purely suxing.
The rst section of this chapter, then, discusses the possible inections
176 CHAPTER 7. VERB INFLECTION AND NUMBER

which occur in Awa Pit in a very cursory fashion, with cross-references to the
major sections or chapters where they are discussed in detail. The nal part
of the section discusses very briey the relative ordering of the various axes.
Following this, the number-marking system, which appears to be derivational
rather than inectional, is discussed.

7.2 Inections
The verb inections in Awa Pit form into a number of dierent groups, based
on a combination of semantic and structural features. In this section a summary
of the inectional system is given, with references to where these inections are
discussed more fully. The important issue of niteness and the justication for the
distinction between the formally identical Imperfective aspect and Imperfective
Participle inectional forms was discussed in section 3.3.2.

7.2.1 Tense inection

There are two formal markers of tense, the Past and the Future. There is an
additional covert category, which expresses the Present: in any particular string
of inections, in any given structure, if either the Past or Future marker could
appear, but does not, this indicates Present:
(352) ku-mtu-ata-w
eat-impf-past-locut:subj
`I was eating.'
(353) ku-mtu-an1-s
eat-impf-fut-locut
`I will be eating.'
(354) ku-mtu-s
eat-impf-locut
`I am eating.'
That is, in pure structuralist terms, the Present is marked by a zero morpheme.
The tense markers appear after any aspect or mood suxes, but before person
marking. Tense is discussed in section 9.2.

7.2.2 Aspect inection

As has been noted elsewhere, aspectual meaning in its widest sense is expressed in
a variety of ways in Awa Pit. There are two aspectual derivations, the Inceptive
(section 6.4.2.3) and the Prospective (section 6.4.2.2). Perfective aspect is also
expressed using a Serial Verb construction (section 11.6).
In addition, however, there are three aspectual inectional suxes, the Im-
perfective, the Terminative and the Completive, discussed in section 9.3. These
7.2. INFLECTIONS 177

aspect markers occur before any tense marking (if it is present in a clause), and
may combine with each other, in the order Imperfective, Completive, Termin-
ative. They do not cooccur with mood marking (but see below). The aspect
inections are the only inections which can occur before the subordinating suf-
x ka `when'.

7.2.3 Mood inection

The mood inections convey a wide range of meanings, and form a less coher-
ent group than others, both semantically and syntactically. None of them can
combine with aspect inections,1 while only one may occur with tense; and a
few are sometimes used with person, while others have inherent person. Two of
the mood inections are formally identical with subordinating markers, and are
presumably historically related, although the meanings expressed are distinct.
The mood markers are discussed in section 9.4.
One mood marker which is perhaps worthy of more comment is the Irrealis
marker na, which is only used in past counterfactual clauses (section 10.3.6).
This sux is unusual in two ways: it occurs after any tense marking, while
mood marking normally occurs before tense marking; and it is followed by person
marking, with an s for Locutor and an unmarked form for Non-locutor. As will
be discussed in section 8.4.1, zero Non-locutor forms are associated with non-
verbal elements. This suggests that in origin the Irrealis sux may have been
the Innitive, a non-nite form which has the same `non-real' semantic idea,
but that through its use in the counterfactual construction and its acquisition of
person marking, the Irrealis has separated from the Innitive.

7.2.4 Negative and interrogative inection

Negation and interrogation are often expressed through content question words
or through auxiliary verbs. However there is one negative inectional marker
ma, and a homophonous interrogative marker ma. These are discussed in sec-
tions 12.5.1 and 12.3.1 respectively. There is also a negative nominalizer, kayN,
which is non-nite, and forms a group with the `other non-nite' inections below.

7.2.5 Serial Verb inection

When verb stems are united to form a Serial Verb construction,2 the non-nal
verb stems are suxed by the Serial Verb inection t, if their nal element is a
vowel; for consonant-nal verb stems, the Serial Verb form is identical to the verb
stem itself. The Serial Verb construction is discussed in section 11.5. The Serial
Verb inection is also involved in the Conjoined Clause construction described in
section 11.7.
1This is not quite true. When the Necessitive is used as a marker of future time rather than
modality, it can combine with aspect  see section 9.4.2.2.
2As will be discussed in section 11.5, there are problems with analyzing this construction as
a true serial verb construction  it should be treated as a label only.
178 CHAPTER 7. VERB INFLECTION AND NUMBER

7.2.6 Person inection

Awa Pit has an unusual system of person marking, which has probably formed
from an earlier system of evidentiality, though it has largely been grammatical-
ized. The inections involved do mark person, however not always just person of
Subject or Object, and hence while it is a person-marking system, it is not strictly
speaking a cross-referencing system. This unusual marking system is discussed
in chapter 8.

7.2.7 Subordinating inection

There are a few inections which are used to mark a clause as non-nite and
subordinate. Two of these are formally identical to mood markers, and while
historically related, they can be distinguished, as was discussed in section 3.3.2.
These inections and their uses in adverbial subordinate clauses are discussed in
section 10.3.

7.2.8 Other non-nite inections

The nal category of inections is, in some senses, a residue class, containing
the Innitive, the two participles, and two adjectivizers, one positive and one
negative. However there are common elements in this group. All are used only
in non-nite clauses, and in particular constructions there are often alternations
between subsets of these inections, such as the two participles and the Innitive,
or the two participles and the adjectivizers. Most of these inections can be used
in non-nite main clauses, with the exception of the Innitive. One inection in
this group, the Imperfective Participle, is formally identical to the Imperfective
aspect marker.
The non-nite forms are used in a wide variety of dierent constructions
with dierent uses, and hence it is easiest to give a summary of these suxes
here, as it is impossible to give cross-references to simply one other section as has
been the case with other inections.
7.2.8.1 Innitive
The Innitive marker na (and its allomorph n) has a wide variety of functions,
although it is restricted to appearing in non-nite subordinate clauses. There are
a variety of denitions given for innitives, from traditional ones, the non-nite
form of the verb usually cited as its unmarked or base form (Crystal 1985:157),
through to more complex ones. such as that of Noonan (1985:57):
verb-like entities that do not bear syntactic relations to their no-
tional subjects; i.e. their subjects do not take nominative case mark-
ing or condition verb agreement.. . . But because innitives are verb-
like, the relations that they may establish with their objects . . . are
the same as those established by [other verb forms].
7.2. INFLECTIONS 179

In Awa Pit the verb-form labelled the Innitive denitely does not t the more
traditional denition  it is not unmarked, and the citation form tends to be
the Imperfective. In many constructions in which it is used, the Innitive does
not have an explicit Subject, and when it does, the verb is not marked for per-
son, which accords with Noonan's denition; but then the other non-nite forms
accord equally well. The label `Innitive' has been chosen for this verb form
because of its use to create a clause about which predications can be made.
For all verb stems ending in consonants, and for most verb stems ending
in vowels, the Innitive is na. There are a few verbs ending in vowels which take
the short form of the Innitive, n rather than na. The verbs which have been
found to take the short form are: a- `come', 1- `go', ki- `do', ku- `eat', kata- `bring'
and shaa- `walk'.3
The Innitive is used entirely in subordinate constructions. It is used
in forming indirect questions, where the subordinate clause contains a Subject
(section 10.2.2); in complements of intention and evaluative predicates (sec-
tions 10.2.4.1 and 10.2.4.2); to form a clausal complement to a postposition (sec-
tion 10.2.3.2); to form the same-Subject purposive construction (section 10.3.1);
and in absolute constructions (section 10.3.7).
7.2.8.2 Imperfective Participle
The Imperfective Participle has two quite distinct forms. For stative verbs and
the Serial Perfective aspect verbs, the Imperfective Participle form is simply the
verb stem; for active verbs, the Imperfective Participle is formed with a sux,
which has the form mtu after vowels and tu after consonants.
The major use of the Imperfective Participle is in non-nite subordinate
clauses. It is used to form indirect question complements (section 10.2.2); in-
direct statement complements (section 10.2.3.1); complements of intention and
evaluation predicates (sections 10.2.4.1 and 10.2.4.2); to form complements to
postpositions (section 10.2.3.2); it forms absolute clauses (section 10.3.7); it is
used in the formation of counterfactual clauses (section 10.3.6); and it is used in
nominalizations (section 10.5). In addition, the Imperfective Participle can be
used as a main verb with an auxiliary (section 11.2); and it can also be used in
non-nite main clauses (see section 3.3.2).
7.2.8.3 Perfective Participle
The Perfective Participle is formed with a sux ta with most verbs. The sux
has the short form t with the same verbs that have a short form of the Innitive 
a- `come', 1- `go', ki- `do', ku- `eat', kata- `bring' and shaa- `walk'  and also with
all stative verbs ending in a vowel. When the copula i is used in the Perfective
Participle form, it takes the allomorph a, which is also used with non-Present
inections.
3 Note that these verbs also all take the short form of the Perfective Participle; see below.
180 CHAPTER 7. VERB INFLECTION AND NUMBER

In addition to the normal Perfective Participle, there is also an extended

Perfective Participle, which consists of an (active) verb stem, followed by the Ter-
minative or Completive aspect markers, followed by the t form of the Perfective
Participle. The extended Perfective Participle is used in some constructions, the
Perfective Participle in others.
The Perfective Participle is used as a derivational marker to form change of
state adjectives from verbs (section 5.2.3.1). It is also used to form indirect ques-
tions (section 10.2.2) and statements (section 10.2.3.1); complements to intention
and evaluation predicates (sections 10.2.4.1 and 10.2.4.2); clausal complements
to postpositions (section 10.2.3.2); counterfactual clauses (section 10.3.6); and
absolute clauses (section 10.3.7); as well as occurring in agentive nominaliza-
tions (section 10.5). The extended Perfective Participle is used as a main verb
with an auxiliary in the Resultative (section 11.2.2.2) and Past Anterior (sec-
tion 11.2.2.1) constructions; and it can also be used in non-nite main clauses
(see section 3.3.2).
7.2.8.4 The adjectivizers
There are two adjectivizers in Awa Pit, the (Positive) Adjectivizer mu (with
its allomorph m, occurring after vowels), which is found in a wide variety of
constructions, and the Negative Adjectivizer kayN, which is limited to occurring
in one construction.4
The essential element in all of the uses of the adjectivizers is an idea
of habituality, as well as converting the clause it marks into an adjective-like
or sometimes noun-like clause. The (Positive) Adjectivizer is used in agent-
ive and instrumental nominalizations (section 5.2.1), purposive adjectivizations
(section 5.2.3.2), and as one part of forming agentive clausal nominalizations (sec-
tion 10.5); with the Negative Adjectivizer, it is used in forming habitual relative
clauses (section 10.4).

7.2.9 Order of inections

The inections given in the preceding sections do not, of course, occur in a
random order, but rather are very constrained in their cooccurrence. A general
summary of the order in which inections occur following a verb stem is given
in Table 7.1. It must be strongly stressed that this is simply a very general
outline, and individual axes may be highly constrained in their cooccurrence
possibilities. Any possible combination of inections in Awa Pit ts the outline in
Table 7.1, but the schema given there generates many impossible combinations.
For example, mood is clearly followed by tense, then by person, as can be seen
in a verb such as kwa-npa-ta-w `I needed to eat', where the verb stem is followed
4 It is interesting to note that there is a nominalizer mu in the related language Cha'palaachi
(Abrahamson 1962:231). Unfortunately, not enough information is available on this language
to know how the Cha'palaachi nominalizer is used  the example given shows the verb hi `go'
followed by the nominalizer mu forming a word himu `traveller', but this could be a lexical
nominalization, an agentive nominalization, or a headless relative clause.
7.3. PLURAL MARKING 181

8
>
> Negative  Aspect
>
> Mood  Tense  Person
>
>
>
> Interrogative  Person
>
<
Verb stem  > Aspect  Other non-nite  Subordinating
>
>
>
>
>
> Serial Verb
>
>
: Conjoining

Table 7.1: Schema to cover possible inectional combinations

by the Necessitive mood, Past tense and rst person (Locutor). However while
person follows tense follows mood in the general schema to cover verb forms
such as kwa-npa-ta-w, some moods, such as the imperatives, cannot be used
together with tense or person; one aspect in a particular tense cannot be followed
by person; and so on. Thus the description in Table 7.1 is a quick summary,
but by itself overgenerates inectional combinations. For precise combinatory
possibilities, it is necessary to refer to the description of individual inections.

7.3 Plural marking

The number-marking system of Awa Pit is, in one sense, quite complex, with a
variety of verb suxes being used in some regions for dual and plural, varying in
dierent tenses, aspects and persons (Lee Henriksen, p.c.).
However this complex system of number marking is simply not used in the
Pialapí region. It was sometimes possible to elicit one of these forms from one
informant, but only by being extremely explicit: they hit us never produced
the complex forms, while those two hit all of us would sometimes produce a
complex verb form, although not often enough to allow anything beyond a very
preliminary analysis of the forms.
While the lack of use of the forms in the Pialapí region could be taken
as a sign of language decay, it may simply be a dialect feature. The speaker
who did occasionally produce complex forms, T, had lived for a number of years
in Ecuador, and may perhaps have acquired a passive knowledge of them there.
The forms were clearly not part of his active knowledge of the language  in
one language session, a dual form had been successfully elicited, when two men
walked by, and the informant said spontaneously:
182 CHAPTER 7. VERB INFLECTION AND NUMBER

(355) Planada=ta-s a-mtu-y

Planada=in-from come-impf-nonlocut
`He/she/they are coming from La Planada.'
This spontaneous sentence contains no number marking, even though it was
produced directly after an elicited sentence with dual number marking.
Despite the absence of this complex number-marking system, speakers of
Awa Pit in the Pialapí region at least have a much simpler system available to
them. This system only distinguishes plural (more than one), and contrasts with
no marking, which may indicate any number. Thus this plural-marking system is
completely optional, unlike the much more extensive system which is obligatory
for human referents in some other regions (Lee Henriksen, p.c.). Indeed, even this
simple plural marking is rarely used, unless the number of a human argument is
to be especially emphasized.
While the plural-marking system used in the Awa Pit of Pialapí is ap-
parently a relatively simple system compared with that used in some regions,
there are strong restrictions and constraints on it. It can only mark plural for an
argument which is human, and this argument must be either the Subject or the
Object of an active verb, or the Subject of a stative verb. The subsystems used
for active and stative verbs are quite distinct, and stative plurals are only used
in Present tense; the two subsystems will be discussed separately below.
Perhaps the most interesting feature of the plural-marking system is that
it appears to be derivational rather than inectional. This will be discussed
below, after exemplifying the active and stative systems.

7.3.1 Plural marking in active verbs

With active verbs, the marker of plural Subject is a, and the marker of plural
Object, clearly only with transitive verbs, is na. These suxes are placed directly
after the verb stem  that is, either after the verb root if it carries no derivational
axes, or after any derivational axes that there might be (but see below).
(356) uspa=na Santos=ta t1t-shi-a-mtu-y
3pl.(nom)=top Santos=acc cut-desid-pl:subj-impf-nonlocut
`They want to cut (stab) Santos.'
(357) ampata-tuz=na ku-a-t1-zi
four-nmlzr.pl=top eat-pl:subj-past-nonlocut
`Between the four of them they were eating.'
(358) kayl-a-n1-zi
return-pl:subj-fut-nonlocut
`They will come back.'
(359) pashpa=ta pan kwin-na-ta-w
child=acc bread give-pl:obj-past-locut:subj
`I gave bread to the children.'
7.3. PLURAL MARKING 183

(360) na=na uspa=tuza t1t

1sg.(nom)=top 3pl=(3pl.)acc cut
kyan-na-ta-w
throw-pl:obj-past-locut:subj
`I stabbed them.'
As can be seen from the above examples, a wide variety of tense, aspect and mood
suxes can cooccur with active plural marking; in fact, plural marking is even
possible on some non-nite forms (see below). It is used with any grammatical
person, and does not vary with this feature.
The plural Subject and plural Object markers can be used, at least in the
Pialapí region, to show any number of humans greater than one, including two:
(361) na=na paas pashpa=ta piina-na-ta-w
1sg.(nom)=top two child=acc bathe-pl:obj-past-locut:subj
`I bathed the two children.'
This appears to be dierent in other dialects; the only verb paradigm given in the
pedagogical grammar of Awa Pit (Henriksen & Obando Ordóñez 1985:65), the
Present Imperfective, contains dual forms, which do not have the Plural Subject
marker a, unlike the plural forms. This use of the Plural markers to cover two
people is presumably related to the lack of the dual forms in the language of the
region of Pialapí.
There is only one positional slot for the plural marker, and hence if both
Subject and Object are plural and human, only one of these can be marked. The
marker which is chosen depends on which participant the speaker wishes to focus
on:
(362) uspa=na au=m1za kal
3pl.(nom)=top 1pl=(1/2pl.)acc work(1)
ki-wayn-a-mtu-y
work(2)-help-pl:subj-impf-nonlocut
`They are helping us work.'
(363) uspa=na au=m1za t1t
3pl.(nom)=top 1pl=(1/2pl.)acc cord
ta-na-t1-zi
give-pl:obj-past-nonlocut
`They passed us a cord.'
Even though the Subject and Object are both plural in these two sentences, in
the rst the focus is on the Subject, in the second on the Object.
There are two peculiarities of the plural Subject marking on active verbs,
compared with other axes. The rst was noted earlier, in the discussion of the
Inceptive aspect sux m1z (see section 6.4.2.3). Rather than the plural Subject
sux following this derivational sux, as it follows all others, the two form a
portmanteau morpheme m1taz:
184 CHAPTER 7. VERB INFLECTION AND NUMBER

Locutor Subject Non-locutor Subject

Unspecied i-s i
Plural makpa-s puta-y
Table 7.2: Unspecied number and plural forms of the copula verb i

(364) uspa=na say-m1taz-tu-y

3pl.(nom)=top look:for/nd-incep.pl:subj-impf-nonlocut
`They are about to look for [it].'
The second peculiarity occurs when the plural Subject sux is placed on a
verb stem ending in a stop consonant. The stop consonant is geminated between
the verb stem and the plural marking, thereby retaining its voiceless quality,
rather than being voiced and fricativized, as a single stop between vowels would
be:
(365) uspa=na s1p-pa-ma-t1-zi
3pl.(nom)=top sew-pl:subj-comp-past-nonlocut
`They sewed.'

7.3.2 Plural marking in stative verbs

The marking of plural in stative verbs is quite distinct from that of active verbs.
It is still completely optional, and appears to be even more uncommon than with
active verbs. Once again, only human arguments may be marked as plural. In the
case of stative verbs, only Subjects can be marked as plural; of course, excluding
verb stems formed with the Desiderative or Inceptive aspect suxes, there is only
one stative verb with an Object, so this is perhaps not surprising.
The copula verb i has a set of suppletive plural forms. Unlike with the
active verbs, there are separate forms for Locutor and Non-locutor, shown in
Table 7.2. The nal s and y in the plural copula verb forms mark person. Their
separateness from the stem can be ascertained through the use of the Topic
marker with these forms. When the Topic marker is used on verbs, person mark-
ing is absent (see section 14.2.4). With the plural forms of the stative verbs used
with Topic marking, the nal segment is absent; for example:
(366) yal=ta=na, p1na akkwan awa
house=in=top very many person
puta=na=ma
be.pl:nonlocut:subj=top=temp
`There are loads of people in the house.'
Thus the nal s or y, absent in this sentence, is clearly person marking.
7.3. PLURAL MARKING 185

These suppletive copula forms have only ever been found in the Present
(unmarked) tense. When the copula occurs in other tenses, the forms are always
those of the unspecied form: there is no way (in the verb) of indicating the
plural nature of the Subject for non-Present copula verbs.
Plural marking of the other stative verbs, apart from the copula, is based
on the forms for the copula. There are two suxes, makpa-s for Locutor plural
Subject and puta-y for Non-locutor plural Subject:
(367) aN=ta pana-makpa-s
here=in be:standing-pl:locut:subj-locut
`We are standing over here.'
(368) uspa uz-puta-y
3pl.(nom) be:sitting-pl:nonlocut:subj-nonlocut
`They are sitting.'
Unsurprisingly, these suxes are only used in the Present tense; in other tenses
plural is not indicated with stative verbs.

7.3.3 Plural marking as derivational

While the situation is not at all clear, there are reasons for considering that the
marking of plural in the Awa Pit of Pialapí may be derivational rather than
inectional. The evidence is suggestive, but not conclusive.
As an initial consideration, it should be noted that plural marking is closer
to the root than any other suxes except derivational suxes. While this is not
evidence in itself, it does mean that all (other) inectional morphology is outside
plural marking, and hence plural marking could be derivational. In addition, in
the case of plural Subject marking of active verbs, the plural forms a portmanteau
morpheme with one of the derivational suxes, m1z, the Inceptive aspect marker.
The plural marking is completely optional. While it is not unknown for
inectional morphology to be optional, it is much more common for inection to
be obligatory, and only derivational morphology to be optional.
Perhaps most tellingly, plural marking may occur on certain non-nite verb
forms. Inectional morphology in Awa Pit is not generally found on non-nite
forms, except for specic exceptions (such as t1 or ma found on the Perfective
Participle in the Resultative construction). However the plural marker may be
found on non-nite forms:
(369) izh-na-mtu ka-s
see-pl:obj-impfpart be:permanently-locut
`I know (lit. see) them.'
(370) m1n-a-waza t1t-na-t kyan-ta-w
someone-acc cut-pl:obj-sv throw-past-locut:subj
`I cut some people (lit. someone-plural).'
186 CHAPTER 7. VERB INFLECTION AND NUMBER

Occurring as it does on non-nite forms, the plural marking appears more similar
to derivational marking, which may occur in these circumstances. However there
are restrictions placed on the occurrence of plural marking in non-nite forms
which are not placed on (other) derivational suxes. Plural marking only occurs
on non-nite verb forms which are main verbs and followed by a (nite) auxiliary
 unlike other derivational morphemes, plural marking may not occur on non-
nite verbs in subordinate clauses.
Thus the evidence for the precise status of plural markers is unclear. Their
position in the verb, their optionality and their wider distribution than (other)
inectional axes suggest that they are derivational; but their range is somewhat
more restricted than that of (other) derivational axes.
Chapter 8
Person marking
8.1 Introduction
As has been noted in a number of places, there is an unusual system of person
marking operating in Awa Pit. Person marking is important in examining the
issue of niteness  person marking is never used on verbs in non-nite clauses
(whether main or subordinate), while verbs in nite clauses are generally person
marked.1
There are a few cases of nite clauses where person marking is not ex-
pressed through the usual system  either it is simply unexpressed, or it is
shown inherently through some other marking.
There is only one case where a nite clause does not express person. For
some reason, there is a clash between the Terminative aspect marker t1 and
person marking, so while the Terminative aspect marker can be followed by a
Past or Future tense marker and then person, in the Present tense, which is
shown by the absence of Past or Future marking, person cannot be expressed.
As noted in section 3.3.2, the Obligative and Potential moods are often
used to express universal ideas  `someone must do X/X must be done' and
`someone can do X/X can be done' respectively. In these impersonal uses of
these moods, person marking is never expressed; but it can be used with the same
moods when a personal interpretation is intended (`I must do X', `Demetrio can
do X').2
There is also a small group of verb inections which have inherent person
marking: imperatives and hortatives (see section 9.4.5). The inections indicat-
ing these moods have inherent person  rst person for hortatives and second
person for imperatives. It is interesting to note that, by their very nature, clauses
containing these forms are outside the declarative/interrogative split which is
of extreme importance for the usual person-marking system.
1Or at least one of the verbs is person marked, in the case of clauses involving the Serial
Verb or Conjoined Clause constructions  see sections 11.5 and 11.7.
2This use of impersonal is not to be confused with the impersonal (zero argument) verbs,
such as k1n- `to dawn'  these verbs do have person marking (in the usual contexts), despite
their inherent impersonal nature.
188 CHAPTER 8. PERSON MARKING

In the remainder of this chapter the basic person-marking system of Awa

Pit will be described. That is, non-declarative, non-interrogative clauses (direct-
ives) will be left aside, as will impersonal uses of the Obligative and Potential
mood markers. It will be assumed that nite clauses involving the Terminative
aspect do carry person marking, and it is simply a peculiarity of the system that
in the Present tense this marking is not expressed.3
It is also important to note that the system being described is the reduced
system of marking used in the region of Pialapí. That is, some varieties of Awa
Pit appear to have complex marking involving portmanteau person and number
marking, but this system is not used in Pialapí (see section 7.3 for more details).
It should, however, be mentioned that from what little data is available the
Locutor/Non-locutor system (see below) appears to operate in the more complex
person/number system also.
The remainder of this chapter is thus dedicated to the person-marking
system of Awa Pit. It begins with a discussion of the Locutor/Non-locutor person
marking split; the full Past system is explained, followed by the full non-Past
system; and nally a number of parallels with other languages and possible origins
for the system are discussed.
In the following, a number of terms are used with somewhat unusual mean-
ings, more precise than common usage rather than distinct, and for reasons which
will become clearer below. First, a distinction is made between an entity being
`involved' in an activity or state and an entity being a (grammatical) `participant'
in an activity or state. For example, alu ku- `rain' is a zero-argument verb in Awa
Pit: there are no participants (whether complements or adjuncts). However an
entity can be involved in an activity in which it is not a participant  if it rains,
it could aect me by raining on me, and I am then involved in this activity even
though (in Awa Pit) I cannot be a (grammatical) participant. This distinction
between participants and involved entities then leads to a distinction between
the notions of `cross-referencing' and `person marking'. Cross-referencing is some
marking of information about participants on some word or words other than
those denoting the participant entity (usually on verbs or auxiliaries). It is also
possible, however, to give some indication on, for example, a verb that an entity
is involved in an action. If this marking gives some information about the person
of an involved entity, then this marking can be considered person marking. If
information can only be given about the person of participants, then this person
marking is (a subset of) cross-referencing. But it is also possible that person
marking could refer to non-participant involved entities, and consequently the
person marking is not cross-referencing. This distinction will be taken up again
in more detail after an examination of the person-marking system of Awa Pit.
3In fact, this assumption has no impact on the system, and it could just as easily be con-
sidered that these nite Present Terminative aspect verbs do not carry person.
8.2. LOCUTOR AND NON-LOCUTOR 189

8.2 Locutor and Non-locutor

The complexity of the Awa Pit person-marking system is due in part to the
interaction of two separate systems; or at least, two systems which appear to
be separate, although cross-linguistic comparisons suggest that the two may be
interrelated. While each of these systems is not, of itself, complex, both are
unusual, and the interaction of the two leads to an overall complex system which
is bewildering, on the surface. This section deals with the Locutor/Non-locutor
distinction which is the major person-marking distinction in the system; the
following two sections examine how this system interacts with semantic roles and
grammatical relations to produce the system used to mark person in the Past
tense and non-Past tenses.
As is detailed in section 4.3.3, Awa Pit recognizes the usual three persons
lexically, in its pronouns: rst person na `I', second person nu `you' and third
person us `he, she', to give the singular Nominative personal pronoun forms.
These persons and correspondences to lexical pronouns are invariant  that is,
a notional rst person argument can always be expressed through one of the rst
person pronouns, and so on, regardless of the type of utterance; of course the
common use of ellipsis in Awa Pit means that the personal pronouns are not
often used, except to indicate emphasis or to disambiguate utterances.
The person-marking system, as expressed through verbal axes, does not
have this three-way division, but rather a binary division into Locutor and Non-
locutor. These terms are perhaps not the most felicitous, but have been chosen
as they are used in Vásquez de Ruíz's (1988) analysis of the related language
Guambiano; although it is, in fact, unclear from Vásquez de Ruíz's analysis
whether the system in which the Guambiano Locutor and Non-locutor operate is
the same as the Awa Pit system or not.4 The terms `Locutor' and `Non-locutor'
could perhaps be more accurately replaced by `item of interest is (or is not)
epistemic source for this type of utterance'; this is clearly not appropriate for
glossing or easy reference, however.
In discussion of a similar phenomenon in Tibetan languages, the terms
`conjunct' (corresponding to Locutor) and `disjunct' (Non-locutor) have some-
times been used, but these terms have been avoided for four reasons. Firstly,
they refer only to the verb form, and are not used to refer to entites, which will
be done below with Locutor and Non-locutor. Secondly, the terms were chosen
by Hale (1980) for Kathmandu Newari because of the behaviour of these verb
forms in indirect speech subordinate clauses, and this is precisely where they
dier from Awa Pit Locutor and Non-locutor, which never occur in such clauses.
Thirdly, the most precise denition of a conjunct/disjunct system is that given
by DeLancey (1992:57), and this denition includes features which are not found
in the Awa Pit system. Finally, and perhaps most importantly in practical terms,
4Vásquez de Ruíz (1988) only gives examples of Guambiano person marking in statements,
where Locutor corresponds to rst person, Non-locutor to second and third person. As she
gives no examples of questions, it is unclear whether the Locutor and Non-locutor in Guam-
biano retains these correspondences of marking to person in questions, or undergo the same
reassignment as in Awa Pit.
190 CHAPTER 8. PERSON MARKING

`conjunct' has been used in the analysis of North American indigenous languages
to refer to a very dierent feature of verbs, found in Algonquian languages such
as Fox (see, for example, the use of `conjunct' in Goddard (1996)).
The Locutor/Non-locutor distinction is, as is obvious from its name, a
binary one. In statements, Locutor corresponds to rst person, Non-locutor to
second and third person:
(371) (na=na) pala ku-mtu-s
(1sg.(nom)=top) plantain eat-impf-locut
`I am eating plantains.'
(372) (nu=na) pala ku-mtu-y
(2sg.(nom)=top) plantain eat-impf-nonlocut
`You are eating plantains.'
(373) (us=na) atal ayna-mtu-y
(3sg.(nom)=top) chicken cook-impf-nonlocut
`He/she is cooking chicken.'
In these three statements there are two person markers, with a contrast between
Locutor s (rst person) and Non-locutor y (second and third person).
However the person-marking system is not a straightforward rst versus
non-rst system. In questions the correspondence between Locutor and Non-
locutor and the person system is dierent.5 In questions, Locutor corresponds to
second person, while Non-locutor corresponds to rst and third person:
(374) m1n-a=ma (na=na) ashap-tu-y?
who-acc=inter (1sg.(nom)=top) annoy-impf-nonlocut
`Whom am I annoying?'
(375) shi=ma (nu=na) ki-mtu-s?
what=inter (2sg.(nom)=top) do-impf-locut
`What are you doing?'
(376) m1n=ta-s (us=na) a-mtu-y?
where=in-from (3sg.(nom)=top) come-impf-nonlocut
`Where is he coming from?'
While these three questions are all content-word questions, the same system is
in operation for polar questions.
The system, as explained to this point, can thus be summarized as in
Table 8.1.
In fact, in elicitation occasional examples of Locutor marking have been
found on questions involving a rst person; and Non-locutor marking has very
occasionally been found on questions involving a second person:
5As noted earlier, directives, including imperatives, have inherent person marking and are
consequently outside the Locutor/Non-locutor system.
8.2. LOCUTOR AND NON-LOCUTOR 191

Statement Question
1st person Locutor Non-locutor
2nd person Non-locutor Locutor
3rd person Non-locutor Non-locutor
Table 8.1: The distribution of Locutor and Non-locutor

(377) m1n=ta na=na tu-s?

where=in 1sg.(nom)=top be:in:place-locut
`Where am I?'
(378) nu=na Ricaurte=mal puz-ta ki?
2sg.(nom)=top Ricaurte=loc go:out-pfpart Q.(nonlocut)
`Did you go to Ricaurte?'
There are two possible explanations for this. The rst assumes that the speakers
are in some sense semi-speakers of Awa Pit, and that they have been inuenced
by Spanish, with the strictly assigned person marking of Spanish extending into
speakers' Awa Pit. This does not seem likely, given the strength of the appearance
of the usual complex assignment, and the unusualness of the questions involving
this secondary assignment of marking to person.
A much more likely explanation relies on the peculiarness of elicitation
conditions. While the speaker was asked how to say in Awa Pit `where am I?',
the speaker in fact knew perfectly well, at the time of asking, where he was.
In a parallel fashion, when the speaker was asked how to say `did you go to
Ricaurte?', he was fully aware that the addressee had in fact gone to Ricaurte.
If the person-marking system is based on an idea of `epistemic source', as will
be suggested below, these circumstances are precisely those in which the system
would be expected to deviate from its normal pattern.
This suggestion in fact gains force from a consideration of the best de-
scription of the parallel conjunct/disjunct system of marking in Tibeto-Burman
languages, given in Hale's (1980) description of (Kathmandu) Newari. Hale shows
that (a class of) verbs in Newari have a two-choice person-marking system, with
`conjunct' (Locutor) used for rst person in statements and second person in
questions, while `disjunct' (Non-locutor) is used for second and third person in
statements, rst and third person in questions. However, he notes that while
this applies to true questions, rhetorical questions pattern in the same way as
statements: if a speaker knows the answer to a question, and is not asking it
of another person, then a rst person is indicated by the `conjunct' form, just
as in statements. While this aspect of the person-marking system of Awa Pit
needs further research, it is likely that it parallels the Newari system, with truly
information-seeking questions patterning in one way, in contrast to statements
192 CHAPTER 8. PERSON MARKING

and rhetorical questions.6

While no other clear examples of a Locutor/Non-locutor distinction have
been found in the literature on South American languages, there are suggestions
that a related system may be found in Nambiquara. While unfortunately no
detailed paradigms for the interrogative system are available, in describing the
interaction of person-marking and evidentiality Lowe (1972) notes that there
are gaps in marking possibilities, and that for `observational orientation', rst
person marking is possible in declaratives and second person in interrogatives,
while for `dedication orientation', second person is possible in declaratives and
rst person in interrogatives. However the system is complex, with third person
being possible in either orientation, and the marking of rst person in declaratives
and second in interrogatives (and vice versa) being dierent.
The Locutor/Non-locutor distinction in Awa Pit is, then, essentially that
given in Table 8.1 above, with exceptions for rhetorical questions. The deviation
of rhetorical questions can be explained in a straightforward manner by invoking
the concept of `epistemic authority', and this will be examined in the discussion
after exemplication of the full Awa Pit person-marking system.

8.3 The Past paradigm

Thus far the discussion of the Awa Pit person-marking system has been entirely
theoretical, and skated over the actual realization of the system. In particular,
some sleight of hand has been used, in that it has been implied in examples and
through omission that the Locutor/Non-locutor distinction applies to the Subject
of a clause. In fact, in Awa Pit the system applies to a mix of grammatical
functions and semantic roles, and operates dierently in sentences in the Past
tense and non-Past tenses.
The essential complication in the system is that a Locutor entity (re-
gardless of grammatical function) is marked over a Non-locutor element, with
only one sux available to code participants. In some respects this is similar to
the systems of person marking used in Kamaiurá and Mapudungun, where rst
person is marked preferentially over second, and second over third. In Mapu-
dungun, however, much more information is also marked (Grimes 1985), while in
Kamaiurá there are some portmanteau prexes (Seki 1990:383). Both languages
have a relatively straightforward rst, second and third person system; and both
systems mark the distinction in grammatical roles in dierent tenses, unlike Awa
Pit.
Awa Pit sentences which deal with past time can be marked with a variety
of aspect and mood markers, and may or may not have explicit tense markers.
The Past system of person marking applies only to those person markers which
occur in clauses with an explicit Past tense marker  because of the ordering
of axes, person marking always directly follows any explicit Past tense marker.
Here there are three possible person-marking axes: w, which indicates a Locutor
6 See below for further details on the Newari system.
8.3. THE PAST PARADIGM 193

Locutor ta-w Past tense Locutor Subject

t1-s Past tense Locutor Undergoer
Non-locutor t1-zi Past tense Non-locutor
Table 8.2: Allomorphs of the person markers in the Past system (where t1 and
ta are Past tense markers)

Subject; s, which marks a Locutor undergoer; and zi, which shows an absence of
included Locutor entities (see Table 8.2). Note that the imbalance between the
notions of Subject (a grammatical function) and undergoer (a semantic role) is
intentional, and necessary to an understanding of the system.

8.3.1 Locutor Subject marking

The presence of a nal w indicates that the grammatical Subject of a clause is
a Locutor participant. This Subject is most commonly an agent, but may be an
experiencer, patient, or other semantic role, as long as it is a Subject:
(379) nash-na=kima na=na kal
afternoon-inf=until 1sg.(nom)=top work(1)
ki-mtu-ata-w
work(2)-impf-past-locut:subj
`I was working till late.'
(380) shi ayuk=ta=ma libro ta-ta-w?
what inside=in=inter book put-past-locut:subj
`Under what did you put the book?'
(381) k1n-ka=na, na=na Santos=ta
dawn-when=top 1sg.(nom)=top Santos=acc
izh-ta-w
see-past-locut:subj
`At dawn I saw Santos.'
(382) yawa payu=ma shaa-ta-w?
how:many day=inter be:around-past-locut:subj
`How many days were you there for?'
(383) na=na=ma pit-t1-ta-w
1sg.(nom)=top=temp sleep-term-past-locut:subj
`I fell asleep.'
194 CHAPTER 8. PERSON MARKING

(384) payta-ma-ta-w
sweat-comp-past-locut:subj
`I sweated.'
It should be noted that while a w always indicates a Locutor grammatical Sub-
ject, the reverse is not necessarily the case: a Locutor grammatical Subject does
not always imply a nal w (see below).

8.3.2 Locutor Undergoer marking

A nal s person marker after a Past tense morpheme indicates that a Locutor
entity underwent the action of the verb.7 This Past tense s is the most complex
of the three Past system person markers.
When there is a Locutor participant in the Object grammatical function,
a Past tense verb is always marked with s, and this is the most common usage
of s, suggesting that in fact it marks a Locutor Object in the same way in which
w marks a Locutor Subject:
(385) Libardo (na-wa) pyan-t1-s
Libardo (1sg-acc) hit-past-locut:under
`Libardo hit me.'
(386) Demetrio=na t1t-ma-t1-s
Demetrio=top cut-comp-past-locut:under
`Demetrio cut me.'
(387) nu-wa=na, m1n=ma pyan-t1-s?
2sg-acc=top who=inter hit-past-locut:under
`Who hit you?'
However s is also found in two other cases as well. It can be found in utter-
ances where a Locutor participant is the grammatical Subject of an intransitive
verb:
(388) na=na kwayzh-ma-t1-s
1sg.(nom)=top get:tired-comp-past-locut:under
`I got tired/am tired.'
(389) mayN-ma-t1-s
lose:consciousness-comp-past-locut:under
`I lost consciousness.'
(390) na=na=ma pit-ma-t1-s
1sg.(nom)=top=temp sleep-comp-past-locut:under
`I fell asleep then.'
7It should be noted that there is also an s morpheme in the non-Past person-marking system,
but while this morpheme is still related to Locutor, it has a much wider function than s in the
Past system, as will be discussed below.
8.3. THE PAST PARADIGM 195

This use of s to mark a grammatical Subject only appears in the data with a
handful of verb roots:
kwayzh- `get tired'
mayN- `lose consciousness'
nayn- `fall'
n1jul- `gain consciousness'
pit- `sleep'
s1hppayl- `get thin'
tazh- `go down'
These verbs have a semantic feature in common  while they are all intransitive,
their single argument, the Subject, undergoes the action expressed in the verb.
This marking of Locutor undergoer Subjects by s appears to conict with
the earlier statement that w marks all Locutor Subjects. In fact it is also possible
to mark Locutor Subjects of these verbs using w. Thus corresponding to the three
sentences above, the following are also well-formed Awa Pit sentences:8
(391) na=na p1na kwayzh-ma-ta-w
1sg.(nom)=top very get:tired-comp-past-locut:subj
`I got tired/am tired.'
(392) mayN-ma-ta-w
lose:consciousness-comp-past-locut:subj
`I lost consciousness.'
(393) na=na=ma pit kway-ta-w
1sg.(nom)=top=temp sleep drop-past-locut:subj
`I fell asleep.'
There are four possible explanations for this alternation between s and w, as-
suming that it is not simply random: the alternation may be marking a semantic
distinction, a pragmatic distinction, it may be related to language change, or it
may be related to the origin of the person-marking system itself. Of course, these
explanations are not necessarily mutually exclusive. These various possibilities
will be discussed shortly, after all the relevant data is given.
In the data, only the verbs listed above were found to take s as well as
w, but it seems likely that all intransitive verbs with a single participant who is
an undergoer can take this alternation, and that it is simply a lack in the data
that some verbs in this semantic class  for example n1lta- `shiver'  were never
found with s marking. Certainly all verbs from this class which were tested had
the possibility of s marking as well as w marking. It is also worth pointing out
that all these verbs are intransitive. This is probably a semantic rather than a
syntactic fact, as no transitive verbs were found where the grammatical Subject
was semantically an undergoer.
8 Note that the t1 ta alternation in the Past marker is allomorphy.
196 CHAPTER 8. PERSON MARKING

These examples clearly show that s does not mark a Locutor Object: there
are cases where a Locutor Subject is cross-referenced through a verb sux s
also. Perhaps even more interesting are those cases where the person marker s
is present, but is not cross-referencing any participant at all. This can occur in
sentences with intransitive verbs, either stative or active, and in copula sentences,
and indicates that the action or state aected a Locutor entity, even though that
entity was not, strictly speaking, a participant in the action or state, being some
sort of experiencer of the action or state:
(394) kerosín way-a-s
petrol lack-past-locut:under
`Petrol was lacking to me.'
(395) alu ki-ma-t1-s
rain(1) rain(2)-comp-past-locut:under
`(I was on my way to bathe,) it rained on me.'
(396) aympi p1na 1-ma-t1-s
blood very go-comp-past-locut:under
`(I cut myself,) lots of my blood owed everywhere.'
(397) p1na us a-t1-s
very heavy be-past-locut:under
`I found [the bag] was very heavy.'
It is important to note that these verbs remain one place (intransitive) or two
place (copula) verbs. It is not the case that an extra argument, an Object, is
added to the verb, as can be seen by the fact that such an argument cannot be
expressed:
(398) (*na-wa) alu ki-ma-t1-s
(1sg-acc) rain(1) rain(2)-comp-past-locut:under
`It rained on me.'
This cross-referencing of an aected non-argument is similar in intent to
the ethical dative construction of languages such as Spanish, where an aected
party can be indicated in a sentence by including it as a dative argument:
(399) el perro me muri
o
the dog me.dat died
`The dog [went and] died on me.'
However there are two clear dierences. Firstly, the aected party in Spanish
is, truly, an argument, and can be expressed by any appropriate form, either
pronoun or full np, and can be any person:
(400) el perro le muri
o a Santos
the dog him.dat died to(dat) Santos
`The dog [went and] died on Santos.'
8.3. THE PAST PARADIGM 197

In Awa Pit, only an aected Locutor can be indicated, and only in the verb,
not by an argument. Equally, even if it were an argument in Awa Pit, it would
necessarily be a core argument (being cross-referenced in the verb), while in
Spanish and other languages the ethical dative is at least arguably an oblique
argument, implying no change in the verb valency.9
This s person marking in the Past paradigm is thus clearly not cross-
referencing an Object in the way in which w cross-references a grammatical
Subject. An s indicates that a Locutor entity is aected by the action in some
way, whether a participant in the action or not. This concept of `aectedness' is,
however, very broad, and perhaps better stated as `involved in the action but not
as a (volitional controlling) agent'. In any sentence with a Locutor Object (as
well as in some sentences without a Locutor Object) the verb will be marked with
s regardless of the semantic role of the Object; it is most commonly a patient,
but may also be, for example, a theme, in which case it is not `aected' by the
action in the usual sense of the word:
(401) Juan=na (na-wa) izh-t1-s
Juan=top (1sg-acc) see-past-locut:under
`Juan saw me.'
It is, however, involved in the action, and not as an agent.
It is interesting to return at this point to the possible explanations for the
alternation between s and w with some intransitive verbs given above, now that
other uses of s have been seen. The four explanations suggested there were: a
semantic distinction, a pragmatic distinction, language change, and the origin of
the system. The rst two of these tend to appear less likely, now that it is clear
that there is not simply an alternation between s and w with some intransitive
verbs, but that s has a habit of popping up in unexpected places even when there
is no argument that it can cross-reference. That is, a semantic dierence between
the use of s and w based on some element such as volition or control such
as occurs in many uid-s languages (Dixon 1994:7883); a pragmatic distinction
based on the division between foregrounded and backgrounded activities such as
occurs in the Peruvian Arawakan language Asheninca (Judith Payne & David
Payne 1991); or a marking system such as that of Yagua where so marking is
favoured in contexts where a main character is moved to a new scene or at points
of episodic climax (Doris Payne & Thomas Payne 1990:257) will hardly explain
the use of s in clauses where there is no Locutor participant at all.
The third possibility for alternation suggested above was language change.
It could be assumed that, whatever the original use of s and w was, only one
of them was originally used with verbs such as pit- `sleep', and that language
change is leading to the other appearing there also. The most logical a priori
suggestion here is that these verbs originally only occurred with s, and that w
is extending its range, perhaps from marking agents to marking Subjects, with
9There are also languages which add an aected argument through the use of an applicative
construction  here, however, there is usually an extra applicative morpheme in the verb
complex, indicating the change in the verb valency.
198 CHAPTER 8. PERSON MARKING

these intransitive verbs now taking w by analogy with the many intransitive
verbs which have an agent as their single argument. This would, in fact, make
the postulated original system very similar to the northern Colombian language
Ika, which has a basically nominative/accusative cross-referencing system, but a
very small group of about twelve intransitive verbs cross-reference their subjects
in the same way as transitive verbs cross-reference their objects (Frank 1990:22);
the Ika system is, however, a cross-referencing system, with the person marking
only used to mark participants.
Of course it could be considered that a change in the system is a response
of semi-speakers acting under the inuence of a largely nominative/accusative
system in Awa Pit and a completely nominative/accusative system in Spanish.
There are two reasons to reject this analysis, one language-internal and one theor-
etical. Language-internally it is clear that the system of marking is very unusual
in itself, and there seems to be no reason to assume that this apparent additional
peculiarity is in any way a language death phenomenon, rather than simply a
fact of the system. And while the evidence from changing case-marking systems
in language death situations is rather scanty, this phenomenon has been studied
for Dyirbal, where an originally ergative system is replaced by accusative mark-
ing (Schmidt 1985a, Schmidt 1985b). In this case, however, individuals were not
found to be varying from one form to another, but rather each individual was
highly consistent in his/her response (Schmidt 1985a:382) as to the form and use
of ergative marking. This contrasts markedly with the Awa Pit situation, where
the same speaker alternates between one form and another. The situations are
slightly dierent, of course, in that in the hypothesized original Awa Pit sys-
tem there are only a few verbs with unusual marking, while in Dyirbal an entire
structural system is altering.
For the moment the precise synchronic and historical value of these Awa
Pit markers will be left open. The possibility that these alternations between
s and w are inherent to the system will be taken up again when this system is
compared with similar systems in other languages and in discussing the origin
of the system, where it will be seen that the alternation is a likely synchronic
outcome of a potential historical change.

8.3.3 Non-locutor marking

The third sux in the Past person-marking system, zi, is the default elsewhere
marker, stating that there was no Locutor participant in the activity and that
no Locutor entity was aected by the activity.10 The rst of these options is, of
course, syntactic  if there is a Locutor Subject or Object the verb is necessarily
marked with w or s. The second, no Locutor element aected by the activity,
is more a matter of speaker choice. If it rained, a speaker may state simply this
fact:
10The same marker, zi, is also one of the markers in the non-Past system, and indicates
precisely the same concept.
8.4. THE NON-PAST PARADIGM 199

(402) p1na alu ki-ma-t1-zi

very rain(1) rain(2)-comp-past-nonlocut
`It rained heavily.'
Alternatively, as above, speakers may indicate that this rain aected them:
(403) p1na alu ki-ma-t1-s
very rain(1) rain(2)-comp-past-locut:under
`It rained heavily on me.'
Other examples of the use of the Non-locutor marker are:
(404) anshik kata-t1-zi, cigarrillo
yesterday bring-past-nonlocut cigarette
`Yesterday he/she brought cigarettes.'
(405) nu=na Juan=ta pyan-t1-zi
2sg.(nom)=top Juan=acc hit-past-nonlocut
`You hit Juan.'
(406) na-wa=na m1n=ma pyan-t1-zi?
1sg-acc=top who=inter hit-past-nonlocut
`Who hit me?'

8.4 The non-Past paradigm

In the previous section the Past person-marking system was discussed. That
system is only used together with a Past tense marker. When there is no Past
tense marker, the non-Past system is used, regardless of what actual time is being
referred to.
The non-Past system has only two marking possibilities, as opposed to
the three in the Past system. Essentially, the distinction between the Locutor
Subject and Locutor Undergoer markers is collapsed, so that there is one marker,
Locutor, used when there is a Locutor entity involved in any way in an action or
state or aected by an action; and a separate marker, Non-locutor, when there
is no Locutor in an action or state.
These two markers have a variety of allomorphs, listed in Table 8.3. Some
of this allomorphy is relatively clear. It appears possible that, in origin, the
person markers here are derived from the forms of the copula verb i, which is
i in the Non-locutor and is in the Locutor, although it is unclear then why the
Locutor form after a vowel is s rather than is.11 The forms i, y and zero (after
/i/) can easily be internally reconstructed to i. The zero forms after ma and na
are unclear (but see section 8.4.1 below). The form zi is the most complex. This
form occurs in the non-Past system only following the Future tense marker n1,
and it could consequently be considered to be a specic future allomorph; but
11 The change from VVC to VC is, of course, a common change across languages.
200 CHAPTER 8. PERSON MARKING

Locutor is after a consonant or glide

s after a vowel
Non-locutor i after a consonant or glide
; after /i/, ma `Q.past' and na `irr'
y after /a/ and /u/
zi after /1/
Table 8.3: Allomorphs of the person markers in the non-Past system

this has not been done for two reasons. First, while this marker occurs in the
non-Past system only following n1 (there are no other cases where a non-Locutor
person marker follows /1/), there is an identical form in the Past system, with
precisely the same meaning  and there it follows the Past tense marker t1,
which also ends in /1/. And secondly, while the origin of the z is unknown, it is
worth noting that something unusual must occur after /1/, as the combinations
/1i/ or /1y/ are not phonotactically possible in Awa Pit.12
The Locutor forms can thus be used when the sentence has a Locutor
participant as either Subject or Object:
(407) tuk-shi-s
smoke-desid-locut
`I want to smoke.'
(408) pala ku-mtu-s
plantain eat-impf-locut
`I am eating plantains.'
(409) shi=ma ki-mtu-s?
what=inter do-impf-locut
`What are you doing?'
(410) t1lawa a-n sa-s?
tomorrow come-inf Q:unsure-locut
`Will you come tomorrow?'
(411) na-wa=na Santos t1ttu-mtu-s
1sg-acc=top Santos spy:on-impf-locut
`Santos is spying on me.'
12It is perhaps possible that this goes some way to explaining the absence of person marking
after the Terminative morpheme t1  i or y are not possible, and zi would create ambiguity
between t1-zi (Past Non-locutor) and *t1-zi (Terminative Non-locutor); and in the Locutor
forms between t1-s (Past Locutor Undergoer) and *t1-s (Terminative Locutor).
8.4. THE NON-PAST PARADIGM 201

(412) shi=ma (nu-wa) ish-tu-s?

what=inter (2sg-acc) hurt-impf-locut
`What hurts? How are you sick?'
Clearly, of course, the unusual alternation found in the Past system
between Locutor Subject and Locutor Undergoer marking is not possible in the
non-Past. When a Locutor entity is a Subject and an undergoer in the non-Past,
there is no contrastive marking as there is in the Past system: a form kwayzh-tu-s
`I am getting tired' simply indicates some Locutor involvement, not whether the
Locutor is being treated as a (grammatical) Subject or a (semantic) undergoer.
The Locutor form is also used in the non-Past system in a fashion similar
to the Locutor Undergoer in the Past system, to indicate that a Locutor entity
was aected by an activity or state in which they were not a participant. Once
again this occurs with active or stative intransitive verbs and with copula verbs:
(413) p1na ii ki-mtu-s
very be:hot(1) be:hot(2)-impf-locut
`I feel it's hot.'
(414) nyam way-is
salt lack-locut
`I feel that salt is lacking.'
(415) alizh i-s
angry be-locut
`He is angry at me.'
(Note that while in the context in which it was used this last example meant `he
is angry at me', in other contexts the same form could also mean `I am angry'.)
The Non-locutor form in the non-Past system is, unsurprisingly, used in
all other cases  where there is no Locutor participant, and the speaker does
not wish to indicate that a Locutor entity was aected:
(416) nu=na pala ku-mtu-y
2sg.(nom)=top plantain eat-impf-nonlocut
`You are eating plantains.'
(417) nashka alu ki-n1-zi
later rain(1) rain(2)-fut-nonlocut
`It will rain later.'
(418) t1lawa=na Hugo=na Ricaurte=ta puz-na
tomorrow=top Hugo=top Ricaurte=in go:out-inf
ki?
Q.(nonlocut)
`Will Hugo go to Ricaurte tomorrow?'
202 CHAPTER 8. PERSON MARKING

(419) (us=na) atal ayna-mtu-y

(3sg.(nom)=top) chicken cook-impf-nonlocut
`He/she is cooking chicken.'

8.4.1 Person marking on non-verb constituents

Perhaps one of the most unusual features of the person-marking system is the
appearance of person marking on words which are not, by other criteria, verbs:
the negative particle shi (see section 12.6), the semblative postposition kana
(section 5.4.12), and the question markers sa and ki (section 12.3.2).
As these elements are not verbs, they are never marked with tense, and
consequently the person system found is that of the non-Past, with a Locutor
versus non-Locutor distinction only. It is interesting to note that all these words
have an opposition between a Locutor s form and a Non-locutor zero form 
while this is the expected form for stems ending in i, it is not the form that verbs
ending in a take (normally y), except for the negative sux ma and the irrealis
na. This suggests that these non-verbs may originally have only been marked
with Locutor when a Locutor element was present, rather than strictly marked
for Locutor and Non-locutor.
(420) paas akkwihsh m1j kana-s
two mother have.(impfpart) like-locut
`It's like I have two mothers (I'm like having two mothers).'
(421) na=na 1nkal awa shi-s
1sg.(nom)=top mountain person neg-locut
`I am not a mountain person (ie. I am not an Awa).'
(422) t1lawa a-n sa-s?
tomorrow come-inf Q:unsure-locut
`Would you be coming tomorrow?'
In fact it could be maintained that, historically at least, the other two
forms with no Non-locutor y are also non-verbs, and there are other factors
which seem to support this. The question sux ma (see section 12.3.1) has no
possibility of combination with tense marking, for example, suggesting it may
have earlier been a separate particle, grammaticalizing to sux status. Similarly
the counterfactual marker na (section 10.3.6) appears in an unusual position
(being placed after tense marking), and as was suggested in section 7.2.3 may
have originated as the Innitive sux, which while verbal is non-nite and hence
would not be expected to mark person.
The feature which unites these non-verb elements which can be marked
with person is their occurrence in sentence-nal position. In the case of the
semblative postposition and the negative particle (and the complementizer use of
sa, assuming that the complementizer (section 10.2.2) and the question marker
are the same), these words can be used in sentences where they are non-nal,
8.5. DISCUSSION AND COMPARISONS 203

or in subordinate clauses, and in these cases it is impossible for them to mark

person. However in other constructions they are sentence-nal, and can then be
marked with s to indicate a Locutor element in the sentence.

8.5 Discussion and comparisons

Having exemplied the system of person marking in Awa Pit, we can now turn
to an examination of this system in theoretical terms, and also compare it with
a variety of systems that exist in other languages. There are two factors which
require explanation  the Locutor/Non-locutor distinction itself, and the pref-
erential marking of Locutor entities regardless of their grammatical function or
semantic role. These two factors could be explained by separate devices, but
in fact it will be suggested that they have a unitary source in an experiential
evidential system.
After a return to the importance of maintaining a distinction between
person marking and cross-referencing, elements of the Awa Pit system will be
compared with inverse alignment systems and passive constructions. Then a
variety of similar phenomena in dierent languages will be examined, together
with the proposed explanations for these phenomena. Finally it will be shown
that the best way of accounting for the Awa Pit system (and the similar system
in Lhasa Tibetan) is to consider it as the grammaticalization of an experiential
evidential system.

8.5.1 Person marking, but not cross-referencing

The terms `person-marking system', `cross-referencing system' and `bound pro-
noun system' are often used in an interchangeable fashion, to refer to the same
sort of system, one where some morpheme or portion of a portmanteau morph-
eme axed or cliticized to the verb indicates something about the person (and
perhaps other features) of one of the arguments of the verb. Of course person
marking tends to be used when the information being conveyed is rst, second
or third person only, while cross-referencing systems can convey a much broader
range of information (including person). Thus, for example, in her article on
Person marking and discourse in North Arawak languages, Aikhenvald (1995)
uses `person marking' and `cross-referencing' interchangeably: the section Per-
son marking and discourse organization in Bare begins Here I shall consider the
correlation of cross-referencing and discourse-pragmatic properties in Bare, and
so on.
In most of its uses, the Awa Pit person-marking system appears to be a
cross-referencing system, with clauses containing a verb sux agreeing in person
with one of the verbal arguments. The details are quite complex, as explained
above, but in the following two sentences, ignoring details, the nal w cross-
references the rst person Subject pronoun na, and the nal s cross-references
the rst person Object pronoun nawa; the verb suxes agree with the nominal
arguments:
204 CHAPTER 8. PERSON MARKING

(423) anshik=na na=na maza arroba piya

yesterday=top 1sg.(nom)=top one arroba corn
pay-ta-w
buy-past-locut:subj
`Yesterday I bought one arroba (unit of measurement) of corn.'
(424) Libardo na-wa pyan-t1-s
Libardo 1sg-acc hit-past-locut:under
`Libardo hit me.'
This appears to be a simple case of cross-referencing  the person features
marked on the verb are identical to those of verbal arguments (ignoring the
fact that in the rst sentence the features are those of the Subject while in the
second sentence they are those of the Object).
However as noted earlier there are reasons for considering that, in fact, the
verb suxes indicating person are not acting as cross-referencing. To begin with,
person marking occurs on impersonal verbs  that is, person marking can occur
on verbs which do not have arguments. If this marking is cross-referencing this
is completely unexpected, as there are no arguments for the verb marking to
cross-reference:
(425) wat k1n-ma-t1-zi
good dawn-comp-past-nonlocut
`It dawned well (ie. the weather was ne).'
Equally, there are sentences where person marking occurs parallel to that
in sentence (424) above, but a rst-person Object pronoun (that is, an argument)
is impossible:
(426) (*na-wa) pa mal-tu-a-s
(1sg-acc) sun shine-impf-past-locut:under
`The sun was shining on me.'
This marking is person marking, as it indicates something about the person
involved in the action or state (for example, `on me' in the last sentence), but
is not cross-referencing, as it does not necessarily refer to any complement or
adjunct in the sentence.13
Only one of the four elements of the person-marking system does, in fact,
always cross-reference an argument. If the Locutor Subject marking appears,
then the grammatical Subject of the clause is, necessarily, a Locutor argument.
However the three other markers are not necessarily coreferential with any ar-
gument. The Locutor Undergoer (in the Past system) and the Locutor (in the
non-Past system) can occur to indicate that there is an involved non-participant
entity; and the Non-locutor (in both systems) can occur with zero-argument
13The idea of `on me' in the sentence above cannot be represented through the use of a
complement or adjunct; it could be expressed through a circumlocution, such as `it was shining;
I got warm' or `it was shining; I was working outside'.
8.5. DISCUSSION AND COMPARISONS 205

verbs. This latter phenomenon could, of course, be accounted for by introducing

a default option or a dummy subject, as is often done in analyzing verbs such
as llueve `rain-3sg' in Spanish, but this cannot account for the other cases, and
will be avoided at this stage in favour of an analysis which accounts for all cases.

8.5.2 Inverses and passives

At rst glance, the focus on a Locutor participant, regardless of its semantic
role or grammatical function, puts one in mind of inverse alignment systems (to
use Gildea's (1994) distinction between `inverse alignment' and `inverse voice')
or passive constructions. As will be shown, however, there are too many dif-
ferences between the Awa Pit system and these other phenomena to make the
comparison viable; and, of course, these comparisons give no explanation for the
Locutor/Non-locutor distinction.
Like an inverse alignment system, the Awa Pit system appears to make a
distinction related to speech act participants (saps), and to mark the appearance
of an sap element above other elements, regardless of the grammatical function
of the sap. However there are many points of dierence.
Many denitions of inverse insist that there should be some separate, clear
inverse marker to consider something an inverse system, and Awa Pit does not
have this  the person marker and the inverse marker are fused. Related to
this is the fact that (at least in the Past where there is a distinction) the marker
indicating the presence of a Locutor element is not the same  it varies between s
and w, depending at least in part on the grammatical role of the Locutor element.
Even for those denitions which do not involve a clear inverse marker,
Awa Pit fails as an inverse system. Inverse alignment systems necessarily have a
constant distinction between saps and other elements, which is lacking in Awa
Pit, as either second person (in statements) or rst person (in questions) are
treated on a par with non-saps. And nally while in some inverse alignment
systems rst person may be ranked above second person on the marking hierarchy,
while in others second person is ranked above rst person, this ranking is constant
within each language; this is quite distinct from the Awa Pit system where the
relative ranking depends on the speech act in question.
The same types of argumentation can be brought against considering that
the distinction is somehow related to passive. In recent work, Shibatani (1997)
has considered that active sentences are those where the action goes from the ma-
jor participant (prototypically subject), while passive sentences are those where
the action goes to the major participant. There is a hierarchy associated with
this (linked to volition), where sentences involving a rst person object are
more likely to be able to be passivized than others  however Shibatani notes,
on the basis of examples from, for example, Nepali, that this is rst person in
statements, second person in questions. It is not an enormous step to assume
that the Awa Pit system can be considered a grammaticalization of this system,
with obligatory passives when a Locutor element is (would be) an Object. There
are two main problems with this line of analysis. Firstly, just as with the inverse
206 CHAPTER 8. PERSON MARKING

alignment hypothesis, there should be some independent marker of passive; and

second, passives exist in contrast to actives, and in Awa Pit there is no possibility
of a contrast.
Thus it is clear that while there are points of contact between the Awa
Pit system and inverse alignment systems or passive constructions, it cannot be
analyzed in these terms.

8.5.3 Anticipation and Sherpa

Turning from an examination of the focus of attention on Locutor entities to an
examination of the Locutor/Non-locutor constrast, it can be noted that a less
or more grammaticalized Locutor/Non-locutor contrast is found in a variety of
languages, and dierent explanations of this phenomenon have been given for
dierent languages by dierent authors. Of course it is possible for a variety of
explanations to be appropriate for dierent languages  there may be a variety
of phenomena leading to the same distinctions.
Perhaps the most obvious, but probably the least explanatory, explan-
ation for Locutor/Non-locutor marking treats the distinction as, essentially, a
rst person versus non-rst person split, with a peculiar anticipation occurring
in questions. Thus Woodbury remarks in a footnote that
The rst versus nonrst person distinction is widespread in Sherpa,
but the term `rst person' is something of a misnomer. In the in-
terrogative all so-called rst person phenomena are associated with
second person. This is because second person forms in questions
anticipate the use of rst person in the answer.
(Woodbury 1986:192, footnote 3)
While perhaps descriptively adequate, this notion of anticipation does not ex-
plain anything  what determines whether a language will use anticipation in
its person marking?

8.5.4 Conjunct/disjunct and Kathmandu Newari

Perhaps the rst account to attempt a principled explanation of a Locutor/Non-
locutor distinction was Hale's (1980) description of the conjunct/disjunct system
of (Kathmandu) Newari. The system of marking in Kathmandu Newari has
additional elements not found in the Awa Pit system. The conjunct/disjunct
marking is not found with all verbs  Malla (1985:38) claims that Copula Verbs,
Attributive Verbs, and Impersonal Verbs do not inect for the category of person
at all, although the actual situation is more complex than this (see the discussion
of epistemic authority below). However while the distinction is not made for all
verbs, the contrast can be used in subordinate clauses, unlike in Awa Pit, and
here the system depends on logophoricity.
A full discussion of logophoricity is unnecessary here, and only the es-
sential ideas will be given. Hale (1980) shows that, in Newari, the same verb
8.5. DISCUSSION AND COMPARISONS 207

marking which is used to signal a rst person in statements and a second per-
son in questions, the conjunct form, is used in some subordinate clauses such as
indirect speech to indicate that the subordinate subject is coreferential to the
matrix subject, while the disjunct form is used to signal a lack of coreferentiality,
regardless of the precise grammatical person involved.14 Thus schematically:
(427) I said(-Conjunct) that I went-Conjunct
I said(-Conjunct) that you went-Disjunct
I said(-Conjunct) that he went-Disjunct
You said(-Disjunct) that I went-Disjunct
She said(-Disjunct) that I went-Disjunct
He said(-Disjunct) that she went-Disjunct
Shei said(-Disjunct) that shei went-Conjunct
Hale (1980) took this subordinate use of the conjunct/disjunct system as
the basic distinction (and this is the reason for the terms `conjunct' and `dis-
junct'). He extended this use to main clauses by assuming that there was an
unspoken abstract performative matrix clause above any main clause which
varied for speech act. Essentially, statements and rhetorical questions are in-
troduced by an unspoken I say to you: component, while true questions are
introduced by an unspoken I ask you: component. For declaratives, coreference
between the actor of the quote frame (the speaker) and the actor of the subor-
dinate clause leads to conjunct marking, while for interrogatives the conjunct
shows coreference between the goal of the quote frame (the hearer) and the actor
of the subordinate clause.
While Hale's suggestion is ingenious, it does not, unfortunately, account
for everything. It is not clear why coreference should be marked with actors in
declaratives but goals in interrogatives. And it completely fails in Awa Pit in
any case. The dictum of a Newari indirect speech sentence is a nite clause,
in the sense that it is marked for (logophoric) person and the marking depends
on coreference. In Awa Pit, in contrast, indirect speech involves the use of a
non-nite subordinate clause (see section 10.2.3.1 for details), and consequently
the subordinate clause is not marked for person. In the very few cases where
subordinate clauses are nite in Awa Pit,15 the person system is not logophoric
in Hale's sense: all these subordinate clauses are statements, and a Locutor verb
form indicates rst person, Non-locutor indicates second or third person, just as
in main clause statements. Thus the Awa Pit Locutor form is never used for
third person reference, unlike the Newari conjunct and the introduction of an
unspoken abstract performative matrix would lead to non-nite marking rather
than Locutor/Non-locutor marking.
14 It should be noted that while Hale (1980) considers that this indicates the logophoric nature
of person marking in Newari, others, such as Sells (1987), exclude this type of phenomenon
from logophoricity; however what is important here is the phenomenon, rather than a label.
15Finite subordinate clauses are used for direct speech, complements to verbs of perception
and cognition; see section 10.2.1.
208 CHAPTER 8. PERSON MARKING

8.5.5 Interior states in Japanese and Korean

Evans (1996) has noted that there are a variety of interesting interactions between
person marking and predicate types. One of the least grammaticalized and most
easily understandable of these interactions is found in a variety of languages, in-
cluding Japanese and Korean. In these languages there are patterns of interaction
between person and speech act with particular predicate types  `private pre-
dicates', `internal state predicates' or `predicates of subjective authority', those
predicates with meanings related to cognition, perception, bodily states or sen-
sations. In Japanese, for example, a speaker can use a simple construction with
atu `hot' to refer to themselves, but not to others:16
(428) atu-i
hot-nonpast
`[I] am hot.'
(429) *kare wa atu-i
he top hot-nonpast
To refer to the sensations of others, a more complex form involving gar (in the
examples below, gatteiru) must be used, adding the idea of `appears to be' or `is
behaving like'; this form cannot be used in rst person:
(430) atu-gatteiru
hot-gatteiru
`[He/she] is hot, appears to be hot.'
(431) *watasi wa atu-gatteiru
I top hot-gatteiru
However in questions the simple form is used to refer to second person, while the
complex forms are for rst and third person.
This split is not completely grammaticalized in Japanese and Korean: an
omniscient narrator, for example, can say `he/she is hot' rather than `he/she
appears to be hot'. The system is, in fact, semantically transparent. Speakers
can only truly know their own internal state, not that of someone else; speakers
can know (and hence say) that they themselves are hot, but about others they
can only know (and hence say) that these others appear to be hot, act like they
are hot, or claim that they are hot  unless, of course, the speaker is omniscient,
as the narrator of a story may be within the world of the story.
It is not clear that the semantically transparent system of Japanese or
Korean could develop into the Awa Pit Locutor/Non-locutor division. Not only
would the system need to be fully grammaticalized, but it would need to extend
from a relatively small number of interior state verbs, a well-dened semantic
subclass, to cover all verbs in the language. However it does lead to an interesting
idea, that a Locutor/Non-locutor distinction could be based on privileged access
to knowledge.
16 Data from Aoki (1986); glosses modied.
8.5. DISCUSSION AND COMPARISONS 209

8.5.6 Epistemic authority and Kathmandu Newari

This idea of privileged access to knowledge is formalized and explored more fully
in Hargreaves's (1990, 1991) analysis of Kathmandu Newari. As noted earlier, in
Kathmandu Newari only some verbs have a conjunct/disjunct contrast, but Har-
greaves analyses this system as the interaction of two independent systems. There
is a contrast between those verbs which imply an element of control, which operate
with the conjunct/disjunct system, and those which have no element of control,
which are always in disjunct form.17 This control/non-control distinction must be
set aside as being irrelevant in Awa Pit  Awa Pit uses the Locutor/Non-locutor
distinction for all verbs, controlled or non-controlled. However the interacting
Newari system, that of conjunct/disjunct, needs to be examined. Hargreaves
explains this as being related to `evidential authority' in discourse:
From pragmatic theory we make the assumption that one of the
pre-conditions for a declarative speech act is that the speaker has
evidential authority for the information in the utterance; in con-
trast, one of the pre-conditions for an interrogative speech act is
the assumption that the addressee has the evidential authority for
the information in the utterance (cf. Gordon and Lako 1971).
(Hargreaves 1991:381)
Hargreaves then denes a discourse role, that of `epistemic authority' (Hargreaves
1991:381) or `epistemic source' (Hargreaves 1990:189), which is assumed by the
speaker in declarative utterances and by the addressee in (true) interrogative
utterances.18 He also notes that this explains the dierential marking in true
questions and rhetorical questions (Hargreaves 1990:189)  in a true question, it
is assumed that the addressee has knowledge of the event; in a rhetorical question,
the speaker has (or at least claims to have) knowledge of the event, just as in a
statement.
This analysis, then, explains the mysterious Locutor/Non-locutor split
not as a random peculiarity or anticipation, but as a coherent system. Rather
than marking person in a traditional sense, it singles out one participant in any
speech act as being a source of knowledge for that event, as having the epistemic
authority to make a claim about an event.
This assumption that Locutor/Non-locutor systems are related to (claims
of) knowledge can account for the straightforward binary distinction found in
a language such as Kathmandu Newari, but does not explain the unusual dis-
tribution of markers in Awa Pit, where an entity with knowledge is privileged
above others, regardless of its grammatical function or semantic role. But a
17The notion of `control' is somewhat unusual  thus `I felt/became angry' is controlled,
while `I felt/became surprised' is non-controlled. The precise denition of `control' is unim-
portant here; see Hargreaves (1990, 1991) for details.
18`Epistemic authority' is also assigned to a subordinate participant in an indirect speech act
who is coreferential to the matrix subject, although this is not relevant in Awa Pit, as noted
above.
210 CHAPTER 8. PERSON MARKING

system based on marking knowledge suggests an evidential system, and indeed

Locutor/Non-locutor systems are often found together with complex evidential
systems.

8.5.7 Evidentiality and Akha

A Locutor/Non-locutor system is found in a complex interaction with an evid-
ential system in Akha, a language of the Loloish branch of Tibeto-Burman, dis-
cussed in Thurgood (1986). Evidentiality is indicated in Akha through a series
of sentence-nal particles. There are three major evidential divisions, between
a `general' particle @, a series of `sensorial' particles (which indicate visual/non-
visual evidence, an expected/non-expected event, taking place in the past/non-
past), and a series of `non-sensorial' particles (which indicate the expected/non-
expected status of an event, and whether it is past/non-past). This nal series
has a further distinction between whether the subject is conjunct (rst person
in statements, second person in questions, coreferential in indirect speech) and
disjunct (otherwise). Thus Akha has a conjunct/disjunct (Locutor/Non-locutor)
distinction, for subjects only, and for those sentences which are not marked for
visual or non-visual sensorial evidence or given the `general' marker.
The precise use of the Akha non-sensorial particles is unclear from Thur-
good's (1986) description. Introducing them he states that they do not indicate
the source of evidence for a statement (Thurgood 1986:219), which makes them
sound as though they are not evidentials at all; but later, discussing the expected
conjunct non-sensorial particle, he cites personal communication from Inga-Lill
Hansson that it means `I know by experience that this is true'.
If we take this idea of experience one step further, an interesting pattern
emerges. If there is an evidential meaning (in statements) `I know by experience
that this is true', when could such a particle be used? Clearly, in statements,
only for events in which I was involved in some fashion. In questions, evidentials
request the source of knowledge from the addressee, and hence in a question this
evidential would mean `do you know by experience that this is true?'  and
would be restricted to questions about events in which the speaker believes the
addressee was involved. Thus an experiential evidential of this type, with no
explicit restrictions on its distribution, is necessarily restricted to statements of
events in which the speaker was involved and questions about events in which
(the speaker believes) the addressee was involved. Simply by its very meaning
this evidential is restricted to Locutor contexts.19
In Akha there are a variety of other evidentials, such as visual and non-
visual sensorial particles, which do not have a Locutor/Non-locutor distinction.20
If however these other distictions collapsed into one, or there never had been these
19Note that this is slightly dierent from a rst-hand versus non-rst-hand knowledge evid-
ential split, where one does not need to have participated in an activity  seeing an activity
happen counts in many systems as having rst-hand knowledge.
20While Thurgood's (1986) description does not have a Locutor/Non-locutor distinction for
these particles, it would seem likely that there are such restrictions. The use of a sensorial
evidential in (many) Locutor contexts would seem odd: ?I hit you, I know because I saw it.
8.5. DISCUSSION AND COMPARISONS 211

choices, the system would be left with a simple Locutor/Non-locutor contrast

created by the meaning of the experiential evidential itself, without any arbitrary
person restrictions.
In Awa Pit, of course, the distinction is not quite binary. It is in the
non-Past system, and there a straightforward experiential evidential system can
be posited as the origin:
Locutor: Epistemic authority knows X because they are involved in it
Non-locutor: Epistemic authority knows X, but not because they are in-
volved in it
Note that this accounts not only for the Locutor/Non-locutor split, but also
the focus on the epistemic authority regardless of its grammatical function or
semantic role. The three-way division in the Past can be accounted for by intro-
ducing not simply the idea of `involvement', but some concept of agency:
Locutor Subject: Epistemic authority knows X because they did it
Locutor Undergoer: Epistemic authority knows X because they were in-
volved in it but didn't do it
Non-locutor: Epistemic authority knows X, but not because they were
involved in it
This description of the Locutor Undergoer may seem excessively complex, and
perhaps should be replaced by `epistemic authority knows X because it happened
to them'. However this has not been done because of the use of this marker in
sentences such as Santos saw me. It is not clear here that anything `happened
to me', but I was involved in some fashion, and not agentively.
This evidential system would seem to underlie the Awa Pit person-marking
system; synchronically, however, the system is clearly not evidential. The system
is used whenever an appropriate Locutor/Non-locutor entity is present in a sen-
tence, regardless of the evidential value which could be placed on this. If Santos
had been spying on me, but I was not aware of this until Enma told me the next
day, I could still say later:
(432) Santos (na-wa) t1ttu-mtu-at1-s
Santos (1sg-acc) spy:on-impf-past-locut:under
`Santos was spying on me.'
The Locutor Undergoer can be used here, even though I didn't know this because I
was involved, but because Enma told me. Equally it is unclear what the evidential
status would be of a sentence such as:
(433) pyan-tu-s
know-impf-locut
`I know.'
212 CHAPTER 8. PERSON MARKING

`I know that I know because I am involved'? Or for statements of such things as

ethnic origin:
(434) na=na awa i-s
1sg.(nom)=top person be-locut
`I am an Awa.'
`I know I am an Awa because I am involved in it'? In Awa Pit the evidential
nature of these suxes  `epistemic authority knows because'  has been lost,
leaving only the nal statement from the denitions above. But an origin as an
evidential system of this type explains both the Locutor/Non-locutor distinction,
as well as the distribution of forms.
If we assume an origin in this sort of evidential system, we can suggest
that, in the original system, the Locutor Subject marker was in fact simply a
Locutor Agent marker, and was only used when the epistemic authority had
been truly agentive in the event. However with the loss of the evidential nature
of the system, it has become a person-marking system, and much more akin to a
cross-referencing system. While it makes no sense to talk about the alignment of
an evidential system, a cross-referencing system can be talked about in terms of
alignment  nominative/accusative, ergative/absolutive, direct/inverse, and so
on. If we assume that the Awa Pit person-marking system is now more cross-
referencing like than before, it is more likely to have an alignment. This could
explain the alternations found with non-agentive intransitive verbs: as an eviden-
tial system based on agentivity and involvement, these would have been marked
by the Locutor Undergoer evidential, as the Subject was non-agentive. However
as the system has developed towards a cross-referencing system, it has also moved
towards a nominative/accusative system, in line with the case-marking system
of the language, and this has lead to these verbs at least sometimes taking the
Locutor Subject marker, as would happen in a nominative/accusative system,
rather than the Locutor Undergoer, as would have occurred in the evidential
system.
Evidentiality is, of course, common in many Amazonian languages (see,
for example, Aikhenvald & Dixon forthcoming). Closer to the region in which
Awa Pit is spoken, the situation is less clear. Imbabura Quechua has a system of
`validators' which do distinguish between `rst-hand information', `conjecture',
`doubt' and `question element' (Cole 1985:164), but it is not clear that these are
best treated as evidentials, being optional and attaching to the element to which
they refer, rather than the sentence as a whole. One dialect of Guambiano,
related to Awa Pit, has been analyzed as not having grammatical marking of
evidentiality (Vásquez de Ruíz 1992); however another dialect appears to have a
very unusual system, although without enough detail to make an analysis possible
(Triviño Garzón 1992).
The area in which Awa Pit is spoken thus appears to have some elements
related to evidential systems, though many are by no means straightforward
evidential systems, and given the apparent ease with which evidential systems
spread within linguistic areas, it is perhaps not surprising that Awa Pit should
contain some type of evidential-like system.
8.5. DISCUSSION AND COMPARISONS 213

Volitional action Non-volitional action

Locutor agent Conjunct Disjunct
Locutor undergoer Conjunct Disjunct
No Locutor element Disjunct Disjunct
Table 8.4: Conjunct and disjunct forms for Lhasa Tibetan imperfectives
Volitional action Non-volitional action
Locutor agent Conjunct agent Conjunct undergoer
Locutor undergoer Conjunct undergoer Conjunct undergoer
No Locutor (rst-hand) Disjunct Disjunct
No Locutor (inferred) Inferential Inferential
Table 8.5: Conjunct and disjunct forms for Lhasa Tibetan perfectives

8.5.8 Mirativity and Lhasa Tibetan

In fact an alternative origin for a system very similar to the Awa Pit system
has been oered by DeLancey (1992). DeLancey looks at the systems in Kath-
mandu Newari, Akha and Lhasa Tibetan, and suggests that the conjunct/disjunct
opposition was a grammaticalization of a mirativity distinction, essentially old
information or information well-integrated into someone's world-view versus new
or unintegrated information.21
The system of evidentiality in Lhasa Tibetan is described by DeLancey in
a number of works (e.g. DeLancey 1986, 1992), and is rather complex, involving
not only a conjunct/disjunct system, but also a mirativity distinction and an
inferential evidential. Simplifying the system by leaving aside some of these
additional factors (and hence distorting it somewhat), the system is as follows.
The copulas have a straightforward conjunct/disjunct pattern, with no additional
parameters such as volition involved. Non-copula verbs have inections which
are based on the copulas with the addition of other elements. The imperfective
(akin to a non-past) has a distinction for volitional verbs between conjunct and
disjunct, but unlike in Kathmandu Newari this system is not dependent just on
the subject: Locutor actors and Locutor undergoers of volitional actions lead
to conjunct forms, other participants lead to disjunct forms in the verb (leaving
aside mirativity contrasts), as in Table 8.4. For the perfective (similar to past) the
distinctions are more complex. There is a marker of a Locutor agent of a volitional
activity, another marking a Locutor undergoer of any action or a Locutor agent of
a non-volitional action, and two markers for no Locutor participant, depending
on whether the knowledge is rst-hand or inferential; the system is shown in
Table 8.5.
21 See DeLancey (1997) for a fuller discussion of mirativity.
214 CHAPTER 8. PERSON MARKING

The Lhasa Tibetan system is particularly interesting here. If we ignore the

volitional/non-volitional distinction, looking only at the volitional system, and
collapse the evidential distinction between rst-hand knowledge and inferential
knowledge of actions in which no Locutor participant was involved, we have
precisely the Locutor/Non-locutor system as it exists in Awa Pit  complete
with the two-way marking in non-past (imperfective) and the three-way system
in the past (perfective).
In his discussion of the various Tibeto-Burman conjunct/disjunct sys-
tems, DeLancey (1992:57) considers that three elements dene the true con-
junct/disjunct pattern: there is an evidential system including a mirative dis-
tinction (established versus new information); a grammaticalization of this dis-
tinction leading to the Locutor/Non-locutor system; and the use of this system
to distinguish volitional and non-volitional action. DeLancey (1992:60) goes on
to suggest that while this conjunct/disjunct system in Newari (and Monpa) may
have been inuenced by the system in Tibetan, the Akha system cannot be ac-
counted for in this manner, and that we apparently must accept the conclusion
that the grammatical conjunct/disjunct pattern, peculiar as it may be, did in-
deed develop independently from an original evidential or mirative opposition at
least twice (DeLancey 1992:60).
However it is not clear that the Locutor/Non-locutor system did develop
from a grammaticalization of mirativity. DeLancey relies on the natural tend-
ency for statements about 1st person to represent old, and about non-1st per-
sons to represent new, knowledge (DeLancey 1992:57); but it is not clear that
this is, in fact, a natural tendency.
With events which happened in the recent past, it seems at rst reasonable
to consider that, in general, those with which we were involved would have been
more integrated into our established world-view than those which did not involve
us. However this does not necessarily appear to be the case with more distant
past events  it could be stated, for example, that those events which form part
of our culture's history are more likely to be fully integrated into our world-view,
whereas individual relatively unimportant events which happened to us a long
time ago are not integrated, as they are not important. Even with recent events, it
is not necessarily the case in fact, and not at all clear that it is even prototypically
true: is a landslide which killed my neighbour likely to be less integrated into
my world-view than a landslide which knocked me over and swept me down 100
metres? Equally it is not clear that this is true for future events: why should
knowing that Juan is going to attack me be more integrated into my world-
view, treated more as old knowledge, than the knowledge that Juan is going to
attack Santos?
Indeed, in his later article on mirativity DeLancey (1997) notes that past
events are most commonly old information, and considers that it is this fact
which leads a form signalling new information in the present to often indicate
an inference when used in the past. While an inferential form would not normally
be used with a rst person agent, there is no link between non-inferential and
non-rst person in this system which could lead to a Locutor/Non-locutor split.
8.5. DISCUSSION AND COMPARISONS 215

DeLancey (1992) considers that the marking of Locutor/Non-locutor in

Lhasa Tibetan has grammaticalized from the mirativity distinction, presumably
because the same markers can be used to signal mirativity and this is a less
grammaticalized usage. However in Awa Pit there is no evidence of a mirativity
use of these markings  nor, indeed, is there in Kathmandu Newari or in Akha.
In fact, in Akha there is a distinction between unexpected (unintegrated) and
expected (old) knowledge, but this distinction is entirely separate from the con-
junct/disjunct marking. If the marking developed from a mirativity distinction,
we would not expect to nd both operating independently, with conjunct/disjunct
being marked in non-sensorial contexts, while expected/unexpected is marked
across conjunct non-sensorial, disjunct non-sensorial and sensorial. It thus seems
likely that mirativity is not, in fact, the origin of the conjunct/disjunct system.
A much more likely origin for the Locutor/Non-locutor system is one de-
riving directly from the experiential evidential nature of the system. In fact the
Tibetan system, with the element of volitionality involved in it, can be simply
generated from the earlier Awa Pit suggestions with minor changes. Ignoring
the rst-hand versus inferred knowledge distinction in sentences with no Locutor
involvement, the imperfective and perfective systems could be hypothesized as
originating in:
Imperfective system:
Conjunct: Epistemic authority knows X because they are volitionally
involved in it
Disjunct: Epistemic authority knows X, but not because they are
volitionally involved in it
Perfective system:
Conjunct Agent: Epistemic authority knows X because they volition-
ally did it
Conjunct Undergoer: Epistemic authority knows X because they
were involved in it, but didn't do it volitionally
Disjunct: Epistemic authority knows X, but not because they were
involved
The system would now have developed on from this in two ways. In the imper-
fective, the volitionality element is now attached to the event rather than the
epistemic authority (if a Locutor element is involved non-volitionally in a voli-
tional event, it is marked); and mirativity has developed. Mirativity could be
considered to develop from the volitionality element of this system: if you know
something happened because you wanted it, it is likely to not be unexpected or
new; if, on the other hand, it was not volitional on your part, it is more likely to
be unexpected.
There is one further interesting similarity between the Awa Pit and Lhasa
Tibetan systems. This is the use of a greater number of distinctions in the
216 CHAPTER 8. PERSON MARKING

Past/perfective than in the non-Past/imperfective, and this clearly requires some

comment. DeLancey (1992) does not deal with this issue, as it is only present
in the Lhasa Tibetan system not in the other Tibeto-Burman languages; but
its existence in two (genetically) unrelated systems invites speculation. It would
seem best to analyze this in terms of the determinacy of past events compared
to present and future events. With a past event, the precise agency or other-
wise of a participant can be evaluated: I can clearly and unequivocally state,
for example, that I hit Santos. However with non-completed events, agency is
very much less open to evaluation. I may be about to hit Santos, or planning
to hit Santos, and I can be fairly certain that I will be involved in the activ-
ity  but if I set out to hit Santos, it is possible that, in fact, Santos will hit
me instead or as well; and this may account for the lack of distinction between
Locutor agent/Subject and Locutor undergoer marking in the imperfective/non-
Past systems. It is interesting to note here that it is often considered that
agents of completed (past/perfective) actions are more agentive than those of
non-completed activities, having successfully carried out an action rather than
being potential agents.22 This accords precisely with the greater focus on agen-
tivity in Past/perfective systems in Awa Pit and Lhasa Tibetan than in non-
Past/imperfective.

8.6 Summary
To summarize the Awa Pit person-marking system: There is a distinction between
Locutor and Non-locutor throughout the system. This distinction is based on
epistemic authority, with Locutor essentially corresponding to rst person in
statements and second person in questions, while Non-locutor covers rst person
in questions, second person in statements, and all third person. Following the
Past tense marker there are three person markers. Locutor Subject shows that
a Locutor element is the grammatical Subject of a verb; Locutor Undergoer
shows that the action or state expressed by the verb aected a Locutor entity,
whether this entity was directly involved in the action or state or more indirectly
aected; and Non-locutor marking states that no Locutor element was involved
or aected. Where person marking appears but no Past tense marker, there are
only two person markers available: Locutor, which corresponds to both Locutor
Subject and Locutor Undergoer in the Past system; and Non-locutor, identical
to the Non-locutor marking in the Past system.
Both unusual elements of this system, the Locutor/Non-locutor split and
the unusual distribution of markers with the focus always on a Locutor entity
regardless of its grammatical function or semantic role, can be hypothesized as
having developed simply and directly from an experiential evidential system. If
this system involved sentence-nal particles, the use of person marking on both
verbs and some non-verb sentence-nal elements could be explained. However
the Awa Pit system is no longer an evidential system, but has rather changed into
22 See Dixon (1994:99) or Hopper & Thompson (1980) for ideas along these lines.
8.6. SUMMARY 217

a true person-marking system, although it is still not, strictly speaking, a cross-

referencing system. Together with this change from an evidential system towards
a cross-referencing system has come an alternation in marking. As the system
has moved towards a cross-referencing system, the idea of alignment has become
relevant, and the person-marking system has the option of marking non-agentive
intransitive verbs as it did under the evidential system (producing an alignment
pattern of split-ergativity) or as they would be marked in a nominative/accusative
system, the same alignment system as the case-marking.
Chapter 9
Simple tense, aspect and mood
9.1 Introduction
This chapter is a detailed examination of inectional tense, aspect and mood
in Awa Pit. The relatively straightforward system of three tenses is described
rst, then the three inectional aspects are covered: Imperfective, Completive
and Terminative. The derivational aspects  Prospective and Inceptive  were
discussed in sections 6.4.2.2 and 6.4.2.3 respectively, and a discussion of the
perfective aspects expressed through Serial Verb constructions is held o until
section 11.6, after the general discussion of Serial Verbs. Many of the mood in-
ections are discussed in the present chapter as well, although a full discussion of
counterfactuals is postponed until section 10.3.6, and interrogatives and negatives
are discussed in depth in chapter 12.

9.2 Tense
There are two formal markers of tense in Awa Pit, the Past tense inection
and the Future tense inection. These markers do not quite have a one-to-one
correspondence to the notions of past events or states and future events or states,
however. To begin with, they are only used when nite forms are called for, and
consequently most subordinate clauses are not explicitly marked for tense, even
if semantically they refer to past or future time. Equally, there are a variety
of other ways that nite clauses can refer to time  Imperfective aspect can
refer to a scheduled future, as can Necessitive mood, for example. However if
the inectional tense forms occur, the verb almost always refers to a past or
future event; the only exception is the combination of the Past tense marker
and the Completive aspect on stative verbs, which refers to present time (see
section 9.3.2).
Present tense exists in Awa Pit in contrast to Past or Future, but is not
formally marked, being a covert category  if Past or Future inections could
occur, but do not, this signals Present tense.
220 CHAPTER 9. SIMPLE TENSE, ASPECT AND MOOD

9.2.1 Past tense

The Past tense marker t1, which is obligatorily followed by person marking, has
a number of allomorphs, depending on the morphemes which precede or follow
it:
ata used before the Locutor Subject marker w either on stative
verbs (except those ending in a) or after the Imperfective
aspect marker (m)tu
at1 used on either stative verbs (except those ending in a) or
after the Imperfective aspect marker (m)tu, except before
the Locutor Subject marker w
ta used before the Locutor Subject marker w, except where
the conditions for ata are fullled
t1 used elswhere (not before the Locutor Subject marker w,
when conditions for at1 are not fullled)
The change from 1 to a before w is presumably phonetically motivated: the vowel
1 does not form any diphthongs with vowels or semivowels. The presence of the
additional a after stative verbs and the Imperfective marker has a clear historical
explanation, with these forms deriving from a postposed, tense-marked copula
(see section 3.3.2).
Examples of the various allomorphs are:
(435) anshik=na paas shitshu izh-ta-w
yesterday=top two bird see-past-locut:subj
`Yesterday I saw two birds.'
(436) Ecuador=mal 1-ta-w, ma kutnya año
Ecuador=loc go-past-locut:subj now three year
paa-ma-t1
become-comp-term
`I went to Ecuador three years ago.'
(437) na=na uN=ta pana-ta-w
1sg.(nom)=top there=in stand-past-locut:subj
`I was standing there.'
(438) na=na kayl kway-ka=na, pashpa=na
1sg.(nom)=top return drop-when=top child=top
m1j-at1-zi=ma
have-past-nonlocut=temp
`When I returned, she had a child.'
(439) ashaNpa m1j-at1-zi
woman have-past-nonlocut
`He had a wife.'
9.2. TENSE 221

(440) na=na pueblo=ta-s kayl kway-ka=na, kuzhu

1sg.(nom)=top town=in-from return drop-when=top pig
nak-tu-at1-zi
skin-impf-past-nonlocut
`When I returned from town, he was skinning a pig.'
(441) ashaNpa=na m1j-ata-w, ma=na shi
woman=top have-past-locut:subj now=top neg
m1j ki-s
have.(impfpart) be.neg-locut
`I used to have a wife, now I don't.'
(442) Demetrio a-ka=na, kal
Demetrio come-when=top work(1)
ki-mtu-ata-w
work(2)-impf-past-locut:subj
`When Demetrio came, I was working.'
(443) paas awa=na Santos=ta pyan-t1-zi
two man=top Santos=acc hit-past-nonlocut
`Two men hit Santos.'
(444) anshik=na cigarrillo kaa-t1-zi
yesterday=top cigarette bring-past-nonlocut
`Yesterday he brought cigarettes.'
(445) Jos
e au=m1za=na pyan-na-ma-t1-s
José 1pl=(1/2pl.)acc=top hit-pl:obj-comp-past-locut:under
`José hit us.'
(446) Demetrio yal ma=ki pana-t1-zi
Demetrio house beside=at stand-past-nonlocut
`Demetrio was standing beside the house.'
If the sequence t1 occurs after an a (whether this a is part of the verb
stem, part of another ax or part of the Past marker itself) and before the Non-
locutor person marker zi, the t1 may be elided. The tense is still retrievable, as
the person marker zi occurs only in Past and Future tenses, and the absence of
the Future marker n1 clearly signals Past tense. Thus the sequences at1zi and azi
are in free variation.
(447) Libardo=na m1n-a=ma ta-zi?
Libardo=top who-acc=inter pay-nonlocut
`Whom did Libardo pay?'
(448) Carmen=na ap puns1h=mal pana-zi
Carmen=top my behind=loc stand-nonlocut
`Carmen was standing behind me.'
222 CHAPTER 9. SIMPLE TENSE, ASPECT AND MOOD

(449) Santos=na a-mtu kizh-a-zi

Santos=top come-impfpart say-past-nonlocut
`He said that Santos was coming.'
The Past inection t1 itself carries only an indication of Past tense, with
no aspectual overtones. It may be used together with the Imperfective inection
(m)tu, as in sentences (440) and (442); and when used with stative verbs, which
are inherently imperfective, the meaning is naturally imperfective, as in (437)
and (446). When used with no aspectual inections, as in sentences (436) and
(443), it is usually interpreted as perfective, however this is not inherent to t1,
but is rather the use of this inection on its own contrasting with its use together
with the Imperfective marker.
There is only one condition under which the Past inection t1 does not
indicate a past event. This is the combination of stative verb, Completive aspect
and Past tense, which carries the meaning of present tense. As this peculiarity
appears most closely related to the Completive aspect marker, it will be discussed
with this marker in section 9.3.2 below.

9.2.2 Future tense

In many ways the Future tense marking parallels that of the Past. The allo-
morphy of the Future inection n1 is similar to that of the Past inection t1, with
an additional a appearing after the Imperfective marker, and after most stative
verbs  once again, this shows the origin of those forms as two separate words,
the verb stem plus a following copula. As there are only two person-marking
axes for the Future, s and zi, both of which can phonotactically be preceded
by 1, there is no special form corresponding to ta in the Past:
an1 with stative verbs (except those ending in a) and after the
Imperfective marker (m)tu
n1 otherwise
For example:
(450) t1lawa a-mtu-an1-s
tomorrow come-impf-fut-locut
`I will come tomorrow.'
(451) Demetrio=na 1lapa=na, akkwan pampa
Demetrio=top old=top many grandchild
m1j-an1-zi
have-fut-nonlocut
`When Demetrio is old, he will have many grandchildren.'
(452) t1lawa=na uN=pa pana-n1-zi
tomorrow=top there=in(approx) stand-fut-nonlocut
`Tomorrow he will be standing around there.'
9.2. TENSE 223

(453) t1lawa uN=ta pana-n1-s

tomorrow there=in stand-fut-locut
`Tomorrow I will be standing there.'
(454) kazt1la kayl-n1-zi
day:after:tomorrow return-fut-nonlocut
`He will come back the day after tomorrow.'
(455) Demetrio=ta namna-n1-s
Demetrio=acc follow/catch:up:to-fut-locut
`I will catch up to Demetrio.'
The Future tense is not as clear as the Past with regard to its semantics.
While the Past simply states that an event occurred before the reference time,
by its very nature the future is uncertain, and hence the Future cannot simply
state that an event will occur. The Future in Awa Pit can be used to indicate
intention, especially about one's own actions, as in sentence (450) above. When
used about the actions of others, however, intent is not a necessary part of the
meaning. The speaker is always uncertain about these events, and indeed these
sentences are often translated into Spanish with an additional talvez `maybe' or
quizás `perhaps'. Sometimes an event appears likely to the speaker on the basis
of some evidence:
(456) nash-ka alu ki-n1-zi
afternoon-when rain(1) rain(2)-fut-nonlocut
`This afternoon it may rain.'
On the other hand, sometimes the speaker makes a statement for which he or
she has no justication at all:
(457) t1lawa a-n1-zi
tomorrow come-fut-nonlocut
`(He should have been here yesterday.) (Perhaps) he will come
tomorrow.'
A variety of events in the future may also be encoded using the Imper-
fective (section 9.3.1), the Necessitive (section 9.4.2.2) or the Prospective aspect
(section 6.4.2.2).

9.2.3 Present tense

As noted above, Present tense is indicated in Awa Pit through an absence of
possible Past or Future inection, rather than by an actual form. There are in
fact relatively few situations in which this contrast is possible, although these
few situations are the most commonly used verb forms in the language.
The (unmarked) Present contrasts with Past and Future after the negative
marker ki (see section 12.5.3):
224 CHAPTER 9. SIMPLE TENSE, ASPECT AND MOOD

(458) shi ayna-mtu ki-s

neg cook-impfpart be.neg-locut
`I'm not cooking.'
(459) palanca shi m1l ki-ata-w
shovel neg have.(impfpart) be.neg-past-locut:subj
`I didn't have a shovel.'
In this case there is a complex interaction between the tense in which ki occurs
and the form of the non-nite main verb. Thus a main verb in the non-nite
Perfective Participle form may be followed by a Present tense ki, for example,
with this combination indicating a reference to past time. These interactions are
discussed in section 12.5.3.
After the Necessitive mood inection (section 9.4.2.2), there is a contrast
between the Past tense and the Present tense, with this contrast indicating a
distinction between a past necessity and a current necessity.
The most common nite verb form is either an active verb with Imper-
fective aspect marking followed by Past, (unmarked) Present, or Future tense
inection, or a stative verb (inherently imperfective) followed by one of the three
tense inections. That is, with the exception of negative clauses and Necessitive
clauses, verb forms (un)marked as Present are necessarily imperfective, either
inherently (stative forms) or formally marked with an Imperfective aspect inec-
tion (for active verbs). Thus this use of the Present is discussed together with
the Imperfective aspect below.

9.3 Aspect
As noted earlier, aspect can be marked in Awa Pit clauses in three separate ways.
There are a number of aspect inections in Awa Pit, which will be discussed here.
In addition to indicating aspect through inection, other aspectual distinctions
can be made using a variety of auxiliary verbs, discussed in section 11.6, or
through aspectual derivations, discussed in sections 6.4.2.2 and 6.4.2.3.

9.3.1 Imperfective aspect

The Imperfective aspect is the most important aspect in Awa Pit. It is very
frequently used, and for most speakers the citation form of verbs is imperfective,
using either the Imperfective Present form or the Imperfective Participle. While
other aspects are morphologically optional (although clearly not semantically
optional), the Imperfective aspect is obligatory for indicative Present tense.
There are two main issues surrounding the Imperfective aspect in Awa
Pit. The rst is the distinction between the Imperfective aspect and the Im-
perfective Participle, two forms which are clearly related historically, and this
distinction was dealt with in section 3.3.2. The other issue concerns stative verb
stems. For an active verb stem, there is a distinction between Imperfective aspect
9.3. ASPECT 225

and unmarked/non-Imperfective, and this contrast is formally signalled by the

presence or absence of the Imperfective aspect inection mtu (or its allomorph
tu after consonants). Stative verbs, on the other hand, have inherent imperfect-
ive meaning, and cannot take the Imperfective aspect inection mtu. However
the distributional and semantic properties of active verb stems with the inection
mtu and stative verb stems (without mtu) are identical. Thus to facilitate discus-
sion and avoid repetition, stative verb roots will be considered to be obligatorily
covertly marked for Imperfective aspect. That is, for example, a statement such
as an Imperfective form is followed by tense marking indicates that an active
verb stem suxed with mtu is followed by tense inection, and also that a stative
verb stem (with inherent Imperfective) is followed by tense inection.
It should be pointed out that while the inectional form mtu cannot be
used with stative verb stems, it is a true imperfective, not a progressive, as it can
signal habitual activities, as will be discussed below.
The Imperfective aspect inection has three forms. With stative verb
stems, it is covert, with no formal marking. Following an active verb stem ending
in a consonant the form is tu, while following an active verb stem ending in a
vowel, it is mtu. This last form often undergoes assimilation in faster casual
speech, being realized as [ndu] with a dental rather than a bilabial nasal.
The Imperfective aspect inection and the main clause use of the Im-
perfective Participle (see section 3.3.2) are used to cover four dierent situations,
all of which are some facet of imperfectivity: they encode on-going activities or
states, habitual activities or states, planned future activities, and can be used to
make generic statements.
9.3.1.1 On-going activities and states
The major use of the Imperfective is to encode on-going activities and states.
Any event or state which is taking place at the moment of speech is necessarily
on-going, and marked for Imperfective if active or covertly Imperfective if stative,
with the absence of explicit tense marking showing Present tense:
(460) shi=ma ki-mtu-s?
what=inter do-impf-locut
`What are you doing?'
(461) profesora=ta t1tizh-tu-s
teacher=acc wait:for-impf-locut
`I am waiting for the teacher.'
(462) atal ayna-mtu-y
chicken cook-impf-nonlocut
`(Smoke is rising from the house.) She is cooking chicken.'
(463) Demetrio=na banca=ta tansha-y
Demetrio=top bench=in be:sitting-nonlocut
`Demetrio is sitting on the bench.'
226 CHAPTER 9. SIMPLE TENSE, ASPECT AND MOOD

(464) pala way-i

plantain lack-nonlocut
`There are no plantains.'
The Imperfective can also be used with Past or Future marking to indicate
an on-going state or action in the past or future. These forms can occur in simple
sentences, but are commonly used in complex sentences or situations where a
punctual action is located with respect to an on-going activity:
(465) anshik=na shi=ma ki-mtu-ata-w?
yesterday=top what=inter do-impf-past-locut:subj
`What were you doing yesterday?'
(466) ap aympihsh=ta kal ki-wayn-tu-ata-w
my brother=acc work(1) work(2)-help-impf-past-locut:subj
`I was helping my brother work.'
(467) t1lawa a-mtu-an1-s
tomorrow come-impf-fut-locut
`I will come tomorrow.'
(468) Demetrio a-ka=na, kal
Demetrio come-when=top work(1)
ki-mtu-ata-w
work(2)-impf-past-locut:subj
`When Demetrio came I was working.'
(469) na=na pueblo=ta-s kayl kway-ka=na, kuzhu
1sg.(nom)=top town=in-from return drop-when=top pig
nak-tu-at1-zi
skin-impf-past-nonlocut
`When I returned from town, he was skinning a pig.'
(470) na=na uN=pa-s kayl kway-ka=na,
1sg.(nom)=top there=in(approx)-from return drop-when=top
kal ki-mtu-an1-zi=ma
work(1) work(2)-impf-fut-nonlocut=temp
`When I return from there, [my friends] will already be working.'
9.3.1.2 Habitual activities and states
Imperfective forms can be used to show that a particular activity or state is, was,
or will be habitually performed by someone:
(471) nash-ka=na p1na alu ki-mtu-y
afternoon-when=top very rain(1) rain(2)-impf-nonlocut
`It is raining every afternoon (at the moment).'
9.3. ASPECT 227

(472) mes=ayzhpa Ricaurte=ta puz-tu-s

month=every Ricaurte=in go:out-impf-locut
`Every month I go out to Ricaurte.'
(473) pueblo=ta=na akkwan tabaco
town=in=top many tabacco
pay-nin-a-mtu-y
buy-caus-pl:subj-impf-nonlocut
`In the town they sell many types of cigarettes.'
(474) Carmen=na uN=ta tansha-y
Carmen=top there=in be:sitting-nonlocut
`Carmen sits there (as a habit; ie. that's where Carmen sits).'
(475) anya=na, awa=na shi=kasa=ma payna
before=top person=top what=with=inter deer
nak-tu-at1-zi?
skin-impf-past-nonlocut
cuchillo=kasa nak-tu
knife=with skin-impfpart
`In earlier days (before), what did people skin deer with? They skinned
them with knives.'
To refer to a habitual action, however, the adjectivizer m(u) is also often used
(see section 10.4):
(476) anya=na, awa=na shi=ta=ma ayna-m?
before=top person=top what=in=inter cook-adjzr
wisha payl=ta=yN ayna-t ku-m
white:person pot=in=rest cook-sv eat-adjzr
`In earlier days (before) what did people cook in? They cooked and ate
in clay pots.'
9.3.1.3 Scheduled future
As with the English Progressive, the Awa Pit Imperfective can be used for sched-
uled future activities (this use has not been recorded with stative verbs). As
noted in section 9.2.2, the Future tense often carries an idea of probability or
possibility:
(477) t1lawa a-n1-zi
tomorrow come-fut-nonlocut
`(He should have been here yesterday.) (Perhaps) he will come
tomorrow.'
In contrast to this future of possibility, to ensure the idea that a future action is
planned and therefore likely to occur (although not denite), the Present Imper-
fective may be used, together with the appropriate time reference:
228 CHAPTER 9. SIMPLE TENSE, ASPECT AND MOOD

(478) Demetrio ma payu a-mtu-y

Demetrio now day come-impf-nonlocut
`Demetrio is coming today.'
(479) kutnya payu=wa=na 1-mtu-s
three day=in(approx)=top go-impf-locut
`I'm going in three days time.'
The Past Imperfective may be used in a construction equivalent to the
planned future use of the Present Imperfective. This conveys the meaning that,
at some point in the past, an activity was planned for the future. This activity
was not, however, carried out  had it been carried out in the past, there would
be no reason to indicate a planned future idea; if it was still to be carried out,
then the Present Imperfective could be used, to indicate that it was still planned.
The planned future use of the Past Imperfective, then, is used for an irrealis
statement, and must be followed by an explicit statement showing that it was not
true, as otherwise it would be taken simply as an ongoing activity in the past.
(480) na=na martes=na 1-m1z-tu-ata-w,
1sg.(nom)=top Tuesday=top go-incep-impf-past-locut:subj
miercoles 1-ma-t1-mtu-s
Wednesday go-comp-term-impf-locut
`I was planning to go on Tuesday, (but now) I am going to leave on
Wednesday.'
It is interesting to note this usage of the Imperfective as a scheduled fu-
ture. In their typological study of tense, aspect and modality, Bybee, Perkins &
Pagliuca (1994:249251) discuss four morphemes with the meaning of expected
or scheduled future, and note that three of these are imperfectives with a present
or progressive use. As their examples involve Latin, Modern Greek and Baluchi,
they wonder about the distribution of such a combination:
Whether all the instances of this use of imperfectives are in Indo-
European languages because they tend to enjoy more comprehens-
ive descriptions than nonIndo-European languages is unclear.
(Bybee, Perkins & Pagliuca 1994:251)
Thus it is cross-linguistically interesting to see this combination of meanings in
one morpheme in a nonIndo-European language.
9.3.1.4 Generic statements
Generic statements, that is those statements which describe a general truth
throughout time, can also be marked as Imperfective. Clearly these statements
use Present tense rather than Past or Future.
9.3. ASPECT 229

(481) 1nkal awa=na pala waa-mtu-y

mountain person=top plantain sow-impf-nonlocut
`The Awa grow plantains.'
(482) awa=na piya=na t1=kasa waa-mtu-y
person=top corn=top stick=with sow-impf-nonlocut
`People (the Awa) sow corn with a stick.'
(483) cielo=na kwash=pa tu-y
sky=top above=in(approx) be:in:place-nonlocut
`The sky is above [everything].'
It should be noted that while generic statements may be expressed through
the use of the Imperfective, the more common way is through the use of the
adjectivizer m(u):
(484) 1nkal awa tunya ku-m
mountain person rat eat-adjzr
`The Awa eat rats (lit. the Awa are rat-eaters).'

9.3.2 Completive aspect

The Completive aspect, signalled by the sux ma, is used to indicate that the
speaker is talking about the culmination of an action, the point at which the
action indicated in the clause is complete. However this meaning interacts with
the inherent aspectual qualities of clauses in four ways, depending on whether
the clause is (active) telic, (active) atelic, (active) inchoative or stative.
Some active verbs in Awa Pit are inherently telic, and cannot be used in
an atelic manner. For example, the verb say- `look for/nd' has an inherent end-
point  the nding of the object or the person in question. Other active verbs
are inherently atelic, and cannot be used in a telic manner. For example the verb
paa- `fry' cannot be used in a telic fashion. Many active verbs are not inherently
telic or atelic, and may be used in telic propositions or atelic propositions. For
example 1- `go' is not inherently marked for telicity, but when combined with a
nal locational goal the proposition becomes telic, as the activity then has an
inherent end-point: `go' is not inherently telic or atelic, but `go to Sydney' is
telic. In other clauses the same verb may be atelic: `I am going' does not have
an inherent end-point, and is thus atelic.
Inchoative verbs are verbs like kawi- `grow, ripen', and other verbs such
as paa- `become' can be used in inchoative propositions such as `the sky became
dark'. These inchoative clauses are not telic, in that there is no inherent end-
point. If a fruit ripens, there is no particular point before which it is not ripe
and after which it is ripe. But equally it is not atelic, as it is not the case that
the action is the same at all points  while someone can be said to have run as
soon as they start running, a fruit cannot be said to be ripe as soon as it starts
ripening.
230 CHAPTER 9. SIMPLE TENSE, ASPECT AND MOOD

Stative verbs are dierent again. While these are atelic, in the sense that
there is no inherent end-point, there is in fact no action involved in them.
The Completive aspect inection cannot appear without some other tense
or aspect inection following it. Most commonly it is followed by either the Past
tense or the Terminative aspect inection; however in certain circumstances it
can be followed by the Imperfective aspect inection.
9.3.2.1 Completive aspect with telic clauses
In combination with a telic verb or a telic proposition, the Completive signals
what would be expected of an inection indicating completion  it indicates the
reaching of the nal telic point. In these cases, the Completive is followed by the
Past tense or Terminative aspect:
(485) Santos=ta namna-ma-ta-w
Santos=acc follow/catch:up:to-comp-past-locut:subj
`I caught up with Santos.'
(486) us=na Pialap
 pii kwak-ma-t1
3sg.(nom)=top Pialapí river cross-comp-term
`She has crossed the Pialapí river.'
As would be expected of a form whose meaning indicates the point of
culmination of an activity, verbs with a Completive marker do not combine with
words or phrases expressing a period of time:
(487) mansuh cuchillo say-ta-w
all:day knife look:for/nd-past-locut:subj
`I looked for the knife all day.'
(488) *mansuh cuchillo say-ma-ta-w
all:day knife look:for/nd-comp-past-locut:subj
(489) cuchillo say-ma-ta-w
knife look:for/nd-comp-past-locut:subj
`I found the knife.'
It should be noted that verbs such as say- `look for/nd' and namna-
`follow/catch up to' are not polysemous in Awa Pit. They mean essentially `nd'
or `catch up to' but also include the process leading to this end-point. These verbs
must be translated by two distinct words in English depending on the other aspect
markers found with them, but have unitary meanings; the two translations arise
from the combination of lexical and grammatical aspect in English versus those
found in Awa Pit. In particular, the process leading up to a conclusion may be
looked at in a more perfective fashion, as in sentence (487), or imperfectively:
(490) cuchillo say-mtu-ata-w
knife look:for/nd-impf-past-locut:subj
`I was looking for the knife (when you arrived).'
9.3. ASPECT 231

With telic propositions, then, the Completive aspect inection has the
expected function  it focusses attention on the most signicant point of a telic
activity, the end-point. This is the point at which one can say now the action
has happened.
9.3.2.2 Completive aspect with atelic clauses
Awa Pit also contains atelic verbs and propositions. There are verbs which are
inherently atelic, and cannot be used in a telic fashion, such as paa- `fry': it
is impossible to a create a proposition in which something is frying, but is not
considered to be fried until a particular point. Equally, many verbs which are
unspecied for telicity lexically may be used in atelic propositions. While `go' is
telic in `go to Sydney', it is atelic in `I am going now', and hence 1- `go' may be
used in telic or atelic sentences.
With (active) atelic situations, the activity may be considered to be com-
plete as soon as it has begun. It would be possible for the activity to nish as
soon as it had begun, and yet still consider that the activity had been carried
out fully. Using the Completive with atelic propositions containing active verbs
focusses on the point at which the action may be considered to be complete 
the beginning of the activity. Thus contrasting with the sentences above, where
the Completive focusses on the end-point of telic propositions, the following sen-
tences use the Completive to focus on the starting-point of atelic propositions
(compare, in particular, sentence (486), stating that the entire crossing has just
nished, with sentence (491), stating that the beginning of the activity has just
nished):
(491) ma=na alu ki-ma-t1
now=top rain(1) rain(2)-comp-term
`It has just started raining.'
(492) p1na amta pit-ma-t1-s
very at:night sleep-comp-past-locut:under
`I fell asleep very late.'
(493) pit-shi-ma-ta-w
sleep-desid-comp-past-locut:subj
`I began to want to sleep (ie. I got sleepy).'
(494) yak ki-ma-ta-w
be:hungry(1) be:hungry(2)-comp-past-locut:subj
`I got hungry.'
(495) Demetrio=ta=na m1za pyan-a-ma-t1-zi
Demetrio=acc=top almost hit-pl:subj-comp-past-nonlocut
`They almost beat up Demetrio (almost started to hit).'
232 CHAPTER 9. SIMPLE TENSE, ASPECT AND MOOD

With active verbs, then, whether telic or atelic, the Completive indicates
that the activity has been completely carried out. This explains the very com-
mon use of the Completive together with the Resultative construction (see sec-
tion 11.2.2.2): the current state is caused by an activity (Resultative) happening
completely (Completive):
(496) suna kii-ma-t i
that:one marry-comp-pfpart be.(nonlocut)
`He is married.'
(497) na=na ku-ma-t i-s
1sg.(nom)=top eat-comp-pfpart be-locut
`I have eaten.'
9.3.2.3 Completive aspect with inchoative propositions
In general, the Completive does not combine with Imperfective aspect (except in
combination with the Terminative), however this is clearly semantically rather
than formally motivated. The Completive usually focusses on a point in time,
and a particular instantaneous event is unlikely to be viewed imperfectively. In
fact, however, there are a small number of verbs which can combine with both
Completive and Imperfective, because of their inherently inchoative semantics.
These are the inchoative verbs such as kawi- `grow, ripen' or many uses of the verb
paa- `become'. The semantics of inchoativity allow these verbs to combine both
the Completive (indicating, for example, the ripe stage) and the Imperfective
(indicating, for example, that the stage of becoming completely ripe is on-going):
(498) pala kawi-ma-mtu-y
plantain ripen-comp-impf-nonlocut
`The plantains are getting ripe.'
(499) kumpa paa-ma-mtu-y
ready become-comp-impf-nonlocut
`It is becoming ready.'
That is, while with telic propositions and atelic propositions there is a particular
point at which one can say now the action has happened (the end-point and
beginning point, respectively), inchoative propositions are more complex. At the
beginning of a fruit ripening, it is not ripe. But there is no particular point
of time at which one can say now the fruit is ripe; rather than this being a
particular point in time, it is a whole stage, and hence can be looked at as an
on-going activity, through the use of the Imperfective inection.
9.3.2.4 Completive aspect with stative propositions
Finally, ma may be used with stative verbs, and here the meaning departs from
the previous uses. The Completive may be combined with stative verbs and the
9.3. ASPECT 233

Past tense or Terminative, with the total meaning indicating a present state:1
(500) Carmen uN=ta pana-ma-t1-zi
Carmen there=in be:standing-comp-past-nonlocut
`Carmen is standing there.'
(501) Demetrio=na 1nkal awa a-ma-zi
Demetrio=top mountain person be-comp-nonlocut
`Demetrio is an Awa (mountain person).'
(502) sum-ma-t1
be:crouching-comp-term
`He is crouching down.'
This use of the Completive together with Past or Terminative on stative
verbs to convey a present situation appears unusual, but there is a possible his-
torical explanation, suggested by the discussion of anteriors with stative verbs
found in Bybee, Perkins & Pagliuca (1994:7478). With active atelic verbs, the
Completive focusses on the beginning of an action. If it is assumed that this
was originally the case with stative (atelic) verbs such as pana `be standing' in
Awa Pit, then a verb such as pana-ma-t1-zi would originally have focussed on the
beginning of the standing as being in the past. Through an inferential change,
this could come to mean that, since the beginning of the standing was in the
past, the state of standing continued into the present, and hence eventually come
simply to convey present tense. That this idea is historical rather than synchron-
ically relevant can be seen from sentence (501) above: in this case there was no
initial change of state, as Demetrio has always been an Awa. While it is possible
to hypothesize the origin of this construction, then, synchronically it is simply
the case that together with stative verbs the Completive ma followed by Past or
Terminative indicates a present situation.
9.3.2.5 Summary of the Completive aspect
The Completive aspect inection focusses on that period of time when an action
could be said to be complete. For telic propositions, the focus is on the end-point;
for (active) atelic propositions, the focus is on the beginning-point; for inchoative
propositions, the focus is necessarily on a period of time, not a single point; while
with stative verbs there has been a slight shift in meaning, and a stative verb,
Completive aspect and Past tense or Terminative aspect inection focusses on a
current state.
1One of my informants did not spontaneously produce any forms of this type. When sen-
tences such as these in Awa Pit were suggested to him, he would translate them into Spanish,
then suggest a better way of saying them in Awa Pit, using the (unmarked) Present. However,
his Spanish translation and suggested Awa Pit replacement always covered a present situation,
not a past one.
234 CHAPTER 9. SIMPLE TENSE, ASPECT AND MOOD

9.3.3 Terminative aspect

The Terminative aspect inection t1 is somewhat restricted in its uses, being used
only in atelic and telic propositions, never with inchoative verbs. It focusses on
the end of the action expressed by the verb stem to which it attaches; however
as it can follow the Completive aspect, the end on which it focusses may be the
end (Terminative) of the beginning (Completive) of an atelic proposition. Its
precise function depends on what other tense or aspect inections are present,
and consequently it will be discussed in a number of sub-uses: occurring alone,
with Past, with Future, with Imperfective, and with Completive. The focus on
the end of an activity also explains the use of the Terminative aspect inection
in forming the extended Perfective Participle, used in constructions such as the
Resultative (see section 11.2.2.2).
Before examining its uses, it is necessary to establish the existence of
the Terminative aspect inection as a separate grammatical item. The inection
itself, t1, is formally identical to one of the allomorphs of the Past tense inection.
However there are three formal dierences between these inections. First, and
most strikingly, the two inections can be used together on one verb:
(503) ashaNpa az-t1-t1-zi
woman cry-term-past-nonlocut
`The woman had cried.'
Secondly, the Past tense inection is obligatorily followed by person inection,
while when it occurs with no following tense or aspect inection, person marking
does not appear on the Terminative aspect (see below). And nally, when com-
bined with the Imperfective aspect, the Terminative inection appears before the
Imperfective, while Past tense appears after Imperfective aspect.
9.3.3.1 Terminative aspect as a nal inection
When it occurs without Past or Future tense inection, Imperfective aspect in-
ection, or in an extended Perfective Participle, the Terminative inection is the
nal inection on a word. As in this case an absence of inection contrasts with
Past and Future tense, it is considered to be (un)marked for Present tense. How-
ever, somewhat unusually, in this case there is no person marking found, and
the person of participants must be established through context. The reason for
this lack of person marking is unclear, although it does allow the disambiguation
of the Terminative inection t1 (with no person marking) from the Past tense
allomorph t1 with obligatory person marking.
(504) pwitaylap-t1
tell:a:lie-term
`[I] have lied.'
(505) payl-t1=ma
nish-term=temp
`[I] have nished.'
9.3. ASPECT 235

(506) 12 000=ki pay-nin-t1

12 000=at buy-caus-term
`[He] has sold [it] for 12 000 pesos.'
In this case, Terminative aspect indicates that an action (telic or atelic) has just
concluded  it is equivalent to the immediate past usages of the English Present
Perfect. In fact for reasons that will become apparent below in the discussion
of the Terminative with Future and Imperfective, it seems best to state that the
Terminative itself is focussing on the nal point of the action. As it is basically
impossible to state that something punctual (a nal point) is happening at the
moment of speech, the function has slipped slightly, and Present Terminative
indicates an immediately past action.
9.3.3.2 Terminative aspect and Past tense
In combination with the Past tense, the Terminative gives the expected reading
for a past version of the previous function. It indicates that, at some reference
time in the past, the nal point of an event happened. Just as with the Present
Terminative, there is a slight slippage of time, and the Past Terminative indic-
ates that at the past reference time, the event had just nished.
(507) na=na kwa-t1-ta-w
1sg.(nom)=top eat-term-past-locut:subj
`I had just eaten (so I didn't want any).'
(508) t1 kwaa-t1-t1-zi
stick fell-term-past-nonlocut
`They had just felled the tree.'
9.3.3.3 Terminative aspect and Future tense
It might be expected that the combination of Terminative and Future would
indicate that at some future time, an event would just have nished, however it
does not. It rather conveys the idea that (the end of) something is (probably)
just about to happen, and is best translated using English must, in the sense of
`surely this will happen soon':
(509) say-t1-n1-s
look:for/nd-term-fut-locut
`I must be about to nd it.'
(510) kayl-t1-n1-zi
return-term-fut-nonlocut
`They must be about to return.'
(511) kaa-ma-t1-n1-zi
lose-comp-term-fut-nonlocut
`[The path] is about to be lost (ie. vanish).'
236 CHAPTER 9. SIMPLE TENSE, ASPECT AND MOOD

It would seem that this meaning has developed from the focus of the
Terminative on the end of an activity. By stating that the end of an activity is
in the future, rather than simply saying that the activity as a whole is in the
future, there is a strong implication that the activity is currently underway; and
from this it is an easy step to a meaning that, as the event is currently underway,
it will end soon.
9.3.3.4 Terminative aspect and Imperfective aspect
The Terminative aspect inection can be followed by the Imperfective aspect in-
ection. Following this, of course, comes tense marking, but this tense marking
simply places the on-going end of the activity in time. The Terminative aspect
indicates that the speaker is focussing on the end of an activity, while the Imper-
fective states that, at the reference time indicated by the tense, the end of the
activity was happening, it was in a stage of winding down:
(512) mesa=na sa-t1-mtu-s
table=top make-term-impf-locut
`I am nishing making the table.'
(513) nu=na a-ka=na, na=na
2sg.(nom)=top come-when=top 1sg.(nom)=top
pihshka-t1-mtu-ata-w
sweep-term-impf-past-locut:subj
`When you came, I was just nishing sweeping.'
If the end of the activity described is truly punctual, this combination of
aspects still describes the same type of situation, stating in a sense that all the
factors leading to the end are coming together; thus the end itself is denitely
about to happen:
(514) us=na kayl-t1-mtu-y
3sg.(nom)=top return-term-impf-nonlocut
`He is about to return.'
This contrasts with the Terminative and Future tense, which has a stronger idea
of probability rather than certainty (because of the `probable' future idea of the
Future tense, compared to the `scheduled' future uses of the Imperfective).
9.3.3.5 Terminative aspect and Completive aspect
The Terminative aspect inection often follows the Completive aspect inection.
This combination can then be used in any of the above ways  alone, followed by
Past or Future tense, or the Imperfective aspect inection (with tense marking)
 and clearly, given the dierent functions of each of these combinations, the
Completive plus Terminative has a variety of functions.
The Completive aspect inection focusses on the end-point of telic pro-
positions and the beginning of atelic activities, as explained in section 9.3.2, and
9.3. ASPECT 237

the Terminative then refocusses on this point. Used with no following inection,
the combination states that the end of a telic action has just happened:
(515) verano payl-ma-t1
summer nish-comp-term
`Summer is over (has just nished).'
(516) ap aympihsh=na shutta kaa-ma-t1
my brother=top hat lose-comp-term
`My brother has lost [my] hat.'
Or it states that the beginning of an atelic action has just happened:
(517) ma=miN 1-ma-t1 Carmen
now=rest go-comp-term Carmen
`Carmen has just gone.'
(518) ma=na alu ki-ma-t1
now=top rain(1) rain(2)-comp-term
`It has just started raining.'
One particularly important use of this combination is together with a time
period and the verb paa- `become'. This is used to indicate how much time has
passed since something happened, how long ago it was:
(519) kutnya año paa-ma-t1, Ecuador=mal
three year become-comp-term Ecuador=loc
1-ta-w
go-past-locut:subj
`I went to Ecuador three years ago.'
The combination of Completive and Terminative together with Past and
Future tense give the expected readings. With Past, it is stated that at some past
point the end of a telic or beginning of an atelic proposition had just happened;
while with Future, the end of a telic or beginning of an atelic proposition is about
to happen.
The most complex forms are those with Completive, Terminative and
Imperfective aspect inections. The meaning is a combination of the aspects.
For telic propositions, it states that the end-point of the activity is on-going or
about to happen:
(520) piya wan pak-ma-t1-mtu-s
corn all sow-comp-term-impf-locut
`I am about to nish sowing all the corn.'
(521) na=na nukkul-ma-t1-mtu-s
1sg.(nom)=top stop-comp-term-impf-locut
`I am going to stop here.'
238 CHAPTER 9. SIMPLE TENSE, ASPECT AND MOOD

For atelic propositions, the beginning of the activity is on-going or about to

happen:
(522) na=na 1-ma-t1-mtu-s
1sg.(nom)=top go-comp-term-impf-locut
`I'm going now.'
(523) ap su kayl kway-ka=na, ap 1nkwa
my land return drop-when=top my grandmother
izh-n1-ma-t1-mtu-s
see-prosp-comp-term-impf-locut
`When I go back to my home, I will go and see my grandmother.'
9.3.3.6 Summary of the Terminative aspect
The Terminative aspect thus essentially focusses on the end of an activity, al-
though the precise function of the inection depends on whether it is found alone,
or together with other tense and aspect inections. Alone (in Present tense) or
with Past tense marking, the Terminative indicates that at the reference time
given by the tense, the activity had just nished; with Future or Imperfective,
the activity is just about to nish.
While these meanings are disparate, they could perhaps be explained if in
origin the inection was a verb meaning `nish', although no diachronic evidence
for this has been found. In this case in the Present it would mean `I now nish
Verb-ing', leading to an immediate past; with the Past likewise `I nished Verb-
ing then' would lead to immediate past at the reference time. The Future `I will
nish Verb-ing' is appropriate, although why it should be restricted to immediate
future is not clear. The Imperfective also ts here: `I am/was/will be nishing
Verb-ing'. However this is purely speculative, though the meanings themselves
are clear.

9.4 Mood
There are a number of verb axes in Awa Pit which express mood and modality.
Following Lyons (1977:848) and Palmer (1986:2123), `mood' is taken here as a
label for a grammatical category expressed through inection. As a consequence,
two of the markers of modality in Awa Pit, expressing desire and ability, are not
considered to be markers of mood, as they are derivational rather than inectional
suxes  see sections 6.4.2.1 and 6.4.2.4 for a discussion of these axes.
As well as this restriction on the mood inections discussed here, an ad-
ditional constraint has been used. Only those inections which can be used in
main clauses will be discussed. While the exclusion of subordinate markers from
the category of mood is perhaps arbitrary, the various subordinating inections
are more easily discussed as a group in chapter 10.
This section deals then with inections which are used in main clauses
and give some indication of speaker attitude towards the proposition in question.
9.4. MOOD 239

Apart from the unmarked indicative mood, these inections mark obligation and
necessity, potentiality (root possibility) and negative potentiality, counterfactual
irrealis, and directives (imperatives, prohibitives and hortatives). The dierent
mood inections are mutually exclusive.

9.4.1 Indicative
The Awa Pit Indicative is a completely covert category  there is no special
marking, simply an absence of any other mood inection. The full range of
tense and aspect inections is available for the Indicative, while for other moods
there are very severe restrictions on possible combination with tense and aspect
inections.

9.4.2 Obligation and necessity

There are two inections which are used to indicate obligation and necessity,
among other things. These are the Necessitive npa/napa and the Obligative
tpa/tawa, with the rst form in each case used after a vowel, the second after a
consonant. It seems likely that these two markers are historically related, and
consisted originally of two separate morphemes, the Innitive n(a) and Perfect-
ive Participle t(a) respectively followed by a form pa (perhaps a postposition),
although the short forms of these latter two non-nite inections are found only
with a small group of verbs (see section 7.2.8), not with any vowel-nal stem,
unlike the Obligative and Necessitive. Synchronically, however, the two (histor-
ically complex) morphemes are clearly separate, with dierent ranges of use, and
also the distinction between the /p/ in one case, and the /w/ in the other. It is
not clear why these two morphemes should have developed with these two dif-
ferent phonemes, although the relationship is clear: morphemes beginning with
/p/ undergo a morphosyntactic process changing this /p/ to /w/ when they oc-
cur after a vowel. Presumably at the stage where this morphosyntactic rule was
introduced, the Necessitive was already considered to be one morpheme, while
the Obligative was still considered as two, so the /p/ of the rst was retained as
such, while in the latter the /p/ became /w/.
It should be noted that the Obligative and the Necessitive both have ho-
mophonous subordinating inections, the After marker and the dierent-Subject
purposive marker respectively. While these pairs of inections are quite probably
related historically, they are synchronically quite distinct, and the two subor-
dinating inections are discussed in sections 10.3.3 and 10.3.2 respectively. The
probably historical paths taken by these morphemes are discussed there also.
9.4.2.1 Obligative
The Obligative carries the meaning that something has to be done. Most com-
monly the Obligative is used to refer to universal obligations. In these cases,
the Obligative is used without person marking, indicating that an action must
240 CHAPTER 9. SIMPLE TENSE, ASPECT AND MOOD

be carried out, usually as a social or moral obligation, but with no indication of

who the agent should be:
(524) ayna-tpa
cook-oblig
`Cooking has to be done; someone has to cook.'
(525) ii-ka=na=ma, kam-tawa
die-when=top=temp bury-oblig
`When someone dies, they have to be buried.'
(526) 1-m1z-tawa
go-incep-oblig
`It's time to go.'
While it is rarely used to indicate someone's personal obligation, this can
be done. In these circumstances the verb carries person marking:
(527) t1lawa=na a-tpa-y
tomorrow=top come-oblig-nonlocut
`You must come tomorrow.'
The Obligative can thus occasionally be used with person marking, but is
more common without. It cannot be used with any tense or (inectional) aspect
marking.
9.4.2.2 Necessitive
The Necessitive marker, npa/napa, has a variety of functions, covering a range
of concepts from necessity through to future. It can combine with Past tense,
although not with Future tense; the Necessitive is always found with person
marking; and when used for future time reference, it may be used together with
aspect marking, unlike other moods.
Npa/napa can convey the idea of necessity; here it is always the idea of
personal necessity, rather than the moral or social necessity indicated through
the use of the Obligative (see section 9.4.2.1), and the person who necessarily
will do something is the Subject:
(528) na=na sancocho sa-m1z-napa-s
1sg.(nom)=top meat:soup make-incep-necess-locut
`I have to make a meat soup (or else the meat will go o).'
(529) Demetrio=na alcalde=kasa kwinta
Demetrio=top mayor=with talk(1)
kizh-napa-t1-zi
talk(2)-necess-past-nonlocut
`Demetrio had to talk with the mayor.'
9.4. MOOD 241

This concept of personal necessity can extend to a situation of ordering; thus

it is possible to order someone to do something by saying:
(530) t1lawa=na camisa pat-napa-y
tomorrow=top shirt wash-necess-nonlocut
`You have to wash the clothes tomorrow.'
As well as necessity or orders, the Necessitive can be used to convey the
idea of intention, especially with Locutor Subjects. Here there is no idea of
necessity.
(531) t1lawa=na shi=ma ki-npa-s?
tomorrow=top what=inter do-necess-locut
`What are you going to do tomorrow?'
(532) abril=ta=kima kwinta kizh-napa-s
april=in=until talk do-necess-locut
`I will be talking until April.'
More commonly than the notions of necessity, orders or intentions, the
Necessitive conveys some idea of a predictive future, a future which the speaker
believes is likely, but not certain:
(533) Bogota=ta-s kayl kway-ka, verano payl-ma-t
Bogota=in-from return drop-when summer nish-comp-pfpart
a-npa-y
be-necess-nonlocut
`When I return from Bogota, the summer will have nished.'
(534) Demetrio t1lawa a-npa-y
Demetrio tomorrow come-necess-nonlocut
`Demetrio will come tomorrow.'
(535) alu ki-ma-npa-y
rain(1) rain(2)-comp-necess-nonlocut
`It's going to rain.'
(536) atal camisa ku-ma-npa-y
chicken shirt eat-comp-necess-nonlocut
`(The chicken is pecking along towards the shirt.) The chicken will eat
(ie. peck at) the shirt.'
While this morpheme is probably more commonly used for a predictive
future than to mark necessity, it has been labelled the Necessitive as it appears
likely that this use is diachronically prior. A number of cross-linguistic studies 
for example Bybee & Dahl (1989:9094), Bybee, Pagliuca & Perkins (1991:26
29) and Bybee, Perkins & Pagliuca (1994:258264)  have suggested that a
morpheme covering some idea of obligation often evolves to cover intention, and
from there the idea of future can be developed; this appears to t the range of
uses of the Necessitive in Awa Pit.
242 CHAPTER 9. SIMPLE TENSE, ASPECT AND MOOD

9.4.3 Potential and Negative Potential moods

Depending on the sense intended, the English word can may be translated into
Awa Pit in a wide variety of ways. The expression of ability is reported through
the use of the adjectivizer mu (see section 10.4). A learned skill can be reported
using the derivational sux kam `learn' (see section 6.4.2.4). The Potential mood
inection sina and the Negative Potential mood inection satshi are only used
to express root possibility, which reports on general enabling conditions . . . on
general external conditions, such as social or physical conditions (Bybee, Perkins
& Pagliuca 1994:178).
Most commonly, the Potential and Negative Potential are unmarked for
person, and express in this way a general truth about what is possible in the
world. The Negative Potential cannot be sentence-nal, and if it is not followed
by anything else, it is followed by a sux wal:2
(537) attihsh=ta escuela izh-sina
far=in school see-pot
`The school can be seen from far away.'
(538) maza caballo maza carro tayN-sina
one horse one car pull-pot
`A single horse can pull a car.'
(539) 1-satshi paa-ma-t1
go-negpot become-comp-term
`(It has got dark.) It has become impossible to walk.'
(540) p1na t1lkul=na, camina ki-satshi-wal
very dark=top walk(1) walk(2)-negpot-part
`(When) it is very dark, one cannot walk.'
The Potential and Negative Potential can also be used with person mark-
ing. Here it is not a general truth that is being expressed, but rather some
condition in the world which allows the Subject to perform the action (if he or
she wishes).
(541) nyampi=kasa pishkatu ki-sina-y
hook=with sh(1) sh(2)-pot-nonlocut
`You (the addressee) can sh with a hook (because the river is up).'
(542) na=na azhpizh-satshi-s
1sg.(nom)=top open-negpot-locut
`I cannot open it (it is stuck).'
2 The use of this sux wal, which sometimes occurs elsewhere, is not understood, although
it may be some type of participle form, as it is non-nite.
9.4. MOOD 243

It is interesting to note that the Potential and Negative Potential cannot

be used together with tense marking. They necessarily express either that the
action is always possible or not possible under the specied conditions (no person
marking), or that at the specied time the action is possible or not possible for the
Subject because of external conditions (with person marking), with an optional
time reference being added through adjuncts or adverbial clauses.
The Potential and Negative Potential can also be used in the apodosis
(`then'-clause) of counterfactual clauses:3
(543) akki pana-t=na, izh-sina=ma
here be:standing.(impfpart)-cntrfc=top see-pot=temp
`If he were here, we could meet.'
(544) na=na Ricaurte=mal
1sg.(nom)=top Ricaurte=loc
1-ma-t-at=na=ma, ma kwinta
go-comp-pfpart-cntrfc=top=temp ma talk(1)
kizh-satshi-wal
talk(2)-negpot-part
`If I had gone to Ricaurte, we wouldn't be able to talk now.'

9.4.4 Counterfactual irrealis

The counterfactual irrealis inection na appears to be used only in past coun-
terfactual statements, and consequently will be discussed with counterfactuals in
section 10.3.6. It may be preceded by Past tense and followed by person.

9.4.5 Directives
There are a wide variety of directive inections in Awa Pit: imperatives, prohibit-
ives and hortatives. These forms are only used with active verbs: it is impossible
to use a stative verb as a command.
9.4.5.1 Imperatives
There are four markers of imperative in Awa Pit, each with a distinct function.
Two are simple imperatives, a singular ti and a plural tayN. There is also a special
marker zha which indicates both imperative mood and also that the Object is
rst person. There is also a polite imperative, n(a)ka, with only one form for
singular and plural.
All of the imperatives may be used together with derivational suxes,
including the Inceptive aspect marker m1z. They may also cooccur with the
aspectual Serial Verb auxiliaries. There is no special person marking on the
imperatives, with second person being simply understood from the imperative
forms themselves.
3 See section 10.3.6 for details.
244 CHAPTER 9. SIMPLE TENSE, ASPECT AND MOOD

Singular and Plural Imperative The Imperative Singular ti and the Imper-
ative Plural tayN are used as would be expected, to give commands to one or
more people:
(545) pihshka-ti!
sweep-imp.sg
`Sweep!'
(546) ana izh-ti!
this see-imp.sg
`Look at this!'
(547) tuk kway-ti!
suck drop-imp.sg
`Suck!'
(548) 1-tayN!
go-imp.pl
`Go away!'
(549) 1-m1z-tayN!
go-incep-imp.pl
`Go away (at once)!'
Two verbs have been found which have alternative forms of the Imperative
Singular, with the form ending in t rather than ti. These are the verbs ku- `eat'
and sa- `make' (it is possible that there are other verbs which allow this form).
The two alternatives of the Imperative Singular appear to be in free variation:
(550) aza kwa-ti!
quickly eat-imp.sg
`Eat up quickly!'
(551) shap kwa-t!
ripe:plantain eat-imp.sg
`Eat [some] ripe plantain!'
(552) kwa-t, kwa-t!
eat-imp.sg eat-imp.sg
`Eat up, eat up!'
(553) wat sa-t!
good make-imp.sg
`Look after [it]!'
It appears most likely that a small number of verbs originally had a short form
of the Imperative Singular, t rather than ti, but that the paradigm is now being
regularized, so that the long form can be used with all verbs.4
4Compare this with the short form of the Innitive and Perfective Participle taken by a
small set of verbs, including ku- `eat'; section 7.2.8.
9.4. MOOD 245

First Person Object Imperative Awa Pit has a special form of the imper-
ative, zha, which replaces either of ti or tayN if the Object of the verb is rst
person, regardless of whether the Subject is singular or plural.
It is possible to use zha with an inherently transitive verb:
(554) t1tizh-zha!
wait:for-imp.1obj
`Wait for me!'
(555) kanpa-zha!
accompany-imp.1obj
`Come with me!'
However, for semantic reasons, this type of sentence is unusual. It is much more
common to nd zha indicating the Object of a ditransitive verb, or of a derived
transitive or ditransitive verb:
(556) an kwin-zha!
more give-imp.1obj
`Give me more!'
(557) pyal sap-zha!
money receive-imp.1obj
`Take the money from me!'
(558) pit-nin kway-zha!
sleep-caus drop-imp.1obj
`Let me get to sleep!'
(559) ayna-wayn-zha!
cook-help-imp.1obj
`Help me cook!'
(560) pan pay-nin-zha!
bread buy-caus-imp.1obj
`Sell me [some] bread!'
(561) sun an-zha!
that pass-imp.1obj
`Pass me that!'
As the ax zha clearly indicates the Object of the verb, it is uncommon for
it to cooccur with an explicit Object, however this can be done, and presumably
indicates some extra emphasis on the Object:
(562) na-wa=na pit-nin-zha!
1sg-acc=top sleep-caus-imp.1obj
`Let me sleep!'
246 CHAPTER 9. SIMPLE TENSE, ASPECT AND MOOD

Polite Imperative The nal of the four imperative forms is n(a)ka, the Polite
Imperative. In fact, its use appears not so much to do with the relations between
the speaker and the addressee, but rather with the content of the message. The
Polite Imperative can be used to issue direct commands:
(563) aN=pa a-nka!
here=in(approx) come-plt:imp
`Come here!'
(564) kuhsa-naka!
get:up-plt:imp
`Get up!'
However it is also used to convey ideas much closer to suggestions  for
example, when giving a friend or relative a gift of food which needs to be cooked,
it is normal to say:
(565) ayna-t kwa-nka!
cook-sv eat-plt:imp
`Cook and eat it!'
It is also the form which speakers tend to use when they are issuing a
warning to someone, even to someone to whom they would issue an order using
the normal singular or plural Imperative:
(566) wat=miN shaa-nka!
good=rest walk-plt:imp
`Watch where you're going! (walk properly!)'
(567) n1jul-naka!
remember-plt:imp
`Don't forget! (Remember!)'
The form of this ax is either naka or nka. The latter form is only used
after a few verb stems ending in vowels, including a- `come', 1- `go' and ku- `eat'.
These verbs are precisely those verbs which take the short form n of the Innitive
rather than the usual na (see section 7.2.8.1). It appears likely, then, that the
Polite Imperative naka was originally two morphemes, the rst of which is the
Innitive. It is also possible to speculate about the origin of the second part, ka.
Bybee, Perkins & Pagliuca (1994:211) claim that it is common for the
forms used in imperative sentences to also occur in subordinate clauses, particu-
larly the protases of conditional sentences . It will be seen in section 10.2.2 that
subordinate indirect questions, similar in many ways to the protasis (`if'-clause)
of a conditional sentence, are formed with the Perfective Participle (relative past),
Imperfective Participle (relative present) or Innitive (relative future), followed
by a complementizer, either ka (non-future) or sa (future). In these construc-
tions, the relative future Innitive is only followed by the future complementizer
9.4. MOOD 247

sa. However an imperative is irrealis, similar to future, but also non-future in

intent. It is thus possible to speculate that the Innitive and the non-future
complementizer have fused in main clauses to create an imperative, with an
unexpressed matrix clause I'm wondering if you would . . . . This unexpressed
matrix clause would also explain the more polite, suggestive qualities of the Polite
Innitive.
9.4.5.2 Prohibitives
There are two Prohibitives in Awa Pit: Singular mun and Plural man. These
negative imperatives are used as expected, to direct someone to not do something.
They may be used with intransitive verbs, and with inherent or derived transitive
and ditransitive verbs:
(568) kutku-mun!
tell:a:lie-prohib.sg
`Don't lie!'
(569) man ki-mun!
move(1) move(2)-prohib.sg
`Don't move!'
(570) na-wa=na pyaNta-mun!
1sg-acc=top kill-prohib.sg
`Don't kill me!'
(571) tazh-nin-mun!
fall-caus-prohib.sg
`Don't drop it! (Don't let it fall!)'
(572) kwin-mun!
give-prohib.sg
`Don't give [him any]!'
(573) na-wa pyan-man!
1sg-acc hit-prohib.pl
`Don't hit me!'
(574) suna=kana ki-man!
that=like do-prohib.pl
`Don't do (like) that!'
A special negative imperative form is not unusual in South America, where lan-
guages often have a variety of negation markers; and negative imperatives are
found in a number of languages, for example Retuarã (Strom 1992:137).
248 CHAPTER 9. SIMPLE TENSE, ASPECT AND MOOD

9.4.5.3 Hortatives
There are two suxes which are clearly hortative in Awa Pit: the Hortative
Dual pay and the Hortative Plural shayN. There is also a third sux tu which
is perhaps best considered a singular hortative form, although it varies slightly
from the normal idea of a hortative.
The Hortative Dual and the Hortative Plural The Hortative Dual and
the Hortative Plural are second person, addressed to the listener. The Dual form
inherently includes the speaker as well as the hearer, and is always translatable
by `lets'. The Plural form normally includes the speaker, but it is possible to
nd examples where the speaker is not included. Both hortatives are often used
together with the Prospective aspect derivational sux n1 (see section 6.4.2.2).
(575) ku-pay!
eat-hort.du
`Lets eat!'
(576) 1-n1-pay!
go-prosp-hort.du
`Lets go!'
(577) kwa-shayN!
eat-hort.pl
`Lets eat!'
(578) 1-shayN!
go-hort.pl
`Lets go!'
(579) na-wa=na kwin-shayN!
1sg-acc=top give-hort.pl
`Give me [some]!'
(580) kwa-wayn-shayN!
eat-help-hort.pl
`Help [me] eat!'
First Person Hortative The ax tu is perhaps best considered as a special-
ized variety of hortative, although an ax with a similar meaning in Barasano is
referred to as a rst person present singular imperative (Jones & Jones 1991:79).
It is not commonly used, apart from with a few motion verbs  in part, perhaps,
because it is so easily confused, both formally and semantically, with the Imper-
fective Participle (m)tu in its main clause uses. Indeed Calvache Dueñas (1989)
simply considers 1tu and atu to be special forms of 1mtu `be going' and amtu `be
coming'.
9.4. MOOD 249

However there are clear semantic dierences between the Imperfective Par-
ticiple in (m)tu and the First Person Hortative tu. The Imperfective Participle
has a wide variety of uses, while the Hortative has only one:
(581) 1-mtu
go-impfpart
`I am (in the process of) going; I am going (now); I am going
(tomorrow).'
(582) 1-tu
go-hort.1sg
`I am going now, I'm o.'
The Imperfective Participle may be used to indicate that a process is underway
(I'm on the path walking), or for a scheduled future (I'm about to go; I'll go
tomorrow), among other uses. The First Person Hortative can only be used to
state that an action is about to happen, or at least that that is what is planned
(it is of course possible for someone to say 1tu, then for them to be distracted by
something and stay longer).
For any verb stem ending in a consonant, the Imperfective Participle and
the First Person Hortative are formally identical; and although they are semantic-
ally distinct, the Imperfective Participle (in its scheduled future use) includes the
use of the First Person Hortative:
(583) nukkul-tu
stay-impfpart
`[I] am staying (here at the moment).'
(584) nukkul-tu
stay-hort.1sg
`I'm stopping here.'
The First Person Hortative can be combined with the Completive marker
ma (see section 9.3.2), where the Completive focusses on the commencement of
activity for atelic verbs:
(585) 1-ma-tu
go-comp-hort.1sg
`I'm going now.'
(586) sal-ma-tu
play-comp-hort.1sg
`I'm going to play now.'
In this case, however, it is easily confused with the Completive plus Past tense,
which is phonemically distinct but phonetically often produced as the same string:
250 CHAPTER 9. SIMPLE TENSE, ASPECT AND MOOD

(587) nukkul-ma-tu [nukUlmaru nUkUlmaro]



stay-comp-hort.1sg
`I'm staying/stopping (here).'
(588) nukkul-ma-ta-w [nukUlmaraw nUkUlmaro]


stay-comp-past-locut:subj
`I stayed (here).'
The First Person Hortative is very infrequent, and this may perhaps be
because of the possibility of confusing it with either the Imperfective Participle
or the Past tense. It seems to be used almost exclusively with motion verbs, in
particular in the very common expression 1tu `I'm o now'.
Chapter 10
Subordinate clauses
10.1 Introduction
In Awa Pit, as in all languages, there are constructions in which one clause, a
subordinate clause, is dependent on another clause, a matrix clause. These con-
structions must not be confused with mainauxiliary constructions, with which
they share many features, or with Serial Verb or Conjoined Clause constructions:
these latter three types of construction are dealt with in the following chapter,
where the distinction between mainauxiliary constructions and subordinate con-
structions in particular is examined.
On a theoretical level subordinate clauses, the subject of this chapter,
are often divided into three types: complement clauses, relative clauses and ad-
verbial clauses (Longacre 1985:237), although in Awa Pit as in many other South
American indigenous languages (see, for example, Moore 1989), a fourth type
needs to be introduced, the clausal nominalization. In Awa Pit many of these
distinct clause types share morphology: the two Participles are used in some
complement clauses, adverbial clauses and clausal nominalizations; the Innitive
is used in complement clauses and adverbial clauses; and the Adjectivizer is used
in relative clauses and clausal nominalizations.
Subordinate clauses in Awa Pit fall on a continuum. At one end, the
subordinate clause is identical to a matrix clause, except for the fact that it
is embedded  it retains all clausal properties, with the verb taking the usual
number of arguments expressed in the normal fashion and order, any modiers
can be present, the verb shows the usual range of forms for tense, aspect, mood
and person, and negation is expressed in the normal way. At the other end of
the continuum are clause-types such as the relative clause, which is obligatorily
without a Subject although other arguments and modiers are expressed in the
usual fashion, the verb has a xed form with no indication of tense, aspect, mood
or person, and if negation is present it is expressed outside the clause.
Relative clauses and adverbial clauses are straightforwardly dened, in
that they are clauses which function in the same way as other nominal modiers
such as adjective phrases, or modify verbs or propositions in a similar fashion
to adverb-like adjuncts. These two clause-types are examined in sections 10.4
252 CHAPTER 10. SUBORDINATE CLAUSES
Clause Type Subject Tense Aspect Mood Negation Person Stative
Complement clause 1 C + + + + + + +
Simultaneity A + - + - - - +
Counterfactual A + - - - + - +
Complement clause 2 C + - - - + - +
Absolute A + - - - + - +
Complement clause 3 C + - - - + - +
Dierent Subject purposive A + - - - - - -
After A + - - - - - -
Concessive A + - - - - - -
Complement clause 4 C - - - - - - -
Same Subject purposive A - - - - - - -
Relative clause R - - - - - - -
Nominalization N - - - - - - -

Table 10.1: Subordinate clauses and their type (Complement, Adverbial, Relat-
ive or Nominalization), together with whether they have Subjects, sentence-like
tense, aspect, mood, negation and person, and can be used with stative verbs.
The complement clause types are discussed in the following section.

and 10.3 respectively. However complement clauses and clausal nominalizations

need to be examined and contrasted more fully.
Noonan (1985:42) denes complementation as the syntactic situation that
arises when a notional sentence or predication is an argument of a predicate, al-
though he then immediately follows this by noting that for our purposes, a
predication can be viewed as an argument of a predicate if it functions as the
subject or object of that predicate. Other denitions, such as that of Givón
(1990:515), use a more restrictive denition immediately, with sentential com-
plements being propositions functioning in the role of either subject or object
arguments of the verb. These more restrictive denitions will be argued against
in section 10.2.3.2, where it will be claimed that, for Awa Pit, it is best to consider
that complement clauses can be arguments of postpositions as well as verbs 
but what is important here is that complement clauses are functionally equivalent
to noun phrases, being used as arguments. They are discussed in section 10.2.
In addition to clauses of this type, however, Awa Pit has clauses which
are functionally equivalent to nouns. Internally these clauses lack some clause-
like features: they cannot have Subjects, and tense, aspect, mood, person and
negation are not expressed. However the verb itself has not acquired noun-like
features: it can still be modied by adverbs, and with the exception of the
obligatorily unexpressed Subject, all arguments are shown in the usual way. But
externally, the clause acts like a noun, not a noun phrase: it can be modied by
adjectives, for example. These clausal nominalization structures are dealt with
in section 10.5.
For the purposes of comparison, a quick summary of the various subor-
dinate constructions is given in Table 10.1, together with their possibilities of
taking Subjects, tense, aspect, mood, (clause-level) negation, and person, and
indicating whether they can be used with stative verbs.
10.2. COMPLEMENT CLAUSES 253
Clause Subject Sentence-like TAM Participle Innitive Complementizer
Complement clause 1 + + - - -
Complement clause 2 + - + + +
Complement clause 3 + - + + -
Complement clause 4 - - - + -

Table 10.2: Features of the four dierent complement clause types

10.2 Complement clauses

As noted in the previous section, complement clauses are those subordinate
clauses which are arguments, either arguments of verbs or of postpositions. In
Awa Pit there are four dierent complement clause structures: (1) a sentence-like
complement, identical to a matrix clause; (2) a clause with a Subject, no aspect
or mood, with relative tense shown by the two Participles and the Innitive,
and the usual clausal negation, with a clause-nal complementizer; (3) a clause
identical to the previous type, but without a complementizer; and nally (4) a
clause with no Subject, no tense, aspect, mood or negation, and an Innitive
verb form. The relevant distinctions are shown in Table 10.2.
The majority of complement clauses in Awa Pit function as Objects, either
of verbs or postpositions. There is only one construction with a non-Object
complement clause, where the clause acts as the Subject (s) of a predicate.
In addition to these complement types, there is also another device which is
used to convey meanings in Awa Pit which in other languages can be conveyed by
complementation, and which in some theoretical senses can be considered notion-
ally to fall under the classication of complementation. Awa Pit has a variety of
derivational axes which express meanings often covered through complement-
ation, in particular the Causative nin `make/let', the Auxiliative marker payn
`help' and the Desiderative shi `want'. While in theoretical terms these may
be considered as complementation where the matrix verb has `co-lexicalized'
(Givón 1990:538), been `predicate-raised' (Givón 1980:338), or undergone `lex-
ical union' (Noonan 1985:75) with the subordinate verb, they will simply be
considered as derivational axes here  for further details on their syntax and
semantics, see section 6.4.

10.2.1 Complement clause 1: Sentence-like complement

Sentence-like complements can be used with a number of dierent predicate types.
They are always used to report direct speech; they are one of the options for the
cognition verbs pyan- `know' and min- `think'; and they are always used for
complements to verbs of perception:
(589) Alfonso=na [ ``miimal 1-t1-mtu-s'' ] kizh-t1
Alfonso=top [ Chucunés go-term-impf-locut ] say-term
`Alfonso said, I'm going to Chucunés.'
254 CHAPTER 10. SUBORDINATE CLAUSES

(590) na=na [ mamaz i ]

1sg.(nom)=top [ other be.(nonlocut) ]
min-ta-w
think-past-locut:subj
`I thought it was something else.'
(591) na=na [ Santos=na Demetrio=ta
1sg.(nom)=top [ Santos=top Demetrio=acc
pyan-tu-a-zi ] izh-ta-w
hit-impf-past-nonlocut ] see-past-locut:subj
`I saw Santos hitting Demetrio.'
In Awa Pit the sentence-like complement is identical to a fully formed sen-
tence: from the above sentences it is possible to extract the subordinate clauses
 miimal 1t1mtus `I'm going to Chucunés', Santosna Demetriota pyantuazi `San-
tos was hitting Demetrio' and mamaz i `it is something else'  and use them
as full sentences. That there is truly subordination in the above sentences can
only be seen from the positioning of the two clauses, one being embedded in the
other, with the subordinate clause in the position appropriate for Objects of the
matrix verb.
The correspondence between the tense of sentence-like complement clauses
and their time reference depends on the particular type of complement-taking
predicate. For direct speech and cognition complements, relative tense is used;
for complements to verbs of perception, absolute tense is used.1 That is, for
direct speech and cognition complements, the tense used in the subordinate clause
depends on the time reference of the complement event relative to the time of
the matrix event. Thus in the rst two examples above, the subordinate verb is
in Present tense, showing that the complement event was cotemporaneous with
the matrix event (both in the past, as shown by the Past tense matrix verb). For
complements of verbs of perception, on the other hand, absolute tense is used:
the tense indicates the time reference of the complement event relative to the
moment of speaking. Thus in the third example above, the subordinate verb is
in Past tense, as it occurred prior to the moment of speaking; the hitting clearly
occurred cotemporaneously to the seeing, but this is not indicated by the tense
marking.
While the above examples have the subordinate clause placed in the stand-
ard Object position in the matrix clause (after the matrix Subject and before the
matrix verb), this positioning is in fact quite rare for sentence-like complements.
Much more commonly, the subordinate clause is moved to sentence-nal position,
after the matrix verb, often with a pause between the verb and the complement
(see also section 3.2):
1See Chung & Timberlake (1985:210) for further discussion of the distinction between rel-
ative and absolute tense and its use cross-linguistically.
10.2. COMPLEMENT CLAUSES 255

(592) anshik kizh-t1, [ ``miercoles=na Ricaurte=mal

yesterday say-term [ Wednesday=top Ricaurte=loc
shaa-ta-w'' ]
be:around-past-locut:subj ]
`Yesterday [he] said, I was in Ricaurte on Wednesday.'
(593) na=na izh-ta-w, [ Santos=na
1sg.(nom)=top see-past-locut:subj [ Santos=top
Demetrio=ta pyan-tu-a-zi ]
Demetrio=acc hit-impf-past-nonlocut ]
`I saw Santos hitting Demetrio.'
(594) Demetrio=na pyan-i=ma, [ na=na
Demetrio=top know-nonlocut=temp [ 1sg.(nom)=top
Santos=ta pyan kway-ta-w ]
Santos=acc hit drop-past-locut:subj ]
`Demetrio now knows [that] I hit Santos.'
Extraposition of a subordinate clause from Object position to sentence-
nal position is quite common in SOV languages (Noonan 1985:84), and can
also be understood to aid comprehension in a language such as Awa Pit, which
commonly ellipses arguments. For example, if sentence (593) were modied, so
that Carmen, rather than I, saw the activity in question, it would become:
(595) Carmen=na izh-t1-zi, [ Santos=na Demetrio=ta
Carmen=top see-past-nonlocut [ Santos=top Demetrio=acc
pyan-tu-a-zi ]
hit-impf-past-nonlocut ]
`Carmen saw Santos hitting Demetrio.'
If the subordinate clause is in Object position, it is still be comprehensible:
(596) Carmen=na [ Santos=na Demetrio=ta
Carmen=top [ Santos=top Demetrio=acc
pyan-tu-a-zi ] izh-t1-zi
hit-impf-past-nonlocut ] see-past-nonlocut
`Carmen saw Santos hitting Demetrio.'
However, at least one of Carmen, Santos and Demetrio was almost certain to
have been in the preceding discourse, if this sentence occurred in natural speech,
and one or more of these participants would normally be ellipsed (see section 3.2).
If, for example, Carmen were ellipsed, it would be unclear where the boundaries
of the subordinate clause were:
(597) [ Santos=na Demetrio=ta pyan-tu-a-zi ]
[ Santos=top Demetrio=acc hit-impf-past-nonlocut ]
izh-t1-zi
see-past-nonlocut
`[Carmen] saw Santos hitting Demetrio.'
256 CHAPTER 10. SUBORDINATE CLAUSES

(598) Santos=na [ Demetrio=ta pyan-tu-a-zi ]

Santos=top [ Demetrio=acc hit-impf-past-nonlocut ]
izh-t1-zi
see-past-nonlocut
`Santos saw [X] hitting Demetrio.'
Consequently the extraposition of sentence-like subordinate clauses to sentence-
nal position certainly reduces the possibility of ambiguity, and quite probably
aids in the overall processibility of these sentences.
Complement clause type 1, as noted above, is always used for reporting
direct speech. Other complement clause types can be used for reporting indirect
statements (see section 10.2.3.1) and indirect questions (see section 10.2.2), but
there is no method for reporting indirect commands, and consequently all direct-
ives (see section 9.4.5) are reported using direct speech, with the full morphology
associated with the directive:
(599) [ ``pyan-mun!'' ] kizh-ta-w
[ hit-prohib.sg ] say-past-locut:subj
` Don't hit [me]!, I said.'
The matrix verb for reporting direct statements or commands is always kizh-
`say', as is also the case for indirect statements.2 Direct questions, like indirect
questions, can be reported using m1ma- `ask'.
Complements to verbs of perception are likewise only expressed using com-
plement clause type 1. Complements to some verbs of knowledge and belief, on
the other hand, can be expressed using either type 1 or type 2 complement clauses.

10.2.2 Complement clause 2: Indirect questions

Syntactically, the most complex of the Participle/Innitive complements are
those which contain a Subject and a complementizer. This complementizer is
either ka or sa, and semantically these complements are, very approximately, in-
direct questions. They show that, at the time specied by the matrix clause, the
matrix clause Subject did not know about the truth of the subordinate clause
event. These clauses are the Object of verbs such as m1ma- `ask', min- `think',
mayN- `forget', and pyan- `know'.
(600) na=na mayN-ta-w [ Santos 1-t
1sg.(nom)=top forget-past-locut:subj [ Santos go-pfpart
ka ]
nonfut:cpltzr ]
`I forgot that Santos had gone.'
2 See section 10.2.3.1 for a discussion of the verb kizh-, and of some other speech verbs in
Awa Pit.
10.2. COMPLEMENT CLAUSES 257

(601) shi pyan ki-s [ alcaldía=ta-s

neg know.(impfpart) be.neg-locut [ mayor's oce=in-from
m1n=ma a-t ka ]
who=inter come-pfpart nonfut:cpltzr ]
`I don't know who has come from the mayor's oce.'
(602) na=na min-tu-s [ us=na
1sg.(nom)=top think-impf-locut [ 3sg.(nom)=top
kwa-t ka ]
eat-pfpart nonfut:cpltzr ]
`I am wondering if he has eaten.'
(603) na=na [ kukum ka
1sg.(nom)=top [ possum be:permanently.(impfpart)
ka ] min-ta-w
nonfut:cpltzr ] think-past-locut:subj
`I wondered if it was a possum.'
(604) na shi pyan ki-s
1sg.(nom) neg know.(impfpart) be.neg-locut
[ a-mtu mizha=ma ka
[ come-impfpart how=inter be:permanently.(impfpart)
ka ]
nonfut:cpltzr ]
`I don't know if he is coming or what.'
(605) na=na Marcos=ta=na m1ma-ta-w
1sg.(nom)=top Marcos=acc=top ask-past-locut:subj
[ mizhaka=ma a-mtu sa ]
[ when=inter come-impfpart fut:cpltzr ]
`I asked Marcos when he would be coming.'
(606) kutnya domingo paa-ma-t1, Santos=ta
three Sunday become-comp-term Santos=acc
m1ma-ta-w [ Ricaurte=mal mizhaka=ma puz-na
ask-past-locut:subj [ Ricaurte=loc when=inter go:out-inf
sa ]
fut:cpltzr ]
`Three weeks ago I asked Santos when he would go out to Ricaurte.'
All of the verbs which take this complement type can also be used with at
least one other complement type, depending on meaning; the meanings expressed
through the use of this complement type are those where the matrix clause Sub-
ject does not know the truth of the subordinate clause. Thus min- `think' may
be used with complement clause type 2, in which case it means something close
to `wonder whether'; for the meaning `think that', the verb takes a sentence-like
258 CHAPTER 10. SUBORDINATE CLAUSES

complement; min- `think' may also be used to mean something like `intend', in
which case it takes complement clause type 3. MayN- `forget' takes complement
clause type 2 when it means `forget that'; when it means `forget to', it is used
together with complement clause type 3. M1ma- `ask' can be used with comple-
ment clause type 1 to encode a direct question, or with complement clause type
2 for indirect questions. And the verb pyan- `know' only takes complement type
2 when it is negated or desiderative; when the matrix Subject knows the truth
of the subordinate clause, the sentence-like complement type is used.
The choice of Perfective Participle, Present Participle or Innitive is, es-
sentially, a relative tense choice (as it is for complement type 3). If, at the time
signalled by the tense of the matrix verb, the subordinate event is over, the Per-
fective Participle is used (relative past); if the subordinate event is on-going or is
in a scheduled future, the Imperfective Participle occurs (relative present); if the
complement event is in an unspecied future at the time of the matrix event, the
Innitive is used (relative future). Thus it is possible to contrast sentences (600)
(602) with the Perfective Participle, sentences (603)(605) with the Imperfective
Participle, and sentence (606) with the Innitive.
The two complementizers, ka and sa, are in complementary distribution.
Ka is used as the complementizer when the subordinate clause is non-future,
relative to the matrix clause (sentences (600)(604)); sa is used when the subor-
dinate clause is future, relative to the matrix clause (sentences (605) and (606)).
Essentially, then, the distinction appears to be whether what is not known (the
subordinate clause) can truly be known  if it has happened or is happening 
or whether it can only be hypothesized as possible to know at some time in the
future  if it has not yet happened.
It is important to note that the ka/sa dierence is quite distinct from the
usage of the dierent subordinate verbs to mark relative tense. The Perfective
Participle, indicating relative past, is naturally only ever accompanied by ka, the
non-future complementizer (sentences (600)(602)); and likewise the Innitive,
indicating relative future, is only ever associated with sa, the future complement-
izer (sentence (606)).3 However the Imperfective Participle can cover two times
 either it is indicating that the subordinate action is on-going at the time in-
dicated by the matrix verb, in which case the non-future complementizer ka is
used, as in sentence (603); or it indicates that the event of the subordinate clause
is future, but a scheduled future, in which case the future complementizer sa is
associated with it, as in sentence (605).
Complement clauses type 2 sometimes occur in the usual Object position
 after the matrix Subject, but before the matrix verb  as in sentence (603).
Just as with sentence-like complements, however, they are more commonly extra-
posed, and occur in sentence-nal position, as in all sentences above except (603).
While simple questions have a strong division between polar questions
(those which ask about the truth of a statement) and content questions (those
which ask about a particular item in a sentence), with entirely dierent syntax for
3 But see the discussion of the Polite Imperative in section 9.4.5.1 for a possible combination
of the Innitive and ka.
10.2. COMPLEMENT CLAUSES 259

the two question types (see chapter 12), these indirect question complements do
not make this distinction, with the same syntax used for all complements, whether
what is unknown is the general truth of the subordinate clause (sentences (600),
(602) and (603)), or a particular item (sentences (601), (605) and (606)). Indeed,
an indirect question can even be formed on top of a multiple-choice tag question
(sentence (604)).

10.2.3 Complement clause 3

10.2.3.1 Complement clause 3, complement of a verb
Unlike many Amazonian languages, which tend to use direct speech rather than
indirect speech (Derbyshire & Pullum 1986b:19), Awa Pit has a separate con-
struction used for indirect statements. The complements of these constructions
in Awa Pit are quite similar to indirect question complements (type 2), with the
same morphological possibilities, and marking relative tense in the same fashion.
However unlike indirect question complements, indirect statement complements
do not contain a complementizer.
Awa Pit has a number of dierent verbs which indicate some form of oral
interaction. By far the most frequent of these is kizh- `say', which can be used
as a hyponym for a number of the others. This verb can take an Object and a
Second Object: the person to whom the utterance was made (the addressee); and
also the actual item of speech, usually as a subordinate clause, although a noun
phrase such as an anaphoric or cataphoric ana `this' is possible. In addition to
this more general verb, there are other verbs: m1ma- `ask', pata- `speak', kwinta
kizh- `converse, chat, tell a story', kanta kizh- `sing', kutun- `advise'. However,
with the exception of m1ma- `ask', these verbs have dierent argument structures,
and none of them can be used with a subordinate clause indicating the content
of the original utterance. Pata- `speak' refers to the physical act of speaking,
and can be used intransitively to refer to whether someone spoke or not, or can
be used with a language name as an Object to indicate the language spoken.
Kutun- `advise' may have an Object, the addressee, but the content of what was
advised, if present, is indicated in a separate, juxtaposed, clause. Kwinta kizh-
`converse' and kanta kizh- `sing' are both intransitive.4
M1ma- `ask' and kizh- `say' are used to report direct or indirect questions
or statements respectively. When they are used to report direct questions or
statements, complement clause type 1 is used, as was discussed and exemplied
in section 10.2.1:
4 It is perhaps an interesting cultural fact that the two words specically used for non-
utilitarian modes of speech, kwinta kizh- `converse, chat, tell a story' and kanta kizh- `sing'
both involve borrowings from Spanish: cuenta `he/she is telling a story' and canta `he/she is
singing'.
260 CHAPTER 10. SUBORDINATE CLAUSES

(607) Carmen=na na-wa m1ma-t1-zi [ ``m1n=ta=ma

Carmen=top 1sg-acc ask-past-nonlocut [ where=in=inter
1-shi-mtu-s?'' ]
go-desid-impf-locut ]
`Carmen asked me, Where do you want to go?.'
(608) Gregorio=na [ ``ap ashaNpa=na az-tu-y'' ]
Gregorio=na [ my woman=top cry-impf-nonlocut ]
kizh-t1-zi
say-past-nonlocut
`Gregorio said, My wife is crying.'
When an indirect question is made using m1ma- `ask', the indirect question com-
plement is type 2, with a complementizer (see section 10.2.2):
(609) Carmen=na na-wa m1ma-t1-zi [ m1n=ta=ma
Carmen=top 1sg-acc ask-past-nonlocut [ where=in=inter
1-shi-mtu ka ]
go-desid-impfpart nonfut:cpltzr ]
`Carmen asked me where I wanted to go.'
The only word which can be used for indirect speech, kizh- `say', takes comple-
ment clause type 3, without a complementizer:
(610) Gregorio=na [ ashaNpa=na az-tu ]
Gregorio=top [ woman=top cry-impfpart ]
kizh-t1-zi
say-past-nonlocut
`Gregorio said that his wife was crying.'
Just as with complement clause type 2, complement clause type 3 shows
relative tense through the use of alternative non-nite forms. The Perfective
Participle indicates that, at the reference time established by the matrix verb
kizh-, the event discussed in the subordinate clause had already occurred; in
contrast the Imperfective Participle shows that at the reference time the event
was occurring or was planned to occur. The Innitive indicates that at the
reference time the subordinate event was in the future. This form with the
Innitive is seldom used, possibly because of the chance of confusion with the
complement of intention (discussed in section 10.2.4.1), which can also involve the
matrix verb kizh- and an Innitive verb, although this construction obligatorily
lacks a Subject. For indirect statements with a relative future speech clause, the
Imperfective Participle is usually used instead, in its scheduled future reading.
Further examples of complement clause type 3 are:
(611) Mar
a=na [ miimal puz-ta ] kizh-t1
María=top [ Chucunés go:out-pfpart ] say-term
anshik=na
yesterday=top
`María said yesterday that [X] had gone out to Chucunés.'
10.2. COMPLEMENT CLAUSES 261

(612) [ Demetrio a-mtu ] kizh-t1 Carmen=na

[ Demetrio come-impfpart ] say-term Carmen=top
`Carmen has said that Demetrio is coming.'
(613) Eduardo=na [ t1lawa piya pak-tu ]
Eduardo=top [ tomorrow corn harvest-impfpart ]
kizh-tu-y
say-impf-nonlocut
`Eduardo says that [X] will harvest corn tomorrow.'
A few further points can be made about indirect speech. Both the matrix
verb and the subordinate verb can have Subjects, as occurs, for example, in
sentence (610). If one of the Subjects has already been mentioned in previous
discourse, then it can be ellipsed, and this can lead to ambiguity. Thus sentence
(611) is ambiguous, in a context-free situation. It can either be parsed as above,
with María being the speaker (Subject of kizh- `say'), and a previous discourse
participant (X) being the Subject of the subordinate verb, or it could be analyzed
as below, where the Subject of kizh- `say' has been ellipsed, and María is the
Subject of the subordinate clause:
(614) [ Mar
a=na miimal puz-ta ] kizh-t1
[ María=top Chucunés go:out-pfpart ] say-term
anshik=na
yesterday=top
`[X] said yesterday that María had gone out to Chucunés.'
If the Subject of the matrix verb has been ellipsed, the sentence must stand as it
is, and the two meanings must be disambiguated through context. If, however,
the Subject of the subordinate clause is the ellipsed participant, the sentence
can be disambiguated by placing the Subject of the matrix verb after the matrix
verb itself  sentence (615) can only correspond to the analysis given for sentence
(611), never that of (614).
(615) [ miimal puz-ta ] kizh-t1 Mar
a=na
[ Chucunés go:out-pfpart ] say-term María=top
`María said that [X] had gone out to Chucunés.'
This postposing of the matrix Subject is always an option, in fact, and is done
in sentence (612), where it is not necessary for the purposes of disambiguation.
There is in fact a further ambiguity in sentences with only one Subject,
which can only be resolved by context. Without context, sentences such as (613)
are most likely to be interpreted as having the two Subjects coreferential  that
is, the most likely translation is `Eduardo says that he, Eduardo, will harvest
corn tomorrow'. However this interpretation can be defeated by a suciency of
context, as occurred, for example, in the original context of sentence (611), where
it was clear that the person who went to town was someone other than María
herself.
262 CHAPTER 10. SUBORDINATE CLAUSES

Adverbs are likewise open to rearrangement, and may appear after the
verb to indicate that the adverb belongs to the matrix clause, rather than the
subordinate clause. In sentence (611), once again, the adverb has been moved to
sentence-nal position, to indicate that the saying took place yesterday, rather
than that the going to town occurred yesterday.5 Had the word anshik `yesterday'
occurred directly before the subordinate clause, it would have been unclear which
of the clauses it modied.
Complement types 3 and 4, with non-nite verbs and no complementizer,
are necessarily physically positioned in the usual Object slot, directly before the
verb, when they are Objects. This contrasts quite strongly with complement
types 1 and 2, where the preferred ordering of constituents has the complement
Object being extraposed to sentence-nal position.
Finally, the similarities between the indirect statement construction and
the optional hearsay evidential construction must be noted. The two construc-
tions are closely associated  being translated identically into Spanish by speak-
ers  and clearly the evidential construction has developed historically from the
indirect statement construction. For further discussion, and the evidence that
they are probably separate constructions, see section 11.4.
10.2.3.2 Complement clause 3, complement of a postposition
Awa Pit has a series of postpositions, discussed briey in section 4.6 and exem-
plied more fully in section 5.4. Most commonly, the argument of a postposition
is an np:
(616) Pueblo Viejo=ki=na cruz pana=na
Pueblo Viejo=at=top cross be:standing.(impfpart)=top
`At Pueblo Viejo there is a cross.'
(617) suna=akwa 1-ta-w
that=because go-past-locut:subj
`I went because of that.'
(618) las cinco=kima kal ki-n1-s
the ve=until work(1) work(2)-fut-locut
`I will work until 5 o'clock.'
However in addition to having an np as an argument, the postpositions can take
an argument which is a complement clause of type 3: clausal features such as a
full set of arguments and modiers and clausal negation are expressed, but tense,
aspect, mood and person are not  the verb has the choice of three non-nite
inections, the Perfective Participle expressing relative past, the Imperfective
Participle for relative present, and the Innitive for relative future, as is usual
for type 3 complement clauses:
5 In fact, as only one sentence component can be moved to the sentence-nal position, it seems
likely that María, the Subject of kizh-, has not been moved to sentence-nal position as the
speaker decided that moving anshik `yesterday' was more important for correct interpretation
of the sentence.
10.2. COMPLEMENT CLAUSES 263

(619) [ us ii-ta ]=ki=na cruz pana=na

[ he die-pfpart ]=at=top cross be:standing.(impfpart)=top
`At the place where he died there is a cross.'
(620) [ mamá ish-tu ]=akwa 1-ta-w
[ mother sick-impfpart ]=because go-past-locut:subj
`I went because [my] mother was sick.'
(621) [ Demetrio kayl-na ]=kima kal ki-n1-s
[ Demetrio return-inf ]=until work(1) work(2)-fut-locut
`I will work until Demetrio returns.'
Since in these sentences a subordinate clause is appearing in the place of
an np, these subordinate clauses will be considered to be complement clauses.
At a theoretical level it is in fact possible to consider postpositions as having
predicate features and taking arguments, and hence to say that in this situation
a notional sentence or predication is an argument of a predicate (Noonan's
(1985:42) denition of a complement clause); this underlies the common usage
of something being the object of a preposition. This theoretical issue will not
be taken up here, however: it will simply be noted that as these subordinate
clauses are occurring where an np would be expected, they should be con-
sidered complement clauses, in a similar way to that in which Thráinsson (1979)
treats Icelandic that-clauses and innitival clauses in prepositional phrases as
complement clauses.6
While these clause-plus-postposition elements correspond in meaning to
adverbial clauses in many languages (such as English), this is not a sucient
criterion for considering them as adverbial clauses. In their survey of adverbial
clauses, Thompson and Longacre note:
it is crucial to point out that, although we have tried to identify
the major types of adverbial clauses which we have found in the
languages we have looked at, we are by no means claiming that
a relationship which may be signaled by an adverbial subordinate
clause in one language must be so signaled in every other.
(Thompson & Longacre 1985:174)
Their examples of constructions with meanings corresponding to adverbial clauses
involve juxtaposition and clause-chaining, however the same principle applies
for complement clauses: just because in many languages the meaning intended
corresponds to that of an adverbial clause (that is, in the terms used here, a
non-argument clause), there is no reason that Awa Pit cannot use a complement
clause (that is, a clause which is an argument to a verb or postposition) to encode
this meaning.
6Thráinsson (1979) actually argues that complement clauses are nps; this is not the issue
here, which is simply that for Icelandic, as for Awa Pit, there is no reason not to consider that
complement clauses occur with adpositions.
264 CHAPTER 10. SUBORDINATE CLAUSES

It could be argued that while these words following clauses have the same
form and meaning as the postpositions, they in fact belong to a separate class of
subordinators, in the same way as English until in until 6 o'clock is considered
a preposition but until in until he nished is considered a subordinator (Quirk,
Greenbaum, Leech & Svartvik 1972:318). However in English, unlike in Awa Pit,
there are clearly a variety of words which are subordinators, used in adverbial
clauses, which are not equivalent to prepositions: while, as, if, although, and
so on. In Awa Pit on the other hand, adverbial clauses are always formed with
special verb inections. Equally, in English the form of the subordinate clause
(for example, he nished in until he nished) is not a standard complement clause
structure (*until that he nished), while in Awa Pit the equivalent clause is.
There also seems to be no reason to separate the `subordinator' and `post-
position' uses syntactically. Both uses have the same range of occurrence in
sentence, whenever one occurs the other can, and the Topic marker can follow
a `subordinator' under precisely the same conditions in which it can follow an
identical `postposition' (see section 14.2.2).
It is true that not all postpositions listed in section 4.6 can be found after a
clause. However this appears to be a semantic rather than a syntactic restriction.
All locational and temporal postpositions can occur after a clause (ki, ta, pa, mal,
kima), as can the causal akwa `because (of)' and the similarity marker kana `like'.
Those which cannot be used as `subordinators' are the human Object marker ta,
the possessive pa, the comitative or instrumental kasa, and patsa indicating that
something is like something else in size. These all require their argument to be a
clearly dened person or object, and it appears that subordinate clauses cannot
indicate this.
Although there thus appears to be no formal reason for separating the
class of postpositions into `postpositions' and `subordinators', and it will not
be done here, if this were done the subordinators could perhaps join a single
class together with the two complementizers discussed in section 10.2.2. In the
analysis chosen here, there are dierences between these two groups of words 
for example, the complement clause plus complementizer substitutes for an np,
while the complement clause before a postposition substitutes for an np before
that postposition. However there are similarities, with the same features in the
two types of complement clause: both have Subjects and other arguments, clausal
negation is possible, relative tense is shown by the Participles or Innitive, and
there is no aspect, mood or person marking.
The construction consisting of a clause and a postposition is thus being
treated precisely as such  a complement clause, followed by a postposition:
(622) [ shaa-t ]=kana=yN mitt1 ish-tu-s
[ walk-pfpart ]=like=rest foot hurt-impf-locut
`My feet hurt as though I had walked.'
10.2. COMPLEMENT CLAUSES 265

(623) [ wakata nya ayna-mtu ]=ta=na Carmen=na kuzhu

[ cattle meat cook-impfpart ]=in=top Carmen=top pig
m1l-t1-zi
take-past-nonlocut
`Carmen took a pig to (where) [they] were cooking meat.'
(624) [ ap mamá tayaz-tu ]=akwa=na
[ my mother miss-impfpart ]=because=top
1-ma-t1-mtu-s
go-comp-term-impf-locut
`I'm going [back] because my mother misses [me].'
(625) [ taytta shi m1j ki ]=kana
[ father neg have.(impfpart) be.neg.(impfpart) ]=like
i-s
be-locut
`I am like I don't have a father (It is as though I didn't have a father).'
(626) [ kwata-na ]=kima ku-ma-ta-w
[ vomit-inf ]=until eat-comp-past-locut:subj
`I ate until I vomited.'

10.2.4 Complement clause 4

10.2.4.1 Complement clause 4 as Object: Intention
Verbs such as kizh- `say', min- `think' and mayN- `forget' have all been mentioned
in previous sections. Kizh- `say' and min- `think' can be used with complement
clause type 1 (see section 10.2.1); kizh- `say' also has the option of complement
clause type 3 (section 10.2.3.1); and min- `think' and mayN- `forget' take comple-
ment clause type 2. However all can also take a complement clause of type 4 as
their Object.
Complement clause type 4 is without an explicit Subject, as the notional
Subject is necessarily coreferential with the Subject of the matrix clause. The
subordinate verb is in the Innitive, which appears to be a syntactic rather than a
semantic restriction. The subordinate event is necessarily after the matrix event
in time, which suggests that the restriction to Innitive could just be a semantic
restriction to relative future  however in this case it would be expected that the
subordinate verb could be an Imperfective Participle form in its schedule future
usage (cf. section 10.2.2), and this is not possible, showing that the restriction to
Innitive is syntactic. This type of complement indicates that, at the reference
time of the matrix event, the (matrix) Subject had the intention of carrying out
the subordinate activity:7
7Note that this complement type is identical in internal structure, and similar in meaning,
to the same-Subject purposive (see section 10.3.1).
266 CHAPTER 10. SUBORDINATE CLAUSES

(627) Santos=na [ a-n ] kizh-t1-zi

Santos=top [ come-inf ] say-past-nonlocut
`Santos said that he would come.'
(628) Ángel=na [ t1lawa shappi k11-na ] kizh-t1
Ángel=top [ tomorrow cane:juice mill-inf ] say-term
anshik=na
yesterday=top
`Ángel said yesterday that he would mill cane juice tomorrow.'
(629) na=na [ martes=na 1-n ]
1sg.(nom)=top [ Tuesday=top go-inf ]
min-tu-ata-w, miercoles
think-impf-past-locut:subj Wednesday
1-ma-t1-mtu-s=ma
go-comp-term-impf-locut=temp
`I was planning to go on Tuesday, now I will go on Wednesday.'
(630) na=na [ puerta s1p-na ]
1sg.(nom)=top [ door shut-inf ]
mayN-ma-t1-s
forget-comp-past-locut:under
`I forgot to shut the door.'
The complement of intention necessarily occurs in the normal Object position
before the verb, and it cannot be extraposed to nal position.
10.2.4.2 Complement clause 4 as Subject: Evaluative predicates
The nal use of complement clauses is as the Subject of an evaluative predicate.
A Subject-less Innitive complement, that is a complement clause of type 4, can
act as the Subject of an evaluative predicate consisting of an adjective plus a
copula, such as wat i `is good' or kwashmayN i `is tasty'; as would be expected,
it is also possible to use this complement as the Subject of a non-nite verbless
copula sentence (see section 3.3.2), with the Innitive clause followed by the
evaluative adjective. The verb in the Innitive clause may, of course, have an
Object or other argument, and various modiers may be present. The Innitive
clause, as Subject of the matrix verb, is often followed by the Topic marker na,
as is common for Subjects.
(631) [ ashaNpa=ta pyan-na ]=na wat shi ki
[ woman=acc hit-inf ]=top good neg be.neg.(nonlocut)
`Hitting [one's] wife is not good.'
(632) [ nul kwa-n ]=na kwashmayN
[ chontaduro eat-inf ]=top tasty
`Eating chontaduro (a fruit) is really enjoyable/tasty.'
10.3. ADVERBIAL CLAUSES 267

(633) [ ampu-tuzpa kal ki-n ]=na p1na wat

[ man-coll work(1) work(2)-inf ]=top very good
`Working among men is very good.'
This complement construction is unique, in that it is the only complement
construction found in Awa Pit where the complement is a Subject, rather than
an Object.

10.3 Adverbial clauses

As mentioned before, adverbial clauses are those subordinate clauses which func-
tion as modiers of verb phrases or entire propositions (Thompson & Longacre
1985:172). Cross-linguistically, these have functions such as: time, location,
manner, purpose, reason, simultaneous and conditional (Thompson & Longacre
1985:177). As discussed in section 10.2.3.2, however, while these functions may
be expressed through the use of adverbial clauses, various languages use dierent
options to express some of them; in the particular case of Awa Pit, a number of
them are expressed through complementation to a postposition.
Verb roots in Awa Pit are obligatorily followed by one of a variety of
axes (see chapter 7). In the case of non-subordinate verbs, these axes are
selected from aspect, tense, mood, person and number axes. Verb roots in
adverbial clauses likewise must be formally marked with an ax.8 There are a
variety of axes which are possible in adverbial clauses (discussed in the following
sections): some of these are only used under these circumstances, others are
formally identical to axes in matrix clauses, although there are semantic and
combinatorial dierences, and the axes are treated as distinct (see section 3.3.2).
Adverbial subordinate clause verbs have one common feature  regardless
of which axes are present, all subordinate verb forms are non-nite, and have no
marking of tense, mood, number or person. Unlike languages such as Imbabura
Quechua, where many adverbial clauses have pairs of suxes indicating same or
dierent subject (Cole 1985:6066), in Awa Pit there is only one pair, indicating
same or dierent Subject for purposives. All other adverbial clauses are either
necessarily same-Subject (the After construction) or can be used with both same
and dierent Subjects (all others).
As noted above, some of the markers of adverbial clauses in Awa Pit are
used with a variety of dierent functions, and indeed three can also be used
to mark complement clauses: the Innitive, the Imperfective Participle, and
the Perfective Participle. In adverbial clauses, the Innitive is used for same-
Subject purposives and no-Subject purposives; the Imperfective and Perfective
Participles can be used in an absolute construction.
There are two verbal axes used in adverbial clauses which are formally
identical to two main clause mood axes. The rst of these is npa/napa, which
8The only apparent exception are stative verbs. Stative verbs, formally speaking, cannot be
followed by the Imperfective Participle marker -(m)tu, and the Imperfective Participle form of
a stative verb is formally identical to the verb root: for example, i- `be', i `be.impfpart'.
268 CHAPTER 10. SUBORDINATE CLAUSES

is used as a Necessitive mood marker in main clauses (see section 9.4.2.2), but
as a dierent-Subject purposive in subordinate clauses. The other is tpa/tawa,
used in subordinate clauses to mark subsequent time and marking obligation in
main clauses (see section 9.4.2.1).
Finally there are three verb axes which are only used in subordinate
clauses, never in main clauses. Ka indicates, approximately, that two events are
simultaneous; at marks the protasis (`if'-clause) of a counterfactual sentence; and
kikas marks concessive clauses.

10.3.1 Same-Subject purposive

The Innitive can be used to mark a same-Subject purposive (to form a dierent-
Subject purposive, a dierent ax is used; see section 10.3.2). The clause which
states the purpose for which something is done is marked with the Innitive, and
is obligatorily Subject-less, with the notional Subject of the Innitive verb being
coreferential with the matrix clause Subject;9 the matrix clause is simply marked
with mood, tense, aspect, number and person as any main clause. The purpose
clause is most commonly placed after the main clause:
(634) Carmen piya k11-t kway-zi, [ atal pashpa
Carmen corn mill-sv drop-nonlocut [ chicken dim
kwin-na ]
give-inf ]
`Carmen ground corn to give to the baby chickens.'
(635) na=na t1m=ta-s utka-t
1sg.(nom)=top basket=in-from take:out-sv
kyan-ta-w, [ ayna-t kwa-n ]
throw-past-locut:subj [ cook-sv drop-inf ]
`I took [the plantains] out of the basket in order to cook [them].'
(636) Demetrio=na t1lawa a-mtu-y, [ s1
Demetrio=top tomorrow come-impf-nonlocut [ rewood
pyan-na ]
cut-inf ]
`Demetrio is coming tomorrow to cut rewood.'
While the purpose clause is normally placed after the main clause, it may
be embedded within it.
(637) Carmen=na, [ na-wa kwin-na, ] comida
Carmen=top [ 1sg-acc give-inf ] food
ayna-mtu-y
cook-impf-nonlocut
`Carmen is cooking food in order to give [it] to me.'
9 Compare this structure with that of complement clause type 4 used as an Object (see
section 10.2.4.1).
10.3. ADVERBIAL CLAUSES 269

The Prospective aspect derivational marker n1 developed from this embedded

structure, with an Innitive as purposive followed by the main verb 1- `go': liter-
ally it began as `go in order to' (see section 6.4.2.2 for details).
There is one peculiarity with the use of the same-Subject purposive con-
struction together with the First Person Object Imperative (see section 9.4.5.1).
When an innitival clause is used together with a matrix clause containing a
verb in the First Person Object Imperative, the Subject of the subordinate verb
is coreferential with the Object of the matrix verb; that is, it is rst person.
(638) limonada pay-nin-zha, [ tuk-na ]
lemonade buy-caus-imp.1obj [ suck-inf ]
`Sell me lemonade, in order [for me] to drink!'

10.3.2 Dierent-Subject purposive

The Innitive can be used to form a purposive construction, however this is
necessarily a same-Subject purposive (see section 10.3.1). To form a purposive
construction with a dierent Subject, the sux npa/napa is used. The dierent-
Subject purposive is obligatorily unmarked for person, although the Subject can
be fully expressed in the clause. As with the same-Subject purposive construc-
tion, the dierent-Subject purposive clause most commonly follows the matrix
clause, but it may precede the matrix clause or be embedded within it.
(639) na=na s1 pyan-n1-ma-t1-mtu-s,
1sg.(nom)=top rewood chop-prosp-comp-term-impf-locut
[ Carmen ayna-t kwa-npa ]
[ Carmen cook-sv eat-dspurp ]
`I'm going to chop rewood so that Carmen can cook and eat.'
(640) Jos
e=na s1 pyan-m1z-a-zi,
José=top rewood chop-incep-past-nonlocut
[ na=na ayna-t kwa-npa ]
[ 1sg.(nom)=top cook-sv eat-dspurp ]
`José started chopping rewood so that I could cook and eat.'
(641) [ yal=ta 1-npa, ] carpa nam1la-ta-w
[ house=in go-dspurp ] cape lend-past-locut:subj
`I lent [Demetrio] a rain-cape so that [he] could go home.'
(642) Carmen=na [ paynya nalpihsh kwa-npa ]
Carmen=top [ her brother eat-dspurp ]
ayna-ma-t1-zi
cook-comp-past-nonlocut
`Carmen cooked, so that her brother could eat.'
It seems likely that the dierent-Subject purposive inection has developed
historically from the Innitive (n or na) plus a following pa. Synchronically it
270 CHAPTER 10. SUBORDINATE CLAUSES

cannot be analyzed in this way, however, as the npa allomorph of the dierent-
Subject purposive is used after any verb stem ending in a vowel, while the short
form Innitive n only occurs with a handful of verbs (see section 7.2.8.1). The
existence of this pa as, originally, a separate morpheme is also suggested by the
parallelism between the Necessitive npa/napa and the Obligative tpa/tawa. As
was seen in section 10.2.3.2 above, the Innitive marker n/na and the Perfect-
ive Participle t/ta contrast in a number of subordinate clauses directly before
postpositions. The variation between pa and wa can be explained by assuming
that the markers became unitary morphemes at dierent times  in the current
stage of Awa Pit, any morpheme beginning with p undergoes a morphophonemic
alteration to w when it is suxed or encliticized. If napa was developed before
this became a general rule, while tawa developed afterwards, this would explain
the alternation.
The most likely origin of the pa ending of the dierent-Subject purpos-
ive inection would appear to be the postposition pa, which can be locative or
allative.10 Haspelmath (1989:291295) gives many examples of the development
of purposives from allatives or benefactives (see also Austin (1981) and Bybee,
Perkins & Pagliuca (1994:223224)).
It is interesting to consider the historical link between the dierent-Subject
purposive and the formally identical Necessitive npa/napa (see section 9.4.2.2).
The link between these two concepts, obligation and purpose, is not found only
in Awa Pit: Bybee, Perkins & Pagliuca (1994:229) list four languages with this
particular polysemy, as well as common polysemies between purposive and in-
tention, and purposive and future; and Dixon (1980:458) notes that the same
ax used for purposive subordinate clauses in Australian languages is commonly
used in main clauses to indicate need, obligation or desire (and that this ax is
probably related to the nominal ax marking purposive, dative and genitive). If
the origin of npa/napa is indeed the Innitive followed by the postposition pa, it
would seem likely that the obligation usage developed from the purposive usage,
and the main clause use would derive from the subordinate clause use, and not
vice versa (Bybee, Perkins & Pagliuca 1994:224), as a non-nite verb followed
by a postposition could exist as a subordinate form, but hardly as a matrix verb.
The data from Awa Pit suggests that the obligation use of a marker de-
veloped from the purposive use of this marker; and this is precisely the path
suggested by Bybee, Perkins & Pagliuca (1994:223224) based on Patz's (1982)
data on Gugu-Yalanji. Oddly, in discussing the table of purposive markers con-
taining the Gugu-Yalanji marker a few pages later, Bybee, Perkins & Pagliuca
(1994:229) suggest that agent-oriented [obligation, intention] and future uses are
precursors to the purpose, complement clause and speaker-oriented uses, and the
latter three uses develop in parallel, and hypothesize that agent-oriented uses
reect the earliest meanings of modal grams, and that subordinate and speaker-
oriented uses develop from these. Given the data from Awa Pit, it seems much
more likely that the earliest use of at least some of these markers is as purposive
markers, with the other uses, such as obligation, developing from this use.
10 See section 5.4.4; compare also the discussion of tawa in the following section.
10.3. ADVERBIAL CLAUSES 271

10.3.3 After clauses

There are a variety of ways of expressing temporal sequence in Awa Pit. The most
general of these, ka, will be examined in the following section (section 10.3.4);
here the use of the marker tpa/tawa to mean `after' will be looked at.
Parallel to the dierent-Subject purposive marker, the After inection has
two forms, tpa and tawa. The former is used after vowel-nal verb stems; the
latter after consonant-nal verb stems. Just as the dierent-Subject purposive
seems to be derived from the Innitive plus a following pa, it seems likely that
the After inection is derived from the Perfective Participle plus a following
pa.11 Once again, it seems probable that the After inection developed rst in
subordinate clauses, and only later became a main verb form indicating obligation
(see section 9.4.2.1).
The After marker indicates that the activity of the matrix clause has
occurred or will occur only after the activity expressed in the subordinate clause.
If the matrix clause is past or present in reference, then the subordinate activity
has necessarily occurred. If the matrix clause expresses a future action, then in
fact there is no necessary statement that the subordinate clause activity (and
hence the matrix clause activity) will occur; it is simply that the subordinate
event must occur before the matrix clause activity can occur (see sentence (646)
for an example where the there is no concept of the propositions necessarily
becoming reality), and this is presumably the use which led to the development
of this morpheme as a main clause Obligative.
(643) [ ayna-tpa=na, ] au=na kwa-ta-w
[ cook-after=top ] 1pl.(nom)=top eat-past-locut:subj
`After cooking, we ate.'
(644) [ na=na kwa-tpa=na, ] shihtt1 kit
[ 1sg.(nom)=top eat-after=top ] nger wash
kway-ta-w
drop-past-locut:subj
`After eating, I washed my hands.'
(645) [ camisa pat-tawa=na, ] piikam-ta-w
[ shirt wash-after=top ] swim-past-locut:subj
`After washing [my] shirt, I went for a swim.'
(646) [ na=kas ashaNpa say-tawa, ] na=kas
[ 1sg.(nom)=add woman nd-after ] 1sg.(nom)=add
kii-m1z-tu
get:married-incep-impfpart
`After/if/when I nd a woman, I'll get married too.'
11A similar origin has been suggested for a verbal sux tirã meaning `after' in Retuarã,
which is hypothesized as consisting of the perfect sux ti and a locative postposition rã (Strom
1992:168).
272 CHAPTER 10. SUBORDINATE CLAUSES

A subordinate clause marked with the After marker occurs, obligatorily,

before the matrix clause. This contrasts with the purposive constructions dis-
cussed above, but is similar to the general temporal marker ka, discussed in the
following section and is iconic. The Subjects of the two clauses must be corefer-
ential:
(647) *[ Marcos=na camisa pat-tawa=na, ] na=na
[ Marcos=top shirt wash-after=top ] 1sg.(nom)=top
piikam-ta-w
swim-past-locut:subj
Consequently it is unusual for both Subjects to be explicitly stated, although it is
possible for emphasis (see sentence (646)). More commonly, the Subject is only
stated once, although this may be in either of the clauses (compare sentences
(643) and (644)); or else the Subject may be entirely ellipsed, if understandable
through discourse factors or person marking on the matrix verb (sentence (645)).
In its subordinate use, the After inection always occurs directly after
the verb root plus any derivational suxes, and neither tense nor aspect can be
marked on the subordinate verb. The main clause Obligative is usually followed
by person marking; this contrasts with the subordinate After inection where, as
with all subordinate adverbial clauses, there is no person marking. The subor-
dinate clause as a whole is nearly always marked by a encliticized Topic marker.

10.3.4 Simultaneity clauses

The most commonly used marker of temporal sequence is the subordinate suf-
x ka, usually translatable as `when'. Ka establishes simultaneity between the
propositions of the subordinate clause which it marks and the matrix clause.
However, as ka can be suxed to a verb root, a verb root plus derivational mark-
ers, or a verb root plus aspect markers, and the matrix clause may have any
verbal suxes allowable in a main clause, a wide variety of possible temporal re-
lations between the two clauses can be created by using distinct verbal marking
in the two clauses. The subordinate clause often carries na, the Topic marker,
and necessarily occurs before the matrix clause.
If the matrix clause is Past tense, then the propositions contained in the
two clauses have occurred, and ka is appropriately translated by `when':
(648) [ k1n-ka=na, ] na=na Santos=ta
[ dawn-when=top ] 1sg.(nom)=top Santos=acc
izh-ta-w
see-past-locut:subj
`At dawn (when it dawned), I saw Santos.'
(649) [ Santos a-ka=na, ] kula-ta-w
[ Santos come-when=top ] hide-past-locut:subj
`When Santos came, I hid.'
10.3. ADVERBIAL CLAUSES 273

If the matrix clause is Future or some non-indicative mood, the events

have not occurred, and there is no necessity that the events will ever happen. If
the event in the subordinate clause is necessarily going to happen (for example, it
contains nash- `afternoon'), or if it is likely to happen, it will usually be translated
by `when' in English; otherwise it will be translated by `if'. There is, then, no
distinction in Awa Pit in the encoding of a sequence of future events depending
on the speaker's belief in their probability of occurrence.
(650) [ nash-ka ] alu ki-n1-zi
[ be:afternoon-when ] rain(1) rain(2)-fut-nonlocut
`This afternoon it may rain.'
(651) [ ii-ka=na=ma, ] kam-tawa
[ die-when=top=temp ] bury-oblig
`When someone dies, they have to be buried.'
(652) [ an k1h ku-ka=na, ] shi=ma ki-n1-zi?
[ this leaf eat-when=top ] what=inter do-fut-nonlocut
`If [one] eats this leaf, what will happen?'
(653) [ a-t kway-ka=na, ] ayna-n1-s
[ come-sv drop-when=top ] cook-fut-locut
`When/if she comes, I will cook.'
If the matrix clause is Present tense, there are a variety of meanings,
corresponding to the variety of meanings of the Present (Imperfective). It cannot
mean simple present, since then there would be no reason to indicate simultaneity,
just as in English: *When I am coming, he is cooking. The simultaneity of two
activities may refer to a current habitual occurrence, a scheduled future activity,
or a general truth:
(654) [ nash-ka=na ] p1na alu
[ be:afternoon-when=top ] very rain(1)
ki-mtu-y
rain(2)-impf-nonlocut
`It is raining every afternoon (in this season).'
(655) [ dios izh-ka=na, ] kayl-tu-s
[ god see-when=top ] return-impf-locut
`If God is watching, we will return.'
(656) [ shutshu sa-ka=na, ] ishan-tu-y
[ breast touch-when=top ] laugh-impf-nonlocut
`When [one] touches [women's] breasts, they laugh.'
While the use of ka sets up the simultaneity of the subordinate and matrix
proposition, by including aspectual marking in one or the other clause, a variety
274 CHAPTER 10. SUBORDINATE CLAUSES

of diering temporal relations can be established. An Imperfective in one or the

other clause establishes that the activity encoded in that clause was on-going at
the time at which the event in the other clause occurred:
(657) [ Demetrio a-ka=na, ] kal
[ Demetrio come-when=top ] work(1)
ki-mtu-ata-w
work(2)-impf-past-locut:subj
`When Demetrio came I was working.'
(658) [ na=na uN=pa-s kayl
[ 1sg.(nom)=top there=in(approx)-from return
kway-ka=na, ] kal ki-mtu-an1-zi=ma
drop-when=top ] work(1) work(2)-impf-fut-nonlocut=temp
`When I return from there, [my friends] will already be working.'
(659) [ na=na Ricaurte=ta-s a-mtu-ka=na, ]
[ 1sg.(nom)=top Ricaurte=in-from come-impf-when=top ]
maza kwizha izh-ta-w
one dog see-past-locut:subj
`When I was coming from Ricaurte, I saw a dog.'
The aspect markers ma and t1, the Completive and the Terminative, can be
used to focus more precisely on temporal relations, either individually or jointly.
When it occurs before ka, ma indicates that the activity in the subordinate clause
had either just begun (if atelic), as in sentences (660) and (663), or just nished
(if telic), as in sentence (661). The Terminative t1 indicates that the two events
were closely linked in time, with the subordinate event having just nished when
the matrix event took place, as in sentences (662) and (663).12
(660) [ kutshu=ta puz-ma-ka=na, ] alu
[ San Isidro=in go:out-comp-when=top ] rain(1)
ki-ma-t1-zi
rain(2)-comp-past-nonlocut
`When I had just left to go to San Isidro, it rained.'
(661) [ ap kwankwa ii-ma-ka=na, ] ap aympihsh
[ my grandmother die-comp-when=top ] my brother
kii-ma-t a-zi
get:married-comp-pfpart be-nonlocut
`When my grandmother [nally] died, my brother was [already]
married.'
These meanings are the usual meanings for these suxes; see sections 9.3.2 and 9.3.3 for
12
more details and a closer examination of these meanings.
10.3. ADVERBIAL CLAUSES 275

(662) [ waz tazh-t1-ka=na, ] kwizha siente

[ bowl come:down-term-when=top ] dog hear
ki-t1-zi
do-past-nonlocut
`When the bowl fell, the dog heard it.'
(663) [ ish-ma-t1-ka=na, ] médico izh-n1-tpa
[ be:sick-comp-term-when=top ] doctor see-prosp-oblig
`When [you] get sick, it's necessary to go and see the doctor.'
It is also possible to use various aspect markers in the matrix clause,
of course. In addition, the Resultative or the Past Anterior constructions (see
sections 11.2.2.2 and 11.2.2.1 respectively) can be used to state that, by the time
the action in the subordinate clause occurred, the action in the matrix clause
had already occurred. To state the opposite, that the action in the matrix clause
occurred after the action in the subordinate clause, the verbal sux tpa is used,
rather than ka (see section 10.3.3).
(664) [ na=na pueblo=ta a-t kway-ka=na, ]
[ 1sg.(nom)=top town=in come-sv drop-when=top ]
alcalde ii-ma-t a-t1-zi
mayor die-comp-pfpart be-past-nonlocut
`When I arrived in the town, the mayor was dead.'
(665) [ Bogot
a=ta-s kayl kway-ka, ] verano
[ Bogotá=in-from return drop-when ] summer
payl-ma-t a-npa-y
nish-comp-pfpart be-necess-nonlocut
`When I return from Bogotá, the summer will be over.'
(666) [ na=na namna-ka=na, ]
[ 1sg.(nom)=top follow/catch:up:to-when=top ]
uspa=na wakata kan-ma-t
3pl.(nom)=top cattle tie-comp-pfpart
ma-t1-zi
anter-past-nonlocut
`When I caught up, they had tied up the cattle.'
(667) [ Carmen=na pueblo=ta-s a-t kway-ka=na, ]
[ Carmen=top town=in-from come-sv drop-when=top ]
Santos=ta pyan-ma-t ma-zi Demetrio
Santos=acc hit-comp-pfpart anter-nonlocut Demetrio
`When Carmen arrived back from town, Demetrio had hit Santos.'
276 CHAPTER 10. SUBORDINATE CLAUSES

10.3.5 Concessive clauses

Concessive constructions are those for which one might think, given the subor-
dinate clause, that the matrix clause is not true, but in fact it is. In Awa Pit
these constructions consist of a matrix clause marked in the usual fashion, and
an associated concessive subordinate clause, where the verb root is marked by
the concessive mood sux kikas.13 The concessive clause necessarily precedes
the associated matrix clause.
Neither tense nor aspect can be marked in the subordinate clause. This
does not usually create problems in comprehension, as the concessive clause refers
to the same time as the matrix clause  for example, in sentence (668) the clauses
are past, in sentence (669) the clauses are future.
(668) [ say-kikas, ] shi say-ma-s
[ look:for/nd-concess ] neg look:for/nd-neg-locut
`Although I looked for [it], I didn't nd [it].'
(669) [ Laureano a-kikas, ] kal ki-mtu shi
[ Laureano come-concess ] work(1) work(2)-impfpart neg
ka-s
be:permanently-locut
`Even if Laureano comes, I won't work with him.'
This simultaneity of clauses is the assumption, if there is nothing in the
sentence or context which contradicts this assumption. In order to be more
specic, it is always possible to add a lexical time specication to the subordinate
clause:
(670) [ Carmen ma payu=na a-t kway-kikas, ]
[ Carmen now day=top come-sv drop-concess ]
t1lawa=na 1-t1-mtu-s
tomorrow=top go-term-impf-locut
`Even if Carmen comes today, I'll go [to town] tomorrow.'
(671) [ Carmen anshik=na a-t kway-kikas, ]
[ Carmen yesterday=top come-sv drop-concess ]
t1lawa=na 1-t1-mtu-s
tomorrow=top go-term-impf-locut
`Even though Carmen came yesterday, I'll go [to town] tomorrow.'

10.3.6 Counterfactual clauses

The protasis (`if'-clause) of counterfactual sentences in Awa Pit is marked with
the verb sux at, which takes the form t after a. Counterfactual clauses in Awa
13It is interesting to note that the standard of comparison in a comparative construction
can be signalled by a formally identical morpheme, kikas. The two morphemes are clearly
historically related, as will be discussed in section 13.7.
10.3. ADVERBIAL CLAUSES 277

Pit are unusual in that the morphology marking the protasis is not added to the
verb stem, but rather to the Imperfective or Perfective Participle form.14 The
protasis may either be simultaneous with the apodosis (`then'-clause), in which
case the Imperfective Participle form is used; or the protasis may temporally
precede the apodosis, in which case the Perfective Participle is used.
The apodosis is marked dierently depending on whether the counterfac-
tual statement is something which was not true in the past, or something which
is not true now, will not be true in the future, or is not true in general. In the
rst case, a past counterfactual, the matrix verb is marked with the Past ax
or with a Perfective Serial Verb, followed by the irrealis marker na, followed by
person marking, s for Locutor, zero for Non-locutor (see section 8.4.1). If it is
not a past counterfactual, then the Potential or Negative Potential mood marker
is used on the main verb, followed by person marking. The protasis is always
stated before the apodosis, which is a perfectly iconic ordering, and the protasis
is normally marked with the Topic marker.
(672) [ Laureano=na k11-mtu-at=na, ] panela pay-t
[ Laureano=top mill-impfpart-cntrfc=top ] panela buy-sv
kway-n1-t1-na-s
drop-prosp-past-irr-locut
`If Laureano had been milling, I would have gone to buy some panela
(sugar).'
(673) [ akki pana-t=na, ] izh-sina=ma
[ here be:standing.(impfpart)-cntrfc=top ] see-pot=temp
`If he were here, we could meet.'
(674) [ piya waa-t=na, ] arepa sa-t
[ corn there:is.(impfpart)-cntrfc=top ] pancake make-sv
kway-na-s
drop-irr-locut
`If there were any corn, I would have made pancakes.'
(675) [ Demetrio=na anshik a-t-at=na, ]
[ Demetrio=top yesterday come-pfpart-cntrfc=top ]
ma=na 1-t1-na-s
now=top go-past-irr-locut
`If Demetrio had come yesterday, I would have left today.'
(676) [ k11-ta-t=na ] wat-a-na
[ mill-pfpart-cntrfc=top ] be:good-past-irr.(nonlocut)
`If [he] had milled, it would have been good.'
14 This suggests that, historically, at was probably a separate word (perhaps a postposition)
following a complement clause.
278 CHAPTER 10. SUBORDINATE CLAUSES

10.3.7 Absolute clauses

The nal type of adverbial subordinate clause in Awa Pit is the absolute clause.
These clauses are marked as being subordinate, by the presence of the Imperfect-
ive or Perfective Participle or the Innitive form of the subordinate verb, but no
futher specication is made as to their function in the sentence. They either in-
dicate a temporal, causal or conditional idea (through either Participle), the idea
of an accompanying circumstance (through the Imperfective Participle only),
or a no-Subject purposive (through the Innitive).
It is worth noting that while the Absolute constructions have been treated
here as adverbial clauses, in semantic intent the rst two types are very similar to
T-relatives and NP-relatives in Hale's (1976) interpretation of `adjoined relative
clauses' in Australian languages.
10.3.7.1 Temporal, causal or conditional absolute clauses
Absolute clauses marking a temporal, causal or conditional idea contain either a
Perfective Participle (if the subordinate event occurs prior to the matrix event) or
an Imperfective Participle (if the two events are cotemporaneous). The absolute
clause is identical in structure to a matrix clause (with the exception of the verbal
markings), and may contain an explicit Subject. Most commonly, the (notional)
Subjects of the two clauses are the same, and if there is an explicit reference to
the Subject this can occur in either the main or the subordinate clause. If the
Subjects of the two clauses are coreferential, the absolute clause may be embedded
within the matrix clause, although initial position is more common. It could be
claimed that in fact absolute clauses are never embedded, and that in an example
such as sentence (677) it is simply the case that ampu, the Subject, is expressed
in the subordinate clause rather than in the matrix clause; however that this is
not the case can be seen from sentence (684), where the initial element pueblo
clearly is an argument of the matrix clause, and not the subordinate clause.
As well as same-Subject absolute clauses, it is also possible to have sub-
ordinate absolute clauses which, while they do not have the same Subject as
the matrix clause, do not explicitly have a Subject at all. These clauses have
either an indenite Subject (as in sentence (680)) or an impersonal verb (as in
sentence (681)).
As there is no precise specication of the relationship between the ad-
verbial clause and the matrix clause, it is often dicult to tell whether an abso-
lute clause is intended to be temporal, causal or conditional. This is especially
the case, of course, as the explicit marking corresponding to a temporal subor-
dinate clause, ka, marks future temporal and conditional relations, and causal
relations are often dicult to distinguish from temporal relations. Examples of
absolute clauses used in causal, temporal and conditional ways are:
(677) ampu [ pyan-ta=na ] tayaz-tu-at1-zi
man [ hit-pfpart=top ] be:sorry-impf-past-nonlocut
`Having hit [his wife] the man was sorry.'
10.3. ADVERBIAL CLAUSES 279

(678) [ na=na Bogot
a=ta-s kayl-ta=na, ]

[ 1sg.(nom)=top Bogotá=in-from return-pfpart=top ]
kal ki-na a-mtu-s
work(1) work(2)-inf come-impf-locut
`When I have returned from Bogotá, I will come and work.'
(679) [ wiya ki-mtu ] t1t kway-t1-zi
[ ght(1) ght(2)-impfpart ] cut drop-past-nonlocut
`While ghting they cut [each other].'
(680) [ kal ki-t=na ] pyal waa-y
[ work(1) work(2)-pfpart=top ] money there:is-nonlocut
`When one works, then there's money/After working, you get money.'
(681) alu ki-mtu, k1n-ma-t1-zi
rain(1) rain(2)-impfpart dawn-comp-past-nonlocut
`It dawned raining.'
(682) [ shi 1-t ki=na ] pantalón
[ neg go-pfpart be.neg.(impfpart)=top ] pants
pat-m1z-tu-s
wash-incep-impf-locut
`If I don't go (to Pueblo Viejo), I'll wash my pants.'
As seen in these examples, the absolute clause may be negated in the usual clausal
fashion, and an absolute clause often has an associated Topic marker.
10.3.7.2 Accompanying circumstance absolute clauses
Accompanying circumstance clauses are those clauses which simply add addi-
tional information about the Subject  something else that the Subject was
doing. They are necessarily associated with the Subject of the matrix clause,
and necessarily coreferential, with no explicit Subject. As the two clauses are
cotemporaneous, only the Imperfective Participle is used in these clauses.
(683) [ akkwan libro m1j=na ] pashpa escuela=ta
[ many book have.(impfpart)=top ] child school=in
a-mtu-y
come-impf-nonlocut
`The child came to school carrying/with a lot of books.'
(684) pueblo=ta [ pyal shi m1j
town=in [ money neg have.(impfpart)
ki=na ] a-ta-w
be.neg.(impfpart)=top ] come-past-locut:subj
`I came to town, without (not having) any money.'
280 CHAPTER 10. SUBORDINATE CLAUSES

10.3.7.3 No-Subject purposive absolute clauses

The Innitive na is used in a no-Subject purposive construction. The main
verb in these sentences expresses a proposition which is true under certain con-
ditions, and the conditions under which it is true are expressed by the innitival
clause. These conditions are not simple conditions, however, but rather express
a particular activity which one may wish to carry out:
(685) [ maas yal sa-na=na, ] akkwan t1 waa-y
[ new house make-inf=top ] many stick there:is-nonlocut
`To make a new house, there is much wood (much wood is needed).'
(686) [ palanca sa-na=na, ] chonta wat
[ crowbar make-inf=top ] chonta good
`Chonta wood is good for making crow-bars.'
These no-Subject purposives are distinguishable from same-Subject pur-
posives (section 10.3.1), although both use the Innitive to mark the adverbial
clause. They tend to appear in dierent positions: in same-Subject purposives,
the subordinate clause normally follows the matrix clause; in no-Subject purpos-
ives, it normally precedes. More clearly, they are distinguished by the use of the
topic marker, as same-Subject purposive clauses cannot be marked with the topic
marker, while no-Subject purposives are usually topic-marked. Additionally, the
notional Subject of same-Subject purposives is coreferential with the matrix Sub-
ject; in no-Subject purposives, the notional Subject of the subordinate clause is
indenite.

10.4 Relative clauses

In Awa Pit there are a number of constructions which full the same semantic
role as relative clauses in many languages, but only one or two of these construc-
tions truly t the more common denitions of relative clauses, and even these
constructions have a number of somewhat unusual features.15 Before turning to
an examination of these constructions, the various denitions of relative clauses
developed separately and together by Keenan and Comrie will be looked at.
In their original paper, Keenan and Comrie state that they consider any
syntactic object to be a relative clause
if it species a set of objects (perhaps a one-member set) in two
steps: a larger set is specied, called the domain of relativization,
and then restricted to some subset of which a certain sentence,
the restricting sentence, is true. The domain of relativization is
expressed in surface structure by the head NP, and the restricting
sentence by the restricting clause, which may look more or less like
15See also absolute clauses, section 10.3.7, which could be treated as `adjoined relative
clauses'.
10.4. RELATIVE CLAUSES 281

a surface sentence depending on the language.

(Keenan & Comrie 1977:6364)
Comrie (1981:136137) modies this denition slightly, and extends its
use. He considers that his denition is a characterization of the prototypical
relative clause, rather than a set of necessary and sucient conditions for the
identication of relative clauses, and notes that in order to say that a language
has relative clauses, it should be the case that there is some construction or
constructions correlating highly with the denition:
A relative clause then consists necessarily of a head and a restricting
clause. The head in itself has a certain potential range of referents,
but the restricting clause restricts this set by giving a proposition
that must be true of the actual referents of the over-all construction.
(Comrie 1981:136)
However he believes that English sentences such as non-nite constructions (pas-
sengers leaving on ight 738 should . . . ) and those involving restrictive attributive
adjectives (the good students all passed the examination) are included in the
denition of relative clauses.
Keenan (1985:142) has essentially the same semantic denition as the other
two works. However, the presence of a restrictive clause (Srel) is considered to be
the dening feature of a relative clause. Under this denition,
there are structures we consider relative clauses which lack a do-
main noun, in which case the domain of relativization is the class
of objects of which it makes sense to assert the restrictive clause.
(Keenan 1985:142)
While accepting these as relative clauses, Keenan bases his classication on re-
lative clauses which contain a domain noun, and has no further discussion of
relative clauses without such a noun.
In Awa Pit there are various constructions which full at least some of
the requirements for relative clauses. Three of these  the lexical nominaliz-
ations, the deverbal adjectives and the purposive adjectives  lack so many
clausal features and have such a variety of idiosyncratic restrictions that they are
considered to be derivational processes, forming nouns or adjectives from verbs,
rather than being (relative) clauses, and hence these processes were discussed in
sections 5.2.1, 5.2.3.1 and 5.2.3.2 respectively. There are also similarities between
relative clauses and the agentive clausal nominalizations discussed in section 10.5,
but, as will be explained there, there are sucient dierences to consider this con-
struction as quite distinct from a relative clause.
This leaves two constructions as relative clauses, one formed with the Pos-
itive Adjectivizer m(u), the other with the Negative Adjectivizer16 kayN, although
16The Negative Adjectivizer is similar in function to the habitual negative in Epena Pedee,
discussed in Harms (1994:135136).
282 CHAPTER 10. SUBORDINATE CLAUSES

the two dier in distribution. Examining the Positive Adjectivizer rst, it ap-
pears to full all the requirements for relative clauses (looking only at the more
inclusive clause here):
(687) [ [ kal ki-m ]=ta ta-mu ] awa pyal
[ [ work(1) work(2)-adjzr ]=acc pay-adjzr ] person money
kaa-ma-t1-zi
lose-comp-past-nonlocut
`The person who pays the workers lost the money.'
In this sentence there is a head noun in the matrix clause (awa `person'), plus a
clause which restricts the head noun to a more limited range of referents. The
relative clause occurs in the usual place for nominal modiers, before the head.
There are strong restrictions on the use of these constructions. The head
can be in any position in the matrix clause, but is necessarily coreferential with
the (non-expressed) Subject of the relative clause. This restriction of the co-
referential noun to Subject position in the relative clause is not unknown cross-
linguistically, occuring for example in Malagasy, and being in perfect accord with
Keenan & Comrie's (1977:6970) hierarchy for relativization. However there
is a strong dierence between Malagasy and Awa Pit in this respect. Malagasy,
unlike Awa Pit, also has a system for promoting any major constituent to subject
position, and thus through, for example, passivization, what would otherwise be
an object can be relativized (Keenan & Comrie 1977:69). In Awa Pit this is
simply not possible.
In addition there is a restriction on the tense/aspect concepts expressed
by these constructions in Awa Pit. These subordinate clauses necessarily describe
a general or habitual state of aairs, never a one-o event. That is, the construc-
tions can be used for something like `the man who is always hitting me', but not
for `the man who hit/is hitting/will hit me'. This does not necessarily exclude
these from being relative clauses, as they still t the denition, restricting the
set of possible referents, but it is a strong semantic restriction on their use.
The existence of a single np containing both the relative clause and the
head noun can be seen from the occurrence of the relative clause directly before
the head noun, and the impossibility of Topic-marking the relative clause.17 The
relative clause verb itself can be ditransitive (as in the example above), transitive
or intransitive, and may have any modiers and adjuncts; it is necessarily an
active verb.
(688) na=na [ kal ki-m ] awa
1sg.(nom)=top [ work(1) work(2)-adjzr ] person
`I am a person who works (a worker).'
17 These facts distinguish the relative clause construction from the clausal nominalization 
see section 10.5.
10.4. RELATIVE CLAUSES 283

(689) na=na [ kwisha attihsh=ta shaa-m ] awa

1sg.(nom)=top [ very far=in walk-adjzr ] person
i-s
be-locut
`I am a person who habitually walks to very far (I'm used to walking a
long way).'
Just as with lexical adjectives, the noun head of the noun phrase can
be ellipsed if it is clear from context what is under discussion. This is especially
common with relative clauses, and the noun head is often ellipsed without having
been previously mentioned in the discourse, provided that it is clear to whom the
relative clause refers. Thus, for example, the head noun awa `person' is often left
o following an relative clause where the verb could only refer to a human actor,
as in the rst occurrence of m(u) in sentence (687) above, and in:
(690) [ wakata walku-m ]=ta pyan-ta-w
[ cattle steal-adjzr ]=acc hit-past-locut:subj
`I hit the cattle thief.'
While these examples with no noun head in Subject or Object position
in the matrix clause still appear to t the denition of relative clauses, the situ-
ation becomes less clear when non-headed relative clauses are used as Copula
Complements (either with or without an i or ka copula).
(691) [ shutta put-mu ] ka-y
[ hat plait-adjzr ] be:permanently-nonlocut
`She weaves hats (is a hat weaver).'
In this example it could be considered that there is an ellipsed awa `person' as the
head noun, but this solution becomes less likely with the use of this construction
as a straightforward method of expressing generic statements:
(692) shitshu [ p1l-mu ] i
bird [ y-adjzr ] be.(nonlocut)
`Birds y (birds are yers).'
(693) 1nkal awa [ tunya ku-m ]
mountain person [ rat eat-adjzr ]
`The Awa eat rats (the Awa are rat-eaters).'
Even more interesting is the fact that the Negative Adjectivizer can only
be used in this sort of copula statement, to make either a specic habitual claim
about someone or something, or to make a more generic claim:
(694) na=na [ wakata ii-ta=na shi
1sg.(nom)=top [ cattle die-pfpart=top neg
ku-kayN ]
eat-negadjzr ]
`I don't eat dead cattle (ie. cattle which died mysteriously).'
284 CHAPTER 10. SUBORDINATE CLAUSES

(695) an caballo=na [ shi shaa-shi-kayN ]

this horse=top [ neg walk-desid-negadjzr ]
`This horse doesn't like walking.'
(696) na=na [ Awa Pit shi pata-kayN ]
1sg.(nom)=top [ person language neg speak-negadjzr ]
ka-s
be:permanently-locut
`I don't speak Awa Pit.'
(697) anya=na [ trabaja shi ki-kayN ]
before=top [ work(1) neg work(2)-negadjzr ]
a-ta-w, domingo=na
be-past-locut:subj Sunday=top
`We didn't used to work on Sundays.'
The Negative Adjectivizer simply cannot be used with a head noun. To express
this concept the Postive Adjectivizer and a following (clause-external) negative
are used:
(698) [ pata-m ] shi pashpa=na
[ speak-adjzr ] neg child=top
`She is a child who can't [yet] speak.'
The two Adjectivizers are thus treated here as forming relative clauses, as
they full many of the requirements for relative clauses. However there are very
strong syntactic requirements which the two constructions must have, and the
Negative Adjectivizer in particular cannot be used in the normal relative clause
construction, with a head noun. In addition, there are very strong semantic
restrictions, as these relative clauses must refer to a habitual or generic situation.
The lack of other relative clause constructions in Awa Pit is not, perhaps,
surprising. It could be considered a factor related to language death, as the fre-
quency of relative clauses has been found to reduce markedly in such situations
(Hill 1989:149). However Awa Pit does have a strong and productive system of
nominalizations, discussed in the next section, and a variety of languages have
been found to lack relative clauses but use nominalizations in their place  for
example Macushi (Abbott 1991:7071) and Barasano (Jones & Jones 1991:149
153). Equally, Awa Pit has an absolute subordinate clause construction (sec-
tion 10.3.7), highly reminiscent of Australian `adjoined relative clauses'.

10.5 Nominalizations
As was briey mentioned in section 5.2.6, there are two suxes which may be
attached to adjectives to change them into nouns, one singular, mika, and one
collective plural, tuz, where the collective plural can only be used if the entities
referred to are acting together for a common purpose. The derived noun is
necessarily animate in both cases.
10.5. NOMINALIZATIONS 285

(699) 1lapa-mika
big-nmlzr.sg
`the older one (brother)'
(700) kutnya-tuz kal ki-mtu-y
three-nmlzr.pl work(1) work(2)-impf-nonlocut
`The three are working together.'
(701) yawa-tuz=ma kal ki-mtu-y?
how:many-nmlzr.pl=inter work(1) work(2)-impf-nonlocut
`How many are working together?'
In addition to the use of these nominalizers with adjectives, they can be
used as clitics to form clausal nominalizations. The subordinate clause has the
same argument structure as a main clause except that it is obligatorily without
a Subject; the nominalization may also contain verbal modiers such as adverbs.
The subordinate verb is either in Perfective Participle form, signalling that the
subordinate event was over at the matrix event time; in the Imperfective Par-
ticiple form, showing that the subordinate event was on-going at the time of
the matrix event; or the verb is suxed by the Positive Adjectivizer, used when
the subordinate event is habitual. The nominalized clause carries the meaning
of the one/ones who carry/carried out the action in the subordinate clause.
(702) [ wakata pay-nin-ta=mika ]=na
[ cattle buy-caus-pfpart=nmlzr.sg ]=top
az-tu-y
cry-impf-nonlocut
`The one who sold cattle is crying.'
(703) [ Juan=ta pyan-ta=mika ]=na katsa
[ Juan=acc hit-pfpart=nmlzr.sg ]=top big
`The one who hit Juan is big.'
(704) [ ap aympihsh=ta pyan-ta=mika ]=na na-wa
[ my brother=acc hit-pfpart=nmlzr.sg ]=top 1sg-acc
t1t-ma-t1-s
cut-comp-past-locut:under
`The one who hit my brother cut me.'
(705) [ anshik a-t=mika ]=na wiya
[ yesterday come-pfpart=nmlzr.sg ]=top ght(1)
ki-ma-t ma-t1-zi
ght(2)-comp-pfpart anter-past-nonlocut
`The one who came yesterday had fought.'
(706) [ Juan=ta pyan-tu=mika ]=na katsa awa
[ Juan=acc hit-impfpart=nmlzr.sg ]=top big person
`The one who is hitting Juan is big.'
286 CHAPTER 10. SUBORDINATE CLAUSES

(707) [ Cumbal=ta-s a-mtu=tuz ]=na ``guerrilla

[ Cumbal=in-from come-impfpart=nmlzr.pl ]=top guerrilla
shaa-y'' kizh-a-zi
be:around-nonlocut say-pl:subj-nonlocut
`The ones coming from Cumbal said that the guerrilla are around.
(708) [ kal ki-m=mika ] ka-y
[ work(1) work(2)-adjzr=nmlzr.sg ] be:permanently-nonlocut
`He/she is a worker.'
As seen in these examples, a nominalized clause as Subject is usually Topic-
marked, as is common for lexical Subjects. While these clausal nominalizations
are most often used in Subject role, they can be used in other roles, and in
this case the nominalization is followed by appropriate postpositions to mark the
grammatical role:
(709) na=na [ pishkatu pay-nin-tu=mika ]=ta
1sg.(nom)=top [ sh buy-caus-impf=nmlzr.sg ]=acc
pyan-ta-w
hit-past-locut:subj
`I hit the one who was selling sh.'
These clauses are clearly nominalizations rather than, for example, com-
plement clauses, as the nominalization can be accompanied by modiers such as
adjectives at the level of the matrix clause in exactly the same way as any lexical
noun:
(710) sun [ a-t=mika ]=na ii-ma-t1
that [ come-pfpart=nmlzr.sg ]=top die-comp-term
`That one that came has died.'
Thus the clause forms the functional equivalent of a noun, not a noun phrase.
Given that in Awa Pit it is possible to ellipse a head noun, it could be
maintained that the clauses marked with mika or tuz are in fact modiers rather
than nouns in their own right, and that the head noun has been ellipsed. Indeed,
there are sentences which contain both a mika/tuz clause and a noun, which
would appear to support this analysis:
(711) [ na-wa sula kwa-t=mika ] kwizha nya
[ 1sg-acc bite(1) bite(2)-pfpart=nmlzr.sg ] dog meat
walkwa-t1-zi
steal-past-nonlocut
`The dog which bit me stole the meat.'
However there are a variety of reasons for considering that sentences such as
this contain two separate noun phrases, both referring to the same entity, in an
appositional relationship.18
18The same type of appositional relationship is found between nouns and clausal nominaliz-
ations in Macushi (Abbott 1991:9396) and Barasano (Jones & Jones 1991:149153).
10.5. NOMINALIZATIONS 287

To begin with, the order of the head noun and the modier are not
xed. In contrast to the previous sentence, where the clause marked with mika
occurs before the noun, there are sentences where the noun precedes the clause:
(712) kwizha [ a-t=mika ]=na ii-ma-t1
dog [ come-pfpart=nmlzr.sg ]=top die-comp-term
`The dog which came has died.'
Indeed, when there is both a noun and a clause marked with mika/tuz,
these are most commonly separated, with one or the other occurring after the
matrix verb.19 Thus one of the informants translated and then repeated the
sentence the dog which came died, with the noun and the clause rst being placed
together, and then with one and then the other placed after the verb:
(713) [ a-t=mika ]=na kwizha ii-ma-t1
[ come-pfpart=nmlzr.sg ]=top dog die-comp-term
`The dog which came has died.'
(714) sun kwizha ii-ma-t1=ma, [ a-t=mika
that dog die-comp-term=temp [ come-pfpart=nmlzr.sg
]=na
]=top
`That dog has died, the one that came.'
(715) sun [ a-t=mika ]=na ii-ma-t1,
that [ come-pfpart=nmlzr.sg ]=top die-comp-term
kwizha=na
dog=top
`That one that came has died, the dog.'
These alternate positioning of the elements  together in either order, or
with one or the other placed after the matrix verb  is precisely the situation
which occurs when two separate noun phrases referring to the same entity are in
apposition (see section 5.5).
As well as these distributional reasons for considering the noun and the
clause marked with mika/tuz to form two separate noun phrases in apposition,
there are also morphological reasons for this conclusion. Both the head noun
and the clause marked with mika/tuz may be marked with the Topic marker,
and any postpositional case marking must be present on both items.
(716) [ a-mtu=mika ]=na ashaNpa=na wan
[ come-impfpart=nmlzr.sg ]=top woman=top all
pyan-t1-zi
hit-past-nonlocut
`The woman who was coming hit everyone.'
19 This discontinuity is also common in Macushi (Abbott 1991:9396).
288 CHAPTER 10. SUBORDINATE CLAUSES

(717) [ Santos=ta pyaNta-ta=mika ]=ta

[ Santos=acc kill-pfpart=nmlzr.sg ]=acc
pyan-ta-w, ashaNpa=ta
hit-past-locut:subj woman=acc
`I hit the one who killed Santos, the woman.'
This indicates that there must be two noun phrases, in apposition, as the Topic
marker and case marking can only be marked once on each noun phrase within
the sentence.
Agentive nominalizations are clearly quite close semantically to relative
clauses. However they do not, in fact, comply with the conditions set by many
denitions of relative clauses, as there is no domain noun which is further re-
stricted by a subordinate clause. Even when the domain noun is explicitly
expressed, it is placed in apposition to the nominalization, and thus is not in
the same noun phrase. This construction is thus not a relative clause according
to the denitions of Keenan & Comrie (1977) or Comrie (1981), which both re-
quire a domain noun in the same noun phrase as the restrictive clause. Keenan
(1985:142), on the other hand, takes the dening feature of relative clauses to
be the presence of a restrictive clause, and considers that a domain noun is not
necessary in a relative clause construction  and under this denition, this con-
struction is perhaps a relative clause, although the use of identical morphology
on adjectives is somewhat disturbing in this case.20
There are a variety of languages where a nominalization functions as a
relative clause modifying a head noun. For example, Davis (1973:211) gives the
following sentence from Luiseño, where the presence of locative marking on the
verb `make' shows it has been nominalized:
(718) KuP
a;l-up niv
eP-qa w
w-Na nu-sN
aki pu-l
oPxa-Na
y-pres be:in-pres acorn:mush-loc my-wife her-make-loc
`There is a y in the acorn mush my wife made.'
While this example is not structurally identical to the Awa Pit nominalizations
with an apposed noun (for example, in Luiseño the possessive prex nu is only
possible on nouns), there are clear similarities, with the head noun and the
nominalization both separately case marked. In fact, all elements of an np in
Luiseño are normally case-marked, however in addition to this, while the head
noun is usually found immediately before the nominalization, this is not neces-
sarily the case, and the two elements can be found in a variety of places through
the sentence (Davis 1973:206). It is thus not entirely clear whether the head noun
and the nominalization in Luiseño are in the same np or in apposition.
In their discussion of nominalizations being used as relative clauses, Com-
rie & Thompson (1985) make the following comments which are highly relevant
to Awa Pit:
It is not dicult to understand how a nominalization can function
as a relative clause: the nominalization and the noun with which it
20 See section 10.4 for more explicit descriptions of the various denitions of relative clauses.
10.5. NOMINALIZATIONS 289

is in construction can be thought of as two juxtaposed nominal ele-

ments [nom][nom], the modifying relationship between them being
inferred by the language-users (rather than being specied by the
grammar, as it is in languages with specic relative clause morpho-
logy).
(Comrie & Thompson 1985:394)
While it is possible to consider the construction with mika/tuz to con-
stitute a relative clause construction, at least under some denitions of relative
clauses, it is perhaps unnecessary in terms of developing the simplest possible
grammar of a language. If this structure is considered a nominalization  func-
tioning as it does without a domain noun, to create an agentive nominalization
 then other grammatical structures, such as apposition, can be invoked to al-
low speakers to interpret sentences containing a nominalization and a head noun
in apposition in the same way in which an English speaker would interpret a
domain noun plus restrictive clause construction. Furthermore, by adopting this
analysis of these constructions it is unnecessary to construct a restriction on this
type of clause formation in Awa Pit to explain why the relativized constituent is
necessarily the Subject of the subordinate clause: by analyzing this structure as
an agentive nominalization, the restriction is inbuilt.
The conclusion, then, is that while these agentive nominalization con-
structions can be invoked to cover some of the same areas as relative clauses
in English (and other languages), there is no reason to consider them relative
clauses in their own right: the grammar is simpler and clearer if they are simply
considered nominalizations.
It is interesting to hypothesize about the history of these agentive nominal-
izations. If one simply examines the nominalizations involving mu, and compares
them with the relative clauses involving mu (see section 10.4), it seems that it
would be possible to consider that mika and tuz are, essentially, dummy nouns,
equivalent to the English one, simply on the basis of structural equivalency:
(719) na=na [ kal ki-m ] awa
1sg.(nom)=top [ work(1) work(2)-adjzr ] person
`I am a person who habitually works.'
(720) [ kal ki-m=mika ] ka-y
[ work(1) work(2)-adjzr=nmlzr.sg ] be:permanently-nonlocut
`He is one who habitually works.'
This structural equivalency would suggest that the sequence kal ki-m in
the latter example should be considered as the restrictive clause part of a relative
clause construction, with a head noun mika.21 Pursuing this analysis, on the
basis of a sentence such as
21This is essentially one of the analyses which has been applied to Classical Nahuatl relative
clauses (Langacker 1975); but compare Rosenthal (1972) and Karttunen (1976).
290 CHAPTER 10. SUBORDINATE CLAUSES

(721) [ a-t=mika ]=na ii-ma-t1

[ come-pfpart=nmlzr.sg ]=top die-comp-term
`The one who came died.'
one would expect to be able to form a sentence with a lexical noun in place of
the dummy noun mika:
(722) *a-t kwizha=na ii-ma-t1
come-pfpart dog=top die-comp-term
Unfortunately, informants reject this sentence in favour of the various options
given above, which all involve two noun phrases, one containing kwizha `dog',
and the other containing a nominalized clause.
Historically, it is possible that ungrammatical sentences such as that given
above were grammatical, and the Perfective Participle, the Imperfective Parti-
ciple, and mu could all be used to form true relative clause constructions. How-
ever in modern Awa Pit the two Participles cannot be used in relative clause
constructions, but only in agentive nominalizations, and only the adjectivizer mu
functions to form a true relative clause.
Chapter 11
Complex non-subordinate clauses
11.1 Introduction
The previous chapter dealt with the wide range of subordinate clause types in
Awa Pit. In addition to subordination, however, there are a variety of other
constructions in Awa Pit which involve the use of more than one verb under a
single intonation contour  that is, constructions which use more than one verb
in a sentence, but do not involve a main clause and a subordinate clause. These
constructions are the focus of this chapter.
Probably the most obvious complex construction of this type involves the
use of a main verb and an auxiliary verb within a single clause. In Awa Pit, all
the auxiliary verbs are stative, and are used together with non-nite main verbs.
The syntax and semantics of mainauxiliary constructions are examined rst.
There are a variety of strategies used to negate propositions in Awa Pit,
and these are described fully in the next chapter, together with the related inter-
rogative constructions. However a brief analysis of one of these negative construc-
tions is given in this chapter, showing why this particular construction should be
treated as a mainauxiliary construction.
While there is no clear data showing the existence of a grammaticalized
evidential system in Awa Pit, there is some interesting data suggesting that a
hearsay evidential may be developing from a construction involving a non-nite
lexical verb and a speech verb, and this is examined here.
Under certain conditions, Awa Pit allows two lexical verbs to appear to-
gether within a clause, with neither one being subordinate to the other. This
construction, very similar to a serial verb construction, is discussed here, as is a
formally almost identical construction involving a lexical verb and a verb showing
aspectual information. The same morphology used in the Serial Verb construc-
tion appears in the Conjoined Clauses construction, but with dierent syntactic
properties. This construction, with a meaning somewhat akin to conjunction, is
also described in this chapter.
Finally the discussion turns to juxtaposition of clauses, a very common
technique in Awa Pit, used to show a variety of relationships such as conjunction,
disjunction and setting.
292 CHAPTER 11. COMPLEX NON-SUBORDINATE CLAUSES

11.2 MainAuxiliary constructions

There are a number of constructions which involve a main verb in the Imperfective
Participle form or extended Perfective Participle form, followed by a fully inected
stative verb used as an auxiliary:
(723) Venancio tu ka-y
Venancio be:in:place.(impfpart) be:permanently-nonlocut
`Venancio really is [here].'
(724) Doris pit-t1-t tu-y
Doris sleep-term-pfpart be:in:place-nonlocut
`Doris is (lying down) asleep.'
At rst glance these appear to be absolute adverbial subordinate clauses (see
section 10.3.7), which use the Imperfective Participle or Perfective Participle
form:
(725) ampu pyan-ta=na tayaz-tu-at1-zi
man hit-pfpart=top be:sorry-impf-past-nonlocut
`Having hit [his wife] the man was sorry.'
(726) wiya ki-mtu t1t kwa-t1-zi
ght(1) ght(2)-impfpart cut drop-past-nonlocut
`While ghting, they cut [each other].'
However there are clear dierences between the two constructions.
There is a morphological distinction, to begin with. In the absolute con-
struction, the Perfective Participle (or the Imperfective Participle) is used, while
in the mainauxiliary construction the extended Perfective Participle, with the
addition of either the Completive or the Terminative inection to the Perfective
Participle, always occurs. A second dierence is that auxiliary verbs are always
stative, while in the absolute construction both verbs can be active or stative.
Additionally, the verb of an absolute clause may be followed by the Topic marker
na, which cannot be found attached to a main verb in a mainauxiliary construc-
tion. And most tellingly, the absolute adverbial clause (when it has the same
Subject as the main clause, as here) can be either embedded after the Subject or
placed in initial position before the Subject; the main verb in a mainauxiliary
construction is necessarily only found directly before the auxiliary.1
It seems likely the the origin of the mainauxiliary construction is to be
found in the absolute construction, but the two have diverged syntactically, mor-
phologically and semantically. Syntactically the main verb and the auxiliary verb
cannot be separated, and morphologically the main verb is an extended Perfective
Participle, rather than just the Perfective Participle of the absolute construction.
A mainauxiliary construction is, in fact, treated syntactically and morphologic-
ally as a single verb. The two are always found together with no morphological
1 The negative particle shi may occur between the two.
11.2. MAINAUXILIARY CONSTRUCTIONS 293

material between them (except the negative marker shi). The pair of verbs have
a Subject, and may have an Object or nps in other grammatical functions. The
pair take the suxes appropriate for stative verbs.

11.2.1 Imperfective Participle and auxiliaries

The various uses of Imperfective aspect (expressed through the Imperfective as-
pect inection and the Imperfective Participle) were examined in detail in sec-
tion 9.3.1: on-going activity or state, habitual activity or state, scheduled future
and generic statements. These same notions are expressed by the Imperfective
Participle with auxiliaries, although additional information is also conveyed.
There are two types of auxiliary used with the Imperfective Participle.
These are the pseudo-copula ka and the locational/postural verbs.2 With the
postural verbs, the auxiliaries are sometimes used, as might be expected, to
indicate the physical position of the Subject during the activity in question:
(727) ap aympihsh=na cama=ta pit-tu tala-y
my brother=top bed=in sleep-impfpart be:lying-nonlocut
`My brother is (lying) asleep in bed.'
However the use of the auxiliaries appears to have extended somewhat, so the
following sentence, for example, does not indicate that the Subject is sitting, but
rather seems to imply that the illness is prolonged:
(728) ish-tu uz-is
be:sick-impfpart be:sitting-locut
`I am sick.'
Unfortunately relatively few of these examples were recorded, and the exact se-
mantics of the dierent postural auxiliaries is unknown. While these postural
auxiliaries appear not to be used as widely as in some other languages, it would be
interesting to compare the extensions of meaning of the various dierent postural
verbs with those of other languages. Many other languages in South America use
postural verbs either as auxiliaries, as in Urubu-Kaapor (Kakumasu 1986:386
387); or in compounding structures, as in Barasano (Jones & Jones 1991:47);
and often these have developed extensions of meaning, to mark aspect, as in
Guambiano (Vásquez de Ruíz 1988:126127); or even more diverse meanings, as
in Sikuani (Queixalós 1992), where the auxiliaries have extended to such diverse
uses as an aective marker (`lying'), a permanency marker (`sitting'), a durat-
ive marker (`standing') and a diusive marker (`hanging'). In some languages
these verbs have even developed into verb suxes marking tense, aspect or mood
(Doris Payne 1990b:223226).
Much more common is the use of ka as an auxiliary.3 This auxiliary
implies an emphasis being placed on the event,4 which can either indicate that
2See Table 4.8 in chapter 4 for a list of these.
3It is interesting to note that forms similar to ka are found as copulas and auxiliaries in a
variety of South American languages (David Payne 1990).
4Compare the Emphasis marker ka, section 14.7.
294 CHAPTER 11. COMPLEX NON-SUBORDINATE CLAUSES

the activity is permanent, or else a contrastive idea, that the statement really is
true despite what you might think.
(729) escuela izhkwak=ki cancha tu
school opposite=at sports:eld be:in:place.(impfpart)
ka-y
be:permanently-nonlocut
`There's a sports eld opposite the school.'
(730) maza aympihsh=ma m1j ka-s
one brother=temp have.(impfpart) be:permanently-locut
`I have only one brother.'
(731) cigarrillo=na waa ka-y
cigarette=top there:is.(impfpart) be:permanently-nonlocut
`There are cigarettes.'
Given the use of the pseudo-copula ka as an auxiliary, it might be ex-
pected that the copula i could be used as an auxiliary with the Imperfective
Participle, especially given its use with the Perfective Participle (see below).
This is not possible however:
(732) *a-mtu i-s
come-impfpart be-locut
As was explained in section 3.3.2, rather than the i copula being used as an
auxiliary, which it probably was historically, in the modern language it would
seem that the main verb inection and the auxiliary i have fused together, and
in place of this mainauxiliary pair Awa Pit has the nite Imperfective aspect
forms and inections.

11.2.2 Extended Perfective Participle and auxiliaries

The extended Perfective Participle consists of an active verb stem, the Com-
pletive (ma) or Terminative (t1) aspect inection, and the Perfective Participle
inection (t). The two forms (with Completive and Terminative) appear to be
interchangeable; there is presumably some subtle semantic dierence between the
two, although this has not been determined.
This extended Perfective Participle can be used with auxiliaries in two
entirely dierent constructions. One, the Past Anterior, can be used with all
active verbs, and involves an honorary stative verb ma; this construction does
not aect the valency of the verb. The other construction, the Resultative, is
formed with a variety of stative auxiliaries, but is restricted to those main verbs
which imply a change of state; it reduces the valency of transitive main verbs.
11.2. MAINAUXILIARY CONSTRUCTIONS 295

11.2.2.1 The Past Anterior

The Past Anterior construction consists of the extended Perfective Participle
followed by a Past form of the verb ma. The origin of this auxiliary ma is
unknown, and as in this construction it is always immediately followed by Past
tense inection, it is not even clear whether this auxiliary should be considered
active or stative. However because of the apparent parallelism with other, stative,
auxiliaries, this form is treated as a stative verb. In fact, it is possible that this
auxiliary is related to the active verb ma- `remain', as in their cross-linguistic
study, Bybee, Perkins & Pagliuca (1994:64) list one language, Maithili, an Indic
language, as having an anterior construction using a verb `remain'. Perhaps even
more suggestive is the fact that Maithili's anterior is one of only eight anteriors
in the survey which was restricted to past reference, as is Awa Pit's.
Anteriors are used to establish that the situation occurs prior to reference
time and is relevant to the situation at reference time (Bybee, Perkins & Pagliuca
1994:54). The Anterior in Awa Pit is obligatorily Past,5 and consequently the
reference time is always established as before the time of speaking:
(733) na=na namna-ka=na, uspa=na
1sg.(nom)=top follow/catch:up:to-when=top 3pl.(nom)=top
wakata kan-ma-t ma-t1-zi
cattle tie-comp-pfpart anter-past-nonlocut
`When I caught up, they had tied up the cattle.'
Unlike the Resultative discussed in the following section, the Past Anterior
has few restrictions on its use. It cannot be used with stative verbs, but may be
used with all active verbs, whether they express a change of state or not:
(734) Carmen ishan-ma-t ma-zi
Carmen laugh-comp-pfpart anter-nonlocut
`Carmen had laughed.'
Also unlike the Resultative, the Past Anterior does not aect the gram-
matical relations within a sentence, with semantic roles being assigned to gram-
matical relations in the same way as for the Present Imperfective:
(735) uspa=na miimal 1-t1-t ma-zi
3pl.(nom)=top Chucunés go-term-pfpart anter-nonlocut
`They had gone to Chucunés.'
(736) Carmen=na pueblo=ta-s a-t kway-ka=na,
Carmen=top town=in-from come-sv drop-when=top
Santos=ta pyan-ma-t ma-zi Demetrio
Santos=acc hit-comp-pfpart anter-nonlocut Demetrio
`When Carmen arrived back from town, Demetrio had hit Santos.'
5As discussed in section 9.2.1, the Past inection t1 is often deleted after /a/ and before zi
(although the presence of the morpheme zi allows its recovery), and thus the sequences /at1zi/
and /azi/ are in free variation. This seems especially common in the Past Anterior, with /mazi/
much more frequent than /mat1zi/, although the latter does occur.
296 CHAPTER 11. COMPLEX NON-SUBORDINATE CLAUSES

(737) na=na a-ka=na, Santos=ta=na

1sg.(nom)=top come-when=top Santos=acc=top
t1t-ma-t ma-t1-zi
cut-comp-pfpart anter-past-nonlocut
`When I came, they had cut Santos.'
(738) pyan-t1-t ma-t1-s
hit-term-pfpart anter-past-locut:under
`They had hit me.'
The same idea as is expressed through the Past Anterior construction can
be expressed through the use of a non-nite main clause extended Perfective
Participle:
(739) Laureano paynya kuzhu pyaNta-ma-t
Laureano his pig kill-comp-pfpart
`Laureano had killed his pig.'
While it could be claimed that this is simply a Past Anterior construction with
ellipsis of the auxiliary, this has not been done for the reasons outlined in sec-
tion 3.3.2.
11.2.2.2 The Resultative
The Resultative construction is formed from a main verb in its extended Per-
fective Participle form, followed by a stative auxiliary. The auxiliary can be a
locational/postural verb (see Table 4.8) or the pseudo-copula ka, as with the
Imperfective Participle; however for Resultatives the commonest auxiliary is the
i copula.
The Resultative denotes a state that was brought about by some action
in the past (Bybee, Perkins & Pagliuca 1994:63). The Completive/Terminative
aspect marker plus Perfective Participle denotes that the action has fully taken
place by the reference time, while the tense marking on the copula verb indicates
the reference time at which the state holds. As with the Imperfective Participle,
the postural auxiliaries indicate more complex ideas that simply the physical
position of the Subject.
(740) Doris pit-t1-t tu-y
Doris sleep-term-pfpart be:in:place-nonlocut
`Doris is (lying) asleep (is in a state of having fallen asleep).'6
(741) Carmen=na alizhkul-ma-t tu-y
Carmen=top annoy-comp-pfpart be:in:place-nonlocut
`Carmen is annoyed (is in a state of having got annoyed).'
6Note that the stative verb tu `be in a place' is formally and probably historically related
to the active verb tu- `lie down'; and at times tu `be in a place' has a meaning element of `be
lying down'.
11.2. MAINAUXILIARY CONSTRUCTIONS 297

(742) kwa-t1-t ka-y

eat-term-pfpart be:permanently-nonlocut
`He has eaten (is in a state of having eaten).'
(743) iiN kwi-ma-t i
ame go:out-comp-pfpart be.(nonlocut)
`The ame is extinguished (is in a state of having gone out).'
(744) na=na pishkatu ayna-shi-ka=na,
1sg.(nom)=top sh cook-desid-when=top
p1t-ma-t a-t1-zi
rot-comp-pfpart be-past-nonlocut
`When I wanted to cook the sh, it was rotten (was in a state of having
rotted).'
(745) nash-ka=na, pal-ma-t
afternoon-when=top stop:raining-comp-pfpart
a-n1-zi
be-fut-nonlocut
`By the afternoon it will have stopped raining (will be in a state of
having stopped raining).'
(746) Bogot
a=ta-s kayl kway-ka, verano payl-ma-t
Bogotá=in-from return drop-when summer nish-comp-pfpart
a-npa-y
be-necess-nonlocut
`When I return from Bogota, the summer will be over (will be in a
state of having nished).'
This construction is necessarily only used with verbs which denote some
form of change of state, as the Resultative indicates a current state brought about
by the action covered by the verb. While Bybee, Perkins & Pagliuca (1994:54)
suggest that a resultative may only be used with telic verbs, this is clearly not the
case in Awa Pit, as there are change of state verbs which are not telic, and these
can be used together with the Resultative  for example p1t- `rot' in sentence
(744) above, or ayna- `cook' in:
(747) ayna-ma-t i
cook-comp-pfpart be.(nonlocut)
`It is cooked (is in a state of having been cooked).'
While the non-use of the Resultative with verbs which do not denote a change of
state is a semantic restriction, it also clearly means that this construction never
occurs with stative verbs, as these never indicate a change of state. Equally,
none of the ditransitive verbs recorded in Awa Pit (for example, kwin- `give',
naka- `take from') denote a change of state, and hence no ditransitive verbs may
be used in the Resultative.
298 CHAPTER 11. COMPLEX NON-SUBORDINATE CLAUSES

The Resultative always focusses on the entity which has undergone a

change of state, and consequently this construction causes a change in the as-
signment of semantic roles to grammatical relations for transitive verbs. For
intransitive change of state verbs in non-Resultative constuctions, the entity un-
dergoing the change of state is the grammatical Subject, and this entity remains
the Subject for the Resultative of an intransitive verb:
(748) alcalde=na ii-mtu-y
mayor=top die-impf-nonlocut
Subject V
`The mayor is dying.'
(749) na=na pueblo=ta a-t kway-ka=na, alcalde
1sg.(nom)=top town=in come-sv drop-when=top mayor
Subject
ii-ma-t a-t1-zi
die-comp-pfpart be-past-nonlocut
V Aux
`When I arrived in the town, the mayor was dead (was in a state of
having died).'
However for a transitive change of state verb, in non-Resultative uses the entity
undergoing the change of state is the Object:
(750) Demetrio=na Santos=ta kil-tu-y
Demetrio=top Santos=acc hit-impf-nonlocut
Subject Object V
`Demetrio is hitting Santos.'
When these transitive change of state verbs are placed in a Resultative construc-
tion, the semantic patient is expressed as Subject, and the agent is obligatorily
unexpressed:
(751) Santos=na kil-ma-t i
Santos=top hit-comp-pfpart be.(nonlocut)
Subject V Aux
`Santos has been hit (is in a state of having been hit).'
This change in the correspondence between semantic roles and grammatical rela-
tions of transitive verbs in resultative constructions is common in the languages
of the world (Bybee, Perkins & Pagliuca 1994:54).
A nal point to note is that the semantic notion of resultativity can be
conveyed by a non-nite main clause extended Perfective Participle:
(752) t1m=na wan azhpyan-ma-t
basket=top all destroy-comp-pfpart
`The basket is completely destroyed.'
11.3. NEGATIVES 299

The reasons for not considering this construction as a Resultative with ellipsed
auxiliary are examined in section 3.3.2.
It may be thought that using a main clause extended Perfective Participle
to correspond to either a Past Anterior or a Resultative construction could cause
ambiguity. In fact, for transitive verbs there is clearly no ambiguity  in the Past
Anterior construction there are two core arguments, the agent is Subject and the
patient is Object, while in the Resultative there is only one core argument, a
patient Subject. And for intransitives, the distinction is relatively unimportant:
(753) uspa=na 1-t1-t
3pl.(nom)=top go-term-pfpart
`(When I arrived,) they had gone (Past Anterior)',
`(When I arrived,) they were gone (Resultative)'

11.3 Negatives
One of the constructions which can be used to indicate negation involves the use
of two verbs.7 The construction consists of a main verb conveying the action or
state being negated, which is in a participial or Innitive form, and a following
auxiliary verb ki.
(754) Santos=na shi 1-t ki
Santos=top neg go-pfpart be.neg.(nonlocut)
`Santos hasn't gone.'
It clearly needs to be established that this construction does consist of a main
verb and an auxiliary verb in one clause, rather than a biclausal structure with a
matrix negative verb and a subordinate clause indicating the negated proposition.
To enable an unprejudiced discussion of the constructions, ki will be referred to
as the negative verb, while the verb conveying the action or state being negated
will be referred to as the lexical verb. The issue is thus to decide between two
competing analyses of utterances like those above: either the lexical verb is a
main verb while the negative verb is an auxiliary, parallel to the mainauxiliary
constructions; or the negative verb is a matrix verb, with the lexical verb forming
part of a subordinate complement clause, parallel to the complement clauses of
section 10.2. Both of these possibilities have been found occurring in the world's
languages (John Payne 1985:207222).
To begin with, it is clear that if this construction does involve subor-
dination, it is not simply the negative verb in the matrix clause and an entire
subordinated proposition, with the structure:
(755) [Subject (Object) Verb] ki
The negative verb shows person marking corresponding to the Subject, indicating
that the Subject is the grammatical Subject of the negative verb. However it
7 See section 12.5.3 for details.
300 CHAPTER 11. COMPLEX NON-SUBORDINATE CLAUSES

would still be possible to claim that the Subject has been raised from the
subordinate clause to the matrix clause, but that the remainder of the clause is
still subordinated:
(756) Subject [(Object) Verb] ki
There is a further fact which cannot be accounted for in this way, however.
In these negatives, in addition to the nal negative verb, the negative particle
shi appears. The particle occurs directly before the lexical verb  that is, in the
heart of the supposed subordinate clause:8
(757) Subject [(Object) shi Verb] ki
This shows that the subordinate clause analysis must be discarded in favour of a
mainauxiliary analysis.
However it must be noted that the aspectual notions carried by the main
verb and the tense carried by the auxiliary verb are free to vary independently
and meaningfully (see section 12.5.3), unlike in languages such as Fijian, where
only one of the main verb and the auxiliary verb is marked for concepts such as
tense or person (John Payne 1985:210).
It seems clear that historically the negative construction with ki was
biclausal, with a complement clause being subordinated to a negative copula,
synchronically still ki.9 At some point, however, the construction has come to
be monoclausal. It is interesting to speculate on the role that the change from a
system of evidentiality to a system of person marking may have played in this. It
was hypothesized in chapter 8 that Awa Pit originally had a system of sentence-
nal evidential markers. At that stage the appearance of one of these markers
sentence nally, after ki, would be perfectly coherent with a biclausal analysis 
the markers were showing the evidential status of the utterance. However once
these markers began to be used to show person rather than evidentiality, their
occurrence on ki would be more dicult to account for, synchronically, if the
participants in the activity were all in a subordinate clause, and this could lead
to the blurring of the clause boundaries.
Whatever the reasons for a change from biclausal to monoclausal interpret-
ation in the case of ki, the synchronic situation is that these negative utterances
involving ki are monoclausal, consisting of a main verb in a non-nite form,
together with an inected auxiliary ki.

11.4 The hearsay evidential

In the hypothesized change from an existential evidential system to a person-
marking system (see chapter 8), Awa Pit lost its earlier system of marking evid-
entiality. Synchronically, Awa Pit is without evidential markers, but there is a
construction which could be analyzed as an incipient hearsay evidential.
8 This could perhaps be treated as negative concord; but the fact that ki can never be used
on its own to indicate negation seems to go against this suggestion.
9 An alternative is that this ki is in origin the active verb ki `happen, do'; in this case the
structure would perhaps have meant `it didn't happen that . . . '.
11.4. THE HEARSAY EVIDENTIAL 301

As is discussed in section 10.2.3.1, indirect speech in Awa Pit is signalled

by a matrix clause containing an inected form of the verb kizh- `say', and a
subordinate clause indicating what was said, with its verb being either an Im-
perfective Participle or a Perfective Participle. When the Subject of the matrix
verb kizh- is plural, and the act of speech is in the Past, there are three possible
ways of stating this:
(758) (kutnya ampu=na) [ Enrique kamta pyaNta-ta ]
(three man=top) [ Enrique snake kill-pfpart ]
kizh-t1-zi
say-past-nonlocut
`The three men said that Enrique killed a snake.'
(759) (kutnya ampu=na) [ Enrique kamta pyaNta-ta ]
(three man=top) [ Enrique snake kill-pfpart ]
kizh-a-t1-zi
say-pl:subj-past-nonlocut
`The three men said that Enrique killed a snake.'
(760) (kutnya ampu=na) [ Enrique kamta pyaNta-ta ]
(three man=top) [ Enrique snake kill-pfpart ]
kizh-a-zi
say-pl:subj-nonlocut
`The three men said that Enrique killed a snake.'
The rst of these options is the most common, with the verb unmarked for
the number of the Subject. In the second utterance, the verb is marked for the
plurality of the Subject; as it is in the third, with the additional optional deletion
of the Past tense inection after an /a/ (see section 9.2.1). Naturally, in any of
these sentence, the Subject np can be ellipsed.
When hearsay is reported, that is when people make statements for which
no initiator is known or the speaker does not wish to indicate the initiator of the
knowledge, the third of the forms is invariably used:
(761) Enrique kamta pyaNta-ta kizh-a-zi
Enrique snake kill-pfpart say-pl:subj-nonlocut
`They said/it is said that Enrique killed a snake.'
The other two forms are simply not used in these circumstances. This invariant
choice of one of three apparently possible forms suggests that the form kizhazi is
being used as a synchronically simple marker of hearsay, although it is, of course,
still synchronically analyzable.
If this is indeed the case, Awa Pit parallels the nearby language Imbabura
Quechua, which has developed a hearsay evidential in the same way, although
without the use of a plural marker.10 In Imbabura Quechua, however, there is
10 See Jake & Chuquin (1979) and Cole (1985:14) for full details.
302 CHAPTER 11. COMPLEX NON-SUBORDINATE CLAUSES

a clear synchronic separation between the indirect speech construction and the
hearsay evidential construction  the Object status of the complement clause
containing the reported speech is formally marked through the obligatory use of
the accusative marker ta (examples here and below from Jake & Chuquin (1979)):
(762) huarmi yanu-shca-ta wambra-ca ni-n
woman cook-past.nom-acc boy-top say-3
`The boy says the woman cooked.'
(763) *huarmi yanu-shca wambra-ca ni-n
woman cook-past.nom boy-top say-3
When the verb nin is used to mark hearsay, however, the accusative marker
cannot be used:
(764) huarmi yanu-shca ni-n
woman cook-past.nom say-3
`It is said the woman cooked.'
(765) *huarmi yanu-shca-ta ni-n
woman cook-past.nom-acc say-3
Thus the Awa Pit situation appears to parallel the Imbabura Quechua, with the
important dierence that in Imbabura Quechua there is a clear formal dierence
between the indirect speech construction and the hearsay construction; in Awa
Pit there is no clear formal dierence, but there is the obligatory choice of one
particular construction when expressing hearsay, while this construction is one
of three options for indirect speech.

11.5 Serial Verbs

Awa Pit has a particular construction, called the Serial Verb construction here,
which has specic morphological and syntactic properties. The label `serial verb'
has been applied to a great many dierent constructions (cf. Sebba (1987:136),
Zwicky (1990)), and the Awa Pit construction does not, in fact, t many deni-
tions of serial verb constructions, being in some ways more similar to `conjoined
verb' constructions. The label `Serial Verb construction' has been used to dis-
tinguish this from the Conjoined Clauses construction discussed in section 11.7,
and because of its similarities to serial verb constructions, with two verbs joined
together apparently to form a single predication.
Unfortunately, the Serial Verb construction does not occur very often in
the data. This is probably more indicative of the data collection technique than
the actual frequency of Serial Verbs in Awa Pit, however. The data was generally
elicited, rather than being spontaneous, and as Spanish does not have a serial
verb construction, the elicitation of Serial Verbs relied on either the informants
choosing to encode a single verb in Spanish through a more natural Serial Verb
construction in Awa Pit, or to encode a Spanish sentence involving two verbs
11.5. SERIAL VERBS 303

joined with the Spanish conjunction y `and' with a Serial Verb, and the latter
was often translated by preference using a Conjoined Clauses structure. Some
additional information was collected through asking informants for grammatic-
ality judgements on suggested sentences, but this method is, of course, fraught
with problems.
The exact range and frequency of use of Serial Verb constructions has thus
not been established, and this area of the grammar of Awa Pit requires much more
work. Despite this, a number of facts about the Serial Verb construction have
been ascertained.
The Serial Verb construction unites two verbs to create what is essentially
a single idea. The rst verb, which denotes an event which is either chronologic-
ally prior to or simultaneous with that denoted by the second verb, is expressed
in form of a bare stem, if consonant-nal, or a stem suxed with t, if vowel-nal.
The second verb has the usual inectional possibilities available to a simple verb.
(766) gato=na tunya pizh ku-mtu
cat=top rat grab eat-impfpart
`The cat is grabbing and eating the rat.'
(767) sancocho=na sa-t kwa-tpa
meat:soup=top make-sv eat-oblig
`It is necessary to make and eat soup.'
The two verbs involved in a Serial Verb construction are strongly bound
together, and (with the exception of the t marking a serial construction) no
morphological material can appear between them, except the negative particle
(see below). Of particular interest is the fact that any semantic modication
applies necessarily to both verbs  thus they cannot dier for features such as
polarity. They act, externally, as a single verb, and can be used, for example, in
a subordinate clause:11
(768) s1 pyan-n1-ma-tu pala ayna-t kwa-n
rewood cut-prosp-comp-hort.1sg plantain cook-sv eat-inf
`I'm o to cut rewood to cook and eat food (plantains).'
The fact that the two verbs in a Serial construction necessarily have the
same polarity is even more interesting given the placement of the negative particle
 it comes between the two verbs, but applies to both (ta `give' in this example
will be discussed in the following section; here it can be treated as an aspectual
sux on kwa- `eat'):
(769) ayna-t shi kwa-t ta ki=na,
cook-sv neg eat-sv give.(impfpart) be.neg.(impfpart)=top
ii-mtu
die-impfpart
`Not cooking and eating, [one] dies (ie. if you don't eat, you die).'
11While Serial Verbs can occur in subordinate clauses, they contrast in this with the Con-
joined Clauses construction, which can only be used in main clauses.
304 CHAPTER 11. COMPLEX NON-SUBORDINATE CLAUSES

This is perhaps simply an extension of the fact that all morphology in these
constructions is associated with the second verb.
As noted above, there are strong restrictions on the verbs which can be
used in a Serial construction. Both verbs must be active, and must have identical
subcategorization frames. That is, they must require the same grammatical re-
lations as complements. Consequently, only two active intransitive verbs or two
active transitive verbs can be used in a Serial construction, for example; an
intransitive and a transitive verb cannot be united, even if the Object of the
transitive verb is unexpressed:12
(770) *Carmen=na a-t ayna-t1
Carmen=top come-sv cook-term
(771) *Santos=na k1h kwa-t ii-ma-t1
Santos=top leaf eat-sv die-comp-term
This restriction on shared arguments holds only for complements; adjuncts can
be added even though they (appear to) associate with only one of the verbs:
(772) Juan=na kukum iyaNpa=kasa payNta-t ku-mtu
Juan=top possum shotgun=with kill-sv eat-impfpart
`Juan killed and ate a possum with a shotgun.'
In many ways the Serial Verb construction corresponds to denitions of
serial verbs in the literature. For example, it ts the majority of the criteria
suggested by Aikhenvald (1996a) and Durie (forthcoming): it has the properties
of a single predicate, referring to a single event, with the two verbs semantically
sharing tense, aspect, mood and polarity, sharing all arguments, and apparently
having the intonational properties of a monoverbal clause.13
The issue which separates the Awa Pit construction from serial verbs is
the presence, after a vowel-nal rst verb, of the morpheme t. In general, den-
itions of serial verb constructions suggest that they contain two or more verbs
without overt markers of coordination or subordination (Sebba 1987:86). This
is, however, a little unclear: in Awa Pit many of the Serial Verb constructions
(those in which the rst verb stem is consonant-nal) contain no overt markers;
and in the case of the vowel-nal roots, it is not clear that t should necessarily
be treated as an overt marker of coordination or subordination, since it is not
used elsewhere, except with the Conjoined Clauses construction.
While the Serial Verb construction of Awa Pit would thus probably not be
considered as a true serial verb construction by many, it is perhaps worth indicat-
ing that, if it was treated as such, it would be a symmetrical construction in Foley
& Olsen's (1985) terms,14 in that both verbs are from open classes, with the order
12 This restriction to identical subcategorization frames does not hold for the use of the Serial
Verb construction in forming perfective aspect; see section 11.6.
13The precise intonational properties of Awa Pit clauses have not been analyzed; impression-
istically, the Serial Verb construction has monoclausal intonation.
14Unlike the Serial Perfective construction, which is asymmetrical.
11.6. PERFECTIVE ASPECT THROUGH SERIAL VERBS 305

of verbs being iconic; it would be nuclear rather than core serialization (Crowley
(1987:58); cf. Foley & Olsen (1985)), in that the verbs must share all arguments;
and in Durie's (forthcoming) terms, it is contiguous non-incorporating, with the
verbs always being together, but forming two separate phonological words.
There are presumably strong semantic restrictions also on what types of
verbs can be united in a Serial Verb construction. Unfortunately these cannot
be established on the basis of the available data. It should be noted, however,
that some of the more usual serialization uses  to add an instrument or
benefactive to a clause, for example  are never found in Awa Pit. However
other combinations, such as the ubiquitous `cookeat' in Awa Pit, are found
commonly in languages with clear serial verb constructions.
The Awa Pit Serial Verb construction thus consists of two active verbs with
identical subcategorization frames, which are obligatorily adjacent and occur in
a temporally iconic order. The normal range of morphological and syntactic
possibilities is open to the clause, with any inection occurring on the second
verb stem. The rst verb stem is either bare (if consonant-nal) or suxed with
t (if vowel-nal). Any modication, such as negation, applies semantically to
both verbs, but is indicated formally on the second verb.

11.6 Perfective aspect through Serial Verbs

The Serial Verb construction described in the previous section is used to indicate
that two events happened simultaneously, or one event preceded another, and
there are strong syntactic restrictions on the two verbs used in a Serial construc-
tion  the two verbs must have the same complement structure. However an
apparently identical construction is used with a dierent set of syntactic restric-
tions and a dierent semantic outcome with three rather unusual verbs, kway,
which is used very commonly, and kyan and ta, which are rather rare. These
verbs are used to express an aspectual quality, similar to perfectivity. Unfortu-
nately, because of the few examples with kyan or ta, no dierences between the
three could be determined.
Structurally, the Serial Perfective construction is almost identical to the
Serial Verb construction, although the construction is asymmetrical rather than
symmetrical in the terms of Foley & Olsen (1985), as the second verb necessarily
comes from a very small, closed class. The rst verb, which carries the lexical
content, is either a bare stem (if consonant-nal) or suxed by the Serial Verb
inection t (if vowel-nal), and the second verb (kway, kyan or ta) is fully in-
ected.15 As can be seen in the third and fourth examples below, the Past tense
allophone t1 is often elided following kway and ta.16 The normal range of syn-
tactic and morphological possibilities is open to the Serial Perfective construction,
including, for example, imperatives and subordinate constructions.
15The glossing of the second verb as drop, throw or give will be explained below.
16This is especially interesting following the form kway, as the Past tense marker is usually
only elided after vowels. As will be noted below, however, the word kway is almost alway
pronounced as [kwe].
306 CHAPTER 11. COMPLEX NON-SUBORDINATE CLAUSES

(773) shutta say-t kway-ta-w

hat look:for/nd-sv drop-past-locut:subj
`I found the hat.'
(774) na=na t1=na kwa-t kyan-ta-w
1sg.(nom)=top tree=top fell-sv throw-past-locut:subj
`I felled [this] tree.'
(775) Demetrio=na 1-t kway-zi
Demetrio=top go-sv drop-nonlocut
`Demetrio left.'
(776) pa azh kway-zi
sun shine drop-nonlocut
`The sun shone out.'
(777) ya-t kway-ti
proceed-sv drop-imp.sg
`Continue past!'
(778) pwin ta-naka
hide give-plt:imp
`Hide!'
(779) na a-t kway-ka=na, kal
1sg.(nom) come-sv drop-when=top work(1)
ki-mtu-ata-zi
work(2)-impf-past-nonlocut
`When I arrived, you were working.'
However while the usual verb inections are possible, a further inectional form
not available to active verbs is also possible  Present tense with no Imperfective
aspect.
(780) a-t kway-is
come-sv drop-locut
`I arrived.'
(781) ma=na a-t kway
now=top come-sv drop.(nonlocut)
`She just arrived.'
This behaviour implies that the verbs used in the Serial Perfective construction
are stative verbs;17 but they are also found in combination with the Imperfective
aspect sux, which is impossible for stative verbs, and is perhaps surprising,
given their meaning as indicating perfective aspect:
17The zero-marked Imperfective Participle form of the Serial Perfective verbs (see sen-
tence (769)) also suggests that they are stative verbs.
11.6. PERFECTIVE ASPECT THROUGH SERIAL VERBS 307

(782) kwak kway-mtu-y

cross drop-impf-nonlocut
`He has crossed [the river] (he's on the other side).'
This possibility of contrast is presumably because the meaning of the Serial Per-
fective construction focusses more on the fact that at some point of time (either
end-point or initial point) something happened, and there is some sort of result
from that. In the Imperfective example given above, what is important is that he
is on the other side at the moment  there is a continuing result of the action,
and this result is what is being looked at.
The ability of Serial Perfective verbs to act as either active or stative verbs
is perhaps related to their origin and synchronic status. These verbs are clearly
derived from fully active verbs in Awa Pit (see below), which would explain their
active nature; perhaps their current status, similar to auxiliaries, has allowed
them to also behave as stative verbs, as all other auxiliaries in Awa Pit are
stative (see 11.2).
As can be seen from the examples above, the Serial Perfective construction
does not have the same restrictions on identical subcategorization frames as the
Serial Verb construction. The Serial Perfective verbs can be used following an
active verb with any type of complement structure, zero-transitive, intransitive,
transitive, ditransitive. The only syntactic restriction is that the lexical verb
must be active (although this may be a semantic restriction in reality).
Another way in which the Serial Perfective construction diers from the
Serial Verb construction is that some categories which are semantically associated
with the main verb and marked through derivational marking or syntax, such as
number and polarity, are marked on the main verb rather than the second of
the verbs. Thus a (derivational) plural sux appears on the main verb, and
the negative marker precedes the main verb, while the negative sux and other
inectional material attaches to the nal verb:
(783) m1nat t1t-na-t kyan-ta-w
some cut-pl:obj-sv throw-past-locut:subj
`I cut some people (some-plural).'
(784) verano ap-ma-t1=ma, alu shi ki-t
summer enter-comp-term=temp rain(1) neg rain(2)-sv
ta-ma-y
give-neg-nonlocut
`Summer must have arrived, it hasn't rained.'
A particularly clear example of the distinction in position between features as-
sociated with Serial Verbs and with Serial Perfective markers is found in sen-
tence (769) above, repeated here for convenience:
308 CHAPTER 11. COMPLEX NON-SUBORDINATE CLAUSES

(785) ayna-t shi kwa-t ta ki=na,

cook-sv neg eat-sv give.(impfpart) be.neg.(impfpart)=top
ii-mtu
die-impfpart
`Not cooking and eating, [one] dies (ie. if you don't eat, you die).'
Here the negation is applied to the whole rst clause verb complex, but is marked
on (that is, before) the second of two verbs in the Serial Verb construction, but
on the (second) main verb rather than on the Serial Perfective verb; the negative
auxiliary is placed clause-nally, after the Serial Perfective verb.
The Serial Perfective construction has semantic restrictions placed on it.
The lexical verb involved must have either an inherent end-point (that is, be telic),
or else if atelic must have a sharply dened beginning. Thus the construction
cannot be used with verbs such as nash- `afternoon', where there is no particular
clear-cut initial or nal point. The Serial Perfective then focusses on this point, as
can be seen in the telic situation in sentence (773) and the atelic sentence (775).
A particularly common use of the Serial Perfective construction is together with
the verb a- `come', to form a-t kway, best translated as `arrive'; and when it is
transitive the verb k1z- `break' is always used with either kway or with kyan.
The Serial Perfective construction appears similar to the compound verb
construction of Indo-Aryan languages such as Hindi. In these languages, two
verbs can be used together, with the rst being bare, the second carrying in-
ection, to carry some aspectual distinction. This compound verb construction
has usually been analyzed as indicating perfectivity (Hook 1974, Hook 1991),
although more recently others have considered that its semantics indicate some
sort of resultativity (Agha 1994).
The comparison with Indo-Aryan languages is especially interesting given
the forms of the Serial Perfective verbs. While kway, kyan and ta have been
glossed as grammatical markers rather than lexical verbs, there are homophonous
verbs in Awa Pit:18 kway `drop', kyan `throw' and ta `give'. While there are
many more choices available in most Indo-Aryan languages, verbs meaning `drop',
`throw' and `give' are quite commonly used (Hook 1991).

11.7 The Conjoined Clauses construction

In section 11.5, the Serial Verb construction was described. This construction
united two verbs to form a single predication, with strong syntactic and semantic
restrictions. There is a similar construction, with far fewer constraints placed
upon it: the Conjoined Clauses construction.
The Conjoined Clauses construction unites two clauses rather than two
verbs. Both clauses must be active and have the same (notional) Subject, and
the clauses must be main clauses, but these are the only syntactic constraints
18There are slight dierences here in pronunciation: despite being written kway, the Serial
marker is almost invariably pronounced [kwe], while the verb kway `drop' varies between [kway]
and [kwe]; see section 2.4.1.1.
11.7. THE CONJOINED CLAUSES CONSTRUCTION 309

which hold for this construction. Semantically, it states that one event occurred,
then another. Morphologically, the rst verb stem is bare (if consonant-nal)
or suxed with t (if vowel-nal) and followed by kit, while the second verb
stem carries the usual range of inectional possibilities for main verbs. Any
complements or adjuncts which are associated with either verb stem can appear
in their usual positions, with the exception of the Subject of the second verb.
(786) Mar
a=na tazh kit ii-ma-t1
María=top fall and die-comp-term
`María fell over and died.'
(787) Santos=na k1h kwa-t kit ii-ma-t1
Santos=top leaf eat-sv and die-comp-term
`Santos ate a leaf and died.'
(788) Marcos=na a-t kit pala ku-ma-t1
Marcos=top come-sv and plantain eat-comp-term
`Marcos came and ate plantains.'
(789) na=na shihtt1 kit kit pala
1sg.(nom)=top hand wash and plantain
ayna-ma-ta-w
cook-comp-past-locut:subj
`I washed [my] hands and cooked plantains.'
While the verb of the activity which takes place rst is necessarily followed by kit,
and the verb of the activity which takes place second is inected, the clauses do
not necessarily occur in that order, but can be reordered, with a pause between
the clauses:
(790) kukum pyaNta-t kit kutil-tu
possum kill-sv and skin:with:re-impfpart
`They killed and skinned a possum.'
(791) kutil-tu, pyaNta-t kit
skin:with:re-impfpart kill-sv and
`They killed and skinned [it].'
The origin of this construction is presumably the Serial Verb construction
together with the verb ki, which is the usual word for `do', but is also used as
an auxiliary in the compound verb construction (see section 6.3.3). However
while the origin of the construction may be the rst verb in a Serial construction
with ki (explaining the choice between a bare stem for consonant-nal verbs and
a t for vowel-nal stems), and then ki in a Serial construction with the second
verb (explaining the nal t on kit), it cannot be analyzed synchronically in this
fashion, as there are many restrictions on Serial Verb constructions which simply
do not apply to the Conjoined Clauses construction. Consequently, while the
nal t on the rst verb in a Conjoined construction will be glossed in the same
way as for a Serial construction, the kit is considered an unanalyzable whole.
310 CHAPTER 11. COMPLEX NON-SUBORDINATE CLAUSES

11.8 Juxtaposition of clauses

In addition to all of the complex syntactic possibilities found in Awa Pit for
relating two clauses, there is another much simpler one: juxtaposition. Two fully
nite clauses can be said together under a single intonation contour, forming the
two into one sentence. This construction can be used with a variety of dierent
semantic eects, which must be interpreted by the hearer: conjunction (and),
disjunction (or), setting in time, cause and result. In none of these cases are
there any pivot or coreference restrictions between the clauses.
The most common use of juxtaposition is to indicate conjunction, that is,
to show that several things happened at once, or that one followed another in
time. If the two clauses have the same Subject, the Subject of the second clause
is ellipsed;19 if the actions in the clauses are the same with dierent Subjects, the
second and later Subjects are normally marked with kas, the Additive marker
(section 14.4).
(792) Di
ogenes=na wakata pay-t1-zi, manaz
Diógenes=top cattle buy-past-nonlocut again
pay-nin-n1-ma-t1
buy-caus-prosp-comp-term
`Diógenes bought cattle, and [now] he has gone to sell them again.'
(793) Santos wipu kwa-t kway-t1-zi, na=kas
Santos egg eat-sv drop-term-nonlocut 1sg.(nom)=add
kwa-t kway-ta-w
eat-sv drop-past-locut:subj
`Santos ate an egg, and I ate [one too].'
(794) Laureano Ricaurte=mal puz-t1-zi, Demetrio=kas
Laureano Ricaurte=loc go:out-past-nonlocut Demetrio=add
puz-t1-zi
go:out-past-nonlocut
`Laureano went to Ricaurte, and Demetrio went to Ricaurte.'
Juxtaposition is used very commonly in the languages of the world to
indicate coordination, and many South American languages share this; for ex-
ample, Yagua (Doris Payne & Thomas Payne 1990:294), Canela-Krahô (Popjes
& Popjes 1986:139) and Imbabura Quechua (Cole 1985:79).
Disjunction is done in the same fashion. Either the distinction from co-
ordination is shown as the two propositions are necessarily exclusive, or the second
part of the disjunction is introduced with an explicit negation of the rst:
19Ellipsis would of course normally happen if the clauses were separate sentences also, once
the Subject had been introduced in the rst.
11.8. JUXTAPOSITION OF CLAUSES 311

(795) na=kasa 1-mtu mizha=ma

1sg.(nom)=with go-impfpart how=inter
ka-s, nukkul-tu mizha=ma
be:permanently-locut stay-impfpart how=inter
ka-s?
be:permanently-locut
`Are you going with me, [or] are you staying here?'
(796) ma nashka=na Pueblo Viejo=mal 1-t1-mtu-s, shi
today late=top Pueblo Viejo=loc go-term-impf-locut neg
1-t ki=na pantalón
go-pfpart be.neg.(impfpart)=top pants
pat-m1z-tu-s
wash-incep-impf-locut
`Later on today I'm going to Pueblo Viejo, [or] not going to Pueblo
Viejo I'll wash my pants.'
Two clauses can be juxtaposed to indicate that they occurred together;
essentially coordination. However the same structure can be used to establish
one clause as a setting in which the other takes place. The rst, background
clause is marked as Imperfective, to establish the event of that clause as on-going
when the second event occurs.
(797) na=na p1t t1t-tu-ata-w, mayshti t1t
1sg.(nom)=top grass cut-impf-past-locut:subj machete cut
kway-is
drop-locut
`I was cutting grass, [when] the machete cut me.'
(798) na=na anshik=na
1sg.(nom)=top yesterday=top
pii-n1-mtu-ata-w, alu
bathe-prosp-impf-past-locut:subj rain(1)
ki-ma-t1-s
rain(2)-comp-past-locut:under
`I was going o to bathe yesterday, [when] it rained on me.'
More complex ideas such as cause and result can also be indicated through
juxtaposition, with the hearer being left to make the connection between the two
clauses.
(799) Pueblo Viejo=mal shi 1-mtu ki-s, alu
Pueblo Viejo=loc neg go-impfpart be.neg-locut rain(1)
ki-mtu-y
rain(2)-impf-nonlocut
`I'm not going to Pueblo Viejo, [because] it's raining.'
312 CHAPTER 11. COMPLEX NON-SUBORDINATE CLAUSES

(800) an perol=na p1na katsa, impuhs=ta=na alcanza shi

this pot=top very big replace=in=top t(1) neg
ki-ma-npa-y
t(2)-neg-necess-nonlocut
`This pot is too big, it's not going to t on the stove.'
In language contact situations, some languages which previously used jux-
taposition to indicate interclausal relations borrow coordinating conjunctions,
as has happened for example with Imbabura Quechua (Cole 1985:78) and Pipil
(Campbell 1987:255256): they have borrowed the Spanish conjunctions y `and'
and pero `but', among others.
While uncommon, a few examples of this sort have been found in Awa Pit,
most involving y `and' and one using pero `but'; it should be noted that in Awa
Pit, unlike in Spanish, only clauses can be coordinated in this way, not nps:
(801) na=na Demetrio=ta izh-tu y Carmen=ta
1sg.(nom)=top Demetrio=acc see-impfpart and Carmen=acc
izh-tu
see-impfpart
`I saw Demetrio and I saw Carmen.'
(802) Flavio=kas shaa-zi, y Jaime=kas shaa-zi
Flavio=add walk-nonlocut and Jaime=add walk-nonlocut
`Flavio came and Jaime came.'
(803) verano=na alu shi ki-kayN, pero verano=na
summer=top rain(1) neg rain(2)-negadjzr but summer=top
tam=miN
short=rest
`In summer it doesn't rain, but summer is short.'
Chapter 12
Interrogatives and negatives
12.1 Introduction
This chapter discusses two important and interrelated areas of Awa Pit  inter-
rogatives and negatives. These two semantic ideas cover a wide range of syntactic
phonemena, and are closely linked in Awa Pit, as in many other languages.1 The
concepts of interrogation and negation can be expressed in Awa Pit through `con-
tent' words such as m1n `who, no-one', and the negative particle shi; there are
inherently negative axes such as the Prohibitive suxes; there are suxes such
as ma, indicating a polar question, and the related ma, a negative sux; the neg-
ative copula ki; a negative auxiliary verb; interrogative question particles; and
ways of forming binary and multiple tag questions.
To begin with, interrogative and negative content words are discussed,
and the relationship between these and the indenites is commented on. Then
the variety of polar question constructions is dealt with; followed by the tag
questions. Indirect questions are not discussed in this chapter, but rather with
other subordinate clauses, in section 10.2.2. After dealing with all interrogatives,
the two main types of negative are examined  clausal negation (either with
a negative copula, negative sux or negative auxiliary verb) and non-clausal
negation (which is also used with certain non-nite clauses). There are also two
inherently negative suxes, which are not discussed here: the Prohibitives (see
section 9.4.5.2) and the Negative Adjectivizer (section 10.4).
Answers are not a separate speech-act in Awa Pit, but they require some
discussion. The answer to a question is simply a normal, declarative sentence.
As with any other sentence,2 an answer must contain a predicate.
For content questions, the minimal answer consists of the response to the
questioned element plus a predicate; in response to `Who did you hit?':
(804) Mario=ta pyan kway-ta-w
Mario=acc hit drop-past-locut:subj
`I hit Mario.'
1See Haegeman (1995:70111) for examples of links between negation and interrogatives in a
variety of languages, and a survey of how these links have been treated in generative grammar.
2With the exception of utterances consisting simply of an interjection.
314 CHAPTER 12. INTERROGATIVES AND NEGATIVES

m1n `who, no-one' m1nwaza `someone' Personal Pronoun

shi `what, nothing' shiwaza `something' Personal Pronoun
m1n `which' m1nat `some (unknown)' Adjective
yawa `how much/many' - - Quantier
mizha(pa)ka `when' - - Time Adverb
mizhuta `when, never' - - Time Adverb
m1n= `where, nowhere' - - Place Noun
mizha `how' - - Manner Adverb
shin `why' - - Manner Adverb
Table 12.1: Interrogative/negative content words and indenite words

That is, the minimal answer contains a predicate: it is not possible simply to
answer with the word corresponding to the questioned element (*Mario=ta). Of
course, if the predicate is the element being questioned, the minimal response
consists just of a predicate.
With polar questions likewise, the minimal answer is a complete sentence.
There are no words for `yes' or `no' in Awa Pit, and the usual answer to a polar
question is simply a predicate, containing either the verb or Copula Complement
(with or without a copula), with appropriate morphology, to indicate `yes'; or
the negated verb or Copula Complement plus morphology, to indicate `no'.
One point which must be kept in mind throughout the following discussion
is that in (true) questions the person-marking system is dierent from that in
statements or rhetorical questions. That is, as is discussed and exemplied at
great length in chapter 8, in information-seeking utterances, Locutor forms are
used to refer to second person, with Non-locutor referring to rst and third
person; in declaratives, Locutor refers to rst person and Non-locutor to second
and third person.

12.2 Interrogative/negative content words

Awa Pit has a series of words which are used in content questions (wh-questions,
constituent questions), those questions where the speaker and hearer share the
knowledge of a proposition . . . but the speaker does not know one element in the
proposition (Givón 1990:793); some of these words are also used in negative
content statements, such as `no-one came'. These words, discussed in detail
below, are summarized in Table 12.1 together with the indenite words.
The identity of content words and indenite words is not uncommon in
the world's languages, occurring for example in Macushi (Abbott 1991:104), the
Arawakan languages Amuesha and Campa (Wise 1986:573), Imbabura Quechua
(Cole 1985:131), and many Australian languages such as Yidiñ (Dixon 1977:182).
In discussing this identity in Yidiñ, Dixon suggests that these words are truly
indenite, but a cultural/linguistic feature converts this indeniteness into a re-
quest for information.
12.2. INTERROGATIVE/NEGATIVE CONTENT WORDS 315

However this does not seem to be the case for Awa Pit. Firstly, the in-
terrogative and indenite words are not formally identical; while the indenite
words are based on the interrogative/negative words, they do contain additional
phonological material.3 Additionally, the entire series of words can be used as
interrogatives, but only a few words of the set can be used as negatives or have
associated indenite forms, which suggests that the series is, essentially, interrog-
ative in nature.
When the interrogative or negative content words or the indenite words
are used in utterances, the rest of the sentence remains the same. That is, the
usual syntactic and morphological options  derivations, number, aspect, tense
and mood (with the exception of directives)  are available, although of course in
the case of questions the person marking carries a dierent meaning, and negat-
ive sentences are restricted in the usual fashion for clausal negation. In addition,
the argument (complement or adjunct) of a predicate containing a questioned
constituent normally carries the Interrogative clitic ma (see section 14.5), and
the argument of a predicate containing a negative constituent normally carries
the Additive clitic kas (see section 14.4):
(805) [ shi ]=ma pak-ma-t1?
[ what ]=inter burn-comp-term
`What has burnt?'
(806) [ m1n=pa kwizha=ta ]=ma comida kwin-ta-w?
[ who=poss dog=acc ]=inter food give-past-locut:subj
`Whose dog did you give food to?'
(807) [ shi ]=kas shi kizh-tu ki-s
[ nothing ]=add neg say-impfpart be.neg-locut
`I'm not saying anything.'
In questions, the element containing the question word is most commonly
found at the beginning of the sentence, as in the examples above. This position-
ing, while common, is not obligatory, and questioned elements may be found in
their usual position. For example, subordinate clauses may occur rst:
(808) na=na p1n-tu-ka=na, m1n=ta=ma
1sg.(nom)=top pass-impf-when=top where=in=inter
tu-a-zi?
be:in:place-past-nonlocut
`When I passed by, where was he?'
Other elements may also occur before the questioned element, in which case
they are most commonly topicalized elements. These elements are often, but not
always, set o by a pause:
3Although it must be noted that the interrogative and negative content words are usually
accompanied by the clitics ma and kas  see below.
316 CHAPTER 12. INTERROGATIVES AND NEGATIVES

(809) Libardo=na m1n-a=ma ta-zi?

Libardo=top who-acc=inter give-nonlocut
`Who did Libardo pay?'
(810) wakata=na, yawa=ma m1j?
cattle=top how:many=inter have.(impfpart)
`How many cattle do you have?'

12.2.1 M1n `who, no-one'

The word m1n `who, no-one' is used to question or negate a human noun con-
stituent of a sentence. The form m1n is used in the Nominative, and as the base
before most postpositions to form most other case-forms, such as m1n=pa `whose,
no-one's', m1n=kasa `with whom, with no-one'. However there is a special sux
for the Accusative, m1n-a rather than the expected *m1n=ta.4
(811) m1n=ma ka-s?
who=inter be:permanently-locut
`Who are you? (ie. Who's there?)'
(812) m1n=ma kii-t1-zi?
who=inter get:married-past-nonlocut
`Who got married?'
(813) sun escuela=na m1n=ma sa-zi?
that school=top who=inter make-nonlocut
`Who built that school?'
(814) m1n=kas shi a-ma-y
no:one=add neg come-neg-nonlocut
`No-one came.'
(815) m1n-a pyan kway-ta-w?
who-acc hit drop-past-locut:subj
`Who did you hit?'
(816) mayshti m1n-a=ma m1la-ta-w?
machete who-acc=inter give-past-locut:subj
`Who did you give the machete to?'
(817) m1n-a=kas shi izh-ma-s
no:one-acc=add neg see-neg-locut
`I didn't see anyone.'
4 It is unclear whether a locative form m1n=ta exists or not. When used as a translation
of, for example, to whom did the child go?, it was always translated back as where did the
child go?, using m1n= `where' rather than m1n `who'. The Accusative form m1n-a was not
acceptable in this context.
12.2. INTERROGATIVE/NEGATIVE CONTENT WORDS 317

(818) m1n=kasa=ma 1-ta-w?

who=with=inter go-past-locut:subj
`Who did you go with?'

12.2.2 Shi `what, nothing'

Only humans may be questioned or negated using m1n `who, no-one'; the corres-
ponding question/negative word for non-humans, whether animate or inanimate,
is shi. As would be expected for a pronoun which is used for non-humans, there
is no Accusative form, as there is no special marking to indicate Object status.
Shi may, of course, be followed by other postpositions, and also by relational
nouns.
(819) shi=ma ki-ma-zi?
what=inter do-comp-nonlocut
`What happened?'
(820) shi=ma ka-y?
what=inter be:permanently-nonlocut
`What is it?'
(821) piya=na shi=ma kwa-zi?
corn=top what=inter eat-nonlocut
`What ate the corn?'
(822) shi=kas shi kizh-tu ki-s
nothing=add neg say-impfpart be.neg-locut
`I'm not saying anything.'
(823) shi=ta=ma comida kwin-ta-w?
what=in=inter food give-past-locut:subj
`What did you give [it] food in?'
(824) shi=kasa=ma pyan kway-ta-w?
what=with=inter hit drop-past-locut:subj
`What did you hit [him/her] with?'
(825) shi ayuk=ta=ma libro ta-ta-w?
what inside=in=inter book put-past-locut:subj
`What did you put the book under?'
It should be noted that there is no interrogative verb `do what' in Awa
Pit. Rather, as in English and many other languages, the verb ki- `do' is used
together with shi `what':5
5 Note that ki- `do' is distinguished from both ki `be.neg' and the polar question marker ki
as it is active and is used with the Imperfective, while the latter two do not combine with the
Imperfective aspect sux.
318 CHAPTER 12. INTERROGATIVES AND NEGATIVES

(826) Gerardo=na shi=ma ki-mtu-a-zi?

Gerardo=top what=inter do-impf-past-nonlocut
`What was Gerardo doing?'

12.2.3 M1n= `where, nowhere'

The third interrogative/negative content word, m1n= `where, nowhere', is a place
noun, and is obligatorily followed by a locative postposition. It can only be
followed directly by ta `in' and pa `in(approx)', not by mal `loc' nor ki `at'. This
restriction on the following postpositions is presumably related to the semantics
of these postpositions.
While m1n= `where, nowhere' and m1n `who, no-one' have the same form,
they are clearly distinct words. M1n `who, no-one' refers obligatorily to a human,
while m1n= `where, nowhere' refers always to a place. There is a further homo-
phony within the interrogative/negative words, with m1n `which', an adjective.
This third word is an adjective, rather than a pronoun like the other two, and it
may refer to either humans or non-humans. Hence while homophonous, and all
interrogative/negative words, the three forms can be distinguished both gram-
matically and semantically. It is likely, of course, that all three are diachronically
related.
As noted above, m1n= `where, nowhere' can only be followed directly by
ta or pa, in either their locative or allative meanings, or suxed by s in their
elative use; other postpositions may follow ta or pa.
(827) m1n=ta=ma tu-y?
where=in=inter be:in:place-nonlocut
`Where is he?'
(828) m1n=ta=ma 1-mtu-s?
where=in=inter go-impf-locut
`Where are you going?'
(829) m1n=ta=kas shi 1-ma-s
nowhere=in=add neg go-neg-locut
`I didn't go anywhere.'
(830) m1n=pa=ma 1-mtu-s?
where=in(approx)=inter go-impf-locut
`Where are you going?'
(831) m1n=ta-s a-mtu-y, sun ampu=na?
where=in-from come-impf-nonlocut that man=top
`Where has that man come from?'
(832) m1n=ta=kima=ma 1-mtu-s?
where=in=until=inter go-impf-locut
`How far [up to where] are you going?'
12.2. INTERROGATIVE/NEGATIVE CONTENT WORDS 319

12.2.4 M1n `which'

As well as the interrogative/negative pronouns and nouns, there are two inter-
rogative adjectives, m1n `which' and yawa `how much, how many'. The rst of
these is homophonous with both m1n `who, no-one' and m1n= `where, nowhere',
however they are clearly distinct (see previous section for details). These two
adjectives are only interrogative, never negative.
M1n `which' is used to indicate which item out of a selection. It can be
used together with human or non-human nouns, and the noun phrase may be
case marked.
(833) m1n awa=ma shaa-ma-zi?
which person=inter walk-comp-nonlocut
`Which person was around?'
(834) m1n awa=ta pyan kway-ta-w?
which person=acc hit drop-past-locut:subj
`Which person did you hit?'
(835) m1n camisa=ma pay-shi-ta-s?
which shirt=inter buy-desid-dummy-locut
`Which shirt do you want to buy?'

12.2.5 Yawa `how much, how many'

The interrogative quantier yawa `how much, how many' can be used to ask for
the amount of a noun, or the qualication of an adjective. It is often used to ask
for a number or amount:
(836) awa=na, yawa izhpul=ma m1j-i?
person=top how:many soul=inter have-nonlocut
`How many souls do people have?'
(837) yawa shappi=ma kwa-ta-w?
how:much cane:juice=inter drink-past-locut:subj
`How much cane juice did you drink?'
It can also be used in this way without a noun, that is, as an adjective modifying
an ellipsed noun:
(838) wakata=na, yawa=ma m1j?
cattle=top how:many=inter have.(impfpart)
`How many cattle do you have?'
(839) yawa-tuz=ma kal ki-mtu-y?
how:many-nmlzr.pl=inter work(1) work(2)-impf-nonlocut
`How many [people] were working together?'
320 CHAPTER 12. INTERROGATIVES AND NEGATIVES

Yawa can also be used in an adverbial fashion, modifying an adjective,

asking how much of the adjective quality something has. One of the common
uses of this is to ask how much expensiveness something has, that is, how much
it costs:
(840) wakata yawa mun i?
cattle how:much expensive be.(nonlocut)
`How much does a cow cost?'
(841) kuzhu=na, yawa katsa=ma i?
pig=top how:much big=inter be.(nonlocut)
`How big is the pig?'
(842) Chical=na yawa attihsh=ma ka-y?
Chical=top how:much far=inter be:permanently-nonlocut
`How far away is Chical?'
An alternative way of asking the price of something is by using yawa
together with the postposition ki `at'  ki is used together with numbers to
indicate costs (see section 5.4.6).
(843) yawa=ki=ma pay-ta-w panela=na?
how:much=at=inter buy-past-locut:subj panela=top
`How much did you buy the panela (raw sugar) for?'

12.2.6 Mizha `how'

The question manner adverb mizha `how' is used in a variety of contexts. It can
be used to ask the way in which something was done:
(844) mizha=ma azhpizh-ta-w?
how=inter open-past-locut:subj
`How did you open [it]?'
(845) mizha=ma pyan-is?
how=inter know-locut
`How do you know?'
One of the ways of asking how someone is uses mizha (although it is more
common to enquire if someone is well using a polar question):
(846) ap aympihsh, mizha tu-ma-zi?
my brother how be:in:place-comp-nonlocut
`How is my brother?'
While mizha `how' is a component in the various ways of asking `when',
this will be discussed in the following section. It can also be used in the formation
of multiple-choice tag questions, which will be discussed below in section 12.4.2.
12.2. INTERROGATIVE/NEGATIVE CONTENT WORDS 321

12.2.7 Mizha(pa)ka `when' and mizhuta `when, never'

There are a variety of ways of asking when something happened. Two of these,
mizhaka and mizhapaka, appear to be more traditional, and the other, mizhuta,
appears to combine the word for `how', mizha, together with the Spanish word
hora `hour, time'. One of my informants only used the two traditional words
(for questions), while the others used all three forms. The form mizha(pa)ka has
never been found as a negative; even the informant who didn't use mizhuta for
questions used it in negative contexts.
The Awa Pit expressions for asking when something happened all involve
the word mizha `how'. It would seem that the closest translation to the tradi-
tional constructions is something like `at the time when things were how (did
it happen)', as these constructions use a sequence ka, probably related to the
subordinator ka which sets up the simultaneity of two events (see section 10.3.4).
It is not completely clear what the element pa is, although it seems most likely
that it is the postposition pa `in(approx)', which can be used for time, although
in other uses it has only been found referring to time of year (see section 5.4.4).
The distinction between the two traditional question words is one of time
depth, and corresponds to the two separate ways of asking `when' in local Spanish.
The form mizhaka, or local Spanish cuándo `when', is used for asking a general
`when', and could be answered by yesterday, tomorrow, ve years ago. The
form mizhapaka is more specic, and corresponds to local Spanish a qué hora
`at what time'; it can only be used to ask for a period within a day  it can be
answered by in the morning, in the afternoon, at 10 o'clock.
(847) mizhaka=ma payl-n1-zi?
when=inter nish-fut-nonlocut
`When will they nish?'
(848) mizhaka=ma kal ki-ta-w?
when=inter work(1) work(2)-past-locut:subj
`When did you work?'
(849) mizhapaka kayl-n1-s t1lawa=na?
when return-fut-locut tomorrow=top
`At what time will you be back tomorrow?'
(850) mizhapaka=ma a-ta-w?
when=inter come-past-locut:subj
`When did you come?'
The form mizhuta appears to be in variation with both of the above forms
in questions, for those speakers who use it. It is unusual, in that it is never used
together with the question particle ma.
(851) mizhuta nu=na Kutshu=ta-s kayl-tu?
when 2sg.(nom)=top San Isidro=in-from return-impfpart
`When (at what time) are you coming back from San Isidro?'
322 CHAPTER 12. INTERROGATIVES AND NEGATIVES

(852) mizhuta kayl-n1-zi?

when return-fut-nonlocut
`When (what day) is he coming back?'
The reason for the lack of the clitic ma here is unknown. It is also unclear
why only mizhuta, not mizha(pa)ka, can be used in negative contexts; in this
construction it is never followed by the clitic kas, unlike other negative content
words, and this is parallel to the lack of the clitic ma in questions.
(853) Cali=ta=na mizhuta shi 1-t ki-s
Cali=in=top never neg go-pfpart be.neg-locut
`I have never been to Cali.'

12.2.8 Shin `why'

The English word why can be used to ask about two separate ideas. It can be
used to cover the idea of purpose, answerable with a phrase beginning in order to;
and it can be used to give a reason, answerable with because. In Awa Pit these
two ideas are expressed very dierently. The former is asked using a combination
of shi `what', ki- `do' and the innite marker n(a), forming a purposive `in order
to do what':
(854) shi=ma ki-n a-ta-w?
what=inter do-inf come-past-locut:subj
`Why (in order to do what) did you come?'
(855) shi=ma ki-n 1-ma-t1-mtu-s pueblo=ta?
what=inter do-inf go-comp-term-impf-locut town=in
`Why (to do what) are you going to town?'
Shin `why' is only used to ask the reason for something. It can be used
on its own:
(856) shin=ma shi 1-ma-n1-s?
why=inter neg go-neg-fut-locut
`Why aren't you going to go?'
It can also appear followed by a form ka, which is presumably the pseudo-copula
verb, although this is unclear, and there is apparently no dierence in meaning:
(857) shin=ma ka pyan-ta-w?
why=inter be:permanently hit-past-locut:subj
`Why did you hit [him]?'
12.3. POLAR QUESTIONS 323

12.3 Polar questions

There are two constructions which are used to form polar (yes/no) questions
in Awa Pit, in dierent contexts. The simplest polar question form uses a verb
sux ma, which indicates both the idea of a polar question and the idea of Past
tense. The second type of polar question, the most frequent, involves the use
of question markers ki or sa. It is interesting to note, as will be taken up in
section 12.7, that negative statements use similar morphological material plus
the negative particle shi.

12.3.1 Polar questions marked by a sux

One of the ways of forming a polar question in Awa Pit is through the use of a
special verb sux ma,6 which directly follows a verb stem and cannot be combined
with any other inection except person marking, which here has the form s for
Locutor and an absence of marking for Non-locutor.7 Both of these features show
that the ma used to mark polar questions is distinct from the homophonous ma
used for negation (see section 12.5.1) which can be combined with Future tense
marking and has a Non-locutor sux y, as well as distinguishing it from the
Completive aspect ma (section 9.3.2), which has a wide range of combinatorial
possibilities.
The polar question sux ma can be used with both active and stative verb
stems, and apart from in combination with the Desiderative derivational sux,
it always indicates Past tense as well as a polar question. With the exception of
verb inection, a polar question shown through ma has the same constructional
possibilities as any nite main clause utterance.
(858) anshik=na a-ma-s?
yesterday=top come-Q.past-locut
`Did you come yesterday?'
(859) izh-ma-s?
see-Q.past-locut
`Did you see [it]?'
(860) Hugo=na pueblo=mal puz-ma?
Hugo=top town=loc go:out-Q.past.(nonloc)
`Did Hugo go out to the town?'
(861) up paynkul Libio=ta izh-ma?
your son Libio=acc see-Q.past.(nonloc)
`Did your son see Libio?'
6 Perhaps related historically to the Interrogative clitic ma, discussed in section 14.5.
7 For a discussion of this marking pattern, see section 8.4.1.
324 CHAPTER 12. INTERROGATIVES AND NEGATIVES

(862) pala=na waa-ma?

plantain=top there:is-Q.past.(nonloc)
`Were there plantains?'
The polar question sux ma appears to lose its Past nature when com-
bined with the Desiderative sux shi, questioning a present rather than a past
situation:
(863) nul=na kwa-shi-ma-s?
chontaduro=top eat-desid-Q.past-locut
`Do you want to eat chontaduro?'
However this appears to be a feature of the semantics of desideratives, rather than
anything to do with the sux ma; and indeed as is mentioned in section 6.4.2.1,
it is possible to ask a question in a Past tense form referring to a present situation
in many varieties of English.

12.3.2 Polar questions with ki or sa

The status of the syntactic structure of polar questions with ki or sa is extremely
unclear, although the structures themselves are clear. The most common struc-
ture involves a lexical verb (either active or stative, but never the copula i),
together with all of its normal arguments and modiers, in either the Innitive
form or one of the two Participle forms, followed immediately by either ki or sa,8
with these words having no sux for Non-locutor, or a nal s if there is some
Locutor element in the utterance:9
(864) Hugo=na Ricaurte=ta puz-na ki?
Hugo=top Ricaurte=in go:out-inf Q.(nonlocut)
`Will Hugo go to Ricaurte?'
(865) t1lawa a-n ki-s?
tomorrow come-inf Q-locut
`Are you coming tomorrow?'
(866) t1lawa a-n sa-s?
tomorrow come-inf Q:unsure-locut
`Would you be coming tomorrow?'
(867) ashaNpa m1l ki-s?
woman have.(impfpart) Q-locut
`Do you have a wife?'
(868) Alicia=na shappi k11-mtu sa?
Alicia=top cane:juice mill-impfpart Q:unsure.(nonloc)
`Would Alicia be milling cane?'
8 Once again, only the negative marker shi can intervene.
9 This marking pattern is discussed in section 8.4.1.
12.3. POLAR QUESTIONS 325

(869) nu=na Ricaurte=mal 1-t ki-s?

2sg.(nom)=top Ricaurte=loc go-pfpart Q-locut
`Did you go to Ricaurte?'
(870) Demetrio=na paynya yal=ta 1-t
Demetrio=top his house=in go-pfpart
sa?
Q:unsure.(nonloc)
`Would Demetrio have gone to his house?'
In this construction the form of the non-nite lexical verb expresses tense,
with future shown by the Innitive, present by the Imperfective Participle, and
past by the Perfective Participle. The dierence between ki and sa can perhaps be
understood from the translations (compare especially sentences (865) and (866)).
The use of sa rather than ki implies that the addressee may not know the answer
to the question, a kind of do you happen to know if or do you think that
question.10 Sa occurs much more commonly in the Non-locutor form, which is
to be expected given the meaning, as an addressee is less likely to know about
someone else's action rather than his or her own. Locutor person marking can
be used with sa  however only together with an Innitive lexical verb, referring
to the future, which also ts with its meaning, as it is possible that an addressee
will not be sure of his or her own future actions, but is (normally) aware of his
or her past or present actions.
In addition to this structure with a lexical verb, it is also possible to use
ki and sa following a verbless copula construction, where the ki or sa appears
almost like an interrogative copula, although necessarily untensed and referring
to the present:
(871) an kwizha=na alizh ki?
this dog=top erce Q.(nonlocut)
`Is this dog erce?'
Perhaps the parallelism between these question markers and the copula i is most
clearly seen in one way of enquiring about someone's health, and the answer to
this query:11
(872) nu wat=miN ki-s?
2sg.(nom) good=rest Q-locut
(na) wat=miN i-s
(1sg.(nom)) good=rest be-locut
`Are you well? I am well.'
10 This accords well with the use of the homophonous (and presumably diachronically related)
complementizer sa, used in indirect questions; see section 10.2.2.
11This exchange was never observed in natural speech, with the usual form of greeting con-
sisting of a discussion of the weather, followed by a query about where the parties were going to
and coming from. The above transaction may well be a calque from the standard local Spanish
greeting. However the youngest informant, A, did use it, and if other informants were queried
about their health in the above manner, they responded as above. Thus while not perhaps a
traditional greeting, it does show the parallelism.
326 CHAPTER 12. INTERROGATIVES AND NEGATIVES

These structures suggest an analysis where ki and sa are considered as

either interrogative copulas (when used without a lexical verb) or as auxiliaries
with a main lexical verb, as they appear similar to the auxiliary constructions dis-
cussed in section 11.2, and indeed very similar to the clausal negation structures
discussed in sections 12.5.2 and 12.5.3 below, which are analyzed as involving the
use of a main lexical verb in the Innitive or a Participle form, together with the
negative auxiliary ki. However there are problems with this analysis.
Firstly, the auxiliaries used in Awa Pit are a subset of verbs. But the in-
terrogative words ki and sa never take any of the verbal derivations or inections.
They are necessarily either uninected or inected with s indicating a Locutor
entity, and while person marking is most often associated with verbs, in Awa
Pit there is clear evidence of person marking on non-verb elements, such as the
negative particle and a postposition (see section 8.4.1). It could thus be claimed
that perhaps ki and sa are very defective verbs, with no verbal suxation, but
it seems better to look for alternative analysis if possible. If these words are not
verbs, they cannot be auxiliaries like the auxiliaries in section 11.2, so a new
structure is required in any case.12
Secondly, in mainauxiliary constructions such as the clausal negatives,
the meanings conveyed by the non-nite suxes on the lexical verbs are dierent.
In the polar interrogatives, straightforward tense is conveyed by these forms, so
that, for example, an adjunct referring to past time can be added to a clause
with a Perfective Participle:
(873) anshik=na Pedro=na puz-ta ki?
yesterday=top Pedro=top go:out-pfpart Q.(nonlocut)
`Did Pedro go yesterday?'
An apparently structurally identical negative clause, on the other hand, conveys
the present state (Present auxiliary) of a past event (Perfective Participle), and
a past time adjunct cannot be added to such a clause:13
(874) (*anshik=na) Pedro=na shi puz-ta
(yesterday=top) Pedro=top neg go:out-pfpart
ki
be.neg.(nonlocut)
`Pedro hasn't gone (*yesterday).'
This suggests that ki and sa are not forming an auxiliary construction.
It is also possible to form negative polar questions  expecting a negative
response  by the addition of the negative particle shi after the lexical verb and
before the question word:14
12Note that these could perhaps be considered as non-verb auxiliaries, in the same way as
Gazdar, Pullum & Sag (1982), for example, treat the English word to as an auxiliary (see also
Pullum 1981). In any case, a new structure is required.
13See 12.5.3 for details.
14It is clear that this is the interrogative ki rather than the negative ki, as it cannot carry
tense.
12.3. POLAR QUESTIONS 327

(875) nu=na Ricaurte=ta 1-mtu shi ki-s?

2sg.(nom)=top Ricaurte=in go-impfpart neg Q-locut
`Aren't you going to Ricaurte?'
(876) anshik=na a-t shi ki?
yesterday=top come-pfpart neg Q.(nonlocut)
`Didn't he come yesterday?'
(877) Demetrio=na t1lawa a-n shi ki?
Demetrio=top tomorrow come-inf neg Q.(nonlocut)
`Won't Demetrio come tomorrow?'
Marking negation in this fashion suggests two possibilities. This placement of
negation is appropriate for mainauxiliary constructions (see section 12.5.3), but
arguments against this analysis were given above; alternatively, it is possible that
these question clauses are, in fact, non-nite, which is also suggested by the non-
occurrence of the copula verb i in these structures. That is, it looks as though
these clauses correspond to main non-nite clauses such as:
(878) ashaNpa m1l
woman have.(impfpart)
`I/you/he has a wife.'
(879) Alicia=na shappi k11-mtu
Alicia=top cane:juice mill-impfpart
`Alicia is milling cane.'
(880) an kwizha=na alizh
this dog=top erce
`This dog is erce.'
The only dierence between these non-nite main clauses and the questions (867),
(868) and (871) above is the presence or absence of the sentence-nal ki(s) or
sa(s). However there are problems with this analysis also. The Imperfective
Participle non-nite clauses and the verbless copula clauses correspond, but not
the other forms. In non-nite main clauses, (simple) Perfective Participles are not
used, only extended Perfective Participles; and Innitives have not been found
in non-nite main clauses (see section 3.3.2). These two forms occur only in
mainauxiliary constructions such as clausal negatives or in subordinate clauses,
such as indirect question complements.
The polar question construction using ki or sa thus appears to form a very
unusual structure, half-way between a non-nite main clause with following ques-
tion particle, and a mainauxiliary construction. It is interesting in this respect to
examine the uses of homophonous words ki and sa. The negative verb ki is used
either as a copula or a tensed auxiliary verb in a mainauxiliary construction. Sa,
on the other hand, exists as a complementizer in indirect question complements,
together with Innitive/Participle verbs. The polar interrogative construction
328 CHAPTER 12. INTERROGATIVES AND NEGATIVES

with ki or sa appears to combine some features of each of these other uses. This
suggests that historically there were two words, ki and sa, with dierent gram-
matical features, perhaps similar to those of negative ki and complementizer sa
respectively. However the use of both words to form questions has led to the
separation of ki and sa as question markers from negative ki and complementizer
sa; and the use of these two question words (originally with distinct grammatical
features) in similar contexts has led to them developing towards each other and
away from negative ki and complementizer sa.

12.4 Tag questions

There are two further question forms in Awa Pit, the `binary-choice tag question'
and the `multiple-choice tag question'. These tag questions combine elements
of both question-word questions and polar questions. They are not true tag
questions, at least not under Ultan's (1978a:223) denition: [Tag questions] are
always clitic, usually enclitic, to a sentence, most often declarative. While both
forms of these questions in Awa Pit do occur at the end of a sentence, they form a
matrix clause, and the remainder of the sentence is not declarative, but rather a
non-nite subordinated form. However they have the same sorts of semantics as
tag questions, and consequently have been labelled as such, despite their dierent
structural features.

12.4.1 Binary-choice tag questions

Binary-choice tag questions in Awa Pit are composed of a non-nite subordinate
clause  either a verbless copula construction, a stative Imperfective Participle,
an active verb in Imperfective Participle, extended Perfective Participle or In-
nitive form, or an active verb with Adjectivizer  followed (optionally) by the
cliticized nominal question particle ma, followed by the ka-copula, either un-
marked (Imperfective Participle) or in the Perfective Participle form, and nally
the interrogative marker ki, together with person marking:
8 9
>
> (verbless copula clause) >>
>
< Imperfective Participle > =
Subordinate clause->ext. Perfective Participle > (=ma) ka(t) ki(s)
>
> Innitive >
>
: Adjectivizer ;
The interrogative marker ki naturally does not take tense marking, only person
marking (s for Locutor, unmarked for Non-locutor). Time reference can be indic-
ated on the ka-copula (either unmarked present, or past through the Perfective
Participle) and also, if the subordinated verb is active, through the choice of
non-nite form.
These tag questions expect a positive answer. For example, sentence (881)
below was to be asked at a shop, where it would be expected that there would
12.4. TAG QUESTIONS 329

be cigarettes. Similarly, sentence (884) was translated as seguro vos vas a salir?
`you're denitely going to town?', while sentence (885) was translated as ayer
parece que ha molido? `it seems as though you were milling yesterday?'.
(881) cigarrillo waa=ma ka
cigarette there:is.(impfpart)=inter be:permanently.(impfpart)
ki?
Q.(nonlocut)
`There are cigarettes, aren't there?'
(882) shaa=ma ka-t ki-s?
walk.(impfpart)=inter be:permanently-pfpart Q-locut
`You were around (yesterday), weren't you?'
(883) Miguel=na blanco=ma ka
Miguel=top white=inter be:permanently.(impfpart)
ki?
Q.(nonlocut)
`Miguel is white, isn't he?'
(884) t1lawa=na miimal 1-mtu=ma
tomorrow=top Chucunés go-impfpart=inter
ka ki-s?
be:permanently.(impfpart) Q-locut
`You're going to Chucunés tomorrow, aren't you?'
(885) anshik=na k11-mtu=ma ka-t
yesterday=top mill-impfpart=inter be:permanently-pfpart
ki-s?
Q-locut
`You were milling yesterday, weren't you?'
(886) nu=na=ma pantalón pat-t1-t=ma
2sg.(nom)=top=temp pants wash-term-pfpart=inter
ka ki-s?
be:permanently.(impfpart) Q-locut
`You've just washed your pants, haven't you?'
(887) Demetrio Pueblo Viejo=mal 1-n=ma
Demetrio Pueblo Viejo=loc go-inf=inter
ka ki?
be:permanently.(impfpart) Q.(nonlocut)
`Demetrio will go to Pueblo Viejo, won't he?'
330 CHAPTER 12. INTERROGATIVES AND NEGATIVES

(888) shiya=na ku-m=ma ka

pineapple=top eat-adjzr=inter be:permanently.(impfpart)
ki-s?
Q-locut
`You eat pineapple, don't you?'

12.4.2 Multiple-choice tag questions

Just like the binary-choice tag questions (see previous section), the multiple-
choice tag questions are not truly tag questions, in that the main sentence is
subordinated to the tag. However, they are similar to the multiple-choice tag
questions discussed by Ultan (1978a:226), in that they involve some idea of how
about, and indeed use the interrogative manner adverb mizha `how' in their
formation. Unlike the multiple-choice tag questions discussed by Ultan, the Awa
Pit construction places the tag at the end of the sentence, not at the beginning.
While the binary-choice tag questions ask a specically polar question with an
expected positive reply, these multiple-choice tag questions ask a polar question,
but also ask for more details. They were sometimes translated into Spanish with
an introductory cómo es? `how is it, what's it like?'.
The construction of these multiple-choice tag questions is similar to the
way in which binary-choice tag questions are formed. First there is the sub-
ordinate clause, in the same choices of non-nite verb form as above. For the
multiple-choice tags, this is then followed by the question word mizha `how', fol-
lowed usually by the question particle ma, although this is sometimes ellipsed.
Then nally the ka-copula occurs, with person marking. As with all questions
involving content words, there is no sentence-nal question marker.
(889) Miguel=na blanco mizha=ma ka-y?
Miguel=top white how=inter be:permanently-nonlocut
`Miguel is white, or what?'
(890) na=kasa 1-mtu mizha=ma
1sg.(nom)=with go-impfpart how=inter
ka-s, nukul-tu mizha=ma
be:permanently-locut stay-impfpart how=inter
ka-s?
be:permanently-locut
`Are you coming with me, or staying here, or what?'
(891) us=na kwa-t1-t mizha=ma
3sg.(nom)=top eat-term-pfpart how=inter
ka-y?
be:permanently-nonlocut
`So has he eaten, or what?'
12.5. CLAUSAL NEGATION 331

(892) wanmal-na mizha=ma ka-s?

meet-inf how=inter be:permanently-locut
`So do you reckon you'll meet them, or what?'
The two tag question forms have similar structures, diering in a similar
way to polar and content questions  a sentence-nal question marker in one, a
content question word in the other.

12.5 Clausal negation

As mentioned earlier, negation takes a variety of forms in Awa Pit. The in-
herently negative suxes, the Prohibitives (section 9.4.5.2) and the Negative
Adjectivizer (section 10.4), are discussed elsewhere, and no further mention will
be made of them here. The interrogative/negative content words were discussed
in section 12.2, however when one of these words is intended to be understood
negatively, the clause is also negated in the usual fashion. The remaining negat-
ive contexts all involve the use of the negative particle shi, and divide into two
types, clausal negation and non-clausal negation.
Clausal negation, the subject of this section, involves the preverbal use of
shi, plus some other indication of negation  either a special verb sux, a negat-
ive copula verb, or a negative auxiliary verb.15 Non-clausal negation, discussed in
section 12.6 below, involves only one marking of negation, with shi being placed
after the element to be negated. The terms `clausal' and `non-clausal' negation
have been chosen because of the prototypical uses of these constructions: nite
clauses are negated using the clausal negation strategy, while elements such as
adjectives are negated by the non-clausal strategy. It must be kept in mind, how-
ever, that some non-nite clauses are negated using one construction, some using
the other, and others cannot be negated, with alternative constructions being
used. Those subordinate clauses which are negated using the clausal negation
strategy are marked as + Negation in Table 10.1 in chapter 10. Non-nite main
clauses are negated by the non-clausal strategy.
There are three structures used for clausal negation: a verb sux ma can
be used to indicate negation in either Past or Future; there is a special negative
verb ki used in copula constructions with a copula; and the negative copula ki
can be used as an auxiliary verb to negate either active or stative verbs.
15 The use of two markers of negation is similar to the marking found in Imbabura Quechua
(Cole 1985:8386), although in that language the two markers bracket whichever element is in
the scope of negation  for some speakers the second marker (chu), must be after the verb,
making it parallel to the Awa Pit element ma or ki, however the rst marker (mana) is still
free to mark the scope of negation, unlike the Awa Pit shi, which must appear directly before
the verb in clausal negation.
332 CHAPTER 12. INTERROGATIVES AND NEGATIVES

12.5.1 Negatives with ma

The rst method of marking clausal negation is to place the negative particle shi
before the verb and sux the verb with ma.16 This strategy is used with either
active or stative verbs, but not the copula i.
While the negative sux ma appears related to the Past interrogative
sux ma, there are formal dierences, as well as semantic ones (see section 12.3.1
for details). Negative ma follows a verb stem, and cannot be combined with
inectional aspect. If it is followed directly by person marking, either Locutor s
or Non-locutor y, it shows Past tense:
(893) kwizha na-wa=na sula shi ku-ma-s
dog 1sg-acc=top bite(1) neg bite(2)-neg-locut
`The dog did not bite me.'
(894) pala=na shi waa-ma-y
plantain=top neg there:is-neg-nonlocut
`There were no plantains.'
(895) Santos=na shi 1-ma-y
Santos=top neg go-neg-nonlocut
`Santos did not go.'
When used in this way to indicate Past tense (as well as negation), ma implies
that the activity did not go to completion, suggesting a link with the Completive
sux ma (see section 9.3.2), although synchronically there are clear formal dier-
ences between the suxes. This idea of completion can be seen in the following
sentence, where the ma in the main clause shows that the culmination of the
activity (`nding') was not reached, even though the initial stages (`looking for')
were carried out; note that the verb root in both clauses is the same:
(896) say-kikas, shi say-ma-s
look:for/nd-concess neg look:for/nd-neg-locut
`Although I looked for [it], I didn't nd [it].'
While the sux ma with no tense marking indicates past tense, this is
sometimes further signalled through the use of ta as a Perfective Serial Verb (see
section 11.6 for the use of Serial Verbs to indicate aspect). This then indicates
perfectivity and past tense, and the negative marker ma occurs on the Serial Verb
ta.
(897) verano=na ap-ma-t1=ma, alu shi ki-t
summer=top enter-comp-term=temp rain(1) neg rain(2)-sv
ta-ma-y
give-neg-nonlocut
`Summer has arrived, it hasn't rained.'
16It is interesting to note the widespread use of a form similar to ma as a negative morpheme
in a variety of languages in South America (David Payne 1990).
12.5. CLAUSAL NEGATION 333

(898) na-wa shi pit-nin ta-ma-s

1sg-acc neg sleep-caus give-neg-locut
`They wouldn't let me sleep.'
In addition to being used with no tense marking, indicating Past tense,
ma can be followed by either the Future tense inection or the Necessitive mood
inection (plus person marking) to indicate the expected non-occurrence of a
future activity:
(899) alu ki-mtu-ka=na, shi a-ma-n1-s
rain(1) rain(2)-impf-when=top neg come-neg-fut-locut
`If it's raining, I won't come.'
(900) min-tu-s, Demetrio=na shappi shi
think-impf-locut Demetrio=top cane:juice neg
k11-ma-n1-zi
mill-neg-fut-nonlocut
`I think that Demetrio won't mill.'
(901) akkwan shi a-ma-npa-y
many neg come-neg-necess-nonlocut
`Many [students] won't come.'

12.5.2 The negative copula

There is a negative construction in Awa Pit which corresponds to the armative
copula construction with a Copula Complement and the copula i. As expected,
this construction involves placing the negative particle shi before the verb, but
rather than the copula verb i, the special negative copula ki occurs.17 This
negative copula has the expected tense and person forms, which distinguishes it
from the question marker ki which cannot take tense marking (see section 12.3.2).
(902) na=na taytta=kana shi ki-s, an aynki
1sg.(nom)=top father=like neg be.neg-locut more small
`I'm not like my father, [I'm] smaller.'
(903) ap kwizha=na p1na katsa shi ki
my dog=top very large neg be.neg.(nonlocut)
`My dog is not very large.'
(904) ap gallo shi ki-a-zi
my rooster neg be.neg-past-nonlocut
`It wasn't my rooster.'
17Note that negative copula constructions can also be verbless, and then use non-clausal
negation.
334 CHAPTER 12. INTERROGATIVES AND NEGATIVES

12.5.3 The negative auxiliary construction

The nal clausal negation construction once again uses the negative particle shi
before the verb, but then places the active or stative lexical verb in a non-nite
form  Perfective Participle or Imperfective Participle  and follows it with a
tensed form of the negative copula ki as an auxiliary. The mainauxiliary status
of this construction was examined in section 11.3.
(905) shi pana ki
neg be:standing.(impfpart) be.neg.(nonlocut)
`She is not standing.'
(906) palanca shi m1l ki-ata-w
shovel neg have.(impfpart) be.neg-past-locut:subj
`I didn't have a shovel.'
(907) na=na pueblo=ta puz-ka=na, mama shi
1sg.(nom)=top town=in go:out-when=top still neg
a-t ki-a-zi
come-pfpart be.neg-past-nonlocut
`When I left for town, he still hadn't come.'
(908) mama wan shi kwa-t ki-s
still all neg eat-pfpart be.neg-locut
`I still haven't eaten everything.'
(909) na=na kayl kway-ka=na, piya shi
1sg.(nom)=top return drop=when=top corn neg
pak-ta ki-an1-zi
harvest-pfpart be.neg-fut-nonlocut
`When I return, you will not have harvested the corn.'
(910) shi ayna-mtu ki-s
neg cook-impfpart be.neg-locut
`I'm not cooking.'
Just as with other auxiliary constructions, the time frame of these negative
auxiliary sentences is a complicated interaction of the main and auxiliary suxes.
The tense on the auxiliary indicates the reference time for which something is
being claimed, and the non-nite form of the main verb either claims that at
that reference time something is/was not in a state of having happened (Perfect-
ive Participle) or is/was not on-going or planned (Imperfective Participle). For
example, in sentence (907), the tense of the auxiliary (Past) indicates that the
reference time (when I left for town) is in the past, and that at that reference
time, a state existed of him not having come. Likewise in sentence (908), the
tense of the auxiliary (Present) gives the reference time of now, and shows that at
the present time a state exists whereby I have not eaten everything. Perhaps the
12.6. NON-CLAUSAL NEGATION 335

Perfective Participle Imperfective Participle

Past I hadn't done it I wasn't doing it
Present I haven't done it I'm not doing it
Future I won't have done it I won't be doing it
Table 12.2: Translation equivalents of negative mainauxiliary constructions

easiest way to understand these interactions between the tense of the auxiliary
and the form of the main verb is through their translation equivalents, given in
Table 12.2.

12.6 Non-clausal negation

The distinction between clausal and non-clausal negation was examined at the
beginning of section 12.5, where it was noted that non-clausal negation was pro-
totypically associated with negating non-clausal elements, though it is also used
with some non-nite clauses.
Non-clausal negation consists of placing the negative particle shi after the
element to be negated. This has been found with adjectives (including deverbal
adjectives), noun phrases and postpositional phrases:
(911) na=na tizh shi cuchillo kasa=yN kuzhu
1sg.(nom)=top sharp neg knife with=rest pig
nak-ma-ta-w
skin-comp-past-locut:subj
`I skinned the pig with just a blunt (not sharp) knife.'
(912) 11 shi swil kwa-shi-mtu-s
hot neg chili eat-desid-impf-locut
`I want to eat mild (not hot) chili.'
(913) ap shi ka-y
my neg be:permanently-nonlocut
`It is not mine.'
(914) kii-t shi ka-s
get:married-pfpart neg be:permanently-locut
`I am not married.'
(915) taytta shi=kana i-s
father neg=like be-locut
`(I have a child, but I am not involved in raising the child.) I am like a
non-father.'
336 CHAPTER 12. INTERROGATIVES AND NEGATIVES

(916) kwa-t=kana shi=na ku-ma-t1

eat-pfpart=like neg=top eat-comp-term
`[The chicken] ate not like it was eating (ie. as if it wasn't eating).'
Non-clausal negation is also used to negate non-nite main clauses. These
clauses can be either verbless copula constructions, or involve an Imperfective
Participle (see section 3.3.2). The other non-nite form, using the extended
Perfective Participle, has not been found negated. One interesting feature of
negative non-nite main clauses is that while person cannot be expressed in
armative non-nite clauses, it is expressed in negative non-nite main clauses,
with a distinction between the Locutor (sux s) and Non-locutor (unmarked)
sentence-nal negative particle (see section 8.4.1).
(917) na=na 1nkal awa shi-s
1sg.(nom)=top mountain person neg-locut
`I am not an Awa.'
(918) kwizha=na alizh shi
dog=top erce neg.(nonlocut)
`The dog is not erce.'
(919) ap shi
my neg.(nonlocut)
`It is not mine.'
(920) putsha awa tunya kum-tu shi
white person rat eat-impfpart neg.(nonlocut)
`The white people don't eat rats.'
(921) na=na yak ki-mtu shi-s
1sg.(nom)=top be:hungry(1) be:hungry(2)-impfpart neg-locut
`I am not hungry.'
(922) na=na mun pyan shi-s
1sg.(nom)=top name know.(impfpart) neg-locut
`I don't know [his] name.'
(923) kanpa-shi shi-s
accompany-desid neg-locut
`I don't want to accompany [you].'
Mainauxiliary constructions involving stative verbs except the copula (see
section 11.2) are also negated using non-clausal negation: the non-nite main verb
is negated by a following shi, then the auxiliary verb comes nally:
(924) min-tu-s, Demetrio=na shappi k11-mtu shi
think-impf-locut Demetrio=top cane:juice mill-impfpart neg
ka-y
be:permanently-nonlocut
`I think that Demetrio is not going to mill cane.'
12.7. INTERROGATIVES AND NEGATIVES 337

(925) s1 pulish-ta=na azh-tu shi

rewood get:wet-pfpart=top burn-impfpart neg
ka-y
be:permanently-nonlocut
`Wet rewood doesn't burn.'
In addition to non-nite main clauses being negated by the non-clausal
strategy, relative clauses with an associated head are also negated in this way,
showing their parallelism with adjectives.18
(926) pata-m shi pashpa=na yal=ta-s
speak-adjzr neg child=top house=in-from
tazh-ma-t1-zi
fall-comp-past-nonlocut
`The child who doesn't speak (ie. has not yet learnt to speak) fell from
the house.'

12.7 Interrogatives and negatives

As can be seen in this chapter, there are strong yet complex associations between
negation and interrogation strategies in Awa Pit. At a lexical level, some of
the content question words are also negative content words. Non-clausal nega-
tion and inherently negative suxes have no correspondence with interrogatives,
but clausal negation does. Polar interrogatives can be formed with an interrog-
ative sux ma; there is a homophonous negative sux ma, although the two
suxes have somewhat dierent formal properties. Similarly, while having dif-
ferent formal properties, one of the interrogative markers, ki, is identical to the
Present tense form of the negative copula and negative auxiliary ki. While no
further analysis of these similarities will be attempted here, being outside the
scope of this work, an examination of the possible historical paths leading to this
synchronic outcome would be a fascinating study.

18 As explained in section 10.4, when used as a Copula Complement with no head, negated
relative clauses are formed with the inherently negative Negative Adjectivizer.
Chapter 13
Adjuncts and adverbs
13.1 Introduction
This chapter deals with three main areas of the syntax of Awa Pit. First it
discusses the four types of adjuncts in Awa Pit: temporal adjuncts, circumstantial
adjuncts, locational adjuncts, and manner adverbials. These phenomena are all
related, in that they are optional additions to Awa Pit sentences at the level of
a clause or sentence. While they are linked in this way from an external point of
view, internally they have a variety of dierent structures, consisting of adverbial
(subordinate) clauses, postpositional phrases, various dierent types of adverb,
nouns or adjectives. Those constructions which are not dealt with elsewhere
are examined at length here; for other constructions, references are given to the
appropriate section.
Following the discussion of adjuncts, the degree adverbs and comparative
construction are examined. While not structurally comparable externally with
the adjuncts (in that they do not necessarily occur at clause level), these features
of Awa Pit are dealt with here as they can operate at a variety of levels, including
at a level parallel to the manner adverbials.
It is important to stress that the terms `adverb' and `adverbial' are being
used in entirely dierent fashions. The term `X adverb' is used as a term for sev-
eral dierent word classes (for example, time adverb, manner adverb), established
as separate in chapter 4. In contrast, `manner adverbial' refers to a structural
position in the clause, which can be lled by a variety of constituent types.

13.2 Temporal adjuncts

The temporal adjuncts are, unsurprisingly, those elements which can be added to
an Awa Pit clause to indicate something about the time at which the proposition
of the clause holds true. There are a wide variety of temporal adjunct possibilities:
time adverbs, nouns with temporal meaning (with or without the sux ayzhpa
`each'), a quantier plus the noun pyaji `times', a postpositional phrase, and a
range of subordinate clauses.
340 CHAPTER 13. ADJUNCTS AND ADVERBS

As was discussed in section 3.2, the neutral position for most of the tem-
poral adjuncts is directly following the Subject (if it is expressed), although they
are often found in clause-initial position, making Awa Pit clause ordering appear
similar to that of Canela-Krahô, where time elements obligatorily precede the
subject while all other adjuncts appear between the subject and object (Popjes
& Popjes 1986:136137). The After subordinate clause construction and the
simultaneity subordinate clause construction are only found in initial position,
although absolute subordinate clauses may occur in initial position or after the
Subject. As shown in section 14.2.2, the Topic marker na often occurs on tem-
poral adjuncts referring to a specic time.

13.2.1 Time adverbs

Eleven time adverbs have been found in the corpus of Awa Pit sentences (see
Table 4.11 in section 4.7), although there are quite possibly more in the language.
These time adverbs are distinguished from the nouns with temporal reference (see
next section)  time adverbs can only be used as temporal adjuncts, while the
nouns may be used as temporal adjuncts, or as arguments of postpositions, and
occasionally even as arguments of a predicate.
Two of the time adverbs, the interrogative/negative adverbs, will not be
discussed here  see section 12.2.7 for details and examples of these. Examples
of each of the other adverbs are as follows:
(927) amta azh-tu=na
at:night cry-impfpart=top
`[The Vieja (Old Woman)] cries at night.'
(928) an kwizha anya kaa-t kway-t1-zi
this dog rst be:born-sv drop-past-nonlocut
`This dog was born rst (before the others in the litter).'
(929) na=na kayas yal pihshka-m1z-ta-w
1sg.(nom)=top early house sweep-incep-past-locut:subj
`I began sweeping the house early.'
(930) esta=ta=na k1ns1h ku
party=in=top until:dawn dance(1)
ki-mtu-ata-w
dance(2)-impf-past-locut:subj
`We danced until dawn.'
(931) na=na pueblo=ta puz-ka=na, mama shi
1sg.(nom)=top town=in go:out-when=top still neg
a-t ki-a-zi
come-pfpart be.neg-past-nonlocut
`When I left for town, he still hadn't arrived.'
13.2. TEMPORAL ADJUNCTS 341

(932) mansuh alu ki-mtu-at1-zi

all:day rain(1) rain(2)-impf-past-nonlocut
`It was raining all day.'
(933) an nashka a-mtu ka-y
more late come-impfpart be:permanently-nonlocut
`He will come later.'
(934) shil trabaja ki-mtu-s
every:day work(1) work(2)-impf-locut
`We are working every day.'
(935) na=na t1lawayN kuhsa-t
1sg.(nom)=top in:the:early:morning get:up-sv
kway-ta-w
drop-past-locut:subj
`I got up early in the morning.'
It is important to make a few points about some of these adverbs. The time
adverb amta `at night' looks as though it could be analyzed as a postpositional
phrase, with a time noun am= followed by the postposition ta (compare kwizh=ta
`after, later' in section 13.2.4 below). However this is not done as there is not
possible contrast between amta and *ampa as there is between kwizh=ta and
kwizh=pa.
The time adverbs anya `rst, before, earlier' and kayas `early' appear to
have closely related meanings, but the two have dierent uses. Anya is used
to mark that the predication occurred as the rst of a series of events, or be-
fore a subsequent event. A clause containing anya often contrasts with a clause
containing the time noun kwizh= `after, later':
(936) Roberto anya kaa-t kway-zi.
Roberto rst be:born-sv drop-nonlocut
Juana kwizh=ta kaa-zi.
Juana after=in be:born-nonlocut
`Roberto was born before Juana. (Roberto was born rst. Juana was
born after.)'
This contrasts with kayas `early', which simply implies that the event was early
in the day, or relative to the expected time of the event.
(937) Pueblo Viejo=ta=na kayas pyana-ta-w.
Pueblo Viejo=in=top early arrive-past-locut:subj
profesor=na kwizh=pa pyana-zi
teacher=top after=in(approx) arrive-nonlocut
`I arrived early to Pueblo Viejo. The teacher arrived afterwards.'
342 CHAPTER 13. ADJUNCTS AND ADVERBS

In this last example, what is important is that I arrived early, earlier than would
be expected given that it was earlier than the teacher. Had the rst sentence
contained anya rather than kayas, it would have meant that I arrived before the
teacher, but without regard to any absolute time  we both could have arrived
quite late.
Nashka `late' is also quite complex. Formally it appears to consist of the
verb nash- `be afternoon', and the verb sux ka: that is, `when it is afternoon'.
In fact, the phonetic sequence nash-ka can be used in this way:
(938) nash-ka kayl-tu-s
afternoon-when return-impf-locut
`I'll come back in the afternoon.'
However, nashka `late' is distinct from nash-ka `when it is afternoon', both se-
mantically and syntactically, although clearly the latter is the origin of the former.
The semantic dierence is that when nash- is used with any other verbal sux, it
obligatorily refers to afternoon; nashka, on the other hand, means `late', and can
be used for times which are not afternoon, and thus sentence (933) could be used
in the early morning to mean that he will come later on that morning. Syntactic-
ally, nashka can be used with the comparative marker an `more' (as in sentence
(933) above); however the comparative marker can be used with adverbs, but not
verbs (see section 13.7). Thus while nashka `late' is clearly historically derived
from the verb nash- `afternoon', it has developed into a separate, unanalyzable
word, a time adverb.
Three of the other adverbs, k1ns1h, t1lawayN and mama, also appear to be
formally complex. K1ns1h `until dawn', is certainly related to k1n- `to dawn', but
the second segment of the word, s1h, has not been found elsewhere, and cannot
be analyzed separate from the whole. Consequently k1ns1h has been treated as a
single, unanalyzable time adverb.
Similarly t1lawayN `in the early morning' appears related to the noun
t1lawa `tomorrow', possibly with the addition of the Restrictive marker yN. As
a historical origin, this combination appears quite transparent  the combina-
tion t1lawa=yN would mean, approximately, `only just tomorrow', which would,
of course, imply early tomorrow morning. But synchronically t1lawayN has sep-
arated from its origins, both semantically and formally. Semantically, it is not
restricted to early tomorrow morning, but to early on any morning, as in the ex-
ample above, when it was used to refer to this morning. Formally, the Restrictive
marker does not change the word class of the word to which it is attached, but
although t1lawa is a noun, and can be used as the argument of a postposition,
t1lawayN is restricted to appearing as a temporal adjunct.
The time adverb mama `still' could also be formally complex, and once
again the justication for analyzing it as a whole is distributional. It would appear
to consist of the noun ma `now, at this time' and the Temporal marker ma. The
combination of meanings of these two would give, approximately, `already at this
time', which, taken as two separate time frames rather than as a unit, is quite
similar in meaning to `still'. However, as with the case of t1lawayN, mama is only
13.2. TEMPORAL ADJUNCTS 343

used as a temporal adjunct, while ma is a noun with temporal reference, and

may occur followed by a postposition.
There is also one other expression of time which is perhaps best considered
to be on the path between a postpositional phrase and a time adverb. This is
kwizhta, kwizhpa `after, later', and will be discussed in section 13.2.4 below.

13.2.2 Nouns with temporal reference

The temporal adjunct slot may be lled by an unmarked noun with temporal
reference, as well as by a time adverb. The two are distinct classes of words 
while a time adverb can only appear as an unmarked temporal adjunct, a noun
referring to time can appear in this way, but can also be followed by a postpos-
ition, or can be used as an argument of a predicate.1 It must be stressed that
these nouns with temporal reference are distinct from the time nouns, a subclass
of nouns which are only found together with postpositions (see sections 4.3.2
and 13.2.4).
The nouns used unmarked as temporal adjuncts fall into three classes:
relative time, absolute time, and periods of time. The rst two classes establish
on which day or at which time an event occurred, while the nal class discusses
over how many days, weeks, months or years an event occurred.
The rst class contains the following words:
(939) anshik `yesterday'
ma `now, at the moment'
ma payu `today'
t1lawa `tomorrow'
kazt1la `day after tomorrow'
In general terms these nouns require little explanation, except to note that ma
payu `today' is a two-word compound, literally `now day', indicating that the
event occurred some time today, rather than specically now. In fact, ma is a
little more exible than English now, allowing combination with the Terminative
aspect to mean `just (now), recently'. It can also mean `today', even without
being combined with payu.
(940) anshik=na wat k1n-t1-zi.
yesterday=top good be:dawn-past-nonlocut
ma=na=ma wat k1n-t1
today=top=temp good be:dawn-term
`It dawned ne yesterday. And now it's dawned ne today.'
(941) kazt1la kayl-n1-zi
day:after:tomorrow return-fut-nonlocut
`He will come back the day after tomorrow.'
1 In a sentence such as ma=na martes `today [is] Tuesday'.
344 CHAPTER 13. ADJUNCTS AND ADVERBS

(942) Pacho=na ma=na Pialap
=ta tu-y

Pacho=top now=top Pialapí=in be:in:place-nonlocut
`Pacho lives in Pialapí now.'
(943) ma=na alu ki-ma-t1
now=top rain(1) rain(2)-comp-term
`It just started raining.'
The second group of nouns referring to time consists of the days of the
week. These are borrowed straight from Spanish  in fact, for most speakers
they have not truly entered the language, and their pronunciation is on a con-
tinuum between Spanish phonetics and Awa Pit phonetics and phonology. The
days of the week are used unmarked, to indicate either a punctual event, or an
event which occurs on every day of that type.
(944) domingo=na pishkatu put-tu-ata-w, nyampi=kasa
Sunday=top sh sh-impf-past-locut:subj hook=with
`On Sunday I was shing with a hook.'
(945) lunes nap-ma-ta-w
Monday enter-comp-past-locut:subj
`I came in on Monday.'
(946) sábado yal pihshka-tu-s
Saturday house sweep-impf-locut
`I sweep the house on Saturdays.'
Finally, to indicate a period of time, a word expressing a period of time
is used, with a numeral before it indicating the number of those periods of time
the event covered. There is no explicit marking indicating that a period of time
is covered. Only one of the words involved is not borrowed from Spanish 
payu `day'. The Spanish word domingo `Sunday' is used for `week' (as well as
`Sunday'); other words are mes `month' and ano `year'.
(947) na=na kutnya payu ma-ta-w,
1sg.(nom)=top three day stay-past-locut:subj
Pasto=ta=na
Pasto=in=top
`I stayed in Pasto for three days.'
(948) akki=na maza año ma-mtu-s
here=top one year stay-impf-locut
`I'm staying here for a year.'
In addition to the use of unmarked time nouns, nouns which indicate a
period of time can be suxed with ayzhpa `each', to indicate that an event occurs
every time period. The noun can be modied by a number.
13.2. TEMPORAL ADJUNCTS 345

(949) awa=na año-ayzhpa piya kaa-mtu-y

person=top year-every corn sow-impf-nonlocut
`The Awa sow corn every year.'
(950) ampata payu-ayzhpa piya pay-mtu-s
four day-every corn buy-impf-locut
`I buy corn every four days.'
It should be noted that ayzhpa is only used with nouns indicating a period of
time  with days of the week, the unmarked noun can be used to indicate an
activity which occurs on every day of that type (see sentence (946) above).

13.2.3 Number of times

To indicate the number of times which something has occurred, Awa Pit has a
noun pyaji, which is used together with the numerals. This can be used together
with either traditional Awa Pit numerals (for one to four), or the borrowed
Spanish numerals (for ve on).
(951) na=na maza pyaji Cuaiquer Viejo=ta
1sg.(nom)=top one time Cuaiquer Viejo=in
1-ta-w
go-past-locut:subj
`I have been to Cuaiquer Viejo once.'
(952) maza pyaji=N pyan-ta-w
one time=rest hit-past-locut:subj
`I hit [Santos] just once.'
It is unclear whether this word pyaji is native to Awa Pit or borrowed from
Spanish. While it is not the standard Spanish technique for indicating the number
of times something has happened (which uses the noun vez), speakers of Awa Pit
translate pyaji into Spanish using the Spanish word viaje, which in standard
Spanish means `journey, trip'. There are clearly three possibilities here. Pyaji
(which, after a vowel, the most common numeral ending, is usually pronounced
[Bjaxe]) may be a loan from Spanish viaje (pronounced [Bjaxe] after a vowel), with
an extension in meaning from `journey, trip' to `times'; presumably this extension
occurred via sentences where the action was a journey, and the correspondence
between `I've been there ve times' and `I've been there on ve journeys'. Another
possibility is that pyaji is a traditional word, and Spanish viaje began to be used
in local Spanish because of its formal and semantic similarity to the Awa Pit word.
Finally, perhaps, there has been a combination of the preceding two possibilities,
with a phonetically similar word in Awa Pit having altered its pronunciation on
the basis of the similarities with Spanish viaje. While the possibilities are clear,
the origin is undecidable at this stage, with no comparative evidence available.
346 CHAPTER 13. ADJUNCTS AND ADVERBS

13.2.4 Temporal postpositional phrases

The temporal adjunct slot may be lled by a postpositional phrase. Three post-
positions in Awa Pit may be used in a temporal fashion  ta `in', pa `in(approx)'
(or its allomorph wa) and kima `until'. The precise semantics of these three post-
positions will not be examined here, as the meanings of these postpositions is
discussed in sections 5.4.3, 5.4.4 and 5.4.7 respectively, and the reader is referred
there for details. Examples of the three postpositions acting to form temporal
adjuncts are:
(953) mes=ta=na ap su=mal 1-ma-t1-mtu-s
month=in=top my earth=loc go-comp-term-impf-locut
`In a month I will go to my home.'
(954) enero=wa=na na=na Bogot
a=ta
January=in(approx)=top 1sg.(nom)=top Bogotá=in
1-ta-w
go-past-locut:subj
`I went to Bogotá in January.'
(955) noviembre=kima ma-mtu-s mama
November=until stay-impf-locut still
`I am still going to be here until November.'
In addition to postpositional phrases containing nps, it is also possible to
form a temporal adjunct from a postposition used with a clause. For example:2
(956) kwata-na=kima ku-ma-ta-w
vomit-inf=until eat-comp-past-locut:subj
`I ate until I vomited.'
In terms of their use as temporal adjuncts, these constructions function in pre-
cisely the same fashion as any other postpositional phrase (except that, as often
occurs with heavy clausal complements, they are almost always found clause-
initially).
There is one postpositional phrase which needs some additional discussion
 kwizh=ta, kwizh=pa `after, later'. These two words, formally speaking, are
postpositional phrases, consisting of a word followed by one of the postpositions.
Indeed, there is a choice between postpositions, with either ta or pa being ac-
ceptable. However the word kwizh= is only found followed by postpositions, and
is never found on its own. In this it parallels the place nouns aN= `here' and
uN= `there' (see section 13.4). Thus kwizh= has been classed as a special type
of noun, a time noun, obligatorily occurring in a postpositional phrase paral-
lel to the place nouns (see section 4.3.2); however this word-class assignment is
somewhat problematic.
2 For details of the syntax of these clause-plus-postposition constructions, see section 10.2.3.2.
13.3. CIRCUMSTANTIAL ADJUNCTS 347

13.2.5 Temporal adverbial clauses

There are a variety of subordinate adverbial clauses which can function as tem-
poral adjuncts. These express the ideas of `after', with the verb sux tpa/tawa;
simultaneity between two clauses, expressed through the verb sux ka; and a
temporal use of the absolute construction. All of these subordinate clauses devi-
ate in position from other temporal adjuncts, in that they are nearly always in
sentence-initial position, rather than after the Subject, although the absolute can
occur in the normal temporal adjunct position. These clause types are discussed
in sections 10.3.3, 10.3.4 and 10.3.7, and will not be discussed further here.

13.3 Circumstantial adjuncts

Most circumstantial adjuncts occur in the position immediately following tem-
poral adjuncts in the basic clause structure (section 3.2), although they can
be moved to initial or nal position; indeed the concessive and counterfactual
constructions must occur clause-initially. Two construction types are used as cir-
cumstantial adjuncts: postpositional phrases and subordinate adverbial clauses.

13.3.1 Circumstantial postpositional phrases

The circumstantial postpositional phrases give additional information to a pre-
dicate, indicating either an additional person who accompanied the Subject in
their actions; an instrument used to assist in the process of carrying out the
action; or a reason for the main action.
Formally, a comitative/instrumental adjunct is a postpositional phrase,
using the postposition kasa `with'. It is only possible to use kasa in one of its
senses in any adjunct position  thus if a sentence contains an accompanying
actor, it may not also contain an instrument in the adjunct position. It is, how-
ever, possible to ll the adjunct position with two dierent postpositional phrases,
both indicating additional accompanying actors; two instrumental postpositional
phrases do not seem to be possible.
The semantics of the postposition kasa `with' are discussed in section 5.4.9;
here only a few examples will be given.
(957) na=na Santos=kasa Demetrio=wa yal=ta
1sg.(nom)=top Santos=with Demetrio=poss house=in
1-ta-w
go-past-locut:subj
`I went to Demetrio's house with Santos.'
(958) na=na Santos=kasa Demetrio=kasa miimal
1sg.(nom)=top Santos=with Demetrio=with Chucunés
1-ta-w
go-past-locut:subj
`I went to Chucunés with Santos and Demetrio.'
348 CHAPTER 13. ADJUNCTS AND ADVERBS

(959) pyalpiN=kasa puerta azhpizh kway-zi

axe=with door open drop-nonlocut
`He opened the door with an axe.'
The other circumstantial postpositional phrase involves the postposition
akwa `because', and gives the reason for the action in the main clause having
taken place. The complement to the postposition is normally a clause, although
it can be an np. This construction is discussed in section 5.4.10.

13.3.2 Circumstantial adverbial clauses

There are a variety of subordinate adverbial clauses which are used as circum-
stantial adjuncts, and these are discussed in section 10.3. Same-Subject purpos-
ives (section 10.3.1), dierent-Subject purposives (section 10.3.2) and absolute
clauses (section 10.3.7) may appear in the usual circumstantial adjunct position,
though they are often fronted; concessive (section 10.3.5) and counterfactual (sec-
tion 10.3.6) clauses are always in sentence-initial position.

13.4 Locational adjuncts

The third type of adjunct phrase are the locational adjuncts, which normally
occur in a clause after any circumstantial adjuncts. The locational slot can only
be lled by a postpositional phrase, although as mentioned below the status of
akki `here' is unclear. The locational adjunct position may be lled by a variety
of postpositional phrases, using the postpositions ta `in', pa `in(approx)', ki `at',
mal `loc' and kima `until'. A few examples will be given here, but the precise
semantics of each of these locational postpositions is discussed in section 5.4.
As well as following a noun phrase, these postpositions may be used following a
clause, and then the entire postpositional phrase is used as a locational adjunct.
The syntax of postpositional phrase containing a subordinate clause is discussed
in section 10.2.3.2.
(960) Demetrio=na reunión=ta kwinta
Demetrio=top meeting=in speak(1)
kizh-m1z-tu-a-zi
speak(2)-incep-impf-past-nonlocut
`Demetrio began to talk in the meeting.'
(961) paynya pil=pa kal ki-n1-ma-t1
his/her dirt=in(approx) work(1) work(2)-prosp-comp-term
`He/she went to work on his/her land.'
(962) p1na akkwan awa=na a-zi, yal=ki=na
very many person=top be-nonlocut house=at=top
`There were too many people in the house.'
13.5. MANNER ADVERBIALS 349

(963) Santos=pa=mal nya waa-y

Santos=poss=loc meat there:is-nonlocut
`There's meat (for sale) at Santos's place.'
(964) aN=ta-s escuela izh-tu-s
here=in-from school see-impf-locut
`From here I (can) see the school.'
(965) pueblo=ta=kima 1-ma-ta-w
town=in=until go-comp-past-locut:subj
`I went as far as the town.'
Although semantically it is not perhaps a true locational adjunct, the
use of postpositional phrases in ki to indicate the price of something appears to
parallel the use of ki postpositional phrases indicating location.
(966) doce mil=ki pay-ta-w
twelve thousand=at buy-past-locut:subj
`I bought it for twelve thousand [pesos].'
There are a few postpositional phrases which contain a place noun which
can never occur on its own, but is obligatorily followed by a postposition. As
discussed in section 4.3.2, where a list of these words is given, they are being
treated as nouns, as they can be followed directly by two distinct postpositions,
either ta or pa; thus aN=ta `right here' and aN=pa `around here'.
In addition, however, there does appear to be one locational adverb, akki
`here', which can ll the locational adjunct position. This word and its analysis
as either a place noun or a place adverb is discussed at length in section 4.8.

13.5 Manner adverbials

The position (almost) directly before a verb in a sentence in Awa Pit may be lled
by a manner adverbial  only the negative marker shi and the degree adverb p1na
intervene between a manner adverbial and the verb. While manner adverbials
do, in some fashion, state something about the manner in which the verb was
carried out, this semantic idea is somewhat vague, and perhaps a better term
would be simply `adverbial'. The term `manner adverbial' is chosen to maintain
a terminological distinction between, for example, `manner adverbs' (otherwise
`adverbs'), `time adverbs' and `degree adverbs'.
The manner adverbial slot can be lled by words from three distinct word
classes. Perhaps the most basic of words to ll this position are the manner
adverbs. It can also, however, be lled by an adjective (or degree adverb or
comparative marker plus adjective), provided that this is semantically reasonable.
There are also cases where this slot is lled by a noun, although it is unclear in
these cases whether this is a productive use of nouns as manner adverbials, or
whether these are lexically xed compound verbs. The use of nouns as manner
350 CHAPTER 13. ADJUNCTS AND ADVERBS

adverbials will not be discussed here, then, but rather in section 4.5.6, dealing
specically with compound verbs.
While both manner adverbs and adjectives can ll the manner adverbial
position, these two word classes can be distinguished, in the same way in which it
is possible to distinguish time adverbs and nouns with temporal reference. While
adjectives can be used within a noun phrase to modify a noun as well as in the
manner adverbial slot to modify a verb, the manner adverbs can only be used in
manner adverbial position and cannot modify a noun within a noun phrase:
(967) wat ampu
good man
`good man'
(968) ma=na wat k1n-t1
today=top good be:dawn-term
`It dawned ne today.'
(969) na=na aza a-t kway-ta-w
1sg.(nom)=top quickly come-sv drop-past-locut:subj
`I came quickly.'
(970) *aza carro
quickly car
Very few manner adverbs have been found in Awa Pit, with the majority
of words used in this position being adjectives, and having a function within noun
phrases also. The ten words which cannot occur within the noun phrase, and
hence are classed as manner adverbs, are listed in Table 4.12 in section 4.9. In
addition to the example of aza above, and examples of the interrogative manner
adverbs in sections 12.2.6 and 12.2.8, other examples of the use of manner adverbs
are:
(971) na impata=yN a-mtu-ata-w
1sg.(nom) slowly=rest come-impf-past-locut:subj
`I just came slowly.'
(972) na-wa=na m1za kil-ma-t1-s
1sg-acc=top almost dry-comp-past-locut:under
`[The sun] almost dried me.'
A variety of adjectives can be found functioning as manner adverbials, for
example wat `good' in sentence (968) above, t1nta `strong', or 1nkwa `old':
(973) na=na an t1nta kwa-ta-w
1sg.(nom)=top more strong eat-past-locut:subj
`I ate more strongly/much more.'
13.5. MANNER ADVERBIALS 351

(974) ap kwankwa 1nkwa ii-ma-t1-zi

my grandmother old die-comp-past-nonlocut
`My grandmother died [when she was] old.'
Two adjectives, however, are perhaps of special interest: kayN and wan.
As an adjective, kayN means something like `any old, whichever, some':
(975) kayN ashaNpa=na Pueblo Viejo=mal tazh-tu-s
any:old woman=top Pueblo Viejo=loc lower-impf-locut
`Some woman is coming down through Pueblo Viejo (ie. not anyone
special, like the teacher).'
When kayN is a manner adverbial, it gives a meaning of doing something for no
particular reason, just because someone felt like it:
(976) kayN=miN tu ka-s
any:old=rest be:located.(impfpart) be:permanently-locut
`I'm just hanging around.'
Wan `all' occurs as an adjective, and can be used to modify a noun within
a noun phrase; or it may be used in a noun phrase with an ellipsed head, leaving
it as the only constituent within the noun phrase:
(977) wan awa a-t kway-zi
all person come-sv drop-nonlocut
`All the people came.'
(978) wan kii-ma-t
all get:married-comp-pfpart
`Everyone (all my siblings) is married.'
Similar to this use is its use in a noun phrase to mean `everything', in which case
the head is never expressed:
(979) dios wan pyan-i
God all know-nonlocut
`God knows everything.'
In addition to its occurrence in nps, wan can also be used in the manner adverbial
position. In some cases it appears to have the same meaning here, appearing to
have oated from the np as some quantiers do in such diverse languages as
English and the Uto-Aztecan language Pima (Munro 1984):
(980) awa=na wan a-t kway-mtu-y
person=top all come-sv drop-impf-nonlocut
`The people all came.'
352 CHAPTER 13. ADJUNCTS AND ADVERBS

However the distinction in meaning between its use as an adjective within the
noun phrase and as a manner adverb can be seen in sentences where the verbal
arguments are singular, and cannot be modied by wan; it conveys a meaning of
the action being carried out fully and completely:
(981) Lisena=na Florinda=ta=na wan k1hshpizh-ma-t1
Liseña=top Florinda=past=top all scratch-comp-term
`Liseña scratched Florinda all over/to pieces.'
(982) piya=na wan pak-ma-t1-mtu-s
corn=top all sow-comp-term-impf-locut
`I'm going to nish o planting the corn (completely plant the corn).'

13.6 Degree adverbs

There are two degree adverbs or intensiers in Awa Pit: kwisha and p1na. These
degree adverbs appear directly preceding the element which they modify, regard-
less of the word class or structural position of this modied element.
Kwisha is the standard modier for adjectives, and indicates a greater de-
gree of that adjective. While kwisha has, very occasionally, been found modifying
a verb, informants in general will reject this if suggested to them, replacing it
by p1na. Thus kwisha appears to correspond, essentially, to `very', used together
with an adjective; it cannot be used to modify verbs or manner adverbs.
(983) kukum [ kwisha ak ] i
possum [ very shy ] be.(nonlocut)
`Possums are very shy.'
(984) ap anya=na [ kwisha katsa ]
my brother=top [ very big ]
`My brother is very tall.'
(985) [ kwisha akkwan ] awa waa-y
[ very many ] person there:is-nonlocut
`There are very many people [here].'
(986) *na=na [ kwisha kayas ] a-t
1sg.(nom)=top [ very early ] come-sv
kway-ta-w
drop-past-locut:subj
In contrast to kwisha, p1na can be used to modify words from a number of
dierent word classes. It is most commonly found modifying adverbs and verbs,
for which it is the standard intensier:
13.7. COMPARATIVES 353

(987) na=na [ p1na kayas ] a-t

1sg.(nom)=top [ very early ] come-sv
kway-ta-w
drop-past-locut:subj
`I arrived very early.'
(988) [ p1na azhap-pa-mtu-a-s ]
[ very annoy-pl:subj-impf-past-locut:under ]
`They were annoying me a lot.'
P1na can also be used to modify nouns, where it contrasts with the adject-
ive akkwan `many', and to modify adjectives, where it contrasts with the degree
adverb kwisha `very'. In these cases, p1na appears to indicate an excessive de-
gree of something, and was often translated into Spanish by demás `too much'
or bastante `lots of'.
(989) Bogot
a=ta=na [ akkwan awa ] azh-ma-t
Bogotá=in=top [ many person ] grow-comp-pfpart
`Many people have grown up in Bogotá.'
(990) Bogot
a=ta=na [ p1na awa ] azh-ma-t
Bogotá=in=top [ very person ] grow-comp-pfpart
`Loads of people have grown up in Bogotá. (Bogotá has lots of people
in it.)'
(991) up su=na [ kwisha katsa ]
your earth=top [ very big ]
`Your land is very big.'
(992) up su=na [ p1na katsa ]
your earth=top [ very big ]
`Your land is huge!'
(993) [ p1na katsa ] pueblo, Bogot
a=na
[ very big ] town Bogotá=top
`Bogotá is an extremely big city.'

13.7 Comparatives
Awa Pit has a comparative construction, which uses the comparative marker an
`more'. In fact, however, this comparative construction is not commonly used to
compare two objects  the most normal frame for comparing two objects uses
two separate clauses, with dierential adjectives:
(994) Pasto=na aynki pueblo, Bogot
a katsa pueblo
Pasto=top small town Bogotá big town
`Pasto is a small town, Bogotá is a big town. (ie. Pasto is smaller than
Bogotá.)'
354 CHAPTER 13. ADJUNCTS AND ADVERBS

The use of a two clause construction rather than a standard of comparison is

quite common across languages, even in those which have a construction using
a standard of comparison, like Pirahã (Everett 1986:221); and Awa Pit has a
standard of comparison construction as well. The standard of comparison is
marked by kikas, and placed directly before the comparative marker an.
(995) ap anya=na [ Marcos=kikas an katsa ]
my brother=top [ Marcos=than more big ]
`My brother is bigger than Marcos.'
This form kikas appears related to the verb sux kikas marking a concessive
clause (see section 10.3.5), and the relationship between the two is clear; the
meaning of the previous sentence could be just as easily stated as `even though
Marcos is big, my brother is bigger'. Speakers are reluctant to produce sentences
with a standard of comparison, preferring two separate clauses as above, and it
is not clear if this construction is traditional or a recent calque.
The comparative marker an `more' can be used to modify adjectives.
These adjectives may be used adjectivally, either as predicates (see sentence (995)
above), or modifying a noun in a noun phrase, as in sentence (996); or they may
be used as manner adverbials, as in sentence (997). The comparative can also be
used to modify time adverbs, as in sentence (998).
(996) [ an katsa ] ashaNpa na-wa kwinta
[ more big ] woman 1sg-acc talk(1)
kizh-tu-at1-s
talk(2)-impf-past-locut:under
`The bigger woman spoke to me.'
(997) na=na [ an t1nta ] kwa-ta-w
1sg.(nom)=top [ more strong ] eat-past-locut:subj
`I ate more strongly/much more.'
(998) Demetrio=na [ Santos=kikas an kayas ] pyana-zi
Demetrio=top [ Santos=than more early ] arrive-nonlocut
`Demetrio arrived earlier than Santos.'
The comparative marker an can also be used to modify nouns. Very often
the noun itself is ellipsed in this case, with the comparative marker remaining as
the only element of the noun phrase, indicating more of whatever was at issue.
(999) na=na [ an pyal ] m1l
1sg.(nom)=top [ more money ] have.(impfpart)
`I have more money.'
(1000) Marcos=na [ an ] m1l-i
Marcos=top [ more ] have-nonlocut
`Marcos has more [money].'
13.7. COMPARATIVES 355

(1001) [ an ] kwin-zha!
[ more ] give-imp.1obj
`Give me more [food]!'
Equative constructions are formed using the postpositions patsa `like' (see
section 5.4.11) or kana `like' (see section 5.4.12).
Chapter 14
Discourse clitics
14.1 Introduction
Six discourse markers have been found in Awa Pit and are listed in Table 14.1.
These words have been grouped together as they share a variety of formal at-
tributes which show their common nature and their dierence from elements of
other word classes, although it is clear that semantically this is a diverse group.
Formally, the discourse clitics are all monosyllabic elements which en-
cliticize to the preceding word. While each marker has particular distributional
restrictions, all occur on (some subset of) clause-level features: most commonly
on complements or adjuncts, although some occur clause-nally attached to the
predicate syntactically, and in this last case they are semantically associated with
the entire clause rather than just the predicate. These discourse clitics have no
syntactic eect, in the sense that any element which is associated with a discourse
marker maintains its nature  for example, an np followed by the Topic marker
retains its normal internal structure, and can be used in exactly the same way as
an np without the Topic marker.1
As their name implies, the discourse particles are important within dis-
course. Unfortunately, as noted in chapter 1, the majority of data on which this
1Except that, of course, it can only be used as an immediate clausal constituent, and not
as the object of a postposition, for example.

na Topic marker
miN Restrictive marker
kas Additive marker
ma Interrogative marker
ma Temporal marker
ka Emphasis marker
Table 14.1: The discourse clitics
358 CHAPTER 14. DISCOURSE CLITICS

thesis is based consists of elicited sentences. It is perhaps not surprising that

precise usages of several of the discourse particles could not be established. The
use of these particles in a variety of discourse genres is an area of research in
which a great deal of work remains to be done.
In the following sections, there is a general discussion of each of the dis-
course particles. For some of these particles (for example, ma, the Interrogative
marker), a fairly exhaustive indication of its use is made; for others (such as the
Topic marker), only a very general discussion is possible.
Relatively few of the discourse particles can be combined, and those which
can have a strict ordering. The Topic marker na may be followed by the Temporal
marker ma, but by no other discourse particle (see examples (1060)(1062) be-
low). The Restrictive marker miN can be followed by either the Additive marker
kas or the Emphasis marker ka (example (1076)). No other combinations appear
to be possible. This non-combinability is presumably for semantic rather than
formal reasons  for example the Emphasis marker follows an unexpected ele-
ment in a sentence (see section 14.7), and it is not surprising that this cannot be
combined with the Topic marker, as if a referent were to be a topic, the speaker
would hardly assume that the hearer would nd the information unexpected.

14.2 The Topic marker

The Topic marker, na, is perhaps the most complex of the discourse particles. An
attempt at explaining the more obvious uses of this particle will be made here;
however, even more than the other discourse particles, a full examination of na
would rely on extensive analyses of texts of various genres. Some examination of
the use of na in narrative texts has, in fact, already been done, with Henriksen
(1978) and Henriksen & Levinsohn (1977) looking at the use of this particle
attached to conjunctions, dependent clauses and nouns.2
The one form, na, may occur cliticized onto a wide variety of words or
phrases. Most commonly, it occurs on noun phrases (usually on nouns or pro-
nouns), but it may occur on postpositional phrases, adverbs of various types,
predicate adjectives, subordinate clauses, and sometimes even on (non-nite)
main verbs. These elements may be acting as complements or adjuncts, that
is, as core constituents of a clause or as (syntactically) optional additions to a
clause; or of course acting as predicates.
Most commonly, na-marked complements and adjuncts occur in initial
position in the clause in which they occur, or preceded only by other na-marked
elements. However, while this is an obvious tendency, it is clearly not required,
as occasionally an element marked with na may occur after other elements which
are not marked with na. No examples have been found of adjuncts marked
with na being placed after complements which are unmarked with na, however
2The marker discussed in these papers has the form ne rather than na, which may be a
dialect dierence, an orthographic dierence, or a dierence of phonological analysis. In fact,
the Topic marker na is sometimes pronounced [ne] in the data on which this study is based,
with the variation appearing to be speaker dependent (see section 2.4.1.1).
14.2. THE TOPIC MARKER 359

it appears likely that this is a combination of two distinct factors, rather than
an obligatory rule. Those elements marked with na are more topical, and cross-
linguistically (depending somewhat on the denition of topic) topical elements
tend to come in initial position in a clause (Li & Thompson 1976:465). Non-initial
complements, such as Topic-marked Objects following an unmarked Subject, then
have competing constraints on them  rst position because of topicality, or
following the Subject with the usual word order. However adjuncts may generally
occur in two positions in unmarked word order, either after the Subject, or in
sentence-initial position (see section 3.2). An adjunct which is Topic-marked will
then have two possibilities of occurrence  either after the Subject, which is a
natural position for an adjunct, but not for a topic; or in initial position, which
is a natural position both for adjuncts and for topics. Consequently, it is not
surprising that no examples have been found of topical adjuncts occurring after
non-topical Subjects.
The theoretical notion of topic has been a hotly contested one in lin-
guistics. As Tomlin notes,
We still cannot say clearly what a clause level theme or topic is,
despite decades of trying and despite relatively sympathetic toler-
ance among our colleagues for the denitions ultimately employed.
We end up relying on vague denitions whose application in spe-
cic data analyses requires too much dependence on introspection
or indirectly permits the use of structural information in the iden-
tication of instances of the key category.
(Tomlin 1995:519520)
However, in this section my intention is to make some beginning of an account of
the distribution of the morpheme na which occurs in Awa Pit. While this will rely
on previous, often theoretical, discussions of topic, the intention is not to show
how topic is encoded in Awa Pit, but rather to explain where the Topic marker
na occurs. Thus it follows there will be an explicit use of structural information
to establish this distribution, rather than structural information being indirectly
permitted.
In order to discuss the semantics and pragmatics of the Topic marker na,
it is necessary to divide the elements marked with na into four groups. Many of
the denitions or discussions of topic in the literature only consider noun phrases
as topics  for example, Givón (1990:740) states that topicality is a property
of the nominal participants (`referents') . . . of clauses. As noted above, in Awa
Pit the Topic marker may occur on much more than just nominal participants.
However it seems to be useful to consider that there are dierent groups marked
with na: (1) there are those na-marked elements which are complements (that
is, required by the predicate); (2) there are those which are adjuncts (that is,
syntactically optional); (3) there are external topics; and (4) there are topical
clauses. Those elements which are complements are most commonly noun phrases
or postpositional phrases, and can be reference-tracked (cf. Givón 1990:902
910) through the discourse; in contrast, many adjuncts marked with na cannot
360 CHAPTER 14. DISCOURSE CLITICS

be tracked in this way, there being no entity, even notionally, to which they
refer.
In fact, there is also some language-internal evidence of a distinction
between Topic marking on complements and adjuncts, aside from the mean-
ing distinction which will be discussed below. With the exception of pronouns,
which are nearly always marked with the Topic marker, only one complement
in any sentence may be marked with na; in contrast, as many adjuncts (of the
appropriate types) as there are may be na-marked.
In what follows, then, the uses of the Topic marker on complements will
be discussed, then its uses on adjuncts. Following this the two more problem-
atic cases (the Topic marker on external elements, and on predicates) will be
considered in light of the complementadjunct distinction.

14.2.1 The Topic marker on complements

The Topic marker na is found on syntactic complements, that is, on elements
which are (semantically) specied for by the predicate of a clause. These com-
plements are necessarily noun phrases, postpositional phrases or nominalized
subordinate clauses, and the Topic marker is cliticized onto the nal element of
these phrases, regardless of the word class of this nal element.
(1002) wam=na yal kwash=pa p1l-t1-zi
sparrowhawk=top house above=in(approx) y-past-nonlocut
`The sparrowhawk ew above the house.'
(1003) aN=pa kwaka=na awa su paa-ma-t1
here=in(approx) side=top person earth become-comp-term
`Here, [this] side [of the river] has become farms.'
(1004) na=na kwizha=ta=na comida kwin-ta-w
1sg.(nom)=top dog=acc=top food give-past-locut:subj
`I gave food to the dog.'
(1005) nul kwa-n=na kwashmayN
chontaduro eat-inf=top tasty
`Eating chontaduro (a fruit) is really enjoyable/tasty.'
The majority of elicited sentences without ellipsed complements contain
some complement expressed as a noun phrase or postpositional phrase which
is marked with na. However, each clause contains at most one Topic-marked
complement consisting of a full noun, a postpositional phrase containing a full
noun, or a subordinate clause.3
The vast majority of na-marked complements in elicited sentences are
the Subject of their clause  when a sentence is produced out of context, it is
assumed that the Subject is the topic in a language such as Spanish, which does
3 In addition to potentially containing Topic-marked pronouns, and Topic-marked adjuncts.
14.2. THE TOPIC MARKER 361

not mark topics explicitly. However other elements of a sentence can be marked
as topical, if this is reasonable. Discussing chickens, and following a sentence
stating what chickens eat, an informant noted:
(1006) awa atal=na ku-m
person chicken=top eat-adjzr
`Chickens, people eat them.'
As the discussion was about chickens, atal `chicken' is marked as topic, even
though it is the Object in the sentence. Likewise, the following came after another
sentence about how there is a big tree-stump (t1) up behind the house:4
(1007) na=na t1=na kwa-t kyan-ta-w
1sg.(nom)=top tree=top fell-sv throw-past-locut:subj
`I felled [that] tree.'
Here the Object, t1, is marked with na as it is the topic of the current discourse.
It is even more uncommon for (complement) locational phrases to be
marked as topic, but once again this is not a rule, but rather due to pragmatics.
It is possible for a locational complement to be marked as topic:
(1008) ap yal=ta=na kwizha tu-y
my house=in=top dog be:in:place-nonlocut
`In my house, a dog is there.'
Under those circumstances where the location is topic, a locative complement
can be marked with the Topic marker na.
Within connected text, such as narratives or explanations, there appear
to be far fewer examples of Topic-marked complements than in elicited sentences,
although the general lack of such text in the corpus makes any discussion dif-
cult; many sentences in connected text contain no examples of Topic-marked
complements. This is relatively easy to understand, given the facility with which
speakers of Awa Pit ellipse referents which are retrievable from context. Once a
particular entity has been established as the topic, it is, in some senses, foremost
in the mind of a hearer. Consequently, references to that particular entity can be
freely ellipsed, as it is highly retrievable. Equally, an entity which is established
as a topic remains as a topic for at least some stretch of the discourse  as noted
below, topic is a discourse-level phenomenon, not a sentence-level one. Thus after
a participant has been introduced as a topic, the following sentence will probably
contain the same participant as a topic, and if any participant were to be marked
with na it would be this participant. However this participant can be ellipsed,
being easily retrievable from context, and consequently no explicit complement
is marked with the Topic marker. This process occurs in the majority of clauses,
leading to a low use of the Topic marker on complements in connected text.
Given data such as the preceding, especially contextualized examples like
(1006), it seems highly likely that a complement marked with na is stating what
4 The use of two Topic markers in this sentence will be discussed below.
362 CHAPTER 14. DISCOURSE CLITICS

the speaker is talking about. This idea of signalling aboutness is one of the
common ideas of topic, and is often expressed as though occurring at the level of
the sentence: the topic of a sentence is the thing which the proposition expressed
by the sentence is about (Lambrecht 1994:118). As Givón (1990:902) notes,
however, at the level of a single event/state, `topic'  `what is talked about' or
`what is important'  is meaningless. What is in fact meant is that, at the level
of the discourse, a referent has certain discourse properties having to do with the
degree of cognitive and pragmatic accessibility it has in the discourse (Lambrecht
1987:375), and then at the sentence level this referent is taken as topic, with the
remainder of the sentence saying something about that referent. A topic marker
in these terms is some morphological material or syntactic construction which is
used to turn the attention of the hearer to some identiable participant in the
discourse, and then to assert something of that participant (Aissen 1992:50).
A similar, though slightly dierent, formulation is suggested by Doris
Payne (1995), on the basis of the experimental work on comprehension of Gerns-
bacher (1990) and Gernsbacher & Hargreaves (1992). They claim that compre-
hension requires a stage of laying a foundation for a mental structure, then other
information is mapped onto this foundation to develop the structure. Payne
considers that it is not
too far aeld to see Gernsbacher's experimentally-validated found-
ation of a mental representation as roughly what the Functional
Sentence Perspective school and Creider himself were getting at in
talking about the theme or topic of a discourse section. I sug-
gest that whatever a comprehender takes as the foundation for a
mental structure can linguistically be referred to as the thematic
concept or referent of that structure  this is the concept onto
which other information is mapped.
(Payne 1995:451)
This idea of a foundation, while in many ways similar to the aboutness claim, is
important later in contrasting na-marked complements with na-marked adjuncts.
Looking at complements with full nouns in noun phrases and postposi-
tional phrases, these ideas of aboutness or the idea of a foundation appear to
explain this use of the Topic marker in Awa Pit: it shows which (if any) of the
expressed complements of a predicate are the one the discourse is about, the one
onto which other information in the sentence is mapped.
The distribution of the Topic marker on some pronouns is dierent from
that occurring with full nouns. It appears that the third-person pronouns, us
`he/she' and uspa `they', show the same distribution of Topic marker as the
nouns, but the rst- and second-person pronouns are quite dierent: when explicit
rst- or second-person pronouns are used in Awa Pit, they are almost always
accompanied by the Topic marker. Explicit pronouns are not, in fact, often used
in Awa Pit, with a combination of context and person-marking in the predicate
usually being sucient to determine who the participants in any given event
14.2. THE TOPIC MARKER 363

are. If explicit rst- or second-person pronouns are used, this is usually done
to introduce a change in topic, stress the participant in question, or establish a
contrast between one participant and another. In these cases, the participants
indicated through the use of the pronouns are highly topical, and consequently
the pronouns are marked with na. Non-topical pronouns are only used when
there would otherwise be confusion about the participants in an event; and in
these cases, the pronouns are not marked with na. For example, when a speaker
uses a subordinate clause of time involving an action performed by him- or herself
to establish the time of another event, the speaker is not a topical participant
in the major action, but a pronoun must be used to clarify the meaning, as the
subordinate clause verb is not marked for person:
(1009) na a-ka=na, kal
1sg.(nom) come-when=top work(1)
ki-mtu-at1-zi
work(2)-impf-past-nonlocut
`When I came, you were working.'
In this example, the speaker (I) does not participate at all in the main action,
and is simply a minor participant in a subordinate clause. This contrasts with
very similar examples, where, however, the speaker is involved in some way in
the main action, and is marked by na in the subordinate clause:
(1010) na=na a-t kway-ka=na, kuzhu
1sg.(nom)=top come-sv drop-when=top pig
kutil-m1z-t1-zi
skin-incep-past-nonlocut
`When I arrived, they began to skin the pig.'
Here the speaker is involved in the main clause action, in that it was his or her
arrival which triggered that action.
First- and second-person pronouns, then, appear to be special attractors
of the Topic marker na. Indeed, the only circumstances under which two comple-
ments in a clause are marked with the Topic marker is in those cases where one of
them is a pronoun, such as sentence (1004) above. It appears that in a sentence
such as this, the dog is the most topical element, but the pronoun referent retains
some topicality.
This inherent topicality of rst- and second-person pronouns is explainable
in many theories of topic. First- and second-person always have fully accessible
referents (at least in the singular) in the discourse act itself  the speaker and
the hearer. Hence they are high in inherent topicality, if a topic is an entity
whose existence is agreed upon by the speaker and his [or her] audience (Haiman
1978:585), a requirement which all denitions for topic have in some fashion.
It appears that in Awa Pit the inherent topicality of rst- and second-person
pronouns is such that even if another participant in the discourse is more of a
discourse topic, the pronouns often retain enough topicality to be marked with
the Topic marker, as in sentence (1004) above.
364 CHAPTER 14. DISCOURSE CLITICS

14.2.2 The Topic marker on adjuncts

In addition to (optionally) appearing on one of the complements to a verb, the
Topic marker na may occur on one or more adjuncts  nouns, time adverbs,
postpositional phrases or subordinate clauses. Whether the Topic marker occurs
on an adjunct or not does not seem to be a grammatical phenomenon, but rather
semantic (which is unsurprising, given its status as a discourse particle).
Na nearly always occurs on adjunct postpositional phrases using ta and
ki, giving a specic location, and pa, when it is used to give a specic time at
which something happened or during which something happened.
(1011) pueblo=ta=na akkwan tipo pay-nin-a-mtu-y
town=in=top many type buy-caus-pl:subj-impf-nonlocut
`In town they sell many types [of cigarettes].'
(1012) Santos=na ii-ta=ki=na, yal sa-t1-zi
Santos=top die-pfpart=at=top house build-past-nonlocut
`They built a house where Santos died.'
(1013) año=wa=na payl-tu-an1-s
year=in(approx)=top return-impf-fut-locut
`I will come back in a year.'
In contrast to this, the Topic marker never occurs on adjunct postposi-
tional phrases with pa or mal giving a general location, nor on phrases marked
with kima `until', nor on kasa `with', marking accompaniment or an instrument.
The Topic marker usually occurs on causal phrases and conditional phrases
or clauses (including counterfactuals), as well as on clauses indicating time,
whether After clauses, simultaneity clauses or absolute clauses; in contrast, it
never occurs on purposive clauses, whether same-Subject or dierent-Subject.
(1014) ap mamá tayaz-tu=akwa=na,
my mother miss-impfpart=because=top
1-ma-t1-mtu-s
go-comp-term-impf-locut
`I'm going [back] because my mother misses [me].'
(1015) akki pana-t=na, izh-sina=ma
here be:standing.(impfpart)-cntrfc=top see-pot=temp
`If he were here, we could meet.'
(1016) na=na Bogot
a=ta-s kayl-ta=na, kal
1sg.(nom)=top Bogotá=in-from return-pfpart=top work(1)
ki-na a-mtu-s
work(2)-inf come-impf-locut
`After I have returned from Bogotá, I will come and work.'
14.2. THE TOPIC MARKER 365

(1017) nu a-ka=na, na=na

2sg.(nom) come-when=top 1sg.(nom)=top
pihshka-t1-mtu-ata-w
sweep-term-impf-past-locut:subj
`When you came, I was just nishing sweeping.'
In summary, the Topic marker na is almost obligatory for phrases or
clauses which give a specic time or location for an event, a cause or a con-
dition for an event; it cannot occur with phrases or clauses which indicate a
general place or time, accompaniment or an instrument, or the purpose of an
event. In other terms, those phrases or clauses which are used as adjuncts to
establish a specic framework for the matrix predication are marked with the
Topic marker; those which are general, or give additional information which is
not used to establish a framework, cannot be accompanied by the Topic marker.
Here, clearly, the element marked with the Topic marker is not marking
what the sentence is about. In no sense in the above examples are the sentences
or the discourse about the time or location of the event, or the cause or condition
for an event. This use of the Topic marker appears to be quite distinct.
Occurring as it does on non-nominal, non-entity, elements of a sentence,
as well as on more nominal elements, much of the theoretical literature on top-
icality does not apply to this use of the Topic marker  for example, Givón's
reference-tracking cannot be performed.5 However, this use of the Topic marker
corresponds well to Chafe's conception of what a topic is:
What the topics appear to do is to limit the applicability of the main
predication to a certain restricted domain. . . . Typically, it would
seem, the topic sets a spatial, temporal, or individual framework
within which the main predication holds.
(Chafe 1976:50)
In Awa Pit, the Topic marker almost always appears on those adjuncts which are
being used to establish such a spatial, temporal or individual framework; while
those adjuncts (such as purpose clauses or accompaniment phrases) which do not
establish such a framework cannot be marked with the Topic marker.
Chafe's framework also appears to correspond to Doris Payne's concept
of a delimiting phrase, which she establishes in contrast to a foundation (ie.
topic in the sense of what a sentence is about):
A foundation in Gernsbacher and Hargreaves' sense diers from
a delimiting phrase in that a foundation is that concept to which
subsequent information is specically mapped. A delimiting phrase
is more likely to express orienting circumstances relative to which
the foundation and information related to the foundation are to be
interpreted . . . [it] tells us within which universe of discourse the
5 Compare the discussion of this issue in Givón (1990:844846).
366 CHAPTER 14. DISCOURSE CLITICS

subsequent participants and events are to be interpreted.

(Payne 1995:460461)
While Chafe explicitly considers elements similar to adjuncts marked with na in
Awa Pit as topics, Payne specically contrasts them with topics (foundations).
However, while she explicitly excludes them from being topics, the fact that she
feels compelled to do so suggests that she is aware of the potential confusion
between the two  although it is unclear whether this is at a theoretical level
(similar meanings or uses), or at a descriptive level (similar marking in languages).
At a theoretical level it has been suggested that conditional clauses, at
least, are indeed topics. Starting from an assumption that similarities in form
reect similarities in meaning, Haiman (1978) has attempted to show why a
number of unrelated languages use identical morphology to mark topics and con-
ditionals, eventually developing a denition of topic which he feels covers both
the more traditional notion of topic and conditional clauses, arguing that both
are presuppositions of their sentences. Unfortunately it is not clear that all of
the other types of Topic-marked adjuncts in Awa Pit could equally be considered
to be presuppositions.
In fact, Awa Pit is not the only language to mark a variety of adjuncts in
the same way in which topics are marked. Perhaps the most interesting language
to compare here is Mandarin, as it is probably one of the major languages ap-
pealed to in any discussion of topic.6 While many discussions of Mandarin topics,
for example that contained in Li & Thompson (1976), only discuss noun phrases
as topics, there is a clear parallelism in Mandarin between these noun-phrase
topics and other sentence elements. While Chao (1968) unfortunately conates
what would more commonly now be called topics and subjects in Mandarin, he
notes the exact parallelism between certain adverbial clauses and subjects (ie.
topics):
Because (a) such clauses are usually followed by the same pause or
pause particles as after subjects, (b) they occur at the beginning of
a sentence unless they are an afterthought, (c) the so-called subor-
dinate conjunctions can always follow the subject and modify the
verb in the clause, (d) complex sentences shade into compound sen-
tence or simple sentences with complex predicates  for all these
reasons we prefer to treat a dependent clause simply as a clause
subject and the principal clause as the predicate [which can be a
full sentence].
(Chao 1968:113)
Thus in Mandarin (noun phrase) topics and dependent clauses are treated in
the same way, suciently so that Chao considers them to be the same sentence
element. Even more interesting are the various types of subordinate clauses
6 See Thompson & Longacre (1985:229232) for examples of other languages in which some
adverbial clauses may be marked in the same way as topics.
14.2. THE TOPIC MARKER 367

that Chao discusses in this section as being identical to topics: In sum, all the
concessive, causal, conditional, temporal, and spatial clauses are in the last resort
subjects [ie. topics] (Chao 1968:120). These clauses correspond precisely to those
adjunct clauses, adverbs and postpositional phrases which can be marked with
the Topic marker na in Awa Pit.
The similarity between about topics and framework topics will not be
dealt with further here. However it does appears clear that there is some common
core of meaning between the two dierent types: perhaps only at the level of both
being things in a sentence which are not actually encoding the information that
the speaker really wants to convey, with the about topic being what the speaker
wants to convey information about, and the framework topic being background
information. Whether the term topic should be used to cover both uses, or only
one or the other, the morphological marker na is used for both in Awa Pit, and
will be referred to here as the Topic marker, treating it as a unitary morpheme.

14.2.3 External topics

Probably the most well-known phenomenon in languages with a topic construc-
tion is the double-subject construction, a somewhat unfortunate term, demon-
strated by Li & Thompson (1976:468) with the Lahu sentence:
(1018) hO O na-qhO y ve yo
elephant topic nose long prt. declar.
`Elephants (topic), noses are long.'
In sentences such as these, the topic element has no selectional relationship with
the verb (Li & Thompson 1976:468), or in Aissen's (1992) more formal terms,
the `external topic' has no coindexed trace in the basic clause schema. That is,
the referent of the topic element is not a complement of the predicate.
Given that the Awa Pit data here were largely gathered through elicita-
tion from Spanish, it is not surprising that there are few sentences containing an
external topic  in Spanish it is not generally possible for a sentence to con-
tain a non-complement topic, and consequently there is no topic in the Awa Pit
translation, apart from one of the arguments in the Spanish sentence. However a
small number of sentences were found which did contain an external topic, such
as:
(1019) Demetrio=wa caballo=na, m1ta=na p1na akkihsh
Demetrio=poss horse=top tail=top very long
`Demetrio's horse, the tail is so long.'
(1020) Demetrio=wa caballo=na, m1ta=na kwisha akkihsh
Demetrio=poss horse=top tail=top very long
`Demetrio's horse, the tail is very long.'
(1021) Demetrio=na, maza ampu wakata walkwa-t1
Demetrio=top one man cattle steal-term
`Someone stole a cow on Demetrio.'
368 CHAPTER 14. DISCOURSE CLITICS

In the case of the rst two, the original sentence asked was `Demetrio's horse
has a long tail'; the last sentence was asked using the Spanish `dative of aect'
construction.
In these sentences, then, there is an initial, non-complement np, marked
with the Topic marker na. This non-complement element is in some way involved
in the main predication (in all of the above sentences, as an owner of one of the
complements). In all cases this initial element was set o from the remainder
of the sentence by a pause, indicated with a comma above. This setting o of
the na marked element is identical to the use of the Topic marker with adjuncts,
discussed in the previous section, and the meaning developed to cover the usage
of na with adjuncts is, in fact, equally relevant here. This initial Topic-marked
element establishes a framework for the remainder of the predication: We're
discussing something to do with Demetrio's horse; the tail is long. It is important
to note that in the rst two sentences above, there is a second Topic-marked
element, m1ta, the tail, which is the actual about topic, and the Subject of the
predication.
Awa Pit thus seems to have a double-subject construction. Few ex-
amples have been found of it, however this may be more related to the elicitation
technique than the nature of the language. In this construction, there is a main
predication, which may contain an explicit, Topic-marked, Subject, stating what
the predication is about; and there is an initial Topic-marked element, set o by
an intonation indicating that it is an adjunct, and establishes a framework within
which the following predication can be understood.
The presence of an external topic constrution suggests, following Li &
Thompson's (1976) typology, that Awa Pit is at least in part a topic-prominent
language, as all [topic-prominent] languages have sentences of this type, while
no pure [subject-prominent] languages do (Li & Thompson 1976:468). Various
other factors also suggest the topic-prominent nature of Awa Pit (the lack of a
passive and dummy subjects, for example); however nominativeaccusative case-
marking would tend to suggest a subject-prominent language. It would thus
appear that Awa Pit is a mixed topic-prominent and subject-prominent system,
although a great deal of further work would need to be done on the basicness or
otherwise of these constructions with external topics to establish this with any
certainty.

14.2.4 The Topic marker on complete predications

As well as occurring on complements and adjuncts (including external topics),
the Topic marker can also occur syntactically on a variety of main predicates in
Awa Pit, where semantically they have scope over the entire predication. There
are a restricted number of construction types which can take this marking  the
common theme is that they are all non-nite clauses (see section 3.3.2). It is not
surprising to nd the Topic marker on (non-nite) subordinate clauses, of course,
as these are acting as complements or adjuncts. However the Topic marker can
also occur on main non-nite clauses, whether the verb is an extended Perfective
14.2. THE TOPIC MARKER 369

Participle or an Imperfective Participle (including statives):

(1022) Bogot
a=ta=na p1na awa azh-ma-t=na
Bogotá=in=top very people grow-comp-pfpart=top
`In Bogotá many people have grown up.'
(1023) kul=na cava ki-ma-t=na
hole=top dig(1) dig(2)-comp-pfpart=top
`[The armadillo] dug a hole.'
(1024) anshik=na Isabel Demetrio=kasa pueblo=mal
yesterday=top Isabel Demetrio=with town=loc
1-mtu=na
go-impfpart=top
`Yesterday Isabel went to the town with Demetrio.'
(1025) kwizha=na Santos=ta man ki-mtu=na=ma
dog=top Santos=in move(1) move(2)-impfpart=top=temp
`The dog moved to Santos.'
(1026) kwizha uN=ta uz=na
dog there=in sit.(impfpart)=top
`The dog is sitting over there.'
(1027) na=na akkwan libro m1j=na
1sg.(nom)=top many book have.(impfpart)=top
`I have many books.'
In terms of the previous ideas of an about topic and a framework topic,
there are two possible analyses of the Topic-marked predications. Clearly, it is not
possible to consider that the predication is about the predication, but it could
be considered that these sentences are what Lambrecht (1987) calls sentence
focus  rather than the speaker wanting to say something about some entity,
the speaker is interested in the relationship itself. In fact, however, the alternative
analysis of these sentences as a type of framework topic seems more justied.
While a main non-nite sentence cannot be establishing a framework for
itself, sentences do not occur in isolation, but rather in context, and particu-
lar sentences can establish a framework for other sentences. For example, sen-
tence (1022) follows another sentence:
(1028) Bogot
a=kikas an watsal ka-y,
Bogotá=than more beautiful be:permanently-nonlocut
akki=na
here=top
`Here is more beautiful than Bogotá.'
This sentence is, in fact, the main sentence in the interaction, and the following
sentence, (1022), is simply an explanation of why this preceding statement is
true. Likewise, sentence (1023) is preceded by:
370 CHAPTER 14. DISCOURSE CLITICS

(1029) ulam=na pil kul=mal 1-t1-zi

armadillo=top earth hole=loc go-past-nonlocut
`The armadillo went into a hole in the ground.'
That the armadillo vanished is important; sentence (1023) simply explains how
this happened.
Thus the use of the Topic marker on entire predications appears to be a
further example of the framework topic. Unlike earlier cases, however, here the
entire sentence is used to establish a framework for another sentence.
While examples so far have given the Topic marker occurring on a (non-
nite) verb, it is also possible to have non-nite copula clauses (see 3.3.2), which
are verbless, and here the Topic marker occurs on a noun or adjective.
(1030) ap wakata t1l=na
my cattle black=top
`My cattle are black.'
(1031) pala al=na
plantain unripe=top
`The plantains are unripe.'
(1032) Libardo wat awa=na
Libardo good man=top
`Libardo is a good man.'
(1033) ampat1nkwa amta shaa-m=na
Vieja at:night walk-adjzr=top
`The Vieja walks at night (is an at-night-walker).'
It could perhaps be thought that the Topic marker here was (semantically)
on the Copula complement rather than on the predication as a whole, however
this is not the case. When a copula verb is present, the Topic marker can never
occur on a Copula complement:7
(1034) ap wakata t1l(*=na) i
my cattle black(*=top) be.(nonlocut)
`My cattle are black.'
Copula complements cannot be Topic-marked for semantic reasons  in a sense
they are predicates rather than complements, and a sentence cannot be about `is
a person' or `is black', nor can this be used as a framework for the predication,
since it is the predicate.
These Topic-marked non-nite copula clauses are used in the same fashion
as other Topic-marked predications, establishing a framework for other sentences.
7 Equally the Topic marker cannot appear on the copula, as it is always nite.
14.3. THE RESTRICTIVE MARKER 371

Word ends in Restrictive marker

i =N
a, u =yN
other =miN
Table 14.2: Forms of the Restrictive marker

14.3 The Restrictive marker

Turning from the Topic marker, the Restrictive marker is another quite common
discourse particle in Awa Pit. The most common form is miN, but it has three
forms, depending on the nal segment of the word to which it cliticizes, and these
are given in Table 14.2.
The Restrictive marker carries ideas such as `and that's the only thing
involved', or `that's the only quality involved'  it often, but not always, cor-
responds to English just or only; it corresponds very well to the Nariño Span-
ish expression no más, and appears similar to the restrictive or restrictive fo-
cus markers in languages such as Damana (Trillos Amaya 1989:58), Achagua
(Meléndez 1989:53), Barasano (Jones & Jones 1991:176), and the related lan-
guage Guambiano (Vásquez de Ruíz 1988:75).
This discourse particle most commonly attaches to noun phrases, as in
example (1035) and postpositional phrases, examples (1036)(1037). It may also
occur with other elements, especially predicate adjectives, as in example (1038),
and various adjuncts, such as adjectives modifying verbs in example (1039), and
temporal and manner adjuncts in examples (1040) and (1041).
(1035) mitt1=miN izh-ta-w
foot=rest see-past-locut:subj
`I only saw footprints.'
(1036) pueblo=ta mitt1=wa=yN 1-ma-ta-w
town=in foot=in(approx)=rest go-comp-past-locut:subj
`I went to town just by foot (ie. not by horse).'
(1037) na=na tizh shi cuchillo=kasa=yN kuzhu
1sg.(nom)=top sharp neg knife=with=rest pig
nak-ma-ta-w
skin-comp-past-locut:subj
`I skinned the pig with just a blunt (not sharp) knife.'
372 CHAPTER 14. DISCOURSE CLITICS

(1038) nu wat=miN ki-s?

2sg.(nom) good=rest Q-locut
wat=miN i-s.
good=rest be-locut
`Are you well? I am well.'
(1039) wat=miN shaa-nka
good=rest walk-plt:imp
`Walk well (ie. watch where you're going)!'
(1040) ma=miN 1-ma-t1 Carmen
now=rest go-comp-term Carmen
`Carmen has just gone.'
(1041) na impata=yN a-mtu-ata-w
1sg.(nom) slowly=rest come-impf-past-locut:subj
`I was coming slowly.'
The Restrictive marker may also be used on verbal elements, but, just as
with the Topic marker, only on verbal elements which are non-nite. Very few
examples have been found of this use, all involving the non-nite Imperfective
Participle form. Most common is simply a clause in which no person or tense
marking occurs:
(1042) shitshu=na izh-tu=yN
bird=top see-impfpart=rest
`I am just seeing a bird.'
However it is possible to mark tense and person, provided that this is done not
directly on the main verb, but through one of the auxiliary constructions:
(1043) na=na a-mtu=yN ka-s
1sg.(nom)=top come-impfpart=rest be:permanently-locut
`I am only coming (ie. I'm not doing anything else).'

14.4 The Additive marker

The Additive marker, kas, which cannot be combined with the Topic marker na,
most commonly includes another noun phrase into a discourse, in a similar way
to the additive markers in Guambiano (Vásquez de Ruíz 1988:77) and Damana
(Trillos Amaya 1989:59). With the exception of having a dierent noun phrase,
the clause is identical to another earlier clause or earlier idea in the discourse. In
these contexts, it is usually translatable as English too, or sometimes and. For
example, in sentence (1044) the rst clause states that Santos ate an egg, while
the second states that I too ate an egg:
14.4. THE ADDITIVE MARKER 373

(1044) Santos wipu kwa-t kway-zi, na=kas kwa-t

Santos egg eat-sv drop-nonlocut 1sg.(nom)=add eat-sv
kway-ta-w
drop-past-locut:subj
`Santos ate an egg, and I did too.'
While the element containing the dierent noun phrase can be expressed in an
entirely separate clause, as above, if the two actions in the clauses are identical,
the second participant is often added on as an afterthought, following a pause:
(1045) Laureano Ricaurte=mal puz-t1-zi, Demetrio=kas,
Laureano Ricaurte=loc go:out-past-nonlocut Demetrio=add
Clara=kas
Clara=add
`Laureano, Demetrio and Clara went to Ricaurte.'
The new noun phrase can be a Subject, but it can also be in another
grammatical relation. If the noun phrase is within a case-marking postpositional
phrase, the Additive marker kas occurs after the postposition:
(1046) Juan=ta=kas Jos
e pyan-shi-mtu
Juan=acc=add José hit-desid-impfpart
`José wants to hit Juan too (as well as wanting to hit someone else).'
In the above examples, by the time the kas-marked noun phrase was used,
there was already an expressed noun phrase to which it could be notionally added,
either within the same sentence, as in sentence (1044), or earlier in the discourse.
However the Additive marker can also be used when a noun phrase in the current
sentence is merely notionally in a larger group. For example, there are sentences
such as:
(1047) awa=na tunya=kas ku-m
Awa=top rat=add eat-adjzr
`The Awa eat rats (among other things).'
Here the speaker used the Additive marker to mark tunya `rat', to say that this
was one of many things that the Awa eat.8
When used in a negative proposition, the same idea of `one of a group' is
expressed, although in reverse. There are examples such as
(1048) pala=kas shi kwa-t ki, trabaja
plantain=add neg eat-pfpart be.neg.(impfpart) work(1)
ki-mtu
work(2)-impfpart
`Having not eaten even a plantain, I'm working.'
8It was perhaps particularly important for the speaker to point out that this is only one of
many things eaten by the Awa, as the local non-indigenous people nd this one of the more
primitive features of the Awa, and the Awa are aware of this.
374 CHAPTER 14. DISCOURSE CLITICS

Plantains are one of the major bases of the Awa diet. Thus here the speaker is
noting that, of all the things which he could have eaten, he has not even eaten a
plantain.
The Additive marker is also used commonly, though not always, with
the interrogative/negative pronouns (section 12.2) when these are being used as
negative pronouns. That is, words such as m1n `who, no-one' can either be used as
interrogatives or negatives  in the former case, they are usually accompanied
by the Interrogative marker ma (see section 14.5), while as negatives they are
usually accompanied by the Additive marker:
(1049) m1n=kas shi a-ma-y
no:one=add neg come-neg-nonlocut
`No-one came.'
(1050) m1n-a=kas shi izh-ma-s
no:one-acc=add neg see-neg-locut
`I didn't see anyone.'
(1051) shi=kas shi kizh-tu ki-s
nothing=add neg say-impfpart be.neg-locut
`I'm not saying anything.'
(1052) m1n=ta=kas shi 1-ma-s
nowhere=in=add neg go-neg-locut
`I didn't go anywhere.'
The connection between this usage and the earlier function is unclear, perhaps
through some idea such as `of all the people who could have come, no-one came'
(and similar). Its usage may well be frequent in this context as it aids in dier-
entiating the negative usage of these pronouns from their interrogative usage.9

14.5 The Interrogative marker

The Interrogative marker has the form ma. It occurs in content questions, both
direct and indirect  in the former it is optional though common, in the latter
it is obligatory. This discourse particle occurs cliticized onto that immediate
constituent of the clause which contains the content question word, whether this
is a noun phrase, postpositional phrase or adverb. Thus in example (1054) what
is being questioned is the owner of the dog, but the entire postpositional phrase
`to whose dog' is marked with the Interrogative marker, as it is an immediate
constituent of the clause.
9 The combination of an additive marker and an interrogative pronoun to form a negative
pronoun is also found in Imbabura Quechua (Cole 1985:8687); and, apparently, in Guambiano,
based on an example in Vásquez de Ruíz (1988:67).
14.5. THE INTERROGATIVE MARKER 375

(1053) an k1h ku-ka=na, shi=ma ki-n1-zi?

this leaf eat-when=top what=inter do-fut-nonlocut
`If [one] eats this leaf, what will happen?'
(1054) m1n=pa kwizha=ta=ma comida kwin-ta-w?
who=poss dog=acc=inter food give-past-locut:subj
`Whose dog did you give food to?'
(1055) m1n=ta=ma pana-y?
where=in=inter be:standing-nonlocut
`Where is he/she (standing)?'
(1056) na=na shi pyan ki-s,
1sg.(nom)=top neg know.(impfpart) be.neg-locut
mizha=ma ka-t ka
how=inter be:permanently-pfpart nonfut:cpltzr
`I don't know how it was (ie. what it was like).'
(1057) Carmen=na na-wa m1ma-t1-zi, m1n-ta=ma
Carmen=top 1sg-acc ask-past-nonlocut where-in=inter
1-shi-mtu ka
go-desid-impfpart nonfut:cpltzr
`Carmen asked me where I wanted to go.'
While the Interrogative marker is normally only found in content ques-
tions, it is also used in tag questions (section 12.4) on subordinated clauses:
(1058) t1lawa=na miimal 1-mtu=ma
tomorrow=top Chucunés go-impfpart=inter
ka ki-s?
be:permanently.(impfpart) Q-locut
`You're going to Chucunés tomorrow, aren't you?'
The Interrogative marker ma cannot combine with the Topic marker na.
This is unsurprising, as the information being sought by a speaker, contained
within the immediate clausal constituent marked with ma, cannot be the topic
of the clause, as it is not presupposed knowledge.
The Awa Pit Interrogative marker shares features with the Imbabura
Quechua interrogative enclitic taj (Cole 1985:1719). However with the exception
of tag questions such as (1058), the clitic has only been found on the immedi-
ate clausal constituent containing the questioned element; unlike in Imbabura
Quechua's clause-fronting strategy (Cole 1985:19), the clitic has never been
found associated with the last word of a subordinate clause which contains a
questioned element. This also distinguishes this Awa Pit marker from the Napo
Quichua ta, which has similar behaviour to the Imbabura Quechua marker (Orr
& Levinsohn 1992).
For more examples of the Interrogative marker ma, see section 12.2 for
questions, and section 10.2.2 for indirect questions.
376 CHAPTER 14. DISCOURSE CLITICS

14.6 The Temporal marker

The Temporal marker, ma, is formally and semantically similar to the noun
ma (discussed in section 13.2.2). They are, however, distinct. Structurally, the
former cliticizes either to the verb or the rst element in a sentence,10 while the
latter is a free word. Semantically, the noun means, approximately, `now, at that
moment', while the Temporal marker has a more general meaning.
The Temporal marker has two positions of occurrence, as noted above. It
most commonly occurs cliticized after the main verb, whether this is nite or
non-nite:
(1059) na=kas shi a-ma-n1-s=ma
1sg.(nom)=add neg come-neg-fut-locut=temp
`I won't come either.'
It can also occur on the end of the rst constituent of the clause, regardless of
its syntactic status; for example, it may be a noun or pronoun (example (1060)),
a temporal element (example (1061)), or a subordinate clause (example (1062)):
(1060) nu=na=ma pantalón pat-t1-t=ma
2sg.(nom)=top=temp pants wash-term-pfpart=inter
ka ki-s?
be:permanently.(impfpart) Q-locut
`You've just washed your pants, haven't you?'
(1061) siete de julio=wa=na=ma akki tu-an1-s
seven of July=in(approx)=top=temp here be-fut-locut
`I will be here on the 7th of July.'
(1062) kuzhu chambusca ki-ka=na=ma, alu
pig skin(1) skin(2)-when=top=temp rain(1)
ki-ma-t1-zi
rain(2)-comp-term-nonlocut
`While he was skinning the pig, it started raining.'
The Temporal marker can be used with verbs in all tense, aspect and mood
combinations. When used to describe non-past events, ma indicates that the
proposition is true at the time being spoken about (usually the speech moment)
rather than being a universally true statement; or else it implies that there has
recently been a change. For example:
(1063) ayna-mtu=na=ma
cook-impfpart=top=temp
`She is cooking.'
10It should also be noted that the Temporal marker ma appears to combine with the noun
ma `now', to form mama `still'. While this is probably the historical source of mama, here it is
treated simply as a unitary time adverb, and thus is discussed in section 13.2.1.
14.6. THE TEMPORAL MARKER 377

(1064) yal=ta=na, p1na akkwan awa

house=in=top very many person
puta=na=ma
be.pl:nonlocut:subj=top=temp
`There are loads of people in the house.'
(1065) Maximino cruz=ta tu-y=ma
Maximino cross=in be:in:place-nonlocut=temp
`Maximino lives in Las Cruces.'
(1066) na=na min-tu-s=ma, t1lawa
1sg.(nom)=top think-impf-locut=temp tomorrow
a-mtu-y
come-impf-nonlocut
`I think he/she will come tomorrow.'
With the rst two examples here, the presence of ma clearly indicates that these
propositions are predicated about now. Without ma, the rst could be taken
to mean `she cooks', that she is the one who does the cooking. The second,
without ma, would almost certainly be taken to mean that many people lived
in the house, rather than that there were many people inside the house at this
particular point in time. The second pair of examples also state that the propos-
itions are true at precisely this moment, although in these two, because of the
lexical items involved, the suggestion is that they were not true until recently 
they imply respectively that Maximino moved to Las Cruces quite recently, and
that I previously thought he or she was coming back some other time. Equally,
in example (1059) above, ma indicates that prior to this time the speaker was
intending to come, but now that someone else (the hearer) is not coming, he will
not come either. In example (1061), the ma indicates that I will return on or just
before the 7th of July; I will be here then, but this will be a change.
With single activities which occurred in the past, ma indicates that they
were in some recent past, as in example (1060) above. It thus often co-occurs
with the Terminative aspect, but even when combined with other tense and aspect
markings, such as Past tense, or the Past Anterior construction, it indicates that
the event was recent:
(1067) verano=na ap-ma-t1=ma, alu shi ki-t
summer=top enter-comp-term=temp rain(1) neg rain(2)-sv
ta-ma-y
give-neg-nonlocut
`Summer has arrived, it hasn't rained.'
(1068) na=na Pasto=ta-s
1sg.(nom)=top Pasto=in-from
a-ma-ta-w=ma
come-comp-past-locut:subj=temp
`I came recently from Pasto.'
378 CHAPTER 14. DISCOURSE CLITICS

(1069) na=na kuzhu kutil-kam-ma-t=ma

1sg.(nom)=top pig skin-learn-comp-pfpart=temp
`I have learnt how to skin a pig.'
When used on a temporal subordinate clause as in example (1062), ma
indicates the simultaneity of the two events more strongly that would otherwise
be the case. It also often indicates a change in situation, similar to the examples
with non-Past verbs above. Thus in the rst sentence below ma implies a strong
temporal link between the two clauses; and in the second, it also implies that,
while my friends will be working when I return, they are not working now:
(1070) ii-ka=na=ma, kam-tawa
die-when=top=temp bury-oblig
`When someone dies, they have to be buried.'
(1071) na=na uN=pa-s kayl kway-ka=na,
1sg.(nom)=top there=in(approx)-from return drop-when=top
kal ki-mtu-an1-zi=ma
work(1) work(2)-impf-fut-nonlocut=temp
`When I return from there, [my friends] will already be working.'

14.7 The Emphasis marker

The Emphasis marker, ka, is used to mark a contrastive focus, that is, an element
of a sentence which a speaker feels is dierent from what the hearer would be
expecting (cf. Givón 1990, chapter 16).11 To mark an unexpected element, ka
may be cliticized after a noun, pronoun or postpositional phrase, whether these
are complements or adjuncts:
(1072) Santiago=ka pyan-t1-t1-s
Santiago=emph hit-term-past-locut:under
`It was Santiago who hit me.'
(1073) nu=ka pyan-t1-t1-s
nu=emph hit-term-past-locut:under
`It was you who hit me.'
(1074) us=ka azhap-tu
3sg.(nom)=emph annoy-impfpart
`He/she is the one who was annoying [someone].'
(1075) us-a=ka pyan-t1-t1-zi
3sg-acc=emph hit-term-past-nonlocut
`It was him who got hit.'
11 This marker is formallysimilar to the focus marker ka in Inga Quechua (Levinsohn 1976:96).
14.7. THE EMPHASIS MARKER 379

(1076) ma=miN=ka pyan kway-ta-w

now=rest=emph hit drop-past-locut:subj
`It was just now that I hit [someone].'
Thus sentence (1073), for example, was used after the sentence:
(1077) Carmen na-wa pyan-shi-mtu
Carmen 1sg-acc hit-desid-impfpart
`Carmen was wanting to hit me.'
In this case, while Carmen wanted to hit me, it was you who did hit me.
The element which is marked by the Emphasis marker is always found in
initial position in its clause; indeed, when the element in contrastive focus is a
complement, it is often the only complement expressed in the clause.
Cross-linguistically it has been noted that contrastive focus markers are
often related historically to copulas (Givón 1974). Given this, it is interesting
to note the formal identity between the ka pseudo-copula and the Emphasis
marker ka.12 Givón (1990:722) suggests that such similarities in languages stem
from the origin of contrastive focus markers in cleft constructions, using relative
clauses; however in Awa Pit, synchronically at least, the Emphasis marker cannot
be analysed as a cleft construction. Aside from the apparent non-existence of
true relative clause constructions of this type in Awa Pit (see section 10.4), in
a cleft construction the copula is the main verb while the verb expressing the
true predicate is subordinate  but in the Awa Pit construction in question, the
Emphasis marker is non-nite, with the true predicate being a main verb. Thus
while historically the Emphasis marker could, potentially, have developed from
the ka pseudo-copula used in a cleft construction, synchronically it is simply a
discourse particle.
If the Emphasis marker were indeed to have developed from the ka pseudo-
copula, it is also interesting to compare this construction with that involving
ka as an auxiliary verb discussed in section 11.2. As noted in that section,
the distribution of this construction, and the dierences from either the nite
Imperfective or the standard Resultative construction, are unclear. The Emphasis
marker can only be used with complements or adjuncts, never with verbs, and
it is interesting to speculate that constructions with ka as an auxiliary are used
when the speaker wishes to place a verb in contrastive focus. However extensive
discourse studies would have to be carried out to test this hypothesis.
The Emphasis marker ka is thus used as a clitic to complements and
adjuncts which a speaker wishes to place in contrastive focus  that is, when
the speaker believes that the hearer has a particular idea in mind, and wishes to
indicate that this is not true but that rather some other complement or adjunct is
involved in a proposition, the speaker marks this unexpected element by cliticizing
the Emphasis marker to it.

12 And also the complementizer ka; see section 10.2.2.

Bibliography
Abbott, Miriam (1991). Macushi. In Desmond C. Derbyshire & Georey K.
Pullum (eds), Handbook of Amazonian languages, vol. 3. Berlin: Mouton de
Gruyter. pp. 23160.
Abrahamson, Arne (1962). Cayapa: Grammatical notes and texts. In Elson
(1962). pp. 217247.
Adelaar, Willem F. H. (1991). The endangered languages problem: South Amer-
ica. In Robert H. Robins & Eugenius M. Uhlenbeck (eds), Endangered
languages. Oxford: Berg. pp. 4591.
Agha, Asif (1994). The semantics and pragmatics of verb classiers in Urdu-
Hindi. Berkeley Linguistics Society 20:1427.
Aikhenvald, Alexandra Y. (1995). Person marking and discourse in North Arawak
languages. Studia Linguistica 49(2):152195.
Aikhenvald, Alexandra Y. (1996a). The typology of serial constructions and verb
compounding: Evidence from Tariana (North Arawak). Seminar, Depart-
ment of Linguistics, Australian National University.
Aikhenvald, Alexandra Y. (1996b). Words, phrases, pauses and boundaries: Evid-
ence from South American Indian languages. Studies in Language 20(3):487
517.
Aikhenvald, Alexandra Y. (forthcoming). Minor language families and isolates of
Colombia. In R. M. W. Dixon & Alexandra Y. Aikhenvald (eds), Amazonian
languages. Cambridge: Cambridge University Press.
Aikhenvald, Alexandra Y. & Dixon, R. M. W. (forthcoming). Evidentials and
areal typology: A case study from Amazonia. To appear in Language Sci-
ences.
Aissen, Judith L. (1992). Topic and focus in Mayan. Language 68(1):4380.
Anderson, Stephen R. (1985). Inectional morphology. In Shopen (1985), vol. 3:
Grammatical categories and the lexicon. pp. 150201.
Anderson, Stephen R. (1992). A-morphous morphology. Cambridge: Cambridge
University Press.
382 BIBLIOGRAPHY

André, Edouard (1884). América equinoccial. In Carlos Wiener, Doctor Crevaux,

D. Charnay et al. (eds), América pintoresca: Descripción de viajes al nuevo
continente. Barcelona: Montaner y Simon. pp. 477859. (Facsimile edition,
1982, Carvajal, Cali).
Andrews, Avery (1985). The major functions of the noun phrase. In Shopen
(1985), vol. 1: Clause structure. pp. 62154.
Andrews, Avery D. (1996). Semantic case-stacking and inside-out unication.
Australian Journal of Linguistics 16:155.
Annamalai, E. (1978). The anglicised Indian languages: A case of code mixing.
International Journal of Dravidian Linguistics 7(2):239247.
Aoki, Haruo (1986). Evidentials in Japanese. In Chafe & Nichols (1986). pp. 223
238.
Aragón, Luis Eduardo (1974). Cultural integration of the Kwayker Indians,
Colombia: A geographical analysis. PhD thesis, Michigan State University.
Armagost, James L. (1986). Three exceptions to vowel devoicing in Comanche.
Anthropological Linguistics 28(3):255265.
Armagost, James L. (1988). Some places where Comanche's vowel devoicing rule
doesn't work. In William Shipley (ed), In honor of Mary Haas: From the
Haas Festival Conference on Native American Linguistics. Berlin: Mouton
de Gruyter. pp. 113.
Austin, Peter (1981). Switch-reference in Australia. Language 57(2):309334.
Baker, Mark C. (1988). Incorporation: A theory of grammatical function chan-
ging. Chicago: University of Chicago Press.
Barrett, S. A. (1925). The Cayapa Indians of Ecuador, Indian Notes and Mono-
graphs 40. New York: Museum of the American Indian, Heye Foundation.
Bavin, Edith & Shopen, Tim (1985). Warlpiri and English: Languages in contact.
In Michael Clyne (ed), Australia, meeting place of languages. Canberra:
Pacic Linguistics. pp. 8194.
Beuchat, Henri & Rivet, Paul (1910). Anités des langues du sud de la Colombie
et du nord de l'équateur. Le Mouséon 11:3368, 141198. (Reprinted as
monograph, J.-B. Istas, Louvain).
Blake, Barry J. (1994). Case. Cambridge: Cambridge University Press.
Borgman, Donald M. (1990). Sanuma. In Derbyshire & Pullum (1990), vol. 2.
pp. 15248.
Borman, M. B. (1962). Cofan phonemes. In Elson (1962). pp. 4559.
BIBLIOGRAPHY 383

Branks, Thomas & Branks, Judith (1973). Fonología del guambiano. Sistemas
fonológicos de idiomas colombianos 2:3956. (Published by the Ministerio
del Gobierno, Bogotá).
Brinton, Daniel G. (1891). The American race: A linguistic classication and eth-
nographic description of the Native tribes of North and South America. New
York: N.D.C. Hodges. (Reprinted in 1970 by Johnson Reprint Corporation,
New York).
Bybee, Joan L. & Dahl, Östen (1989). The creation of tense and aspect systems
in the languages of the world. Studies in Language 13(1):51103.
Bybee, Joan L. & Pagliuca, William (1985). Cross-linguistic comparison and
the development of grammatical meaning. In Jacek Fisiak (ed), Historical
semantics and historical word formation. Berlin: de Gruyter. pp. 5983.
Bybee, Joan L., Pagliuca, William & Perkins, Revere D. (1991). Back to the
future. In Traugott & Heine (1991), vol. 2: Focus on types of grammatical
markers. pp. 1758.
Bybee, Joan L., Perkins, Revere D. & Pagliuca, William (1994). The evolution
of grammar: Tense, aspect, and modality in the languages of the world.
Chicago: University of Chicago Press.
Caldas, Antonio José (1946). Palabras del idioma kuaiquer recogidas por A. J.
Caldas en Ricaurte. Revista de Historia (Pasto) 2(7/8):136137.
Calderón Rivera, Álvaro, Trillos Amaya, María, Reina, Leonardo & R. de Montes,
María Luisa (1987). Informe acerca de las labores desarrolladas en el Labor-
atorio de Fonética Experimental del Instituto Caro y Cuervo sobre el Awapit.
Appendix 9 of Centro Experimental Piloto Nariño (1990).
Calvache Dueñas, Rocío (1987a). Patrones de comportamiento de un grupo in-
dígena (Kwaiker) en Nariño en relación al desarrollo y evolución de sus
niños de 0 a 7 años: Estudio psicoantropológico. Master's thesis, Facultad
de Psicología, Universidad Santo Tomás de Aquino, Bogotá.
Calvache Dueñas, Rocío (1987b). Primer seminario taller para la unicación de
alfabetos awá. Glotta (Bogotá) 2(2):4346.
Calvache Dueñas, Rocío (1988). Awa Pit: Informe de fonología. Fonologías, vol.
2: Lenguas del Sur Andino. Bogotá: Universidad de los Andes. pp. 206251.
(Unpublished collected working papers, Centro Colombiano de Estudios en
Lenguas Aborígines, Universidad de los Andes, Bogotá).
Calvache Dueñas, Rocío (1989). Fonología e introducción a la morfosintaxis del
Awa Pit. Master's thesis, Departamento de Antropología, Universidad de
los Andes, Bogotá.
384 BIBLIOGRAPHY

Calvache Dueñas, Rocío (1991). La etnoeducación: Una alternativa para la situa-

ción actual de educación en la zona indígena awa de Nariño, Municipio de Ri-
caurte. Ministerio de Educación Nacional, Ocina de Etnoeducación, Pasto.
(Unpublished).
Calvache Dueñas, Rocío & Cerón Solarte, Benhur (1988). Condiciones en que se
desarrolla la lengua indígena awa pit. Instituto Colombiano de Antropología,
Pasto. (Unpublished).
Campbell, Lyle (1985). The Pipil language of El Salvador, Mouton Grammar
Library 1. Berlin: Mouton.
Campbell, Lyle (1987). Syntactic change in Pipil. International Journal of Amer-
ican Linguistics 53(3):253280.
Campbell, Lyle & Muntzel, Martha C. (1989). The structural consequences of
language death. In Dorian (1989). pp. 181196.
Canonge, Elliott D. (1957). Voiceless vowels in Comanche. International Journal
of American Linguistics 23(2):6367.
Carrasco Andrade, Eulalia, Contreras Ponce, Enrique, Espinoza V., María Au-
gustina & Moncayo Román, Ramiro (1984). Diagnóstico socio-económico de
la etnia Awa-Kuaiker. Ministerio de Relaciones Exteriores, Departamento
de Desarrollo Fronterizo, Quito. (Unpublished).
Centro Experimental Piloto Nariño (1990). Informe de la reunión sobre el
proyecto de etnoeducación y lingüística para la zona awa de Nariño. Centro
Experimental Piloto Nariño, Ministerio de Educación Nacional, Pasto. (Un-
published).
Cerón Solarte, Benhur (1986). Los Awa-Kwaiker: Un grupo indígena de la
selva pluvial del Pacíco Nariñense y el Nor-Occidente Ecuatoriano. Quito:
ABYA-YALA.
Cerón Solarte, Benhur (1987). Denominación étnica y lengua indígena Kwaiker.
Paper presented to Seminar on Denominaciones étnicas y lenguas indígenas
de Colombia, Instituto Caro y Cuervo, Bogotá, 1012 March 1987.
Cerón Solarte, Benhur & Calvache Dueñas, Rocío (1990). Condiciones sociales en
que se desarrolla la lengua indígena de los Awa-Cuaiquer (Awa Pit). Pueblos
Indígenas y Educación (ABYA-YALA, Ecuador) 4(13):752.
Chafe, Wallace L. (1976). Givenness, contrastiveness, deniteness, subjects, top-
ics, and point of view. In Li (1976). pp. 2555.
Chafe, Wallace & Nichols, Johanna (eds) (1986). Evidentiality: The linguistic
coding of epistemology, Advances in Discourse Processes 20. Norwood, NJ:
Ablex.
BIBLIOGRAPHY 385

Chao, Yuen Ren (1968). A grammar of spoken Chinese. Berkeley: University of

California Press.
Cho, Young-Mee Yu (1993). The phonology and phonetics of `voiceless' vowels.
Berkeley Linguistics Society 19:6475.
Chung, Sandra & Timberlake, Alan (1985). Tense, aspect, and mood. In Shopen
(1985), vol. 3: Grammatical categories and the lexicon. pp. 202258.
Colcultura [Ann Osborn] (ed.) (1991). Estudios sobre los indígenas Kwaiker de
Nariño. Bogotá: ICBF and Colcultura.
Cole, Peter (1985). Imbabura Quechua, Croom Helm Descriptive Grammars.
London: Croom Helm.
Comrie, Bernard (1975). Causatives and universal grammar. Transactions of the
Philological Society 1974:132.
Comrie, Bernard (1981). Language universals and linguistic typology. Oxford:
Basil Blackwell.
Comrie, Bernard & Thompson, Sandra A. (1985). Lexical nominalization. In
Shopen (1985), vol. 3: Grammatical categories and the lexicon. pp. 349
398.
Constenla Umaña, Adolfo (1981). Comparative Chibchan phonology. PhD thesis,
University of Pennsylvania.
Constenla Umaña, Adolfo (1985). Clasicación lexicoestadística de las lenguas
de la familia Chibcha. Estudios de Lingüística Chibcha 4:155189.
Constenla Umaña, Adolfo (1991). Las lenguas del área intermedia: Introducción
a su estudio areal. San José: Universidad de Costa Rica.
Contreras Ponce, Enrique (1984). Propuesta de alfabeto para la lengua Kuaiker.
In Carrasco Andrade, Contreras Ponce, Espinoza V. & Moncayo Román
(1984).
Costales, Piedad & Costales, Alfredo (1983). Amazonia: Ecuador, Perú y Bolivia.
Quito: Mundo Shuar.
Craig, Colette G. (forthcoming). Classiers. To appear in C. Lehmann (ed),
Morphology: A handbook on inection and word formation, Article 97.
Crowley, Terry (1987). Serial verbs in Paamese. Studies in Language 11(1):3584.
Crystal, David (1985). A dictionary of linguistics and phonetics. 2nd edn. Oxford:
Basil Blackwell.
Cubillos, Julio César (1955). Tumaco: Notas arqueológicas. Bogotá: Minerva.
386 BIBLIOGRAPHY

Curnow, Timothy Jowan & Liddicoat, Anthony J. (forthcoming). The Barbacoan

languages of Colombia and Ecuador. Submitted.
Dahlstrom, Amy (1987). Discontinuous constituents in Fox. In Paul D. Kroe-
ber & Robert E. Moore (eds), Native American languages and grammatical
typology: Papers from a conference at the University of Chicago, April 22,
1987. Bloomington: Indiana University Linguistics Club. pp. 5373.
Davies, W. (1981). Choctaw clause structure. PhD thesis, University of California,
San Diego.
Davis, John F. (1973). A partial grammar of simplex and complex sentences in
Luiseño. PhD thesis, University of California, Los Angeles.
Dayley, Jon P. (1989). Tümpisa (Panamint) Shoshone grammar, University of
California Publications in Linguistics 115. Berkeley: University of California
Press.
DeLancey, Scott (1986). Evidentiality and volitionality in Tibetan. In Chafe &
Nichols (1986). pp. 203213.
DeLancey, Scott (1992). The historical status of the conjunct/disjunct pattern
in Tibeto-Burman. Acta Linguistica Hafniensia 25:3962.
DeLancey, Scott (1997). Mirativity: The grammatical marking of unexpected
information. Linguistic Typology 1(1):3352.
Dench, Alan & Evans, Nicholas (1988). Multiple case-marking in Australian
languages. Australian Journal of Linguistics 8:148.
Derbyshire, Desmond C. & Pullum, Georey K. (eds) (1986a). Handbook of
Amazonian languages, vol. 1. Berlin: Mouton de Gruyter.
Derbyshire, Desmond C. & Pullum, Georey K. (1986b). Introduction. In Derby-
shire & Pullum (1986a), vol. 1. pp. 128.
Derbyshire, Desmond C. & Pullum, Georey K. (eds) (1990). Handbook of
Amazonian languages, vol. 2. Berlin: Mouton de Gruyter.
Dirección Nacional de Educación Indígena Intercultural Bilingüe (1989). El alfa-
beto awá unicado (fase experimental). Quito: MEC, CONAIE & UTEPA.
Dixon, R. M. W. (1977). A grammar of Yidiñ. Cambridge: Cambridge University
Press.
Dixon, R. M. W. (1979). Ergativity. Language 55:59138.
Dixon, R. M. W. (1980). The languages of Australia. Cambridge: Cambridge
University Press.
BIBLIOGRAPHY 387

Dixon, R. M. W. (1988). `Words' in Fijian. In T. L. Burton & Jill Burton (eds),

Lexicographical and linguistic studies: Essays in honour of G. W. Turner.
Wolfeboro, NH: D. S. Brewer. pp. 6571.
Dixon, R. M. W. (1991). A new approach to English grammar, on semantic
principles. Oxford: Clarendon Press.
Dixon, R. M. W. (1994). Ergativity. Cambridge: Cambridge University Press.
Dorian, Nancy C. (ed.) (1989). Investigating obsolescence: Studies in language
contraction and death. Cambridge: Cambridge University Press.
Downing, Pamela & Noonan, Michael (eds) (1995). Word order in discourse,
Typological Studies in Language 30. Amsterdam: John Benjamins.
Duque Gómez, Luis (1955). Prológo. In Cubillos (1955). pp. 59.
Durie, Mark (forthcoming). Grammatical structures in verb serialization: Some
preliminary proposals. In A. Alsina, J. Bresnan & P. Sells (eds), Complex
predicates. Stanford: CSLI Press.
Ehrenreich, Jerey D. (1989). Contacto y conicto: El impacto de la aculturación
entre los Coaiquer del Ecuador. Quito: ABYA-YALA & Instituto Otavaleño
de Antropología. (Translation of his 1985 thesis).
Elson, Benjamin (ed.) (1962). Studies in Ecuadorian Indian languages: I, SIL
Publications 7. Norman, Oklahoma: Summer Institute of Linguistics.
Evans, Nicholas (1996). Grammaticizing the knower: Towards a partial typology
of person eects on predicates. Plenary session, Third Australian Linguistics
Institute, Australian National University, 2 July 1996.
Everett, Daniel (1986). Pirahã. In Derbyshire & Pullum (1986a), vol. 1. pp. 200
325.
Fernández, Ana María (198990). Denominación general de las etnias indígenas
de Colombia. Revista Colombiana de Antropología (Bogotá) 27:284287.
First Binational Meeting (1987). Problems, conclusions and recommendations of
the rst ColombianEcuadorian binational meeting on the unication of the
Awa Pit alphabet, 1721 May 1987. Appendix 7 of Centro Experimental
Piloto Nariño (1990).
Foley, William A. & Olsen, Mike (1985). Clausehood and verb serialization.
In Johanna Nichols & Anthony C. Woodbury (eds), Grammar inside and
outside the clause: Some approaches to theory from the eld. Cambridge:
Cambridge University Press. pp. 1760.
Frank, Paul (1990). Ika syntax, Studies in the Languages of Colombia 1. Ar-
lington, Texas: Summer Institute of Linguistics and University of Texas at
Arlington.
388 BIBLIOGRAPHY

García Ortiz, J. (1949). Breve reseña de la raza Coaiquer. Revista de Historia

(Pasto) 4(2325):278284.
Gazdar, Gerald, Pullum, Georey K. & Sag, Ivan A. (1982). Auxiliaries and
related phenomena in a restrictive theory of grammar. Language 58(3):591
638.
Gernsbacher, Morton (1990). Language comprehension as structure building. Hill-
sdale, NJ: Erlbaum.
Gernsbacher, Morton & Hargreaves, David (1992). The privilege of primacy:
Experimental data and cognitive explanations. In Doris L. Payne (ed), The
pragmatics of word order exibility. Amsterdam: John Benjamins. pp. 83
116.
Gibson, J. (1980). Clause union in Chamorro and in universal grammar. PhD
thesis, University of California, San Diego.
Gildea, Spike (1994). Semantic and pragmatic inverse: `Inverse alignment' and
`inverse voice' in Carib of Surinam. In T. Givón (ed), Voice and inversion,
Typological Studies in Language 28. Amsterdam: John Benjamins. pp. 187
230.
Givón, T. (1974). Syntactic change in Lake-Bantu: A rejoinder. Studies in
African Linguistics 5(1).
Givón, T. (1980). The binding hierarchy and the typology of complements. Stud-
ies in Language 4(3):333377.
Givón, T. (1982). Logic vs. pragmatics, with human language as the referee:
Toward an empirically viable epistemology. Journal of Pragmatics 6:81
133.
Givón, T. (1990). Syntax: A functional-typological introduction, vol. 2. Amster-
dam: John Benjamins.
Goddard, Ives (1996). First- and second-person participles in Fox (Mesquakie).
Paper given at the 35th Conference on American Indian Languages (The
Society for the Study of the Indigenous Languages of the Americas), at
the 95th Annual Meeting of the American Anthropological Association, San
Francisco, 2123 November 1996.
Gordon, David & Lako, George (1971). Conversational postulates. Chicago
Linguistic Society 7:6384.
Greenberg, Joseph H. (ed.) (1978). Universals of human language. Stanford:
Stanford University Press.
Greenberg, Joseph H. (1987). Language in the Americas. Stanford: Stanford
University Press.
BIBLIOGRAPHY 389

Grimes, Barbara F. (ed.) (1996). Ethnologue: Languages of the world. 13th edn.
Dallas, Texas: Summer Institute of Linguistics. (Electronic edition, found
at: http://www.sil.org/ethnologue).
Grimes, Joseph E. (1985). Topic inection in Mapudungun verbs. International
Journal of American Linguistics 51(2):141163.
Gutiérrez, Runo (1920). Monografías, vol. 1. Bogotá: Biblioteca de Historia
Nacional.
Haegeman, Liliane (1995). The syntax of negation. Cambridge: Cambridge
University Press.
Haiman, John (1978). Conditionals are topics. Language 54(3):564589.
Hale, Austin (1980). Person markers: Finite conjunct and disjunct verb forms
in Newari. In Ronald Trail (ed), Papers in South-East Asian linguistics,
number 7, Pacic Linguistics A-53. Canberra: Pacic Linguistics. pp. 95
106.
Hale, Kenneth (1976). The adjoined relative clause in Australian languages.
In R. M. W. Dixon (ed), Grammatical categories in Australian languages.
Canberra: Australian Institute of Aboriginal Studies. pp. 78105.
Hargreaves, David (1990). Indexical functions and grammatical sub-systems in
Kathmandu Newari. Chicago Linguistic Society 26:179193.
Hargreaves, David (1991). The conceptual structure of intentional action: Data
from Kathmandu Newari. Berkeley Linguistics Society 17:379389.
Harkins, Jean (1996). Desire in language and thought: A study in crosscultural
semantics. PhD thesis, Department of Linguistics, Australian National Uni-
versity.
Harms, Phillip L. (1985). Epena Pedee (Saija): Nasalization. In Ruth M. Brend
(ed), From phonology to discourse, Language Data, Amerindian Series 9.
Dallas, Texas: Summer Institute of Linguistics. pp. 1318.
Harms, Phillip Lee (1994). Epena Pedee syntax, Studies in the Languages of
Colombia 4. Arlington, Texas: Summer Institute of Linguistics and Univer-
sity of Texas at Arlington.
Haspelmath, Martin (1989). From purposive to innitive: A universal path of
grammaticization. Folia Linguistica Historica 10(1/2):287310.
Haug, Eugen (1994). Los nietos del trueno: Construcción social del espacio,
parentesco y poder entre los Inkal-Awá. Quito: ABYA-YALA.
Henriksen, Lee A. (1978). Algunas observaciones sobre un texto cuaiquer. Cultura
Nariñense 109:3139. (Reprinted 1979 in Artículos en Lingüística y Campos
Anes (SIL Colombia) 6:917).
390 BIBLIOGRAPHY

Henriksen, Lee A. (1985). Algunas observaciones sobre la gramática cuaiquer. In

Cerón Solarte (1986). pp. 285289.
Henriksen, Lee A. & Henriksen, Lynne A. (1978). APK, Kalpa, Zhitzhu, Pak zha,
Pishkatu. Lomalinda: Townsend. (A series of readers in Awa Pit).
Henriksen, Lee A. & Henriksen, Lynne A. (1979). Fonología del cuaiquer. Sis-
temas Fonológicos de Idiomas Colombianos 4:4962. (Published by the Min-
isterio del Gobierno, Bogotá).
Henriksen, Lee A. & Henriksen, Lynne A. (1986). Mane p1nk1h kamshimtus.
Pasto: Laferoy. (A series of primers in Awa Pit).
Henriksen, Lee A. & Levinsohn, Stephen H. (1977). Progression and promin-
ence in Cuaiquer discourse. In Robert E. Longacre & Frances Woods (eds),
Discourse grammar: Studies in indigenous languages of Colombia, Panama
and Ecuador, vol. 2. Dallas: Summer Institute of Linguistics. pp. 4367.
(Reprinted with minor changes in 1978 as: Progresión y prominencia en el
discurso cuaiquer. Estudios Chibchas 2:50-80).
Henriksen, Lee A. & Obando Ordóñez, Pedro V. (1985). Mane p1nk1h kam-
shimtus: Una gramática pedagógica del Cuaiquer con ejercicios prácticos,
vol. 1. Pasto: Universidad de Nariño.
Hidalgo, Tomás (1913 [1894]). La historia del Ecuador por el Dr. González Suárez.
La Revista Ecuatoriana 6:295299.
Hill, Jane H. (1989). The social functions of relativization in obsolescent and
non-obsolescent languages. In Dorian (1989). pp. 149164.
Hinds, John (1986). Japanese, Croom Helm Descriptive Grammars. London:
Routledge.
Hook, Peter Edwin (1974). The compound verb in Hindi, The Michigan Series in
South and Southeast Asian Languages and Linguistics 1. Michigan: Center
for South and Southeast Asian Studies, University of Michigan.
Hook, Peter Edwin (1991). The emergence of perfective aspect in Indo-Aryan lan-
guages. In Traugott & Heine (1991), vol. 2: Focus on types of grammatical
markers. pp. 5989.
Hopper, Paul J. & Thompson, Sandra A. (1980). Transitivity in grammar and
discourse. Language 56(2):251299.
Hopper, Paul J. & Thompson, Sandra A. (1984). The discourse basis for lexical
categories in universal grammar. Language 60(4):703752.
Huber, Randall Q. & Reed, Robert B. (1992). Vocabulario comparativo: Palabras
selectas de lenguas indígenas de Colombia. Bogotá: Instituto Lingüístico de
Verano.
BIBLIOGRAPHY 391

Jake, Janice L. & Chuquin, Carmen (1979). Validation suxes in Imbabura

Quechua. Chicago Linguistic Society 15:172184.
Jaramillo, María Fernanda & Acosta Pinzón, Carlos (1990). Uso de recursos
naturales en la comunidad awa de Pialapí, Nariño: Informe nal de invest-
igación. Report for the Fundación para la Educación Superior and the World
Wildlife Fund, Cali. (Unpublished).
Jijón y Caamaño, Jacinto (1941). El Ecuador interandino y occidental, antes de
la conquista castellana. vol. 1. Quito: Ecuatoriana. Chapter: La Lengua
Coayquer, pp. 152225.
Jijón y Caamaño, Jacinto (1943). El Ecuador interandino y occidental, antes de
la Conquista castellana, vol. 3: Las lenguas del sur de Centroamérica y el
norte y centro del oeste de Sud-América. Quito: Ecuatoriana.
Jones, Wendell & Jones, Paula (1991). Barasano syntax, Studies in the Lan-
guages of Colombia 2. Arlington, Texas: Summer Institute of Linguistics
and University of Texas at Arlington.
Kakumasu, James (1986). Urubu-Kaapor. In Derbyshire & Pullum (1986a),
vol. 1. pp. 326403.
Karsten, Rafael (1988 [1925]). Los indios Colorados del oeste ecuatoriano. In
José E. Juncosa (ed), Tsachila: Los clásicos de la etnografía sobre los Col-
orados (19051950). Quito: ABYA-YALA. pp. 5578. (Originally published
as `The Colorado Indians of western Ecuador', in Ymer 1924, the Journal of
the Swedish Society for Anthropology and Geography, Stockholm, 1925).
Karttunen, Frances (1976). Uto-Aztecan and Spanish-type dependent clauses
in Nahuatl. Chicago Linguistic Society 12 (Parassession, Diachronic
syntax):150158.
Keenan, Edward L. (1985). Relative clauses. In Shopen (1985), vol. 2: Complex
constructions. pp. 141170.
Keenan, Edward L. & Comrie, Bernard (1977). Noun phrase accessibility and
universal grammar. Linguistic Inquiry 8(1):6399.
Kempf, Judith (1982). The dynamics of culture and health: Disease curing among
the Ecuadorian Coaiquer indians under the impact of acculturation. PhD
thesis, State University of New York at Albany.
Kiefer, Ferenc (199091). Noun incorporation in Hungarian. Acta Linguistica
Hungarica 40(1/2):149177.
Koehn, Edward & Koehn, Sally (1986). Apalai. In Derbyshire & Pullum (1986a),
vol. 1. pp. 33127.
392 BIBLIOGRAPHY

Kumashiro, Toshiyuki (1994). On the conceptual denitions of adpositions and

case markers: A case for the conceptual basis of syntax. Chicago Linguistic
Society 30:236250.
Lambrecht, Knud (1984). A pragmatic constraint on lexical subjects in spoken
French. Chicago Linguistic Society 20:239256.
Lambrecht, Knud (1987). Sentence focus, information structure, and the thetic-
categorical distinction. Berkeley Linguistics Society 13:366382.
Lambrecht, Knud (1994). Information structure and sentence form: Topic, focus,
and the mental representations of discourse referents. Cambridge: Cam-
bridge University Press.
Langacker, Ronald W. (1975). Relative clauses in Classical Nahuatl. International
Journal of American Linguistics 41(1):4668.
Lass, Roger (1984). Phonology: An introduction to basic concepts. Cambridge:
Cambridge University Press.
Lehmann, Henri (1963). Contribution a l'etnographie kwaiker, Colombie. Journal
de la Société des Américanistes 52:255270.
Levinsohn, Stephen H. (1976). The Inga language. The Hague: Mouton.
Li, Charles N. (ed.) (1976). Subject and topic. New York: Academic Press.
Li, Charles N. & Thompson, Sandra A. (1976). Subject and topic: A new typo-
logy of language. In Li (1976). pp. 457489.
Liddell, Scott K. & Johnson, Robert E. (1985). American Sign Language com-
pounds: Implications for the structure of the lexicon. Georgetown University
Round Table on Languages and Linguistics 1985. Washington, DC: Geor-
getown University Press. pp. 8697.
Lindskoog, John N. & Brend, Ruth M. (1962). Cayapa phonemics. In Elson
(1962). pp. 3144.
Longacre, Robert E. (1985). Sentences as combinations of clauses. In Shopen
(1985), vol. 2: Complex constructions. pp. 235286.
Lowe, Ivan (1972). On the relation of formal to sememic matrices with illustra-
tions from Nambiquara. Foundations of Language 8:360390.
Lyons, John (1977). Semantics. Cambridge: Cambridge University Press.
MacEachern, Margaret R., Kern, Barbara & Ladefoged, Peter (1997). Wari'
phonetic structures. Journal of Amazonian Languages 1(1):328.
BIBLIOGRAPHY 393

Malla, Kamal P. (1985). The Newari language: A working outline, Monumenta

Serindica 14. Tokyo: Institute for the Study of Languages and Cultures of
Asia and Africa, Tokyo University of Foreign Studies.
Martínez, E. et al. (1984). Diagnóstico prelimiar de la comunidad indígena
Kwaiker. Pasto. (Unpublished).
Martínez Santacruz, Yolanda (1992). Situación actual de las comunidades Awá
de Colombia e intervención institucional en la región Awá. Report presented
to Asuntos Indígenas, Ministerio de Gobierno, Colombia. (Unpublished).
Matiso, James A. (1981). The grammar of Lahu, University of California Pub-
lications in Linguistics 75. Berkeley: University of California Press.
Matthews, P. H. (1991). Morphology. 2nd edn. Cambridge: Cambridge University
Press.
Meléndez, Miguel Ángel (1989). El nominal en achagua. In Elsa Gómez Im-
bert, Jon Landaburu & Francisco Queixalós (eds), Lenguas de la Orinoquía:
Achagua y guayabero, Lenguas Aborígenes de Colombia, Descripciones 4.
Bogotá: Centro Colombiano de Estudios en Lenguas Aborígenes, Univer-
sidad de los Andes. pp. 166.
Miller, Wick R. (1980). Preaspirated consonants in Central Numic. In Kath-
ryn Klar, Margaret Langdon & Shirley Silver (eds), American Indian and
Indoeuropean studies: Papers in honor of Madison S. Beeler, Trends in Lin-
guistics, Studies and Monographs 16. The Hague: Mouton. pp. 151157.
Mithun, Marianne (1984). The evolution of noun incorporation. Language
60(4):847894.
Mithun, Marianne (1989). The incipient obsolescence of polysynthesis: Cayuga
in Ontario and Oklahoma. In Dorian (1989). pp. 243257.
Montaluisa Chasiquiza, Luis Octavio (1991). Estudio fonológico del awapit:
Hacia la estandarización de su escritura. Cuenca. (Unpublished).
Moore, Bruce R. (1962). Correspondences in South Barbacoan Chibcha. In Elson
(1962). pp. 270289.
Moore, Bruce R. (1966). Diccionario castellanocolorado, coloradocastellano.
Quito: Instituto Lingüístico de Verano.
Moore, Denny (1989). Gavião nominalizations as relative clause and senten-
tial complement equivalents. International Journal of American Linguistics
55(3):309325.
Moravcsik, Edith A. (1978). On the case marking of objects. In Greenberg (1978),
vol. 4: Syntax. pp. 249289.
394 BIBLIOGRAPHY

Mosel, Ulrike (1984). Tolai syntax and its historical development, Pacic Lin-
guistics B-92. Canberra: Pacic Linguistics.
Mougeon, Raymond & Beniak, Edouard (1989). Language contraction and lin-
guistic change: The case of Welland French. In Dorian (1989). pp. 287312.
Munro, Pamela (1984). Floating quantiers in Pima. In Eung-Do Cook &
Donna B. Gerdts (eds), Syntax and semantics 16: The syntax of Native
American languages. Orlando, Florida: Academic Press. pp. 269287.
Murra, John (1963). The Cayapa and Colorado. In Julian H. Steward (ed),
Handbook of South American Indians, vol. 4: The Circum-Caribbean tribes,
Bureau of American Ethnology Bulletin 143. Washington: Smithsonian
Institution. pp. 277291.
Narváez Reyes, Gloria Amparo (1993). La medicina tradicional awa en el cur-
riculo. Master's thesis, Facultad de Ciencias Sociales y Humanas, Univer-
sidad Ponticia Bolivariana, Medellín.
Narváez Reyes, Gloria Amparo (1994a). La medicina tradicional Awá en el cur-
riculo escolar. Salud: Culturas de Colombia (Revista de la Fundación Etnol-
lano) 12:79.
Narváez Reyes, Gloria Amparo (1994b). Los Awa: Su medicina tradicional
desde un proyecto curricular. Ethnia (Instituto Misionero de Antropología,
Medellín) 73:176.
Narváez Reyes, Gloria Amparo & Awa Indigenous Community (1990). Apañando
mediante dibujos nuestra cultura awa. Servicio Seccional de Salud, Pasto,
Nariño. (Unpublished).
Narváez Reyes, Gloria Amparo & Cabildo Cuchilla del Palmar (1992). Traba-
jando pareja: Hombre y mujer awa en el diagnóstico de la salud comunitaria.
Servicio Seccional de Salud, Pasto, Nariño. (Unpublished).
Narváez Reyes, Gloria Amparo & Cabildo Ramos-Mongón-Manchuria-Mirador
(1992). Trabajando unidos en salud comunitaria: Hombres, mujeres y niños
awa en nuestra montaña. Servicio Seccional de Salud, Pasto, Nariño. (Un-
published).
Noonan, Michael (1985). Complementation. In Shopen (1985), vol. 2: Complex
constructions. pp. 42140.
Obando Ordóñez, Pedro (1992). Awa-Kwaiker: An outline grammar of a Colom-
bian/Ecuadorian language, with a cultural sketch. PhD thesis, University of
Texas at Austin.
Orr, Carolyn & Levinsohn, Stephen H. (1992). Clitic placement in content ques-
tions in Napo (Ecuadorian) Quichua. International Journal of American
Linguistics 58(3):299308.
BIBLIOGRAPHY 395

Ortiz, Sergio Elías (1937). Clasicación de las lenguas indígenas de Colombia.

Idearum (Pasto) 1(2):7281.
Ortiz, Sergio Elías (1938). Antroponimia, toponimia y dialectología indígenas del
Suroeste de Colombia. Idearum (Pasto) 1(1112):539559.
Ortiz, Sergio Elías (1954). Estudios sobre lingüística aborígen de Colombia.
Bogotá: Nelly.
Ortiz, Sergio Elías (1965). Lenguas y dialectos indígenas de Colombia. Bogotá:
Lerner. (Book 3 of Prehistoria, vol. 1 of Historia extensa de Colombia).
Osborn, Ann (196972). Alliance at ground level: The Kwaiker of southern
Colombia. Revista de Antropologia (São Paolo) 1720(supplement):209300.
Osborn, Ann (1970). Notas informativas de un estudio sobre los Coaiquer. Sur
(Revista de la Casa de la Cultura de Nariño) 2:6676.
Osborn, Ann (1974). Nomenclatura y parentesco cuayquer. Revista Colombiana
de Antropología 16:262271.
Osborn, Ann (1991a). Compadrazgo y patronazgo: Un caso colombiano. In
Colcultura [Ann Osborn] (1991). pp. 271289. (Originally published in Eco
118).
Osborn, Ann (1991b). Estudio para el conocimiento y rescate de formas autócto-
nas de atención al niño: Los Kwaiker de Nariño. In Colcultura [Ann Osborn]
(1991). pp. 115267.
Osborn, Ann & Melo, Fanny (1991). La nutrición kwaiker en el contexto de su
sistema agrícola. In Colcultura [Ann Osborn] (1991). pp. 53111.
Palmer, F. R. (1986). Mood and modality. Cambridge: Cambridge University
Press.
Parra Rizo, Jaime Hernando (1989). Los indígenas Awa-Kwaiker de Nariño
(Municipio de Ricaurte): Cambio socio-cultural a partir de la década de los
años sesenta. Master's thesis, Universidad de los Andes, Bogotá.
Parra Rizo, Jaime Hernando & Virsano Bellow, Susan (1992). Medicina tradi-
cional del pueblo de Altaquer. Quito: ABYA-YALA.
Patz, Elisabeth (1982). A grammar of the Kuku Yalanji language of North
Queensland. PhD thesis, Australian National University.
Payne, David L. (1990). Some widespread grammatical forms in South American
languages. In Payne (1990a). pp. 7587.
Payne, Doris L. (ed.) (1990a). Amazonian linguistics: Studies in lowland South
American languages. Austin: University of Texas Press.
396 BIBLIOGRAPHY

Payne, Doris L. (1990b). Morphological characteristics of lowland South Amer-

ican languages. In Payne (1990a). pp. 213241.
Payne, Doris L. (1995). Verb initial languages and information order. In Downing
& Noonan (1995). pp. 449485.
Payne, Doris L. & Payne, Thomas E. (1990). Yagua. In Derbyshire & Pullum
(1990), vol. 2. pp. 249474.
Payne, John R. (1985). Negation. In Shopen (1985), vol. 1: Clause structure.
pp. 197242.
Payne, Judith & Payne, David (1991). The pragmatics of split-intransitivity in
Asheninca. Paper presented at the International Congress of Americanists,
New Orleans, July 1991.
Paz y Miño, Luis T. (1946). Lenguas indígenas del Ecuador: La lengua Pasto.
Revista de Historia (Pasto) 2(78):137166.
Pérez T., Aquiles R. (1980). El idioma Cuayker. Quito: Casa de la Cultura
Ecuatoriana.
Petersen de Piñeros, Gabriele (1992). Predicación y distinción verbo-nombre en
uitoto. In Rojas Curieux (1992). pp. 107116.
Popjes, Jack & Popjes, Jo (1986). Canela-Krahô. In Derbyshire & Pullum
(1986a), vol. 1. pp. 128199.
Pullum, Georey K. (1981). The category status of innitival to. University of
Washington Working Papers in Linguistics 6:5572.
Queixalós, Francisco (1992). Auxiliares de postura corporal en sikuani. In Ro-
jas Curieux (1992). pp. 185197.
Quirk, Randolph, Greenbaum, Sidney, Leech, Georey & Svartvik, Jan (1972).
A grammar of contemporary English. Harlow, Essex: Longman.
Rojas Curieux, Tulio (ed.) (1992). Segundo congreso del CCELA, Lenguas Aborí-
genes de Colombia, Memorias 2. Bogotá: Centro Colombiano de Estudios
en Lenguas Aborígenes, Universidad de los Andes.
Romaine, Suzanne (1985). The syntax and semantics of the code-mixed com-
pound verb in Panjabi/English bilingual discourse. Georgetown University
Round Table on Languages and Linguistics 1985. Washington, DC: Geor-
getown University Press. pp. 3549.
Rosenthal, Jane M. (1972). On the relative clauses of Classical Nahuatl. Chicago
Linguistic Society 8 (Parassession, The Chicago which hunt):246255.
Ruhlen, Merritt (1978). Nasal vowels. In Greenberg (1978), vol. 2: Phonology.
pp. 203241.
BIBLIOGRAPHY 397

Schmidt, Annette (1985a). The fate of ergativity in dying Dyirbal. Language

61(2):378396.
Schmidt, Annette (1985b). Young people's Dyirbal: An example of language death
from Australia. Cambridge: Cambridge University Press.
Sebba, Mark (1987). The syntax of serial verbs: An investigation into serialisa-
tion in Sranan and other languages, Creole Language Library 2. Amsterdam:
John Benjamins.
Second Binational Meeting (1988). Problems, conclusions and recommendations
of the second Colombian-Ecuadorian binational meeting on the unication
of the Awa Pit alphabet, 30 May3 June 1988. Appendix 8 of Centro Ex-
perimental Piloto Nariño (1990).
Seki, Lucy (1990). Kamaiurá (Tupí-Guaraní) as an active-stative language. In
Payne (1990a). pp. 367391.
Sells, Peter (1987). Aspects of logophoricity. Linguistic Inquiry 18(3):445479.
Shibatani, Masayoshi (1997). The semantics of voice. Seminar, Department of
Linguistics, Australian National University, 10 March 1997.
Shopen, Timothy (ed.) (1985). Language typology and syntactic description. Cam-
bridge: Cambridge University Press.
Shopen, Timothy & Konaré, M. (1970). Sonrai causatives and passives: Trans-
formational versus lexical derivations for propositional heads. Studies in
African Linguistics 1:211254.
Spencer, Andrew (1991). Morphological theory: An introduction to word structure
in generative grammar. Oxford: Blackwell.
Stark, Louisa R. (1985). Indigenous languages of lowland Ecuador: History and
current status. In Harriet E. Manelis Klein & Louisa R. Stark (eds), South
American Indian languages: Retrospect and prospect. Austin: University of
Texas Press. pp. 157193.
Steever, Sanford B. (1979). Noun incorporation in Tamil, or What's a noun like
you doing in a verb like this? Chicago Linguistic Society 15:279290.
Stolz, Thomas (1996). With whom or with what? comitatives and instrumentals
in Amerindian languages. Paper given at the 35th Conference on American
Indian Languages (The Society for the Study of the Indigenous Languages of
the Americas), at the 95th Annual Meeting of the American Anthropological
Association, San Francisco, 2123 November 1996.
Strom, Clay (1992). Retuarã syntax, Studies in the Languages of Colombia 3.
Arlington, Texas: Summer Institute of Linguistics and University of Texas
at Arlington.
398 BIBLIOGRAPHY

Telban, Blaz (1988). Grupos étnicos de Colombia: Etnografía y bibliografía,

Colección 500 Años 3. Quito: ABYA-YALA.
Thompson, Sandra A. & Longacre, Robert E. (1985). Adverbial clauses. In
Shopen (1985), vol. 2: Complex constructions. pp. 171234.
Thráinsson, Höskuldur (1979). On complementation in Icelandic, Outstanding
Dissertations in Linguistics 23. New York: Garland Publishing.
Thurgood, Graham (1986). The nature and origins of the Akha evidentials sys-
tem. In Chafe & Nichols (1986). pp. 214222.
Tomlin, Russell S. (1995). Focal attention, voice, and word order: An experi-
mental, cross-linguistic study. In Downing & Noonan (1995). pp. 517554.
Traugott, Elizabeth Closs & Heine, Bernd (eds) (1991). Approaches to grammat-
icalization. Amsterdam: John Benjamins.
Trillos Amaya, María (1988). Segundo seminario-taller para la unicación de
alfabetos Awa. Glotta (Bogotá) 3(2):4850.
Trillos Amaya, María (1989). Aspecto, modo y tiempo en damana. In Elsa
Gómez Imbert, Jon Landaburu & Francisco Queixalós (eds), Lenguas de la
Sierra Nevada de Santa Marta, Lenguas Aborígenes de Colombia, Descrip-
ciones 3. Bogotá: Centro Colombiano de Estudios en Lenguas Aborígenes,
Universidad de los Andes. pp. 1142.
Triviño Garzón, Lilia (1992). Sobre la modalidad en la lengua guambiana
(Quizgó). In Rojas Curieux (1992). pp. 219227.
Ultan, Russell (1978a). Some general characteristics of interrogative systems. In
Greenberg (1978), vol. 4: Syntax. pp. 211248.
Ultan, Russell (1978b). Toward a typology of substantival possession. In Green-
berg (1978), vol. 4: Syntax. pp. 1149.
Urban, Greg & Sherzer, Joel (1988). The linguistic anthropology of native South
America. Annual Review of Anthropology 17:283307.
Vásquez de Ruíz, Beatriz (1988). La predicación en guambiano, Lenguas Aborí-
genes de Colombia, Descripciones 2. Bogotá: Colciencias and Universidad
de los Andes.
Vásquez de Ruíz, Beatriz (1992). La modalidad epistémica en guambiano (Guam-
bía). In Rojas Curieux (1992). pp. 209218.
Vittadello, P. Alberto (1988). Cha'palaachi: El idioma cayapa, Miscelánea Antro-
pológica Ecuatoria Serie Monográca 5. Guayaquil: Ponticia Universidad
Católica del Ecuador, Sede de Esmeraldas & Museos del Banco Central del
Ecuador.
BIBLIOGRAPHY 399

Wasow, Thomas (1977). Transformations and the lexicon. In Peter W. Culicover,

Thomas Wasow & Adrian Akmajian (eds), Formal syntax. New York: Aca-
demic Press. pp. 327360.
West, Robert (1957). The Pacic lowlands of Colombia. Louisiana: Baton Rouge.
Wheeler, Alva (1972). Proto Chibchan. In Esther Matteson et al. (eds), Com-
parative studies in Amerindian languages. The Hague: Mouton. pp. 93108.
Wise, Mary Ruth (1986). Grammatical characteristics of Preandine Arawakan
languages of Peru. In Derbyshire & Pullum (1986a), vol. 1. pp. 567642.
Woodbury, Anthony C. (1986). Interactions of tense and evidentiality: A study
of Sherpa and English. In Chafe & Nichols (1986). pp. 188202.
Ziv, Yael (1994). Left and Right Dislocations: Discourse functions and anaphora.
Journal of Pragmatics 22:629645.
Zwicky, Arnold M. (1990). What are we talking about when we talk about serial
verbs? In Brian D. Joseph & Arnold M. Zwicky (eds), When verbs collide:
Papers from the 1990 Ohio State Mini-Conference on Serial Verbs, Ohio
State University Working Papers in Linguistics 39. Columbus: Ohio State
University Department of Linguistics. pp. 113.