Articles in the world’s languages

An der Philologischen Fakultät der Universität Leipzig eingereichte

DISSERTATION

Zur Erlangung des akademischen Grades

Doctor Philosophiae
(Dr. phil.)
von

Laura Becker

Datum der Einreichung:

Datum der Verleihung:

Gutachter: Prof. Dr. Martin Haspelmath

Prof. Dr. Barbara Stiebels
Contents

1 Introduction 1
1.1 Background and objectives . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.2 Overview . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
1.3 Data and methodology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9

2 Articles as a crosslinguistic category 20

2.1 The problem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20
2.2 Defining articles . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
2.2.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
2.2.2 Articles encode referent types . . . . . . . . . . . . . . . . . . . . . . . . . 25
2.2.3 Articles occur in the nominal domain . . . . . . . . . . . . . . . . . . . . . 32
2.2.4 Articles are systematically distributed . . . . . . . . . . . . . . . . . . . . 36
2.3 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44

3 The distinction between articles and similar categories 45

3.1 Demonstratives . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 45
3.2 The numeral ‘one’ . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 54
3.3 Negative polarity items . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 59
3.4 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 64

4 Articles and referentiality 65

4.1 Referentiality and referent types . . . . . . . . . . . . . . . . . . . . . . . . . . . . 65
4.1.1 The discourse model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 65
4.1.2 The link between linguistic expressions and referent types . . . . . . . . . 68
4.1.3 Identifiabilty as a core concept of referentiality . . . . . . . . . . . . . . . 70
4.1.3.1 Defining identifiability . . . . . . . . . . . . . . . . . . . . . . . 71
4.1.3.2 Other approaches . . . . . . . . . . . . . . . . . . . . . . . . . . 73
4.2 Defining the referent types . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 76

i
 4.2.1 Definite referents . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 76
4.2.2 Specific referents . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 82
4.2.3 Nonspecific referents . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 84
4.2.4 Generic referents . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 85
4.3 Defining the article types . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 86
4.3.1 Article types from the definite domain . . . . . . . . . . . . . . . . . . . . 86
4.3.1.1 Definite articles . . . . . . . . . . . . . . . . . . . . . . . . . . . 86
4.3.1.2 Anaphoric articles . . . . . . . . . . . . . . . . . . . . . . . . . . 87
4.3.1.3 Recognitional articles . . . . . . . . . . . . . . . . . . . . . . . . 87
4.3.2 Article types from the indefinite domain . . . . . . . . . . . . . . . . . . . 88
4.3.2.1 Exclusive-specific articles . . . . . . . . . . . . . . . . . . . . . . 88
4.3.2.2 Nonspecific articles . . . . . . . . . . . . . . . . . . . . . . . . . 88
4.3.2.3 Indefinite and presentational articles . . . . . . . . . . . . . . . 88
4.3.3 Domain-crossing article types . . . . . . . . . . . . . . . . . . . . . . . . . 89
4.3.3.1 Inclusive-specific articles . . . . . . . . . . . . . . . . . . . . . . 89
4.3.3.2 Referential articles . . . . . . . . . . . . . . . . . . . . . . . . . 89
4.4 The relation between referent types and article types . . . . . . . . . . . . . . . . 89
4.5 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 93

5 Articles: examples from the world’s languages 95

5.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 95
5.2 Definite articles . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 97
5.2.1 Kaqchikel . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 97
5.2.2 Mokpe . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 100
5.2.3 Definite articles with deictic referents . . . . . . . . . . . . . . . . . . . . 105
5.2.4 Definite articles with absolutely unique referents . . . . . . . . . . . . . . 108
5.2.5 Definite articles with proper nouns . . . . . . . . . . . . . . . . . . . . . . 110
5.3 Anaphoric articles . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 112
5.3.1 Limbum . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 112
5.3.2 Other examples . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 114
5.4 Recognitional articles . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 116
5.5 Exclusive-specific articles . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 119
5.5.1 Q’anjobal . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 119
5.5.2 Other examples . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 122
5.6 Inclusive-specific articles . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 125
5.6.1 Bemba . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 125
 5.6.2 Tongan . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 129
5.7 Nonspecific articles . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 131
5.7.1 Q’anjobal . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 131
5.7.2 Tongan . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 132
5.7.3 Lakota . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 135
5.7.4 The source of nonspecific articles . . . . . . . . . . . . . . . . . . . . . . . 137
5.8 Indefinite articles . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 139
5.8.1 Tz’utujil . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 140
5.8.2 Carib . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 142
5.8.3 Bonan . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 143
5.9 Presentational articles . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 145
5.9.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 145
5.9.2 Examples . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 146
5.9.2.1 Lango . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 146
5.9.2.2 Logba . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 147
5.9.2.3 Aghul . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 148
5.9.2.4 Chatino . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 149
5.9.3 The role of presentational articles for article development . . . . . . . . . 151
5.10 Referential articles . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 151
5.10.1 Rapa Nui . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 152
5.10.2 Halkomelem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 156
5.10.3 Other examples . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 160
5.11 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 162

6 Morphological properties of articles 163

6.1 Article inflection . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 163
6.1.1 Selected patterns . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 163
6.1.2 Crosslinguistic trends . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 174
6.1.3 Factors that condition the article inflection . . . . . . . . . . . . . . . . . . 179
6.1.3.1 Inflection of the noun . . . . . . . . . . . . . . . . . . . . . . . . 179
6.1.3.2 Inflection of the article’s source element . . . . . . . . . . . . . 181
6.2 Article paradigms . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 186
6.2.1 One article or many? . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 186
6.2.2 A complex article paradigm in Mokpe . . . . . . . . . . . . . . . . . . . . 188
6.3 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 193
7 Crosslinguistic variation and universals 195
7.1 Areal trends . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 195
7.2 Unattested and rare article types: two examples . . . . . . . . . . . . . . . . . . . 199
7.2.1 Non-anaphoric definite articles . . . . . . . . . . . . . . . . . . . . . . . . 199
7.2.2 Generic articles . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 203
7.3 A typology of articles and article systems . . . . . . . . . . . . . . . . . . . . . . . 206
7.4 Interactions between articles and selected constructions . . . . . . . . . . . . . . . 213
7.4.1 Definite articles in superlative constructions . . . . . . . . . . . . . . . . . 213
7.4.2 Articles and plurals . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 216
7.4.3 Definite articles co-occurring with demonstratives and possessives . . . . 222
7.4.4 Articles co-occurring with other articles . . . . . . . . . . . . . . . . . . . 229
7.5 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 231

8 Conclusions 233
8.1 Main findings . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 233
8.2 The path forward . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 236

Appendix A Article properties (99-languages sample) 240

Appendix B Questionnaire 263

iv
List of Figures

1.1 Areal coverage of the sample . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10

1.2 Areal distribution of definite and indefinite articles (Dryer 2013a) . . . . . . . . . 10

3.1 Grammaticalization path of demonstrative and definite article . . . . . . . . . . . 46

3.2 Cut-off point between demonstratives and articles . . . . . . . . . . . . . . . . . . 47
3.3 Demonstrative with anaphoric article . . . . . . . . . . . . . . . . . . . . . . . . . 49
3.4 Demonstrative with deictic and anaphoric functions . . . . . . . . . . . . . . . . . 50
3.5 Demonstrative with uniqueness-based definite article . . . . . . . . . . . . . . . . 51
3.6 Demonstrative with definite article . . . . . . . . . . . . . . . . . . . . . . . . . . 52
3.7 Development of the numeral ‘one’ to a nominal marker . . . . . . . . . . . . . . . 55
3.8 Cut-off point between the numeral ‘one’ and indefinite articles . . . . . . . . . . . 56

4.1 Discourse situation and discourse universe . . . . . . . . . . . . . . . . . . . . . . 66

4.2 Structure of a referring word . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 69
4.3 Link between the referents of the kinds ‘book’ and ‘author’ . . . . . . . . . . . . . 80
4.4 Link between the referents of the kinds ‘village’ and ‘church’ . . . . . . . . . . . . 81
4.5 Anchoring of a specific referent . . . . . . . . . . . . . . . . . . . . . . . . . . . . 83
4.6 Nonspecific referent with its concept . . . . . . . . . . . . . . . . . . . . . . . . . 85
4.7 Overview of referent types and article types . . . . . . . . . . . . . . . . . . . . . 90
4.8 Hierarchy of the major referent types . . . . . . . . . . . . . . . . . . . . . . . . . 91

6.1 Article inflection in the six macroareas . . . . . . . . . . . . . . . . . . . . . . . . 175

6.2 Categories of article inflection in the six macroareas . . . . . . . . . . . . . . . . . 176
6.3 Inflectional behaviour of following and preceding articles . . . . . . . . . . . . . . 176
6.4 Inflectional behaviour of different article types . . . . . . . . . . . . . . . . . . . . 177
6.5 Inflectional categories of different article types . . . . . . . . . . . . . . . . . . . . 178
6.6 Number and case marking on the article and the noun . . . . . . . . . . . . . . . . 180

7.1 Article types . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 196

7.2 Article types in the six macroareas . . . . . . . . . . . . . . . . . . . . . . . . . . . 197

v
7.3 Co-occurrence of definite articles with demonstratives and possessives . . . . . . 225
7.4 Position of demonstratives and possessives and co-occurrence with definite articles 227

8.1 Overview of referent types and article types . . . . . . . . . . . . . . . . . . . . . 234

vi
List of Tables

1.1 Summary of referent types and article types . . . . . . . . . . . . . . . . . . . . . 6

1.2 Primary data collection . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
1.3 Languages from the macroarea “Africa” . . . . . . . . . . . . . . . . . . . . . . . . 14
1.4 Languages from the macroarea “Eurasia” . . . . . . . . . . . . . . . . . . . . . . . 15
1.5 Languages from the macroarea “Papunesia” . . . . . . . . . . . . . . . . . . . . . . 16
1.6 Languages from the macroarea “Australia” . . . . . . . . . . . . . . . . . . . . . . 17
1.7 Languages from the macroarea “North America” . . . . . . . . . . . . . . . . . . . 18
1.8 Languages from the macroarea “South America” . . . . . . . . . . . . . . . . . . . 19

3.1 Macedonian definite article paradigm . . . . . . . . . . . . . . . . . . . . . . . . . 52

3.2 Characteristic properties of indefinite articles (Heine 1997: 68) . . . . . . . . . . . 56
3.3 Relevant contexts for adnominal NPIs . . . . . . . . . . . . . . . . . . . . . . . . . 60

4.1 Traditional view of specificity and nonspecificity . . . . . . . . . . . . . . . . . . 83

4.2 Summary of referent types and article types . . . . . . . . . . . . . . . . . . . . . 94

5.1 Examples for different article types . . . . . . . . . . . . . . . . . . . . . . . . . . 96

5.2 Mokpe article paradigm . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 101
5.3 Kaqchikel demonstratives . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 107
5.4 Limbum demonstratives . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 112
5.5 Bemba noun class markers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 127
5.6 Tongan article paradigm . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 129
5.7 Paradigm of Tongan adnominal possessive markers (A-set) . . . . . . . . . . . . . 134
5.8 Paradigm of Tongan adnominal possessive markers (O-set) . . . . . . . . . . . . . 134
5.9 Lakota specific and nonspecific article paradigm . . . . . . . . . . . . . . . . . . . 135
5.10 Halkomelem article exponents . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 156
5.11 Niuean article paradigm . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 161

6.1 German article paradigm . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 165

6.2 Ch’ol numeral classifiers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 165

vii
6.3 Bemba article and noun class prefix . . . . . . . . . . . . . . . . . . . . . . . . . . 166
6.4 Gender marking on Supyire nouns and articles . . . . . . . . . . . . . . . . . . . . 167
6.5 Examples of gender marking on Supyire articles . . . . . . . . . . . . . . . . . . . 168
6.6 Ute definite article paradigm . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 169
6.7 Case marking on nouns and the definite article in German . . . . . . . . . . . . . 170
6.8 Albanian case paradigm . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 171
6.9 Diyari definite article paradigm (core cases) . . . . . . . . . . . . . . . . . . . . . . 171
6.10 Macedonian definite article paradigm . . . . . . . . . . . . . . . . . . . . . . . . . 172
6.11 Numbers for article and noun inflection . . . . . . . . . . . . . . . . . . . . . . . . 180
6.12 Inflection of articles and their source elements (personal/possessive pronouns) . . 183
6.13 Inflection of articles and their source elements (demonstratives) . . . . . . . . . . 184
6.14 Inflection of articles and their source elements (numeral ‘one’) . . . . . . . . . . . 185
6.15 Biak specific article exponents . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 186
6.16 Specific and nonspecific article exponents . . . . . . . . . . . . . . . . . . . . . . . 187
6.17 Mokpe article paradigm . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 189
6.18 The definite article in Mokpe with demonstratives and possessives . . . . . . . . . 191

7.1 Distribution of article types in the sample . . . . . . . . . . . . . . . . . . . . . . . 196

7.2 The relation of third person pronouns and demonstratives in Australian languages 198
7.3 A typology of articles . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 207
7.4 Article systems encoding 1 referent type . . . . . . . . . . . . . . . . . . . . . . . 208
7.5 Article systems encoding 2 referent types . . . . . . . . . . . . . . . . . . . . . . . 209
7.6 Article systems encoding 3 referent types . . . . . . . . . . . . . . . . . . . . . . . 210
7.7 Unattested article systems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 210
7.8 The compatibility of articles with plural nouns . . . . . . . . . . . . . . . . . . . . 217
7.9 Singular and plural forms of indefinite articles . . . . . . . . . . . . . . . . . . . . 219
7.10 Lavukaleve presentational article . . . . . . . . . . . . . . . . . . . . . . . . . . . 220
7.11 Lavukaleve ru ‘big’ . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 220
7.12 Lavukaleve definite article . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 221
7.13 Lavukaleve predicate . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 221

8.1 Systems encoding 1 referent type . . . . . . . . . . . . . . . . . . . . . . . . . . . 235

8.2 Systems encoding 2 referent types . . . . . . . . . . . . . . . . . . . . . . . . . . . 235
8.3 Systems encoding 3 referent types . . . . . . . . . . . . . . . . . . . . . . . . . . . 235
8.4 Unattested systems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 235

A.1 Distribution of article types in the 6 macroareas . . . . . . . . . . . . . . . . . . . 240

viii
A.2 Article properties (Africa) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 241
A.3 Article properties (Eurasia) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 242
A.4 Article properties (Papunesia) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 243
A.5 Article properties (Australia) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 244
A.6 Article properties (North America) . . . . . . . . . . . . . . . . . . . . . . . . . . . 245
A.7 Article properties (South America) . . . . . . . . . . . . . . . . . . . . . . . . . . . 246
A.8 Article inflection by area . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 247
A.9 Bayes proportion test for article inflection by area . . . . . . . . . . . . . . . . . . 247
A.10 Article inflection by article position . . . . . . . . . . . . . . . . . . . . . . . . . . 247
A.11 Bayes proportion test for article inflection by article position . . . . . . . . . . . . 247
A.12 Article inflection by article type . . . . . . . . . . . . . . . . . . . . . . . . . . . . 248
A.13 Bayes proportion test for article inflection by article type . . . . . . . . . . . . . . 248
A.14 Inflectional behaviour of different articles within single languages . . . . . . . . . 249
A.15 Bayes proportion test for article inflection . . . . . . . . . . . . . . . . . . . . . . 250
A.16 Bayes proportion test for noun inflection . . . . . . . . . . . . . . . . . . . . . . . 250
A.17 Distribution of article systems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 251
A.18 Bayes proportion test for different referent types expressed by articles . . . . . . . 252
A.19 Bayes proportion test for 2-referent-types systems . . . . . . . . . . . . . . . . . . 252
A.20 Bayes proportion test for 3-referent-types systems . . . . . . . . . . . . . . . . . . 252
A.21 Definite articles and plural nouns . . . . . . . . . . . . . . . . . . . . . . . . . . . 253
A.22 Anaphoric and recognitional articles and plural nouns . . . . . . . . . . . . . . . . 254
A.23 Indefinite and presentational articles with plural nouns . . . . . . . . . . . . . . . 255
A.24 Exclusive specific and nonspecific articles with plural nouns . . . . . . . . . . . . 256
A.25 Bayes proportion test for compatibility with plural . . . . . . . . . . . . . . . . . . 256
A.26 Bayes proportion test for plural marking . . . . . . . . . . . . . . . . . . . . . . . 256
A.27 Co-occurrence of definite articles and demonstratives . . . . . . . . . . . . . . . . 257
A.28 Co-occurrence of definite articles and possessives . . . . . . . . . . . . . . . . . . 258
A.29 Co-occurring definite articles and demonstratives . . . . . . . . . . . . . . . . . . 258
A.30 Bayes proportion test for definite articles and demonstratives . . . . . . . . . . . . 258
A.31 Co-occurring definite articles and possessives . . . . . . . . . . . . . . . . . . . . 259
A.32 Bayes proportion test for definite articles and possessives . . . . . . . . . . . . . . 259
A.33 Co-occurring definite articles with pre-/postnominal possessives . . . . . . . . . 259
A.34 Bayes proportion test for definite articles and pre-/postnominal possessives . . . . 259
A.35 Co-occurrence of the definite article and the demonstrative (i) . . . . . . . . . . . 260
A.36 Co-occurrence of the definite article and the demonstrative (ii) . . . . . . . . . . . 261
A.37 Co-occurrence of the definite article and the possessive (i) . . . . . . . . . . . . . 261

ix
A.38 Co-occurrence of the definite article and the possessive (ii) . . . . . . . . . . . . . 262

x
Abbreviations

1,2,3 1st, 2nd, 3rd person cd clausal determiner

abl ablative cj conjoint tense
abs absolutive cl class marker, classifier
acc accusative cntg contiguous
actl actual comm common noun
act active compar comparative
addit additive compl completive
addr addressee marker comp complementizer
adj adjective cond conditional
admon admonition conj conjunction
adnom adnominal conseq consequential
advers adversative cont continuous
adv adverbial cop copula
affect affective custom customary verbal suffix
aff affirmative cv converb
ag agentive dat dative
am association marker deb debitive
ana anaphoric decaus decausative
ant anterior decl declarative
an animate def definite
aor aorist dem demonstrative
appl applicative dep dependent aspect
apud apudessive desid desiderative
art article dev devaluation
assert assertive dim diminutive
attr attributive marker dir directional
augm augmentative dist distal
aug augment dj disjoint tense
aux auxiliary verb ds different subject (switch reference)
ben benefactive dur durative
caus causative du dual

xi
emph emphasis narr narrative marker
epist epistemic possibility neg negation
erg ergative nf non-feminine
evid evidential nhum non-human
ev eventive nmlz nominalizer
excl exclamative nm non-masculine
exist existential copula nom nominative
exspec exclusive-specific npi negative polarity item
ex exlusive nposs nonpossessed
foc focus npst nonpast
fut future nspec nonspecific
f feminine ntr neutral aspect
gen gentive num numeral marker
hab habitual nvis non-visible
hesit hesitation n neuter
hon honorific obl oblique
hort hortative orig origin (spatial value)
hum human o object marker
imm immediate part partitive
impers impersonal pass passive
impl implicated perf perfect
imp imperative pfv perfective
inan inanimate pl plural
inch inchoative pm phrasal marker
indef indefinite polit politeness marker
iness inessive pol polarity
infer inferential poss possessive
inf infinitive post postposition
inspec inclusive-specific pot potential
instr instrumental pred predicative
intr intransitive presump presumptive
int intensive pres presentational
in inclusive prog progressive
irr irrealis prom prominence marker
iter iterative prop proper noun marker
lat lative prs present
loc locative pst1 past 1, recent past
med medial pst2 past2 , intermediate past
min minimal number pst3 past 3, remote past
m masculine pst:uw unwitnessed past

xii
pst:w witnessed past sg singular
pst past src source
ptcp participle ss same subject (switch reference)
p patient stat stative
qot quotative sti indirect stance marker
q question/interrogative marker subj subjunctive
rc relative clause marker sub subordinate subject
real realis super super localization
recip reciprocal sup superlative
recog recognitional s subject
recpst recent past temp temporal marker
red reduplication top topic
refl reflexive tr transitive
ref referential vent ventive
rel relativizer vicin vicinity
rempst remote past vis visible
rem remote A agent argument in transitive clauses
rep repetitive P patient argument in transitive clauses
res resultative S sole argument in intransitive clauses
seq sequential

xiii
Chapter 1

Introduction

1.1 Background and objectives

At first sight, linguists seem to have an intuitive understanding of what articles are as a category:
they usually occur with nouns and mark them as definite or indefinite. However, a closer look
at how the label of articles is used in the linguistic literature reveals that we still lack a proper
understanding of what articles are. There are three main problems.
First, articles are often classified as definite or indefinite even if their functions differ consider-
ably from what we would usually call definite or indefinite articles; in many cases simply because
these categories are not defined in a precise way or because no better notions are available. Thus,
we still lack coherent criteria to define articles and also different types of articles. Second, descrip-
tions or analyses of markers that could be articles are sometimes not treated as such based on the
properties of articles in English or other European languages. It is often their similarity to English
articles that is taken as a criterion to treat the marker as an article or not. Third, we find descrip-
tions or analyses of markers as articles based on the fact that they occur together with nouns. For
instance, many Oceanic languages have a marker that occurs with proper nouns (sometimes only
in certain syntactic contexts) and that is traditionally referred to as proper or personal article.
What these three different uses of article as a label show is that it is still not clear which
of the properties of articles in well-known languages are relevant to define articles language-
independently. We still lack a precise definition of their function, which is why markers are
sometimes treated as articles or not based on the functions of articles in e.g. English. In addition,
a principled distinction and definition of different article types are still not available, which is why
we see the labels of definite and indefinite articles applied to different markers across languages
with a wide range of functions. What complicates the picture even more is that their occurrence
with nouns is sometimes taken as sole criterion to treat markers as articles, leading to an even
less coherent use of the notion. Hence, even though most linguists may have an intuition about

1
articles, we still do not have principled criteria that can be applied to any language in order to
decide whether a given marker is an article or not, and if so, which type of article it is.
This is surprising especially given the large amount of literature that has dealt with articles
in single languages and language families. Since articles can be approached from various point
of views, for instance, with respect to their semantic and/or syntactic properties, it is impossible
to provide an appropriate overview of this literature here. The main point is that while all these
approaches may have suitable criteria to discuss articles within single languages or language
families, there still is no good way of comparing these results against each other, since the notion
of articles is understood in so many different ways. This study aims at closing this gap by defining
articles as a crosslinguistic category and analyzing the crosslinguistic trends of their properties.
During the last decades, similar advances have been made for a number of functionally or for-
mally related categories, e.g., Aikhenvald (2008) for classifiers, Diessel (1999) for demonstratives,
Haspelmath (1999) for indefinite pronouns, and Bhat (2004) for pronouns in general. Articles,
on the other hand, have not yet been discussed in detail as a crosslinguistic category. However,
even though they have not been the main point of attention in a typological study, a number of
typologically oriented work that also discusses articles has to be mentioned.
The studies that probably come closest to a detailed typological discussion of articles are:
Dryer (1989, 2013a,b, 2014), Ghomeshi et al. (2009), Himmelmann (1997), Leiss (2000), Schroeder
(2006). Ghomeshi et al. (2009) is a collection of studies on determiners from a variety of languages,
but articles are only one of various phenomena discussed. Both Schroeder (2006) and Leiss (2000)
offer a comparative studies on articles, focusing on articles in Europe. While Schroeder (2006)
provides an overview of different properties of the articles found in various areas of Europe, Leiss
(2000) investigates the relation between aspect and articles in a number of Indo-European lan-
guages, arguing that articles emerged as a consequence of the loss of aspectual distinctions in the
verbal domain.
Dryer (1989, 2013a,b), on the other hand, discusses definite and indefinite articles from a
broader crosslinguistic perspective. Dryer (1989) discusses the correlation between the order of ar-
ticles and nouns with VO and OV word orders. Therefore, his focus is not on articles as such, even
though it is based on probably the largest crosslinguistic sample of articles available until now.
Similarly, Dryer (2013a,b) are two crosslinguistic studies of definite and indefinite articles, respec-
tively. They correspond to two chapters in the World Atlas of Language Structure (WALS, Dryer
& Haspelmath 2013). Since this project aims at an overview of various grammatical phenomena
at a large crosslinguistic scale, these two chapters can only offer a very condensed overview of a
few selected properties of articles. In addition, the distinction of article types is kept as minimal
as possible, differentiating between definite and indefinite articles only. Dryer (2014) is a reply to
criticisms of these two WALS chapters and clarifies a number of methodological questions espe-

2
cially concerning the definition of articles. It is therefore also a condensed discussion of articles,
crosslinguistic trends with respect to different article types, and selected peculiarities of articles.
Himmelmann (1997) offers another crosslinguistic study discussing a topic that intersects with ar-
ticles. He addresses questions concerning referentiality, markers in the noun phrase that express
referential functions, and the relation to syntactic structure in the noun phrase. Therefore, the
scope in Himmelmann (1997) is wider, and articles are only one of several related issues that are
addressed.
Thus, to the best of my knowledge, a detailed and systematic crosslinguistic study of articles
is still not available. I propose a language-independent definition of articles as a crosslinguistic
category based on three main criteria: their domain (the noun phrase), their function (express-
ing referentiality), and their systematic occurrence. The next step is the definition of different
types that reflect the properties of articles that are found in the world’s languages. To do that,
I define different referential functions, called referent types, which are used to define different
types of articles. On the basis of a sample including 99 languages with 131 articles, I show de-
tailed examples from different languages of the world for the nine major article types that can be
distinguished based on their functions. Having defined and illustrated different types of articles,
this study explores the crosslinguistic variation and trends of other properties that articles have.
I focus mainly on article inflection but also show their interaction with certain constructions in
which articles are not required as markers of referentiality. It can be shown that these proper-
ties are relatively independent of their referential functions; their inflectional behaviour and also
their compatibility with certain constructions can be explained to a large extent by the properties
of the articles’ source elements instead of the properties of articles themselves. The last part of
this study proposes a typology of articles and article systems that result in five universals. The
universals reflect the findings regarding crosslinguistic trends of article types and article systems
in this study.
The focus of this study is a crosslinguistic comparison of articles by capturing the variation and
trends. Therefore, it does not aim at an in-depth analysis of articles or article systems of particular
languages: even though I provide a number of case studies, they only serve to illustrate and
contribute to the crosslinguistic picture and they should not be taken as an exhaustive discussion
of the article (system) in the languages in question.

1.2 Overview
Chapter 2: Articles as a crosslinguistic category In Chapter 2, I define articles as markers that
occur systematically with nouns to mark the referential function of the noun. This definition
includes three types of criteria: the function, domain, and distribution criteria. The function

3
criterion allows us to distinguish between the following main types of articles based on the ref-
erential functions they have. The types are: anaphoric, recognitional, definite, inclusive-specific,
exclusive-specific, indefinite, nonspecific, and referential articles. The form criterion ensures that
only markers which occur in the nominal domain together with nouns qualify as articles, and
it excludes, e.g., definiteness-sensitive adjective markers. The distribution criterion requires that
articles occur systematically to indicate the referential function of the noun that they occur with.
Systematically does not mean obligatory in the sense that the article has to occur in all contexts
in which it is expected due to its referential functions. Rather, articles can still be treated as such
even though they do not occur in all contexts in which they would be referentially required, as
long as the exceptions follow systematic rules. This holds in either direction: articles can be absent
in contexts in which their presence would be required, but they can also be present in contexts
in which they are not required to indicate a referential function. I provide a number of examples
for markers that I consider to be articles based on these three criteria, and examples for markers
that I do not treat as articles.

Chapter 3: The distinction between articles and similar categories Chapter 3 follows up on the
definition of articles as a crosslinguistic category and delimits certain article types from other sim-
ilar categories, namely demonstratives, the numeral ‘one’, and negative polarity items. Demon-
stratives and the numeral ‘one’ are difficult to distinguish from definite and indefinite articles
because of the diachronic relation of these elements: definite articles often originate from demon-
stratives, as well as indefinite articles from the numeral ‘one’. Demonstratives that develop into
definite articles usually start out as mainly spatial deictic markers (exophoric deixis) which then
extend to the marking of discourse anaphora (endophoric deixis). The next functional extension
is to definite contexts based on uniqueness, i.e. to referents that are the only salient referents of
their kind in the discourse situation without exophoric or endophoric deixis. I define the cut-off
point between (adnominal) demonstratives and articles in the following way: markers that en-
code exophoric deixis only or extend to endophoric deixis are treated as demonstratives. Markers
that encode endophoric deixis (discourse anaphora) only or extend to uniqueness are viewed as
articles. Also, markers that encode all three functions of exophoric and endophoric deixis as well
as uniqueness-based identifiability are treated as articles. Accordingly, the distinction between
the numeral ‘one’ and indefinite and specific articles is based on the scale from the functions of
individuation, presentational marking, specific marking, and nonspecific marking. Although this
scale is usually taken to describe the diachronic development of indefinite articles, it can be used to
delimit the numeral from articles: I argue that numerals generally have the potential to assume the
function of a presentational marker, i.e. to occasionally mark a referent as non-identifiable. This
is not counted as an article yet. If the marker is used consistently to signal the non-identifiability

4
of the referent, even though it is restricted by discourse prominence or information structure, it is
treated as an article (presentational article). Another article type, nonspecific articles, also needs
to be delimited from a subgroup of elements usually called NPIs (negative polarity items). I argue
that even though nonspecific articles and certain NPIs have the same referential functions, NPIs
lead to pragmatics effects that nonspecific articles do not have, which manifests itself in the op-
tionality of NPIs in the sense that there always is a simpler construction with similar referential
functions available in the language (a bare noun or a noun with an indefinite article). This is not
the case for nonspecific articles; they are used systematically in nonspecific contexts.

Chapter 4: Articles and referentiality In Chapter 4, I define the referential functions, called
referent types, that form the basis for the distinction of different article types.
In order to define the referent types, the following notions related to discourse are necessary:
the discourse-participants which are the hearer and the speaker, the discourse situation, corre-
sponding to the setting in which the discourse takes place and comprising the speaker, hearer,
and the utterance itself. To distinguish forms (nouns) from their meanings (referents), referents
are defined as elements of a different level in the discourse, which is called the discourse universe.
Thus, all referring words from the discourse situation have a corresponding referent in the dis-
course universe. I model the latter as a set of all referents that belong to the discourse in question,
irrespective of their type. I use the notion of identifiability as a basis for the distinction of dif-
ferent referent types. Identifiability can be defined as the discourse participants’ knowledge and
commitment to this knowledge about the referent.
An overview, summarizing the article types that I distinguish based on the referent types that
they encode, is provided in Table 1.1. The referent types are listed and briefly described in the
different rows of Table 1.1, while the article types are represented to the left of the Table with
arrows indicating their functional domains.

5
 Table 1.1: Summary of referent types and article types

Rdei A deictic referent Rdei is identifiable based on its unambiguous

link to an object in the discourse situation.
Do you see the house over there?
recog Rrecog A recognitional referent Rrecog is identifiable based on shared
previous experience or (common) knowledge.
What happened to that dog (we used to have)?
Rabs.u An absolutely unique referent Rabs.u is identifiable because it is
the only referent of its kind.
The Earth is round.

definite domain
Rana An anaphoric referent Rana is identifiable based on shared iden-
def
ana

tity with a previously mentioned referent.

What did you think about the movie last night? Wasn’t it great?
inspec

Rbrid A bridging referent Rbrid is identifiable based on its unambigu-

ous link to another previously mentioned referent.
ref

Eva bought a book. The author is French.

Rsit.u A situationally unique referent Rsit.u is identifiable because it is
the only salient referent of its kind in the discourse situation.
What’s the best way to the center?
Rest An establishing referent Rest is marked as identifiable and thus
constructed as such.
Did you hear the news? They are going to close the museum.
A specific referent Rspec is not identifiable but linked to a partic-

indefinite domain
Rspec
exspec

ular referent of its kind.

indef/pres

I met a strange new neighbour yesterday.

Rnspec A nonspecific referent Rnspec corresponds to a single, but no par-
nspec

ticular (any), referent of its kind.

Do you have a pen? Any pen will do.
Rgen A generic referent Rgen corresponds to the set of all referents of
its kind.
I like cats.

The article types that I distinguish are: recognitional (rec), anaphoric (ana), definite (def), inclusive-
specific (inspec), exclusive-specific (exspec), nonspecific (nspec), indefinite (indef), presenta-
tional (pres), and referential (ref) articles. The difference between indefinite and presentational
articles is that presentational articles are restricted to discourse prominent referents, while indef-
inite articles are not. Even though the articles that mark referents from the indefinite domain
may code all different subtypes of definite referents, I only treat the marking of certain definite
referent types as necessary criterion for definite, inclusive-specific, and referential articles. These
referents are anaphoric and situationally unique referents. The various definite referent types

6
form the definite domain, while the indefinite domain consists of the specific and the nonspecific
referent, as is indicated by the curly brackets on the right side of Table 1.1.

Chapter 5: Article types – examples from the world’s languages In Chapter 5, I discuss exam-
ples from various languages of the world for the 9 article types.
For definite articles, I show that certain subtypes of definite referents can but need not be
expressed by many definite articles from the languages of the world. This concerns deictic refer-
ents Rdei and absolutely unique referents Rabs.u . In addition, for many definite articles, the data
provided in the descriptions does not cover their use for establishing, recognitional, and bridging
referents Rest , Rrecog , and Rbrid . Thus, rather for practical than linguistic reasons, the expression
of these definite referent types was not taken as a necessary criterion for definite articles either.
Hence, a definite article is considered as such if it is used to mark anaphoric and situationally
unique referents Rana and Rsit.u .
As for articles from the indefinite domain, I propose a distinction between indefinite and pre-
sentational articles. Both encode specific and nonspecific referents Rspec and Rnspec , but presenta-
tional articles are more restricted to discourse prominent positions. Discussing nonspecific arti-
cles, I show that they can often be traced back to irrealis markers in the verbal domain that extend
to the marking of nominal predicates in irrealis contexts and may develop into nonspecific articles.

Chapter 6: Morphological properties of articles Chapter 6 discusses the morphological prop-

erties of articles, focusing on the inflection of articles and selected phenomena concerning arti-
cle paradigms. Articles across the world’s languages mainly inflect because they agree with the
noun in gender (including animacy), number, and/or case. In addition, articles can morphologi-
cally mark other categories, which is not attested very frequently and mainly related to deixis. A
look at the crosslinguistic distribution of article inflection reveals that the inflection patterns are
largely independent of the macroarea, of the position of the article, and of the article type. I show
that there are two factors that influence the inflection patterns of articles, namely the inflectional
morphology on the noun that the article occurs with and the inflectional properties of the article’s
source element.
As for the inflection of the noun, we see that frequent number and case marking on noun
negatively correlates to their marking on articles: the latter often inflect for gender, but less so
for number and case. Importantly, all the three categories are rarely marked on both the article
and the noun. What this suggests is that there is a stable trend against inflection on both the
noun and the article for these categories, and it accounts for the crosslinguistic trend against
marking number and case on articles. In addition, I compare the inflectional properties of articles
with those of their source elements (demonstratives, pronouns (personal and possessive), and

7
the numeral ‘one’). I show that articles mostly retain the inflectional properties of their source
elements and that they can lose inflectional distinctions, but almost never show innovations in
their inflectional behaviour. Thus, article inflection is additionally restricted by the inflectional
behaviour of its source element.
The second part of this chapter deals with article paradigms. I discuss the paradigm of the
definite article in Mokpe which is an example for a particularly complex paradigm. It illustrates
how different noun classes can trigger the article to surface as a segmental or as a tonal article
marker, but that they have to be two exponents of a single definite article because of their use in
the same referential contexts.

Chatper 7: Crosslinguistic variation and universals The first part of Chapter 7 shows areal
trends with regard to different article types. Most unexpected perhaps is the high number of
anaphoric articles in Australia. I argue that it may be related to the absence of third person
pronouns that are formally distinct from demonstratives in many Australian languages.
In the second part, I discuss a rarely attested and an unattested article type: non-anaphoric
definite articles and generic articles. I propose two scenarios that can lead to the development
of non-anaphoric definite articles and argue that most of the reported examples are not non-
anaphoric articles in the strict sense. Generic articles, on the other hand, are not attested. I show
a number of examples of other article types that can be used to mark generic referents.
On the basis of the discussion in the previous chapters and on the basis of the 99-languages
sample, I propose a typology of articles and article systems, according to the number of referent
types that are encoded. The types of article systems represent the possible combinations of articles
that encode one, two, or three referent types. I show and discuss the frequency distributions of
the systems and attested gaps. This motivates five universals: The first two universals concerns
two types of articles that are not attested: generic articles and articles that coexpress definite and
nonspecific referents without encoding specific referents. The other universals apply to article
systems. Nonspecific articles are only attested in article systems in which specific referents are
also encoded, systems of articles with overlapping functions are avoided, and article systems that
single out definite referents are more frequent in the languages of the world that single out specific
or nonspecific referents.
The last part deals with the interaction of articles and a number of constructions in which
they are not required as referential markers: definite articles with superlatives, different types of
articles with plural nouns, and definite articles with demonstratives and possessives.
The presence of definite articles in superlative constructions is due to their semantic compat-
ibility, and corresponds to the strongest semantically motivated use of articles outside of their
referential functions. However, I argue that the compatibility should rather be viewed as strong

8
correlation instead of an inherent property of definite articles. The other two contexts show that
the source of the article can result in incompatibilities of the article with certain contexts and de-
termine its distribution on the synchronic level. I show that the trend for indefinite articles to be
incompatible with plural nouns is mostly due to their frequent origin from the numeral ‘one’. It is
thus not a property of the article itself that blocks its use, but the properties of its source element.
The same can be shown for the use of definite articles together with adnominal demonstratives
and possessives. Definite articles and demonstratives only show a trend against co-occurrence
if they are on the same side of the noun. This can be accounted for by the fact that definite ar-
ticles often originate from demonstratives. At least in the initial stages of the emerging definite
article, they are formally and functionally indistinguishable in deictic contexts which can account
for the co-occurrence restriction in these cases. For article-demonstrative pairs that do not share
this twofold overlap in form and function, the co-occurrence restriction is less motivated, which
explains why we do not see any co-occurrence effect for article-demonstrative pairs that occur on
either sides of the noun. In accordance with this, possessives, which develop into definite articles
less frequently, do not show any effect either.

1.3 Data and methodology

The major part of this study is based on a sample of 99 languages with 131 articles in total. Since I
only included languages that have articles, the sample is a convenience sample. Nevertheless, the
sample contains between 12 and 20 languages for each of the six macroareas introduced in the
World Atlas of Language Structure and glottolog (Dryer & Haspelmath 2013, Hammarström et al.
2018).1 The macroareas are: Africa, Eurasia, Papunesian, Australia, North America, and South
America. The languages in my sample are provided in Tables 1.3 to 1.8 at the end of this section,
sorted by macroareas, language families, and listed together with the main source(s). Figure 1.1
shows the areal distribution of the languages in my sample. Each dot represents a language, the
different colors refer to the macroareas.2
1 Thisdivision into six macroareas is not without its downsides, as discussed in Hammarström & Donohue (2014).
Because my sample is a convenience sample constrained by the availability of articles in the world’s languages, and
because areal effects are not the main interest of this study, this issue can be neglected.
2 The map was created using the R package “lingtypology” (Moroz 2017) which imports the data from glottolog (Ham-

marström et al. 2018).

9
 Figure 1.1: Areal coverage of the sample

The two major areas that the sample does not cover are most parts of Russia and Canada. This is
due to the general lack of languages with articles in these areas. The areal distributions of definite
and indefinite articles observed in the WALS chapter on definite articles (Dryer 2013a) show this
trend as well. Figure 1.2 below features the map that belongs to this chapter.3 Blue and red dots
indicate definite articles, white diamonds correspond to languages with an indefinite article but
no definite articles, and white dots stand for languages without articles. Without going into detail
with respect to the different types of articles marked in the map in Figure 1.2, we see that also
based on a larger sample, Russian and Canada seem to lack languages with articles. Another area
in WALS indicating a trend against articles is South America.

Figure 1.2: Areal distribution of definite and indefinite articles (Dryer 2013a)

It also follows from this unequal distribution of articles in the world’s languages that the macroar-
eas do not contain an equal number of languages; South America and Australia are represented
3 The map is available online at http://wals.info/feature/37A.

10
by slightly less languages than the other areas. Since this most probably reflects the natural dis-
tribution of articles in the world, it is not problematic for the purposes of the present study.
For large language families, the sample includes a higher number of languages.4 This is the
case for Niger-Congo, Indo-European, Austronesian, and Pama-Nyungan. Since the different lan-
guages from these families are generally diverse, especially with respect to their article related
properties, it should not bias the sample with regard to the purposes of the present study. My
sample also includes a number of languages from the Arawakan family. The latter is not as large
in terms of numbers of languages as the other families mentioned above, but since Arawakan
languages are mostly spoken in South America, an area with a relatively low number of articles,
I included a higher number of languages from this family due to the lack of alternatives.
In certain cases, I included more than one language from the same genus. Genera with more
than one languages in my sample are: Core Siouan (Crow and Lakota), Western Pama-Nyungan
(Martuthunira and Yankuntjatjara), Oceanic (Maori, Rapa Nui, Siar Lak, Sye, and Tongan), Semitic
(Hebrew, Amharic, and Neo-Aramaic), Kwa (Akan and Logba), as well as Bantu (Limbum, Mokpe,
Bemba, and Runyankore). I included the two Western Pama-Nyungan languages, because the
languages in Australia with articles are relatively rare. Since the Pama-Nyungan family is one of
the largest language families in the world, it is not necessarily over-represented in my sample by
including more than one language. The same reasoning applies to the Oceanic, Bantu, and Indo-
European languages in my sample. In addition, most of these families cover a large geographical
area. Other genera with more than one language include Semitic (3 languages). These languages
have very different properties with respect to their article systems and are not in contact. The rea-
son for including two Siouan languages is that they both have nonspecific articles which are rare
in the languages of the world. Thus, in favor of a better understanding of certain rare article types
(the two article systems are not entirely identical, though), the sample includes these two closely
related languages at the expense of a higher level of balance. Another factor that determined the
choice of languages was to include as many recent grammars as possible. The methodology for
language documentation and grammatical descriptions has made considerable advances over the
last decades; recent descriptions often provide more examples and better glosses, leading to more
transparency in the primary sources.
In addition to the data from the sample described above, I also discuss data that I collected.
Table 1.25 gives an overview of the languages, the number, age, and sex of the consultants, as well
as the locations and dates. The basis for the data collection was a questionnaire that is provided
in Appendix B in its English version. For the data collection from Tikuna, Kaqchikel, Q’anjobal,
Awakateko, and Tz’utujil, I used a corresponding questionnaire in Spanish. The contexts from the
4 Large is meant here as containing a large number of languages (cf.
http://glottolog.org/glottolog/family).
5I thank my colleagues Jude Nformi and Sampson Korsah for their help as informants and providing me with data
from Limbum and Akan.

11
questionnaire where adapted as fit and depending on the articles found in the different languages,
I added contexts for selected areas. In the following Chapters, I indicate primary data as “prim.
data” in the examples. From the languages for which I have primary data, Mokpe, Limbum, and
Akan are also part of the 99 languages sample. Most other languages, i.e. Macedonian, Kaqchikel,
Q’anjobal, Awakateko, and Tz’utujil are not part of the sample but serve as illustrations in various
parts of the study. The two remaining languages Ejagham and Tikuna do not have articles; I use
data from these languages mainly in the first part of the study to illustrate certain other relevant
properties.
The presentation of examples from languages other than English follow the Leipzig glossing
rules.6 All abbreviations used in the glosses are given in the list of abbreviations. To indicate that
the presence of a certain marker is obligatory in a given context, it appears in round brackets with
an asterisk outside of the brackets: *(marker). To indicate that a certain marker is infelicitous in
a given context, it appears in round brackets with an asterisk inside of the brackets: (*marker).
Since this study includes various examples of articles from different languages that occur on either
sides of the nouns, I use square brackets whenever they are helpful to indicate which noun the
article belongs to, e.g. [noun article]. Note that this is only a matter of presentation and not
intended to relate to any particular syntactic structure.
6 Available at https://www.eva.mpg.de/lingua/pdf/Glossing-Rules.pdf.

12
 Table 1.2: Primary data collection
language location time informants
Mokpe Buea (Cameroon) Feb 2017 female, 21
male 26
Ejagham Buea (Cameroon) Feb 2017 male, 28
Limbum Buea (Cameroon) Feb 2017
male, 29
Leipzig (Germany) Feb 2018
Akan Leipzig (Germany) Jun 2017 male, 31
Macedonian Leipzig (Germany) Apr-Jun 2017 female, 26
Tikuna Bogotá (Colombia) Oct 2017 male, 28
male, 25
Kaqchikel Patzún (Guatemala) Nov 2017 female, 36
female, 27
female, 65
Q’anjobal Santa Eulália (Guatemala) Nov 2017 male, 65
male, 41
Awakateko Aguacatán (Guatemala) Nov 2017 female, 65
female, 40
Tz’utujil San António (Guatemala) Dec 2017 female, 35
male, 38

13
 Table 1.3: Languages from the macroarea “Africa”
language genus country reference
Afro-Asiatic
1 Amharic Semitic Ethiopia (Leslau 1995)
2 Dime South Omotic Ethiopia (Seyoum 2008)
3 Hausa West Chadic Nigeria, Niger (Newman 2000)
4 Konso East Cushitic Ethiopia (Orkaydo 2013)
5 Moloko North Biu-Mandara Cameroon (Friesen 2017)
6 Sheko Dizoid Ethiopia (Hellenthal 2010)
Benue-Congo
7 Oko Oko-Eni-Osayen Nigeria (Atoyebi 2010)
Dogon
8 Jamsay Dogon Mali (Heath 2008)
Eastern Sudanic
9 Gaahmg Eastern Jebel Sudan (Stirtz 2011)
10 Lango Nilotic Uganda (Noonan 1992)
Mande
11 Beng Eastern Mande Côte d’Ivoir (Paperno 2014)
Niger-Congo
12 Akan Kwa (Nyo) Ghana (Amfo 2010)
(Arkoh 2011),
(Arkoh & Matthewson 2013),
Sampson Korsah, pc
13 Bemba Bantu (Sabi) Zambia (Givón 1969)
14 Bullom So Mel Guinea, Sierra Leone (Childs 2011)
15 Limbum Bantu (Narrow Grassfields) Cameroon (Fransen 1995),
Jude Nformi, pc
16 Logba Kwa (Na-Togo) Ghana (Dorvlo 2008)
17 Mokpe Bantu (Sawabantu) Cameroon primary data,
(Atindogbe 2013)
18 Runyankore Bantu (East Bantu) Uganda (Asiimwe 2014)
19 Supyire Gur Mali (Carlson 1994)
Sandawe
20 Sandawe Sandawe Tanzania (Steeman 2011)
Songhay
21 Koyra Chiini Songhay Mali (Heath 1999)

14
 Table 1.4: Languages from the macroarea “Eurasia”
language genus country reference
Afro-Asiatic
1 Hebrew Semitic Israel (Glinert 2005)
(Coffin & Bolozky 2005)
2 Neo-Aramaic Semitic Iraq (Khan 2008)
Altaic
3 Bonan Mongolic China (Fried 2010)
4 Turkish Turkic Turkey (Göksel & Kerslake 2005)
Austro-Asiatic
5 Kharia Munda India (Peterson 2011)
Austronesian
6 Sri Lanka Malay Malayo-Sumbawan Sri Lanka (Nordhoff 2009)
Dravidian
7 Tamil Southern Dravidian India (Lehmann 1993)
Indo-European
8 Albanian Albanian Albania (Newmark et al. 1982)
(Buchholz & Fiedler 1987)
(Zymberi 2004)
9 Armenian Armenian Armenia (Dum-Tragut 2009)
(Andonian 2005)
10 Dameli Iranian Pakistan (Perder 2013)
11 Domari Indic Jordan, Iran, Isral (Matras 2012)
12 German Germanic Germany (Hentschel & Weydt 2003)
13 Irish Celtic Ireland (Stenson 2008a)
(Stenson 2008b)
14 Palula Dardic Pakistan (Liljegren 2016)
15 Persian Iranian Iran (Mace 2003)
16 Rajbanshi Iranian Bangladesh, India (Wilde 2008)
Isolate
17 Basque Isolate Spain, France (de Rijk 2008)
(Hualde & Ortiz de Urbina 2003)
Nakh-Dagestanian
18 Aghul Lezgic Russia (Ganenkov et al. 2009)
Sino-Tibetan
19 Ao (Mongsen) Kuki-Chin India (Coupe 2007)
20 Lepcha Lepcha Bhutan, India, Nepal (Plaisir 2006)
Uralic
21 Hungarian Ugric Hungary (Kenesei et al. 1998)

15
 Table 1.5: Languages from the macroarea “Papunesia”
language genus country reference
Austro-Asiatic
1 Papuan Malay Malayo-Sumbawan Papua New Guinea (Kluge 2017)
Austronesian
2 Agta Greater Central Philippine Philippines (Robinson 2008)
(Dupaningan)
3 Biak South Halmahera Indonesia (van den Heuvel 2006)
4 Maori Oceanic (Eastern Polynesian) New Zealand (Harlow 2012)
5 Rapa Nui Oceanic (Centr. East. Polynesian) Chile (Kieviet 2017)
(Weber 2003)
6 Siar Lak Oceanic Papua New Guinea (Rowe 2005)
(New Ireland-Northw. Solomonic)
7 Sye Oceanic (Southern Melanesian) Vanuatu (Crowley 1998)
8 Tongan Oceanic (Nuclear Polynesian) Tonga (Völkel 2010)
(Churchward 1985)
(Morton 1962)
9 Yapese Yapese Micronesia (Yap) (Jensen 1977)
Oksapmin
10 Oksapmin Oksapmin Papua New Guinea (Loughnane 2009)
Sepik
11 Alamblak Sepik Hill Papua New Guinea (Bruce 1979)
Skou
12 Skou Western Skou Indonesia (Donohue 2004)
Solomons East Papuan
13 Lavukaleve Lavukaleve Solomon Islands (Terrill 2003)
Timor-Alor-Pantar
14 Abui Greater Alor Indonesia (Kratochvı́l 2007)
Trans-New Guinea
15 Menya Angan Papua New Guinea (Whitehead 2006)

16
 Table 1.6: Languages from the macroarea “Australia”
language genus country reference
Bunuban
1 Gooniyandi Bunuban Australia (McGregor 1990)
Gunwinyguan
2 Wubuy Nunggubuyu Australia (Heath 1984)
Mangarrayi-Maran
3 Mangarrayi Mangarrayi Australia (Merlan 1989)
Maningrida
4 Guragone Bureran Australia (Green 1995)
Mirndi
5 Jingulu Djingili Australia (Pensalfini 1997)
Nyulnyulan
6 Bardi Nyulnyulan Australia (Bowern 2004)
7 Warrwa Nyulnyulan Australia (McGregor 1994)
Pama-Nyungan
8 Arrernte (Mparntwe) Central Pama-Nyungan Australia (Wilkins 1989)
9 Diyari Central Pama-Nyungan Australia (Austin 2011)
10 Duuŋidjawu Southeastern Pama-Nyungan Australia (Kite & Wurm 2004)
11 Martuthunira Western Pama-Nyungan Australia (Dench 1994)
12 Yankuntjatjara Western Pama-Nyungan Australia (Goddard 1985)
13 Yolŋu Yolŋu Australia (Wilkinson 1991)
Yangmanic
14 Wardaman Yangmanic Australia (Merlan 1994)

17
 Table 1.7: Languages from the macroarea “North America”
language genus country reference
Algic
1 Blackfoot Algonquian Canada,USA (Taylor 1969)
(Frantz 1991)
Hokan
2 Chimariko Chimariko USA (Jany 2009)
Mayan
3 Ch’ol Mayan Mexico (Vázquez Alvarez 2011)
Misulmalpan
4 Sumu Misumalpan Nicaragua (Norwood 1997)
Mixe-Zoque
5 Zoque Mixe-Zoque Mexico (Johnson 2000)
Muskogean
6 Koasati Muskogean USA (Kimball 1985)
Oto-Manguean
7 Chatino (Teotepecan) Zapotecan Mexico (McIntosh 2011)
Salishan
8 Halkomelem Central Salish Canada (Suttles 2004)
Siouan
9 Crow Core Siouan USA (Graczyk 2007)
10 Lakota Core Siouan USA (Van Valin 1977)
(Williamson 1984)
(Rood & Taylor 1996)
(Ingham 2003)
Totonacan
11 Tepehua Totonacan Mexico (Kung 2007)
Uto-Aztecan
12 Cupeño California Uto-Aztecan USA (Hill 2006)
13 Tarahumara Tarahumaran Mexico (Cohen 1998)
14 Ute Numic USA (Givón 2011)
Wakashan
15 Nuuchahnulth Southern Wakashan Canada (Davidson 2002)
Xinkan
16 Xinkan Xinkan Guatemala (Rogers 2010)

18
 Table 1.8: Languages from the macroarea “South America”
language genus country reference
Arawakan
1 Arawak Caribbean Arawakan Suriname (Pet 2011)
2 Asheninka Perené Pre-Andine Maipuran Peru (Mihas 2010)
3 Bare Inland North. Arawakan Brazil, Venezuela (Aikhenvald 1995)
4 Baure Bolivia-Parana Bolivia (Danielsen 2007)
5 Parecis Central Arawakan Brazil (Barros Brandão 2014)
6 Wayuu Wayuu Colombia (Álvarez 2017)
Cariban
7 Carib Cariban Suriname (Courtz 2008)
Choco
8 Epena Pedee Choco Colombia (Harms 1994)
Macro-Ge
9 Apinayé Ge-Kaingang Brazil (Cunha de Oliveira 2005)
10 Karajá Karajá Brazil (Ribeiro 2012)
Nambikwaran
11 Sabanê Sabanê Brazil (Antunes de Araujo 2004)
Tucanoan
12 Cubeo Tucanoan Colombia (Chacon 2012)

19
Chapter 2

Articles as a crosslinguistic category

What is an article and how might articles be described in a maximally general way such
that the difference between a language that has these elements and one that does not can be
characterized in a meaningful manner? (Moravcsik 1969: 85)

2.1 The problem

Although the question of how articles can be defined for crosslinguistic comparison is not a new
one, we still lack a good answer to it. This might be due to the fact that especially in European
languages, identifying articles is not a very complicated matter. With no established tradition of
what counts as an article and what does not, however, the classification of markers from other
languages is often less obvious. A simple example of a well-established article system is shown
in (1), illustrating the definite and indefinite articles in English in opposition to each other.

(1) a. The woman got off the train.

b. The woman got off a train.

Glossing over the details for now, we can say that the definite article in (1a) marks the referent of
train as identifiable, known, given, or accessible, and also as prominent in the scenario described,
while the indefinite article in (1b) marks it as not uniquely identifiable and also as less prominent.
We find this kind of opposition of adnominal markers that encode identifiability or non-identi-
fiability of discourse referents in many European languages. Although there are crucial differ-
ences in the actual distributions of these articles in different languages, generally, we are still im-
posing the labels and characteristics of a few well-known article systems to other, less described
languages. Therefore, it is not surprising that markers of such languages are often not regarded
as articles. This is problematic, since without having a clear picture of crosslinguistically relevant

20
properties of articles, we do not know which properties are category-defining and which ones are
less central, and we might end up excluding and including markers for no good reason.
In the following paragraphs, I will present a number of examples that illustrate this problem.
Examples (2) and (3) below contrast two functionally similar if not equivalent expressions: in the
case of German, we would traditionally label ein as an indefinite article; the equivalent Chinese
expression consisting of the numeral ‘one’ and a classifier has similar referential functions in this
context, but one may not regard yi zhuang in (3) as an indefinite article.

(2) Ich habe letztes Jahr ein Haus gekauft.

I have last year art:indef house bought
‘I bought a house last year.’ German (prim. data)
(3) ta qunian mai le [(yi) zhuang fangzi]
3sg last.year buy pfv one cl house
‘She/he bought a house last year.’ Chinese (Chen 2003: 1171)

In the case of Chinese, it is not as clear as for German whether this is a construction that functions
as indefinite article or not, and whether this is enough to label it as such. In the literature, this
construction with yi ‘one’ and a classifier has been argued to be used in various contexts corre-
sponding to those of indefinite articles in other languages (cf. Chen 2003). It is even attested in
predicative contexts which are no longer referential:

(4) ta shi [(yi) ge maimairen]

3sg be one cl businessman
‘She/he is a businessman.’ Chinese (Chen 2003: 1171)

Another example of a marker whose classification is problematic is the possessive marker in a

number of Uralic languages, which has functions very similar to definite articles. This is what we
find in, for instance, Udmurt: although still recoverable as third person singular possessive marker,
-ez, was argued to indicate definiteness even in contexts with no possessor being semantically
present or recoverable (e.g. Gerland 2014):1

(5) so-len eš-ez

he-gen friend-poss:3sg
‘his friend’ Udmurt (Edygarova 2009: 101)
1 I do not discuss here whether or not this possessive marker in Udmurt is an article. In general, there is no consensus
to what extent possessives in Uralic classify as definite articles or express definiteness based on possession. Fraurud
(2001), Nikolaeva (2003) argued against an account as a definiteness marker because they take the marking to be
optional, while Csúcs (2003), É. Kiss & Tánczos (2017), Gerland (2014) base their analyses on possessives as indicators
of definiteness and analyse definiteness marking as generally required but allowing for exceptions.

21
(6) piosmurt-ez baš’t-iz baton
guy-poss:3sg take-pst.3sg loaf
‘The guy took a loaf.’ Udmurt (Arkhangelskiy & Usacheva 2016)
(7) guždur vïl-ïn turïn-ez čeber
clearing surface-iness grass-poss:3sg beautiful
‘The grass in the clearing is beautiful.’ Udmurt (Csúcs 1998: 285)

In (5), we see -ez in its original function, marking the possessor on the possessum eš ‘friend’. In
(6) and (7), on the other hand, the context does not feature a potential possessor, and the same
marker -ez is used to indicate that the referents guy and grass are marked as identifiable and hence
as definite. Again, this shows that one needs language-independent criteria to classify a marker
as an article.
Another entirely different phenomenon that touches upon the definition of articles as a crosslin-
guistic category is usually referred to as “nominal tense”. This is found in different languages of
South America (i.e. Arawakan, Cariban, Movima, Matacoan, Tupi-Guarani) and refers to temporal
and/or deictic marking by grammatical elements in the nominal domain. In some cases, nominal
tense combines with temporal marking on the verb, and sometimes it is the only grammatical
indicator for temporal relations (Campbell 2012: 285). What is relevant for articles and definite-
ness is that in some languages, these temporal markers on the noun can lead to an anaphoric or
nonspecific interpretation of the referent. Example (8) from Nivaclé (Matacoan, Paraguay) shows
how different past markers on the noun mark the referent as anaphoric and thus as definite. In
(8a), -naxi indicates yesterday’s past and -mati is used in (8b) to mark today’s past.

(8) a. mansana Ø-tolu Ø-łile-naxi hohnat wit hi-kwes

apple 3-come.from poss:3-tree-pst2 ground conj 3-split
‘The apple fell from the tree (that we saw yesterday) and split.’
b. sinox-mati atana Ø-yił-łi
dog-pst1 now 3-die-iter.sg
‘The dog (from earlier today) is sick now’.
Nivaclé (Campbell 2012: 287)

Since these two markers have additional referential functions, encoding the referent as anaphoric,
they may not fundamentally differ from markers that we treat as articles in other languages.
The last three examples from Chinese, Udmurt, and Nivaclé showed that it is not sufficient to
take the properties of the “European” article as a comparative concept of articles. This is especially
important if one takes into account other non-referential properties of (European) articles as well.
For instance, we find definite articles in generic contexts (which do not involve definiteness) in
some languages, but in other languages, articles are not required in such contexts, leading to

22
variation even within European languages. Example (9) and (10) show German and English, which
do not use a definite article with a generic referent in the plural, while (11) and (12) show that
Spanish and Hungarian require a definite article:

(9) I like (*the) cats. English (prim. data)

(10) Ich mag [(*die) Katzen].
1sg like.1sg art:def cats
‘I like cats.’ German (prim. data)
(11) Me gustan [los gatos].
1sg.dat please.3pl art:def cats
‘I like cats.’ Spanish (prim. data)
(12) Szeretem [a macskákat].
like.1sg art:def cats
‘I like cats.’ Hungarian (prim. data)

There are two main problems in the perception of articles. On the one hand, we see that even in
European languages, the distribution of articles varies, and a distributional criterion for articles
must take this into account. On the other hand, the fact that definite articles occur with generic
referents also shows that the use of articles is not necessarily restricted to their referential values.
The aim of this chapter is to identify the properties of articles that define them as a crosslin-
guistic category, i.e. not as a language-specific category. In section 2.2, I present the criteria that
define articles. I argue that we need three main criteria that specify their function as a referen-
tial marker, their domain as restricted to the noun phrase, and their distribution as systematic.
Sections 2.2.2, 2.2.3, and 2.2.4 present the three criteria, respectively.

2.2 Defining articles

2.2.1 Introduction
A suitable approach to define articles for crosslinguistic comparison is the one of comparative con-
cepts, proposed in (Haspelmath 2010):2 Comparative concepts “are not part of particular language
systems and are not needed by descriptive linguists or by speakers. They are not psychologically
real, and they cannot be right or wrong. They can only be more or less well suited to the task
of permitting crosslinguistic comparison […] Comparative concepts are universally applicable,
2 Thecomparability of linguistic categories (mostly syntactic categories, syntactic operations based on syntactic cate-
gories, and parts of speech or lexical categories) across languages is not a new topic and was extensively discussed
by a number of authors (e.g. Cristofaro 2008, Croft 1991, Culicover 1999, Dryer 1997, Haspelmath 2008, Lazard 2005,
Newmeyer 2007, Slobin 1997, 2001, Song 2001, Stassen 1985, Wierzbicka 1998).

23
and they are defined on the basis of other universally applicable concepts: universal conceptual-
semantic concepts, general formal concepts, and other comparative concepts.” (Haspelmath 2010:
665). In the present study, I will adopt this approach to define a crosslinguistic category of articles.
That we are still in need of a detailed discussion is shown by previous definitions of articles. For
instance, even though the definition of articles in Dryer (2014: 234) is similar to what I propose
in this study, it is formulated in rather vague way.

[…] I use the term ‘article’ in a somewhat nonstandard sense, one that is more semantic than
common uses of the term. Specifically, although I restrict it to words or morphemes that occur
in noun phrases, I otherwise make no assumptions as to the syntactic status of the words or
morphemes in question. Namely, they must code something in the general semantic domain
of definiteness or indefiniteness […].

A different example comes from Lyons (1999). In his approach to articles, he emphasizes that the
function of articles should not be restricted to the marking of (in)definiteness.

We will see, however, that articles can encode more than definiteness or indefiniteness, and
that they have been argued to have a quite different principal function, at least in some lan-
guages. (Lyons 1999: 6)

In either cases, articles are defined (or characterized) based on their referential functions. How-
ever, the latter are formulated in a vague way that does not make it explicit how they can be
applied to markers in practice.
In contrast to these two definitions that do not distinguish between different types of articles,
Schroeder (2006) provides two separate definitions for definite and indefinite articles, respectively,
also referring to their formal and functional properties:

Definite articles will be defined as a morphological device (free morpheme, enclitic morpheme,
or affix) with the primary function to denote the definiteness of noun phrases with anaphoric
and/or nonanaphoric reference. […]
“Indefinite article” will be defined as a morphological device (free morpheme, enclitic mor-
pheme, or affix) with the primary function to denote the indefiniteness of noun phrases.
(Schroeder 2006: 553)

Since no explicit and applicable definition of articles seems to be available, the goal of this section
is to define articles as a crosslinguistic category following the approach of comparative concepts.
The definition of articles as a crosslinguistic category is based on criteria that can be applied to
any marker in any given language, which allows us to discuss and compare articles in the world’s
languages. I propose three main criteria to define articles. The first criterion is their function:

24
articles are defined as referential markers. The second criterion refers to their domain in that
articles have to occur together with nouns. The third criterion is concerned with the distribution
of articles and requires them to occur systematically (allowing for exceptions) in the contexts
matching their referential functions. The definition of articles is given in (13).

(13) Articles as a crosslinguistic category

An article is a marker that systematically occurs with a noun and whose primary function
is to indicate the referential function of the noun.

In the following sections, I will discuss these three criteria in detail with examples from different
languages.

2.2.2 Articles encode referent types

Articles encode the referential functions of the noun that they occur with. I call these functions
referent types. In this section, I give an overview of the different referent types that articles can
mark, and I provide examples for the main article type in this section (for a detailed discussion of
referent types and article types, see Chapter 4).
We can distinguish four main referent types: definite, specific, nonspecific, and generic refer-
ents (for details, see Chapter 4). Definite referents are unambiguously identifiable by the speaker
and the hearer; there are different subtypes of definite referents which differ in the way identifi-
ability is established. Deictic definites are identifiable by ostension, i.e. physical perception. This
requires a link between the referent to an object that is present in the discourse situation:

(14) Do you see the house over there? (deictic)

Anaphoric referents, on the other hand, are fully identifiable by the speaker and the hearer due
to previous mentioning in the discourse. In (15), the referent of shoes is used anaphorically and
can be referred to by the pronoun they:

(15) A: Do you like my new shoes?

B: I think they look good on you. (anaphoric)

Similar to anaphoric contexts are bridging contexts which have also been labelled associative
anaphoras because the two referents do not share full identity but are linked to each other in an
unambiguous way. A typical example is the relation between a book and its author:

(16) I bought a book yesterday. The author is French. (bridging)

25
Another type of definite referents are recognitional definites. Such referents are unambiguously
identifiable by shared experience between the speaker and the hearer, or by common (personal or
cultural knowledge).

(17) What happened to that dog (we used to have)? (recognitional)

Unique referents represent another type of definite referents because they are the only salient
referent of their kind. This can either hold for the discourse situation; in this case the referent
is situationally unique as in (22), in which the referent of kitchen is situationally unique due to
the context of a flat. In (18b), on the other hand, the referent of moon is absolutely unique in our
default perception of the world, which also makes it definite.

(18) a. Where is the kitchen in your flat? (situationally unique definite)

b. The moon is beautiful tonight. (absolutely unique definite)

Another type of definite referents that can be distinguished is the so-called establishing referent.
Establishing referents go back to Hawkins (1978) and correspond to referents that are marked as
identifiable by the hearer, even though the referent is not yet identifiable at the time of utterance.
An example is given in (24).

(19) London has been buzzing with the rumour that the Prime Minister is going to resign.
(Hawkins 1978: 102)

Specific referents are no longer unambiguously identifiable by both the speaker and the hearer.
However, either of them can nevertheless link the expression to a particular referent of its kind.3
Nonspecificity, in contrast to definiteness and specificity, does not involve a particular referent
that is linked to the noun. Any referent of a given kind, existing or non-existing, can be linked
to the expression used. The restriction with respect to nonspecific referents is that there can only
be one referent linked to the expression, without any restriction on the choice of the referent
that is linked. In other words, the discourse situation does not provide enough information to
decide which of all the referents of a kind is linked to the noun. We typically find such referents
in interrogative, negative, irrealis contexts and with intensional verbs. A classic example of the
distinction between specific and nonspecific referents is provided in (20), taken from Givón (1984:
441) and going back to Quine (1953):
3 Wecan further distinguish two types of specific referents: referents that are only nonidentifiable for the hearer but
not for the speaker, and referents that are nonidentifiable for either discourse participants. Since this distinction
does not seem to be relevant for articles, I do not consider it in this study.

26
(20) John wanted to marry a rich woman …
a. …but she refused him. (specific)
b. …but he couldn’t find any. (nonspecific)

As English does not obligatorily mark the difference between specific and nonspecific referents on
the expression itself, the referent a woman, marked by the indefinite article, remains ambiguous
between a specific and nonspecific interpretation without further context. Such a contextual
disambiguation is provided in the continuations in (20a) and (20b). In (20a), the referent is specific;
even though the hearer might not be able to unambiguously identify the referent linked to a rich
woman, the only valid interpretation is that there is a single referent of a rich woman the speaker
refers to. In (20b), on the other hand, the referent is marked as nonspecific, since there is no single
referent linked to the expression a rich woman. Any referent, existing or nonexistent, that fulfills
the criteria of being a rich woman can be linked to the expression used, but the context implies
that neither the speaker nor the hearer have more information about which referent to link.
In other languages, this ambiguity does not arise because they formally distinguish between
specific and nonspecific referents. One example is Bemba (Atlantic-Congo, Zambia). As is shown
in (21a), the referent of ‘book’ is necessarily interpreted as specific because it is marked by i-,
which is a specific article (cf. section 5.6.1). If the referent is nonspecific, we see in (21b) that the
specific article i- does not occur.

(21) a. n-dee-fwaaya i-ci-tabo

1sg-prs-want art:spec-cl7-book
‘I want the/a (certain) book.’ (specific)
b. n-dee-fwaaya ci-tabo
1sg-prs-want cl7-book
‘I want a (any kind of) book.’ (nonspecific)
Bemba (Givón 1984: 65)

Generic expressions, similarly to nonspecific ones, do not refer to an unambiguously linked par-
ticular referent either. In contrast to nonspecific expressions, however, generic expressions make
reference to all referents that belong to a certain kind. Therefore, generics pattern with nonspe-
cific expressions in that there is no unambiguous link to any referent. On the other hand, generic
referents differ from all other referent types, since they do not make reference to a single referent
(whether unambiguously linked or not), but they refer to the entire set of referents that belong
to a certain kind. In other words, while definites, specifics, and nonspecifics refer to individuals,
generics are kind-referring:

(22) The peregrine falcon is the fastest animal on Earth. (generic)

27
The next paragraphs give a brief overview of the markers that we can classify as different types
of articles based on their referential functions. Here, I only present the major article types to
provide an overview. The article types are defined in Chapter 4 and presented with more detailed
discussions in Chapter 5.
Definite articles mark various subtypes of definite referents. Here, I only show a subset of the
functions of the well-known definite article in German (cf. section 5.2 for a detailed discussion
of definite articles). The examples below show that the definite article is used to encode different
types of definite referents, namely anaphoric (23), bridging (24), and recognitional referents (25),
respectively:

(23) S: Wollen wir heute Abend einen Film sehen?

want.1pl we today evening art:indef movie watch
‘What do you think about watching a movie tonight?’
H: Ich dachte, wir gehen ins Theater.
I thought.1sg we go.1pl in.art:def theatre
‘I thought we’d go to the theatre.’
S: Ich möchte heute zuhause zu bleiben. Und [der Film] soll wirklich
I want.1sg today at.home inf stay and art:def movie is.supposed really
gut sein.
good be
‘I’d like to stay at home tonight. And the movie is supposed to be really good.’
German (prim. data)
(24) Peter fuhr zur Arbeit. [Das Lenkrad] war eiskalt.
Peter drove to.art:def work art:def steering.wheel was freezing
‘Peter was driving to work. The steering wheel was freezing.’ German (prim. data)
(25) Kannst du dich noch an [die Katze] erinnern, die wir damals hatten?
can.2sg you refl still prep art:def cat remember rel we then had.1pl
‘Can you remember the cat we used to have?’ German (prim. data)

Another type of definite articles that needs to be distinguished is the anaphoric article. Anaphoric
articles are used to encode anaphoric definites only (cf. section 5.3). An example of such an article
from Bare (Arawakan, Brazil) is given in (26). The noun tʃiñu ‘dog’ occurs as bare noun when it
is mentioned for the first time in the discourse, but it appears with the anaphoric article da when
being mentioned again.

(26) tʃiñu yawi mehésa me-wát’uka [da tʃiñu] i-báraka

dog angry 3pl.want 3pl-beat art:ana dog 3sg.nf-run
‘The dog was angry, they wanted to beat the dog, it ran away.’ Bare (Aikhenvald 1995: 24)

28
Specific articles are also common in the world’s languages. There are two different types of specific
articles. In most cases, they are used to mark specific referents only and cannot encode definite
referents. I call these exclusive-specific articles. Such an article can be found in Crow (Siouan,
USA). Example (27) shows the exclusive-specific article, marking the referent of dakáak-kaata
‘bird’ as being a particular bird of its kind without being identifiable. As can be seen in (28),
the exclusive-specific article -m is not used with nonspecific referents which are marked by the
nonspecific article -eem instead. In (28), the referent linked to baláxxiikaashe ‘bow’ corresponds
to two bows that are nonspecific because they do not yet exist, so that no particular set of two
bows can be referred to. Example (29) shows that the exclusive-specific article -m cannot encode
definite referents, as Crow uses the definite article -sh in such contexts.

(27) dakáak-kaata-m húu-laa híi-ka

bird-dim-art:in.spec come-ss reach-decl
‘A bird came, it reached him.’ Crow (Graczyk 2007: 228)
(28) axée baláxxiikaashe dúup-eem4 alúutkaashe áappaa día-a-wa-ku-he
father bow two-art:nspec arrow along.with make-cont-1-give-aff-q
‘Father, will you please make me two bows and arrows?’ Crow (Graczyk 2007: 230)
(29) káalee-sh dáawi-kawe-h he-m
old.woman-art:def go.on-polit-imp say-ds
‘The old woman said, “go on”. Crow (Graczyk 2007: 239)

The second type of specific articles includes the marking of definite referents. I call these inclusive-
specific articles. These articles are crosslinguistically rather rare. The Bemba example presented
in (21a), repeated here as (30), showed such an inclusive-specific article, as it only marks that the
referent is not nonspecific, but it does not disambiguate further between a definite and a specific
referent.

(30) n-dee-fwaaya i-ci-tabo

1sg-prs-want art:spec-cl7-book
‘I want the/a (certain) book.’ Bemba (Givón 1984: 65)

Articles that exclusively mark a referent as nonspecific are relatively rare as well. We find such an
article in Q’anjobal (Mayan, Guatemala), which uses formally distinct articles to express specific
and nonspecific referents. As is shown in (31), a specific referent which is a particular referent
of its kind is marked by the specific article jun. If the referent is nonspecific, i.e. a single but
4 The segmentation of the specific and nonspecific markers in Crow is more complex than presented in (27) and (28).
The specific marker -m is added to the stem form of a noun, usually ending in the vowel -a. The nonspecific marker,
on the other hand, attaches to what Graczyk (2007: 30f) labels the “citation form” of the noun. The latter features
a different final vowel, mostly -ee. Since this nevertheless leads to a formal opposition in specific and nonspecific
contexts, I regard these markers as two distinct articles.

29
not particular referent of its kind, the marker junoq has to be used instead. As I will show in
section 5.7.1, junoq is a nonspecific article. An example for its use is the question in (32), also
showing that the specific article jun cannot be used to encode a nonspecific referent.

(31) ay [jun yatut thioxh] b’ay yich calle.

exist art:exspec house God at back street
‘There is a church at the end of the road.’ Q’anjobal (prim. data)
(32) ay mi [junoq/ *jun yatut thioxh] b’ay yich calle?
exist q art:nspec art:exspec house God at back street
‘Is there a church at the end of the road?’ Q’anjobal (prim. data)

Indefinite articles represent another major type of articles in terms of frequency. They mark
both specific and nonspecific referents. Although being very common in European languages,
they are also found in other parts of the world. The following examples shows the indefinite
article in Awakateko (Mayan, Guatemala). In contrast to the specific article jun in Q’anjobal, the
article jun in Awakateko is an indefinite article because it marks both specific and nonspecific
referents. In (33), the article encodes a specific referent, since only the speaker and not the hearer
can unambiguously identify the church, but it is a particular church. As (34) shows, the same
article jun is used for nonspecific referents which no longer correspond to any particular referent
from their kind.

(33) at [jun tiox] stzi’ a’

exist art:indef church river.bank water
‘There is a church at the river banks.’ Awakateko (prim. data)
(34) nawaj [jun moy]
want.1sg art:indef car
‘I want to have a (any) car.’ Awakateko (prim. data)

The last major article type that we find in the world’s languages does not distinguish between
definite, specific, and nonspecific referents but encodes all three referent types. I call these articles
referential articles. One example comes from Rapa Nui (Austronesian, Chile). In Rapa Nui, nouns
occur together with the marker te in most contexts except, e.g., as predicates or instruments, and
te can be shown to encode definite, specific, and nonspecific referents. In (35), the noun kai ‘food’
occurs with te and has a definite referent. In (36), the article is used with taŋate ‘man’, whose
referent is specific because it is a particular but not (yet) identifiable referent. Example (37), on
the other hand, shows te together with a nonspecific referent.
(35) ’ina he mā’eha mo u’i iŋa i [te kai]
neg pred light for see nmlz acc art:ref food
‘There was no light to see the food.’ Rapa Nui (Kieviet 2017: 93)

30
(36) ’i te noho iŋa tuai era ’ā [te taŋata e tahi] te ’īŋoa ko
at ref stay nmlz ancient dist ident art:ref man num one ref name prom
Tu’uhakararo
Tu’uhakararo
‘In the old times (there was) a man called Tu’uhakararo.’ Rapa Nui (Kieviet 2017: 238)
(37) ko mate atu ’ana ki [te vai] mo unu
prf die away cont to art:ref water for drink
‘I’m dying for water to drink.’ Rapa Nui (Kieviet 2017: 237)

In some cases, adnominal markers may have referential functions, but their main function is a
different one. Such markers are not considered as articles. The next paragraphs provide examples
of such article-like but nevertheless distinct markers.
In Ughele (Austronesian, Solomon Islands), for instance, we find adnominal markers that qual-
ify as articles based on their domain of occurrence. Most probably for this reason, Frostad (2012: 6)
distinguishes between three article types: the personal article e which occurs with proper nouns
referring to a person (Frostad 2012: 89), the common article na, and the focal article ai. The two
markers e and na, labelled personal and common article, are illustrated in (38) below:

(38) a. meke naghe [e Peni] …

and say prop Peni
‘and Peni said …’ Ughele (Frostad 2012: 100)
b. Zioni [na viu] le-lea-na
Zioni comm bird red-good-attr:3sg
‘Zioni is a very nice bird.’ Ughele (Frostad 2012: 97)

Since these two markers occur with different types of nouns, it seems plausible to call them articles.
However, their function is different: e occurs only with proper nouns referring to persons as in
(38a) and na is used only with common nouns as in (38b). This suggests that their function is not
linked to referentiality, which is why I do not treat these markers as articles.
Other elements that occur with nouns but do not have primarily referential functions are case
markers. An example is the accusative-partitive alternation in Finnish. In Finnish, case markers
do not directly encode the referent type of the noun that they occur with, but they can influence
the referential interpretation of nouns in certain syntactic positions. The accusative-partitive
case alternation can influence the referential interpretation of the object of a transitive clause. An
object marked by the accusative case as in (39a) is interpreted as having a definite referent, while
the referent of an object in the partitive as in (39b) is interpreted as specific. Needless to say,
such markers are not considered as articles because marking referentiality is not their primary
function.

31
(39) a. Silja joi maidon.
Silja drank milk.acc
‘Silja drank the milk.’
b. Silja joi maitoa.
Silja drank milk.part
‘Silja drank (some) milk.’
Finnish (Karlsson 1999: 85)

A different type of marker that involves referentiality and occurs adnominally is found in Koasati
(Muskogean, USA). It encodes an absent or deceased referent, which also makes the referent spe-
cific as it necessarily involves a particular referent of its kind. Example (40) and (41) illustrate this:

(40) yalí am-aw-óːto-k cimpónc-oːto-k ónti-toy

here poss:1sg-grandfather-exist.pst-s Jim Poncho-exist.pst-s come.sg-pst
‘Here my grandfather Jim Poncho (who is now deceased) came.’
Koasati (Kimball 1985: 349)
(41) …waciná nałihilk-óːto-n ca-sobáy-ko-k
…English language-exist.pst-o 1sg.stat-know-neg.3-ss
‘…at that time the English language was unknown to me.’ Koasati (Kimball 1985: 349)

In (41) we can see that the lexical meaning of “deceased” is extended to mark a former state of
the referent, ‘the English language’. The marker óːto is treated as an article in the grammar for
language-specific reasons.5 One could argue that it is an article because it encodes specific refer-
ents. However, since this seems to be an effect of the lexical meaning of that marker “deceased,
no longer present”, I do not regard óːto as a specific article.

2.2.3 Articles occur in the nominal domain

The second criterion for articles restricts their domain; they have to occur in the nominal domain
together with nouns. Note that I use “nominal domain” and “noun phrase (NP)” interchangeably,
ignoring framework-dependent concepts, e.g. noun phrase vs. determiner phrase.
The position and form of the article in the noun phrase can be described by partially orthogonal
parameters. It can be a separate or affixed morpheme, as the separate article in English, e.g. the
cat, or the suffixal definite article in Armenian, e.g. katu-n ‘the cat’. If the article is affixal, we can
further distinguish different types. Affixed articles are often suffixes, but we also find examples
5 The marker is part of a paradigm of other adnominal markers that involve the marking of referential functions.
Most of these markers are not real articles either, since they are participial forms indicating temporal relations, the
referential interpretations rather being an additional effect (Kimball 1985: 344). These markers are probably called
articles in the grammar based on their occurrence together with nouns.

32
of prefixed articles. In Sumu (Misumalpan, Nicaragua), the article affix appears as an infix with
a number of nouns that used to be compound forms. An example is waki-ni-sa ‘the banana’, in
contrast to the form without the article wakisa ‘(a) banana’ (Norwood 1997: 22).
Although there might be sufficient language-dependent reasons to classify an article as an affix,
a clitic, or a free word form, such a distinction can hardly be motivated in a comparative study.
Therefore, I do not aim at a distinction between affixed, cliticized, or free articles here. I propose
a distinction between articles that are separable from nouns in the presence of another nominal
element, e.g. an adjective, and articles that are not separable from nouns in such contexts. Thus,
the article can either stay in its original position adjacent to the noun (“N-anchored”), or surface in
a different position at the edge of the entire noun phrase (“NP-anchored”) in the presence of other
nominal attributes. This distinction is more comparable across languages and more workable than
the distinction between free forms, clitics, and affixes.
Another parameter concerns the linear order. Articles can precede or follow the noun, e.g.
German das Haus vs. Icelandic hús-ið ‘the house’. Paring these two parameters, examples (42) to
(45) give an overview of the possible orders of the noun, the article, and the adjective:

• Preposed and N-anchored

(42) büyük bir oda
large art:indef room
‘a large room’ Turkish (Göksel & Kerslake 2005: 145)

• Preposed and NP-anchored

(43) juu maqaniyaa lapanak
art:def old people
‘the old people’ Tepehua (Kung 2007: 388)

• Postposed and N-anchored

(44) kìl tìcɛ́ tìlòlìyɛ́n
houses art:def low
‘the low houses’ Bullom So (Childs 2011: 71)

• Postposed and NP-anchored

(45) kavayi ságharumu ’uway
horse white art:def
‘the white horse’ Ute (Givón 2011: 195)

As these examples show, the classification of articles into N-anchored and NP-anchored ones is
not entirely unrelated to the classification of articles into free forms, clitics, and affixes. Articles
that are classified as affixal are generally anchored with respect to the noun, also in the presence

33
of other elements in the NP. Articles are often classified as clitics if they are able to appear on
different elements in the NP. Usually, they occur in a fixed position with respect to the entire NP
(mostly at the edge), which corresponds to NP-anchored articles. What is traditionally viewed as
a free marker, however, also often falls into the category of NP-anchored for staying at the edge
of the entire NP.
We do also find languages with a flexible order of the article and the noun, which is the case in Sri
Lanka Malay (Malayo-Sumbawan, Sri Lanka). Although described as free variation in Nordhoff
(2009: 319-320), the position of the article may be conditioned by the presence or absence of other
elements that occur in the NP, as e.g. the additive marker in (46b).

(46) a. se hatthu=aade
1sg art:indef=younger.sibling
‘I am a younger sibling’
b. giini criitha=hatthu=le aada
like.this story=art:indef=addit exist
‘There is also a story like that.’
Sri Lanka Malay (Nordhoff 2009: 319)

Another formal type of articles can be distinguished if we consider whether or not they are seg-
mental morphemes. Although articles with a nonsegmental exponent are rare in the languages
of the world, they are attested. For instance, in Mokpe (Bantu, Cameroon), the definite article
corresponds to a tonal process on nouns of certain noun classes. If the noun occurs with a def-
inite article, its first tone-bearing unit appears with a high tone that overrides lexical low tones,
observable in indefinite contexts and in the citation form (cf. sections 5.2.2 and 6.2.2 on the defi-
nite article in Mokpe). This is shown in (47) for a number of nouns in isolation, and for the noun
lìkândò ‘roof’ in a bridging context in (20).

(47) Mokpe (prim. data)

a. mòlélí ‘food’ vs. mólélí ‘the food’
b. lìkàlà ‘(a) bridge’ vs. líkàlà ‘the bridge’
c. ßètáŋgùlɛ̀ ‘(some) lizards’ vs. ßétáŋgùlɛ̀ ‘the lizards’
(48) ì mɛ̀ nɛ́ ŋmánɛ́ βòβá ndáwò ó mbówà. líkândò/ *lìkândò lí mà kúwú
we see.pst.1pl small ugly house in village roof.cl5.art:def roof.cl5 cl5 pst finish
lì βáŋgèjà
inf get.damaged
‘We saw a small ugly house in the village. The roof had been damaged.’
Mokpe (prim. data)

34
In the following paragraphs, I address some markers which are article-candidates based on their
referential function, but which do not meet the domain criterion.
Nyaturu (Bantu, Tanzania) has a referential marker that indicates definite referents, but it
surfaces in the verbal complex. In (49b), the agreement marker in the verbal complex triggers a
definite interpretation of the referent of the object.

(49) a. n-a-onaa mwalimu

sm:1sg-pst1-see cl1.teacher
‘I saw a teacher.’ Nyaturu (Hualde 1989: 182)
b. n-a-mʊ-onaa mwalimu
sm:1sg-pst1-om:cl1-see cl1.teacher
‘I saw the teacher.’ Nyaturu (Hualde 1989: 182)

Since there is no additional marker in the noun phrase to indicate the referentiality of the nominal
expression, the function of the object marker is comparable to a definite article. In accordance to
the intuition that this marker in Nyaturu should not be treated as an article, it can be excluded
based on the domain criterion here.
Besides referential markers in the verbal domain, we also find referential markers in the noun
phrase that do not qualify as articles because they do not primarily occur with simple nouns
and require another element to be present in the noun phrase as well. A well-known example
is the German adjective agreement, which is sensitive to definite and indefinite contexts in the
nominative:

(50) a. ein groß-er Hund

art:indef big-nom.m.sg.indef dog
‘a big dog’
b. der groß-e Hund
art:def big-nom.m.sg.def dog
‘the big dog’
German (prim. data)

Also in Khwarshi (Tzezic, Russia), we find a definiteness marker that only combines with adjec-
tives and similar expressions. Accordingly, the marker so, as shown in (51) and (52), is not treated
as a definite article.

(51) žik’w e [miq’e-so baydan] m-ež-i

man.obl.erg far.away-def field.cl3 cl3-plant-pst.w
‘The man planted the furthest field.’ Khwarshi (Khalilova 2009: 133)

35
(52) ø-oλoλ’o-so-ho y-oλoλ’o-so y-ez-un
i-in.middle-def-apud ii-in.middle-def ii-take-pst.uw
‘The middle (brother) married the (other) middle (sister).’ Khwarshi (Khalilova 2009: 133)

While the marker so in Khwarshi seems to have other non-referential functions as well, Colloquial
Slovenian features a similar marker that is treated as a definite article in Marušič & Žaucer (2014).6
Example (53) illustrates the use of this marker ta.

(53) en od ta hitrih avtov

one.nom of def fast.gen cars.gen
‘one of the fast cars’ Coll. Slovenian (Marušič & Žaucer 2014: 193)

However, ta can only occur together with attributive adjectives in the noun phrase as in (54a); it
cannot mark the referent as definite if no adjective is present (54b). Therefore, I do not regard ta
as a definite article.

(54) a. ta velika knjiga

def big book
‘the big book’
b. *ta knjiga
def book
intended meaning: ‘the book’
Coll. Slovenian (Marušič & Žaucer 2014: 185)

2.2.4 Articles are systematically distributed

The aim of this section is to propose a crosslinguistically applicable criterion with respect to the
use and distribution of articles based on what I will call systematicity.
Often, obligatoriness is evoked as a criterion for articles, also with the reference to English or
“European” articles:

On the basis of what we have seen in English, we may expect articles more widely to act
as default members of larger categories of definite or indefinite expressions, to be obligatory
(except perhaps under certain generally specifiable conditions) in the absence of other such
expressions, and to be unstressed and perhaps phonologically weak. (Lyons 1999: 47)

Another example showing this issue is found in Terrill (2003) concerning Lavukaleve (Lavukaleve,
Solomon Islands). Lavukaleve has an indefiniteness marker which is not considered as an article
6 Colloquial Slovenian has almost completely lost the old distinction between short and long adjective forms that used

to mark the definiteness of the nominal referent. In the standard language, however, we do still find this difference,
e.g. dober pesnik ‘a good poet’ vs. dobri pesnik ‘the good poet’.

36
in the grammar. The marker is estimated to be too infrequent and not obligatory in the sense that
not all indefinite noun phrases occur with this marker. According to the author, the strongest
argument against treating the marker ro as an article is that it can co-occur with the definite
article in the language.

[…] ro is a marker of indefiniteness, but one would not want to call it an indefinite article,
for a number of reasons. Firstly, it has the syntactic distribution of an adjective. Secondly,
it does not have the frequency or obligatoriness which one might expect of an article; it is
certainly far less frequent than the definite article. Thirdly, it is not obligatory in indefinite
noun phrases, again unlike the definite article, which is obligatory in all definite noun phrases
(unless there is a demonstrative to express the definiteness). Fourthly, and crucially, it can,
under certain constrained circumstances, co-occur with the definite article. (Terrill 2003: 81)

There are at least three problematic assumptions in these quotes presented. Firstly, the properties
of an article in a single language (e.g. English) cannot serve as basis for a crosslinguistic com-
parison. Secondly, articles are not “obligatory” in any language if one understands obligatory as
occurring in all contexts in which they are expected according to their referential function. Often
we find that the use of articles is restricted in many contexts in which they should mark the ref-
erent type of the noun (e.g. mass nouns, plural nouns, with certain adpositions, with certain case
markers). Thirdly, although definite and indefinite articles usually do not co-occur in European
languages, this cannot be taken as an argument as such for or against treating another marker as
an article. In fact, there are co-occurring articles attested in a number of languages (see section
7.4.4).
As mentioned above, it is problematic to characterize the distribution of articles as obligatory if
obligatoriness cannot allow for exceptions. I will illustrate this with what I regard as an anaphoric
article in Hausa (Chadic, Nigeria). Hausa has a marker -n, shown in example (55), which started
out as an anaphoric article but which is more and more used in other definite contexts as well.
Example (56) shows a noun which is marked by -n and which has a definite referent because of
the following relative clause.

(55) a. bákâ-n ‘the bow’

b. manòmā-n ‘the farmer’
Hausa (Newman 2000: 144)
(56) yaarò-n dà ya tàfi
boy-art:ana/def rel 3sg.m.perf.rel leave
‘the boy who left’ Hausa (Newman 2000: 145)

In his grammar, Newman (2000: 143) mentions that the label “definite article” is problematic, and
that “previous reference marker” would be a better label, indicating that we are dealing with an

37
anaphoric marker that has started to extend its functional domain to other definite contexts. The
argument used against treating this marker as an article (e.g. in Newman 2000: 143, Lyons 1999:
52, Zimmermann 2008: 419) always is that a bare noun can still be interpreted as definite:

(57) tùlū yā fashè

pot 3sg.perf break
‘The / a water pot broke.’ Hausa (Newman 2000: 143)

In example (57), tùlū ‘pot’ is not necessarily interpreted as indefinite even in the absence of -n.
Hence, the presence of the definiteness marker -n is not obligatory to mark the noun as definite,
so the argument goes.
What the concept of obligatoriness entails is that there is no room for exceptions. However,
the behaviour of articles in many languages is always also influenced by factors other than refer-
entiality; the distribution of articles is never perfect in that sense and always interacts with other
parts of the grammar or extra-linguistic factors, as I will show below.
Therefore, I propose to base the distributional criterion for articles on the notion of system-
aticity instead of obligatoriness, since the latter does not allow for exceptions and suggests that
the decision between treating a marker as an article or not can be made on the basis of single
examples. However, the presence and absence of articles can be conditioned by so many other
syntactic, semantic, morphological, and prosodic properties of the construction, so that individual
examples are not sufficient to classify a marker as an article.
On the other hand, the concept of systematicity only requires articles to be distributed sys-
tematically in language use, which is not evaluated on the basis of single examples, but on the
overall distribution of the marker in the language. This automatically allows for an article to be
absent in certain contexts in which its presence would be expected based on its referential func-
tion, because other, e.g. syntactic, factors may block the use of the article. The same holds for
articles that can appear in contexts in which they are not required as a referential marker, but in
which they occur nevertheless, for instance, definite articles that co-occur with demonstratives
in certain languages.
In the following paragraphs, I discuss a number of markers that should be treated as articles
although they are systematically absent in contexts in which their presence is expected from a
referential point of view.
A simple and uncontroversial example of a morphosyntactic restriction is the indefinite article
in German. It does not have a plural form and cannot combine with nouns in the plural. Example
(58a) shows Katze ‘cat’ marked as nonidentifiable by the indefinite article. Its plural counterpart in
(58b) is expressed as a bare plural noun Katzen ‘cats’ because the indefinite article is not available
in this context.

38
(58) a. Ich sehe [*(eine) Katze] auf dem Dach.
I see art:indef cat.sg on art.def roof
‘I see a cat on the roof.’
b. Ich sehe Katzen auf dem Dach.
I see cat.pl on art.def roof
‘I see cats on the roof.’
German (prim. data)

The relevant argument to treat this marker in German as an indefinite article here is that even
though nouns can be interpreted as nonidentifiable in the absence of the marker, its distribution
is systematic and the contexts in which is does not occur can be predicted. Thus, systematically
distributed means that exceptions are allowed as long as they can be formulated as rules or strong
correlations.
This argumentation seems to be trivial with respect to German; however, in less studied lan-
guages, such distributional gaps are often pointed out to show that the marker in question cannot
be treated as an article. One example is the “definiteness marker” in Sheko (Hellenthal 2010: 153),
which signals the definiteness of the referent as in (59), but does not occur with plural nouns,
shown in (60):

(59) ék-ká fáád-k’à-bààb íǹ-ʃ yī-ərá k’ōdām-m̀-s yīs

there-loc body-in-father wood-art:def.m dist.m-acc.cont qodama-art:def-m dist.m
há=gīīşū-kōb-t=á kēs-ā-m-ə
s:3m=pull-take-ss=s:3m go.out.caus-put-irr-sti
‘that qodama pulls out the wood which is there in the body’ Sheko (Hellenthal 2010: 141)
(60) akǹ sàm-àb fyāānū-s-kǹ ʒéénʃ kī-ít-ə ge-tə
here remain-rel frog-pl-dat well exist-addr:pl-sti say-ss
‘(she) said “stay well!” to the frogs who remained there’ Sheko (Hellenthal 2010: 163)

Since the marker -ʃ is only available for singular nouns, Hellenthal (2010: 136) analyzes it as
being complementary to number marking and does not regard it as an article. However, like the
indefinite article in German, it fulfills the distributional criterion which is why I treat it as an
article in this study.
Another type of systematic restriction on the use of the article can be found in Mokpe (Bantu,
Cameroon). In Mokpe, the definite article has different exponents depending on the class of the
noun. Relevant here is the marker è, surfacing with nouns of class 1. This marker, as example
(61a) shows in opposition to (61b), does not occur with nouns whose first segment is a vowel:

(61) a. è ɲaŋga ‘the sand’

b. ∅ ìkpá ‘(the) salt’
Mokpe (prim. data)

39
Since this restriction of the definite article can be formulated as a simple rule (è → ∅ / _V), it does
not pose a problem for treating the marker as an article.
Articles can also be restricted in a systematic way by processes outside of phonology or mor-
phosyntax, e.g. by semantic effects. In many languages, articles do not combine with proper
nouns or mass nouns. Example (62a) shows this for the definite article in English which does
not combine with proper nouns, while example (62b) illustrates the incompatibility of the English
indefinite article with mass nouns.

(62) a. (*The) Steve baked (*the) Hannah a cake.

b. He is drinking (*a) milk.
English (prim. data)

We also find languages with stylistic restrictions on articles. For instance, in Konso (East-Cushitic,
Ethiopia), the definite article can be omitted in certain anaphoric contexts according to the gram-
mar (Orkaydo 2013: 97), while it is generally used in anaphoric and other definite contexts (Orkaydo
2013: 95-97). Example (63) illustrates the use of the definite article for referents that have already
been mentioned in the discourse:

(63) kaasa-sit-n=in karmaa-siʔ ʔiʃʃ-ay

gun-art:def-instr=1 lion-art:def kill-pf.3.m
‘I killed the lion with the gun.’ Konso (Orkaydo 2013: 97)

In the same story about the lion, the latter does not need to be marked by the definite article every
time it is mentioned. Example (64) shows an utterance that comes shortly after the one in (63);
the referent of karmaa ‘lion’ is definite, but the definite article is not used in this instance.

(64) karmaa ka ʛapaleeta-asiʔ ʛaraa kaassuma=i kaassaɗ-ay

lion and monkey-dem on question=3 ask-pf.3.m
‘And the lion asked this monkey a question.’ Konso (Orkaydo 2013: 97)

Since Orkaydo (2013: 97) argues that this is only allowed if the definite referent is highly promi-
nent in the discourse, we can speak of systematic absence of the marker -siʔ, which is why I treat
the marker as an article.
Another interesting pragmatic restriction on articles is found in Crow (Siouan, USA). In the el-
evated narrative genre, specific and nonspecific articles are replaced by a single indefinite marker
(Graczyk 2007: 233). In examples (27) and (28) , repeated below as (65) and (66), we saw the
“standard” specific and nonspecific indefinite articles -m and -eem, respectively.
(65) dakáak-kaata-m húu-laa híi-k
bird-dim-art:spec come-ss reach-decl
‘a bird came, it reached him’ Crow (Graczyk 2007: 228)

40
(66) axée baláyyiikaashe dúup-eem alúutkaashe áppaa día-a-wa-ku-hee?
father bow two-art:nspec arrow with make-cont-1-give-aff-q
‘father, will you please make me two bows as well as (some) arrows?’
Crow (Graczyk 2007: 230)

In certain narrative contexts, these markers are not used and are replaced by -dak/-lak (-dak fol-
lows consonants while -lak is used after vowels). Note that in this genre, the difference between
specific and nonspecific referents is no longer marked, as (67a) and (67b) illustrate. Example (67a)
features a specific referent (bachee-lák ‘a man’), a referent that is not identifiable but unambigu-
ously linked to the noun. Example (67b) shows the noun iisashpít-dak ‘a rabbit’ with a nonspecific
referent, i.e. not a particular but a single referent of the kind ‘rabbit’ that is linked to the noun.

(67) a. bachee-lák baa-aash-dée-k

man-art:indef indef-hunt-go-decl
‘A man went hunting.’ Crow (Graczyk 2007: 230)
b. éehk bal-héelee-n iisashpít-dak baappeé-k b-aliat-bee-m
that wood-among-loc rabbit-art:indef 1a.kill-decl 1a-think-1a-ds
isáa-kaashi-k
large-augm-decl
‘There in the woods I thought I killed a rabbit, but to my surprise, it was something
very large.’ Crow (Graczyk 2007: 231)

The concept of systematicity not only allows for the systematic lack of articles in contexts in
which they are expected to be present based on their referential functions, but it also accounts for
articles that occur in other contexts.
An example for articles that are used for reasons other than marking referentiality is the in-
definite article in Coptic. It combines with infinitives to form a different non-verbal expression.
Together with the marker hn- ‘in’, the indefinite article ou- derives adverbial expressions from
infinitives:

(68) a. hn-ou-orch
in-art:indef-become.secure
‘securely’
b. hn-ou-ouōnhebol
in-art:indef-show.forth
‘openly’
Coptic, Sahidic dialect (Plumley 1948: 40)

41
Another syntactic use of articles can be illustrated with indefinite articles from Q’anjobal (Mayan,
Guatemala). The language has an exclusive-specific article and classifiers, but no definite article.7
The exclusive-specific article as well as classifiers precede the noun. In order to use an adnominal
demonstrative which follows the noun, Q’anjobal requires this prenominal position to be filled.
The default marker filling that position is the classifier, as is illustrated in examples (69) and (70).

(69) cham icham tu’

cl:man man dem:dist
‘that man’ Q’anjobal (prim. data)
(70) te na tu’
cl:tree house dem:dist
‘that house’ Q’anjobal (prim. data)

If no classifier is available for the noun in question, the indefinite article is used instead. Exam-
ples (71)-(73) show the use of the exclusive-specific article together with the demonstrative. The
prenominal position can also be filled by a combination of the article and the classifier, as in (74).

(71) jun ora tu’

art:exspec hour dem:dist
‘at that hour’ Q’anjobal (prim. data)
(72) jun b’eqan ti’
art:exspec moment dem:prox
‘at this moment’ Q’anjobal (prim. data)
(73) jun tz’uy ti’
art:exspec bag dem:prox
‘this bag’ Q’anjobal (prim. data)
(74) jun winaq unin ti’
art:exspec cl:man child dem:prox
‘this boy’ Q’anjobal (prim. data)

Without going into detail with regard to the distribution of the exclusive-specific article and clas-
sifiers, the important point is that the noun cannot combine with the demonstrative without any
element filling the prenominal position. The default marker being the classifier, the exclusive-
specific article can also take over this function without marking referentiality.
The last example I discuss in this section shows a marker that qualifies as a definite article but
cannot be treated as such due to its unsystematic distribution. Tz’utujil (Mayan, Guatemala)
7 The classifiers do not have (primary) referential functions.

42
has a marker ja(la), which can occur in different definite contexts. Example (75) shows ja in
an anaphoric context; in (76), we see it with a situationally unique referent.

(75) A: taq xinipoon p-jaay xintzu’ [jun ixoq] k’in ik’ee’ ak’alaa’
when arrive.pst.1sg loc-house see.pst.1sg art:indef woman with two children
etz’b’ula’ chuchii’ jaay
seated in.front.of house
‘When I came home, I saw a woman with two children sitting on my doorstep.’
B: awtaqiin ruwach [ja ixoq]?
know.pst.2sg eyes def woman
‘Did you know that woman?’
Tz’utujil (prim. data)
(76) b’anitzra k’ool wa [ja q’atb’al] tziijchpam jawa tinamet?
where exist q def town.hall inside dem town
‘Where is the town hall in this town?’ Tz’utujil (prim. data)

In addition, ja does not occur in indefinite contexts, as is illustrated for specific referents in (77)
and for nonspecific referents in (78). In both of the contexts, Tz’utujil uses the indefinite article
jun instead.

(77) k’oola [jun sepaneem] awyin tat

exis.1sg art:indef gift
‘I have a gift for you.’
(78) k’ool [jun atz’ib’ab’el]?
exist art:indef pen
‘Do you have a pen?
for.2sg you
Tz’utujil (prim. data)
Tz’utujil (prim. data)

Treating ja as a definite article in Tz’utujil is problematic because its use differs to a great extent
among speakers. According to my informants, age and dialectal variation may influence the use of
this marker as a definite article. However, also single speakers use this marker very inconsistently.
Often, both a bare noun and a noun marked by ja can occur in definite contexts, and according
to my informants, there seems to be neither a difference in meaning nor a preference for one
version or the other. Perhaps the marker ja is used systematically and we only do not understand
the factors that determine its use. However, based on the situation as presented above, I do not
treat ja as a definite article, even though it occurs in the nominal domain to mark definiteness. The
fact that speakers have no clear preference in favor or against the use of ja with definite referents,

43
I conclude at this point that ja in Tz’utujil is not distributed in a systematic way, at least not as a
definite article.8

2.3 Summary
In this chapter, I presented the necessary criteria to define articles as a crosslinguistic category,
namely their referential function, their occurrence together with nouns, and their systematic
distribution. In section 2.2.2, I showed different referential functions of articles, giving a brief
overview of the major article types that we find across the languages of the world. These article
types are: definite, anaphoric, inclusive-specific, exclusive-specific, indefinite, and referential ar-
ticles. I also argued that only markers whose main function it is to encode referentiality should
be treated as articles. In section 2.2.3, I argued that articles have to occur together with nouns in
the noun phrase, and I showed that further requirements with regard to their form or position
are not necessary. Section 2.2.4 was concerned with the distributional criterion of articles as a
crosslinguistic category. I argued that systematicity is a better criterion than obligatoriness, since
the latter does not allow for exceptions. However, I showed that we commonly find two types
of exceptions concerning the distribution of articles: articles can either be absent in contexts in
which their presence is expected based on their referential function or they can be present in refer-
entially unexpected contexts. Systematicity can capture this distribution, allowing for exceptions
that can be described as rules or strong correlations.

8 Inthe grammar of the Guatemalan Academy of Mayan languages, ja is treated as a definite article without further
discussion regardings its distribution (Tz’utujil Tinaamitaal 2007: 87-90). Also Dayley (1985: 254-256) treats ja as
a definite article, providing some contexts in which the marker can be used, noting that it also often occurs with
pronouns in topical positions. Thus, ja may not be an article but a topic marker which is highly correlated with
definite referents.

44
Chapter 3

The distinction between articles and similar

categories

It is difficult to delimit certain types of articles from other similar elements, especially if the arti-
cles develop from these. The three major cases that I will discuss in this chapter concern definite
articles and demonstratives, indefinite and exclusive-specific articles and the numeral ‘one’, as
well as nonspecific articles and negative polarity items (NPIs). The former two cases are due
to the grammaticalization of demonstratives and the numeral ‘one’ into definite and exclusive-
specific/indefinite articles, respectively. Their development being a gradual process, it is difficult
to motivate a clear cut-off point between the source elements and articles. Therefore, I propose
criteria for a distinction between definite articles and demonstratives in section 3.1, and for indef-
inite/specific articles and the numeral ‘one’ in 3.2. With respect to nonspecific articles and NPIs,
the similarity is not due to a diachronic relation. However, since the nonspecific article is not a
well-known type of articles, I will point out the differences between this article type and NPIs in
section 3.3.

3.1 Demonstratives
It is widely recognized that most definite articles in the world’s languages originate from demon-
stratives. This grammaticalization path was already proposed by Christophersen (1939) and Green-
berg (1978b: 246), and was revisited in various later studies (De Mulder & Carlier 2011, Diessel
1999, Himmelmann 1997, Leiss 2000, Lyons 1999). The grammaticalization path includes the de-
velopment from the demonstrative to the definite article and the development from the latter to
the inclusive-specific article following Greenberg (1978b: 246) is shown in Figure 3.1:

45
 stage 0 stage I stage II stage III

demonstrative definite article exclusive-specific article noun marker

Figure 3.1: Grammaticalization path of demonstrative and definite article

Relevant here is the development between stage 0 demonstratives and stage I definite articles.
As Schroeder (2006: 554) notes, the contexts of (spatial) deixis, anaphora, and uniqueness of the
referent can be viewed as a continuum on a scale from demonstrative to article. The demonstrative
is understood as a deictic element, typically used for spatial deixis, indicating the relative position
of an object with respect to the discourse situation and to draw the hearer’s attention towards the
referent of this object. This use is called exophoric (cf. Diessel 1999: 94-95). Demonstratives with
a primarily exophoric function can often be extended to anaphoric functions (Diessel 1999: 96), i.e.
the encoding of anaphoric referents. This use is also called endophoric (cf. Diessel 1999: 95-105).
Exophoric and endophoric uses are thus two different types of deixis. For the sake of clarity, I
use the term “deixis” only to refer to (spatial) deixis and treat exophoric uses (or text deixis) as
“anaphora”. To be correct, this use of deixis is not completely restricted to spatial deixis but refers
to physical perception and attention drawing in a broader sense.
Through this extension of the deictic demonstrative to anaphoric, and at a later stage also
other, uses, the marker may become more frequent, which in turn leads to a weaker association
of the marker with deictic referents, until its function includes the systematic encoding of situ-
ationally unique referents, making it a definite article. Due to this strong connection between
demonstratives and definite articles, it is not always clear how to classify markers that are some-
where in between a “typical” demonstrative and a “typical” article situated on the two ends of a
scale. The traditional cut-off point between demonstratives and articles seems to be between the
marking of anaphoric and situationally unique referents. Lyons (1999: 54) notes: “[…] it may be
unclear whether a determiner specialized in anaphoric use is an article or a demonstrative”. Also
in other studies concerned with demonstratives, definite expressions, or articles, anaphoric mark-
ers are usually treated as demonstratives rather than as articles (e.g. Diessel 1999, Himmelmann
1997, Lyons 1999).
Assuming that the development of definite articles from demonstratives involves their func-
tional extension from markers of deictic referents to anaphoric, and finally situationally unique
referents, we can expect to find markers that encode different combinations of functions along
this scale. Figure 3.2 shows the functional domains of such markers.1
1 (Himmelmann 1997: 93-97) proposes a different path for the development of definite articles from demonstratives
through establishing uses. Since I only make use of these main functions and their scalar arrangement that is
synchronically reflected in the functional domains of different types of demonstratives and articles in the world’s
languages, I do not discuss alternative grammaticalization paths here.

46
 deixis anaphora uniqueness

I demonstrative

II demonstrative

III article

IV article

V article

Figure 3.2: Cut-off point between demonstratives and articles

Breaking down the relevant functions to deixis, anaphora, and uniqueness, we usually find these
five combinations of their expression. As is shown in Figure 3.2, I consider markers as demon-
stratives if they are restricted to expressing deixis, and if they encode both deixis and anaphoric
referents (type I and II). As was mentioned above, anaphoric markers (type III) are often treated as
demonstratives, rather than as articles (e.g. De Mulder & Carlier 2011, Diessel 1999, Dryer 2013b,
2014, Himmelmann 1997). Although Himmelmann (1997: 66) also discusses anaphoric markers
on a par with other demonstratives, he notes that this categorization of an anaphoric marker in
Wubuy (Nunggubuyu, Australia) might be problematic, as it cannot be used to mark spatial deixis
and thus differs considerably from the functions usually ascribed to demonstratives.
Primarily concerned with the definite article in English, Hawkins (1978) defines demonstra-
tives as markers of spatial deixis (“immediate situation use”) and anaphoric uses (“direct anaphoric
use”). The demonstrative contrasts in this way with the definite article in English, which is used
in different definite contexts based on the uniqueness of the referent. He distinguishes between
two types: bridging (“associated anaphora”) and (situational) uniqueness (“larger situation use”).
While this distinction may be sufficient for English, it cannot be applied to languages with markers
that encode different combinations of functions.
Thus, in order to compare articles across the world’s languages, I treat markers that only en-
code anaphoric referents (type III) as well as markers that encode both anaphoric and situationally
unique referents (type IV) as articles, as long as they comply with the other criteria discussed in
Chapter 2. Markers of type V probably correspond to the most common type of definite articles
and their status as definite articles is unproblematic. This is not the case with anaphoric markers
of type III. Therefore, the following paragraphs address the question concerning anaphoric mark-
ers being treated as articles, which requires them to be sufficiently distinct from demonstratives.
The prototypical deictic use of demonstratives can usually be accompanied by a pointing gesture

47
(Diessel 2013b: 243). This link between pointing gestures and demonstratives is well known in the
linguistic literature (e.g. Brugmann 1904, Bühler 1934, Clark 1996, Eriksson 2009, Levinson 2004)
and was often tied to the spatial deictic function of demonstratives. Diessel (2013b: 243) points
out that the function of such demonstratives is not only spatial location, but more importantly,
drawing the hearer’s attention towards another object in the discourse situation. He notes:

But a deictic pointing gesture is not just a guidepost for spatial orientation, it also serves to
create what psychologists call a joint focus of attention (cf. Butterworth 1998, Eilan et al. 2005,
Tomasello 1999). […] Deictic pointing is the most basic communication device that people of
all cultures use to establish or manipulate joint attention (cf. Kita 2003). (Diessel 2013b: 243)

Hence, I will take the expression of joint focus of attention as a basic communicative function that
is universal in human language. This is supported by the fact that demonstratives are generally
taken as a universal part of speech (e.g. Evans & Levinson 2009; Diessel 2013b: 245). A purely
anaphoric marker is different. Whether or not one wants to argue that it represents a more ab-
stract, extended, or grammaticalized function, an adnominal anaphoric marker is clearly less basic
in the sense that it is not present in all languages. Most often, it is the deictic demonstrative that
can be functionally extended to mark anaphoric referents. For this reason, I treat markers that
only encode anaphoric referents as articles here.
Entirely deictic demonstratives (type I) have to be infelicitous in anaphoric contexts based
on their definition. It is not clear whether such demonstratives are attested at all in the world’s
languages: even with a distinct anaphoric marker existing in the language, they often seem to
be applicable in certain anaphoric contexts. Nevertheless, we find contexts in which the deictic
demonstrative cannot be used to mark an anaphoric referent. This is illustrated with examples
from Tikuna (Tikuna-Yuri, Colombia) and Limbum (Bantu, Cameroon) below. Examples (1) and
(2) below show the Tikuna demonstrative in deictic contexts. The same demonstrative cannot be
used to mark (at least certain types of) anaphoric referents, as we can see in (3):

(1) nukü dau [ɲema putüra] na meechi

s:2sg.o:3sg see dem flower cop beautiful
‘Look, this flower is beautiful.’ Tikuna (prim. data)

(2) [ɲema buun] na yaichi ineküwa nucha dau iranichi

dem child 3sg.m grow.up before s:1sg.o:3sg see little
‘This child has grown since I have saw him last time.’ Tikuna (prim. data)

(3) A. i-pata-wa nucha dau [wii güena] tare buun ɲema cuchia-wa
1sg-house-loc s:1sg.o:3sg see one woman two child there kitchen-loc
‘When I came home, I saw a women with two children in the kitchen.’

48
 B: nukü wa [*ɲema / (lliema) güena]?
s:2sg.o:3sg know dem / 3sg.f woman
‘Did/do you know this woman?’ Tikuna (prim. data)

Instead of the demonstrative, an adnominal personal pronoun can be used with the anaphoric
referent in (3). At this point, it is not clear how systematic the use of the pronoun as adnominal
anaphoric marker in Tikuna is. However, it is plausible that anaphoric articles in general develop
from constructions like the one shown in (3). The demonstrative in Limbum shows a similar
behaviour to its counterpart in Tikuna; the two languages differ in that Limbum additionally has
an anaphoric article. Example (4) below shows the Limbum distal demonstrative čà in a deictic
contexts. An example of an anaphoric referent is given in (5); it requires the use of the anaphoric
article fɔ̄ (cf. section 5.3.1).

(4) wɛ̀ rı̄ŋ [ndjíŋwɛ̌ čà/ *fɔ̄] ā?

2sg know woman dem:dist art:ana q
‘Do you know that woman (over there)? Limbum (prim. data)
(5) A: mú mfèʔ mɛ̀ mū bāʔ mɛ̀ mū yɛ́ ndjíŋwɛ̀ bá bōō báā ó čútèh mí
when time 1sg pst2 arrive 1sg pst2 see woman and child.pl two 3pl sit in
kíʃı̀n
kitchen
‘When I came home, I saw a woman with two children in the kitchen.’
B: wɛ̀ mū rı̄ŋ [ndjíŋwɛ̌ fɔ̄/ *čà] ā?
2sg pst2 know woman art:ana dem:dist q
‘Did you know that woman?’
Limbum (prim. data)

Figure 3.3 sketches a system of the Limbum type with a demonstrative marker that is used for
deictic referents and an anaphoric article to encode anaphoric referents.

deixis demonstrative

anaphora anaphoric article

uniqueness

Figure 3.3: Demonstrative with anaphoric article

In the simplified system shown in Figure 3.3, the demonstrative is mainly used to mark deictic
referents, while the anaphoric article is used in contexts with anaphoric definite referents. Nei-
ther marker codes uniqueness-based definite referents systematically. This does not necessarily

49
exclude the occurrence of a referential marker with situationally unique referents, but the latter
do not require systematic encoding in this type of systems.
Demonstratives of type II which encode both deictic and anaphoric referents are crosslinguis-
tically common. The following examples from Ejagham (Bantu, Cameroon) show such a demon-
strative used for spatial deixis (6) and anaphora (7):

(6) ma ŋgɔle gɨbe ka [ečo ndjik] elogo

1sg beg hold for.me bag.cl5 dem:prox.cl5 small
‘Please, can you hold this bag for a second?’ Ejagham (prim. data)
(7) A: ewugu m-ba ngi mɔninki na aβɔ aβi ka kiʃən
when 1sg-came saw woman.cl1 with 2 children in kitchen
‘When I came home, I saw a woman with two children in the kitchen.’
B: bo məɲɛ wi [mɔninki ŋu]?
pst know.2sg cl1 woman.cl1 dem:prox.cl1
‘Did you know this woman?’
Ejagham (prim. data)

The marker in Ejagham is a demonstrative and not an article, since it cannot mark situationally
unique referents. Example (8) shows this for the referent of ndzuk oβasi ‘church’ which is situa-
tionally unique in the context of a village:2

(8) [ndzuk oβasi (*ŋi)] erifa ka etek ndjik?

house.cl9 god dem:prox.cl9 where in village.cl5 dem:prox.cl5
‘Where is the church in this village?’ Ejagham (prim. data)

Such a system with a demonstrative that encodes deictic and anaphoric but not situationally
unique referents is illustrated in Figure 3.4 below.

deixis demonstrative

anaphora

uniqueness

Figure 3.4: Demonstrative with deictic and anaphoric functions

Purely anaphoric markers are anaphoric articles (type III), like the anaphoric article in Limbum
shown in example (5). Definite articles of type IV seem to be rather rare crosslinguistically. Such
2 The assumption of demonstratives being generally unable to mark uniqueness-based definite referents is not unprob-

lematic, since restrictive relative clauses are often marked by demonstratives on the head noun. I will not include
this special use of demonstratives here, since it is a minor and formally restricted phenomenon.

50
articles combine the functions of marking anaphoric and situationally unique referents, but are
not used in deictic contexts. A candidate for this type of definite articles can be found in Indone-
sian. Indonesian uses the third person possessive marker -nya to mark definite referents, also in
contexts in which no possessor is semantically recoverable (Rubin 2010). Because of its distinct
source,3 this marker was never used to encode deictic referents and was (not yet) extended to
such contexts. Its use to mark situationally unique referents is shown in (9). In addition, we see
in (10) that the marker -nya can also encode anaphoric referents.

(9) kalau mau makan nasi-nya di lemari

if want eat rice-def in pantry
‘If you want to eat, the rice is in the pantry.’ Indonesian (Rubin 2010: 107)
(10) A: jadi gua ntar ketemu dia langsung di salon
so I soon meet her direct in salon
‘So I’m going to meet her directly at the salon.’
B: salon-nya di deket rumah?
salon-def in near house
‘Is the salon near your house?’ Indonesian (Rubin 2010: 109)

Figure 3.5 shows a schematic representation of the Indonesian system. It consists of a demonstra-
tive that primarily marks deixis, but that be used to encode anaphoric referents as well. In addition
to this demonstrative, there is an article whose main function is the coding of situationally unique
definite referents, also being able to mark anaphoric referents.

deixis demonstrative

anaphora

uniqueness uniqueness-based article

Figure 3.5: Demonstrative with uniqueness-based definite article

Definite articles of type V that can encode all three major types of definite referents probably
correspond to the most common type of definite articles. A system with this type of definite arti-
cles typically uses the demonstrative to express deictic but also anaphoric referents. The definite
article on the other hand encodes situationally unique referents, but it can also be used for the
other two referent types. In such a system, all three types of definite referents are systematically
marked, as is shown in Figure 3.6.
3 The situation that we find in many Uralic languages with a possessive marker is very similar in that these definiteness

markers might also be definite articles of type IV.

51
 deixis demonstrative

anaphora

uniqueness definite article

Figure 3.6: Demonstrative with definite article

These were the main functions that are relevant to distinguish between different types of definite
articles and demonstratives. In addition to deixis, anaphora, and uniqueness, two other functions
may be relevant for setting up the criteria to distinguish between demonstratives and definite
articles. These two functions are the expression of distance and contrast.
Since demonstratives often encode the location of the referent with respect to the discourse
situation, the marking of distance was proposed to be a necessary criterion for demonstratives, es-
pecially regarding their distinction from articles (e.g. Anderson & Keenan 1985: 280). However, a
closer look at demonstratives revealed that distance as such is not a necessary criterion. Especially
with regard to adnominal demonstratives, we find languages with demonstratives that do not ex-
press distance (cf. Diessel 1999: 38, Diessel 2013a, Kemmerer 1999, Lyons 1999: 19f). It seems that
only adverbial demonstratives have at least a two-way spatial contrast (‘here’ vs. ‘there’), and in
many languages they can combine with an adnominal demonstrative in order to express distance
(Diessel 1999: 36). The expression of distance or spatial deixis in the strict sense can therefore be
considered to be a typical function of demonstratives in the world’s languages, but not a necessary
one. Therefore, the marking of distance cannot be used to distinguish between demonstratives
and articles. Moreover, we find article systems that formally encode spatial deixis, which makes
the latter even less appropriate as an exclusive property of demonstratives. A well-known ex-
ample is the definite article in Macedonian which marks three levels of deixis or distance,4 as is
shown in Table 3.1 below (Lunt 1952: 41, Kramer & Mitkovska 2011: 211f).

Table 3.1: Macedonian definite article paradigm

deixis singular plural
masculine feminine neuter
neutral ot ta to te
proximal ov va vo ve
distal on na no ne
4 Thedistribution of these three forms is not only conditioned by deixis but also by additional pragmatic and stylistic
factors.

52
Another criterion used to define demonstratives is the expression of contrast. For instance, Hawkins
(1978: 156f) notes that demonstratives always involve a contrast between two referents, either ex-
plicitly or implicitly. He implements the notion of contrast as a matching constraint that applies
to demonstratives, i.e. the referent is matched with another explicit or implicit referent, from
which the former has to be disambiguated. According to Hawkins (1978: 157), definite articles
automatically exclude contrast due to their uniqueness component. Similarly, Diessel (1999: 2)
notes that “all languages have at least two demonstratives that are deictically contrastive […]”,
although this does not necessarily apply to adnominal demonstratives. A traditional example
comes from German, which has the two contrastive adnominal demonstrative forms dies and das.
Functionally, however, the markers do not distinguish between different distance values. Only in
combination with the adverbial demonstratives, a distance contrast can be expressed, as is shown
for das and dies in (11) and (12), respectively:

(11) a. das hier (12) a. dies hier

that here this here
‘this one’ ‘this one’
b. das da b. dies da
that there this there
‘that one’ ‘that one’
German (Himmelmann 1997: 53)

Therefore, Himmelmann (1997: 53-62) argues against contrast as a necessary criterion for demon-
stratives. While contrast in terms of spatial deixis is arguably not a criterion for demonstratives, I
follow Hawkins (1978) in that demonstratives typically have a contrastive component, which dis-
tinguishes them from definite articles. To the best of my knowledge, contrast in connection with
definite articles has only been proposed for German. Schwarz (2009: 34) provides the following
example showing that the definite masculine article der is used to express contrastive meaning:

(13) Hans ist in DEM Auto [pointing at car 1] gekommen, nicht in DEM
Hans is in art:def.m.dat car come.ptcp neg in art:def.m.dat
Auto [pointing at car 2].
car
‘Hans came in that car, not in that one.’ German (Schwarz 2009: 34)

The capitalization of the definite articles in (13) indicates that the articles bear emphatic stress,
while definite articles in German are normally unstressed. This suggests that the elements in
(13) are not normal articles, but other elements that may qualify as demonstratives based on their
spatial deictic and contrastive function. Although treating them as articles, also Schwarz (2009: 34)
notes: “[…] the strong article has demonstrative uses in addition to the anaphoric ones. However,

53
such uses typically involve a pitch accent on the determiner, which suggests that they have a
special status” (Schwarz 2009: 34). Based on this stress difference between regular definite articles
these article-like elements in German, I do not regard them as articles.5 This is in accordance with
the crosslinguistic tendency of definite articles against the encoding of contrast, which in turn
shows that the marking of contrast may serve as an additional criterion to distinguish between
demonstratives and definite articles.

3.2 The numeral ‘one’

Not very different from the situation seen for demonstratives and definite articles, the proper
distinction between indefinite or exclusive-specific articles and the numeral ‘one’ is difficult to
make in many cases, which often leads to confusion about the ontological status of articles. Do
they correspond to a category, a function that certain markers, i.e. the numeral ‘one’, can have,
or are they markers themselves? In this study, I use the notion of articles in two ways: articles
correspond to a crosslinguistic category on a language-independent level, and on a language-
specific level, they correspond to linguistic markers. I do not view articles as functions.
That this is still an open issue can be seen by the different statuses that are ascribed to articles,
which partially results from the fact that articles are so similar to other elements. An example is
the following description from Romero-Méndez (2008: 270) on Ayutla Mixe (Mixe-Zoque, Mex-
ico): “The numeral tu’uk ‘one’ is commonly used as an indefinite article.” This description is then
illustrated with the following two examples, both introducing a not yet identifiable but discourse-
prominent referent:

(14) kuu-ëk-ja’a jä’äy y-men-y [te’n tu’uk tixytyëjk-ëk

when-qot-dem person 3sg-come-dep dem one woman-qot
poop-te’kn-ëk-ja’a] y-men-y
white-intens-qot-dem 3sg-come-dep
‘When the people came, a woman in white came.’
Ayutla Mixe (Romero-Méndez 2008: 271)
(15) jam [tu’uk mëj tsäj]
there one big stone
‘There was a rock.’ Ayutla Mixe (Romero-Méndez 2008: 271)

As the glosses indicate, I do not treat the marker as an article. As I show later in this section, I
view this as a presentational use of the numeral ‘one’, which seems to be generally available in
the world’s languages, if the language does not have a specific or indefinite article.
5 For a more detailed discussion, see e.g. Engel (1988: 535,660); Himmelmann (1997: 50-56); Bisle-Müller (1991: 62-66);
Hoffmann (2009: 312f).

54
 The distinction between indefinite articles and the numeral ‘one’ is problematic because of
the gradual diachronic development from the numeral to an exclusive-specific or an indefinite
article (cf. Heine & Kuteva 2002: 219-221; Himmelmann 2001; Dryer 2013b). Heine (1995: 71-76)
suggests the following stages for with regard to the development of the numeral and the indefinite
article:

stage 0 stage I stage II stage III stage IV

numeral presentative specific nonspecific generalized

‘one’ marker marker marker marker

Figure 3.7: Development of the numeral ‘one’ to a nominal marker

The stages from the scale in Figure 3.7 that are relevant to the discussion of indefinite articles and
the numeral ‘one’ are the stages 0 to III. Stage I markers, called “presentative markers” in Heine
(1997), are markers that signal prominent, topical discourse referents as not (yet) identifiable to
the hearer. Most instances correspond to the introduction of new referents into the discourse.
Specific markers of stage II corresponds to exclusive-specific articles.Nonspecific stage III markers
correspond to indefinite articles, being used to encode both specific and nonspecific referents. The
scale is usually assumed to be implicational if interpreted synchronically:

The model can be interpreted in a dual way. On the one hand, it may be viewed as a synchronic
implicational scale. This means, for example, that an indefinite article of a given stage also
has, or may have, the properties of all preceding stages, but not vice versa. On the other hand,
it is claimed to reflect diachronic evolution, where the initial stage represents the oldest and
the final stage the most recent situation. (Heine 1997: 71)

By analogy to the distinction between demonstratives and different types of definite article ad-
dressed in section 3.1, I argue that the distinction between the numeral ‘one’ and the article can be
made based on a cut-off point on the scale presented in Figure 3.7, using it a synchronic and not
necessarily diachronic sense. This approach is innovative insofar, as most definitions of indefinite
articles that aim at a distinction between indefinite articles and the numeral ‘one’ consist of a list
of typical features that contrast these two elements. To provide an example without discussing
them further, Table 3.2 shows the properties provided in Heine (1997) based on Heine et al. (1995)
that make reference to the numeral ‘one’ but cannot be used to distinguish between the latter and
an indefinite article.

55
 Table 3.2: Characteristic properties of indefinite articles (Heine 1997: 68)
Length Indefinite articles are generally short (≤ 2 syllables).
Stress They are stressless.
Position They are likely to employ the same position in the clause
as the numeral ’one’.
Noun types They tend to be confined to determining the singular of
count nouns.
Nevertheless, there may be exceptions where the article
has been extended to nonsingular referents.
mass > pl If the indefinite article determines mass nouns,
then it is also used for plural nouns.
pl > sg If it determines plural nouns, then it also determines
singular nouns.

Often, distinct forms or stress patterns of the marker and the numeral are taken as sufficient to
treat it as an article. Although the degree of form developments due to grammaticalization usually
correlates with the degree of semantic changes and extensions, the form development is only an
epiphenomenon to the function development and cannot be a criterion as such. The necessary
criteria should be based on the functions of the markers instead.
The different functions of the four stages the numeral ‘one’ and different types of indefiniteness
markers are shown in Figure 3.8. Again, we can distinguish between different markers based on
the combination of functions they encode.

qantification, (0) presentational (I) specific (II) nonspecific (III)

individuation

numeral

numeral

inclusive-specific article

indefinite article

nonspecific article

Figure 3.8: Cut-off point between the numeral ‘one’ and indefinite articles

56
The stage 0 marker, the numeral ‘one’, has the main semantic function of individuation and quan-
tification (cf. Givón 1981: 50), and it appears to be universal. Even though extremely simple
numeral systems are attested, the simplest that are known still consist of a lexeme for ‘one’, some-
times along with words for ‘two’, ‘three’, or ‘few’ vs. ‘many’ (cf. Stampe 1976: 596; Greenberg
1978a: 256; Dixon 1980: 107-108; Heine 1997: 24).6
For the presentational function of stage I, i.e. the introduction of new referents into the dis-
course, we can observe a strong crosslinguistic trend towards using the numeral ‘one’ especially
with single human referents (e.g. Givón 1980: 50; Heine 1997: 72; Dryer 2013b). Hence, this is a
common property of the numeral ‘one’. Crucially, this use of the numeral one is confined to ref-
erents which are prominent and salient in the discourse. A specific but not discourse-prominent
referent is not included in the function of stage I.
One example for a marker that encodes this presentational function is found in Spoken Hebrew.
Givón (1981: 36) notes that a phonetically reduced form -xad of the numeral exad ‘one’ can mark a
referent that is discourse-prominent as nonidentifiable.7 In (16), ish ‘man’ is the subject and topic
of the sentence and its referent is encoded as specific by the marker -xad. In example (17), on the
other hand, ish ‘man’ is contrasted with isha ‘woman’, which makes it a focused constituent. In
this case, ish ‘man’ occurs as a bare noun.

(16) ba hena ish-xad etmol ve-hitxil le-daber ve-hu

came here man-one.pres yesterday and-started to-talk and-he
‘A man came in yesterday and started talking and he …’
Spoken Hebrew (Givón 1981: 36)
(17) ba hena ish etmol, lo isha!
came here man yesterday not woman
‘A man came here yesterday, not a woman.’ Spoken Hebrew (Givón 1981: 36)

Another example to illustrate this presentational use of the numeral ‘one’ comes from Tikuna
(Tikuna-Yuri, Colombia) in (18). Although the use of the numeral in such contexts is possible in
Tikuna, it appears to be optional or conditioned by other factors, as we see in examples (19) and
(20).
6 The famous counter-example to this claim is Pirahã (Mura, Brazil). The two numeral markers hói and hoí were first
analyzed as lexemes for ‘one’ and ‘two’ in Everett (1992). This was supported by Gordon (2004), who claims that
the language does not encode numerals above ‘two’. In a later study, Everett (2005) rejected the analysis of these
markers as exact quantifiers, supported by results from a cognitive experimental study (Frank et al. 2008). Since it
does not seem entirely clear what the most adequate analysis of the Pirahã numeral system is (cf. Everett 2009), I
assume that the world’s generally have a lexeme that corresponds to the meaning of ‘one’.
7 Note that Heine (1997: 72f) also refers to this example from Givón (1984: 36), but discusses it as a specific marker

of stage II. I treat it as a presentational marker, since Givón shows how this marker is restricted to topical specific
referents, or at least not systematically used in other specific indefinite contexts.

57
(18) nücha dau [wii hipata hichi] llea wadiu-wa (naru) napaeru na
s:1sg.o:3sg see one.pres house old dem:dist village-loc poss roof cop
chiee
damaged
‘I see/saw a house in that village; its roof was damaged.’ Tikuna (prim. data)
(19) i-pata-wa nucha dau [(wii) güena] tare buun ɲema cuchia-wa
1sg-house-loc s:1sg.o:3sg see one.pres woman two child there kitchen-loc
‘When I came home, I saw a women with two children in the kitchen.’
Tikuna (prim. data)
(20) nukü dau [(wii) ail]!
s:2sg.o:3sg see one.pres tiger
‘Look! A tiger!’ Tikuna (prim. data)

Because the numeral ‘one’ is widely attested in this function, I view the presentational function
as an extended function of the numeral ‘one’ that is generally available in the world’s languages,
similarly to the anaphoric use of primarily deictic demonstratives. However, if referents that are
discourse prominent and specific are systematically encoded by a marker which may or may not
be formally identical to the numeral ‘one’, I treat this marker as a presentational article. I treat
presetational articles differently from the numeral ‘one’ because the systematic coding of such
referents does not correspond to the functions the numeral ‘one’ generally has (for the discussion
of presentational articles, see section 5.9).
Markers of type II correspond to exclusive-specific articles and are systematically used to mark
a referent as specific, independently of its discourse prominence. Exclusive-specific articles can be
found in e.g. Palula (Dardic, Pakistan). Examples (21) and (22) below show the article áa(k) with
a discourse prominent referent; in (23), the noun baṭ-á ‘stone’ occurs with the exclusive-specific
article although its referent is less prominent (cf. section 5.5.2).

(21) bhuná [áa ɡíri] heensíl-i

below art:exspec rock stay.pfv-f
‘Down below there was a big rock.’ Palula (Liljegren 2016: 309)
(22) miír thaní [áak míiš] heensíl-u de
Mir qot art:exspec man stay.pfv-m.sg pst
‘There was a man called Mir.’ Palula (Liljegren 2016: 309)
(23) eesé zanɡal-í [áa baṭ-á] ǰhulí harí so kuṇaák
rem forest-obl art:exspec stone-obl on take.away.cv def.nom.m.sg child.nom.m.sg
bheešóol-u
seat.pfv-m.sg
‘In that forest he took the child to a stone and seated him.’ Palula (Liljegren 2016: 283)

58
Another type of articles that can be distinguished based on the scale in Figure 3.8 is the indefinite
article, which encodes specific and nonspecific referents. The marking of both referents and the
resulting ambiguity was shown for English in example (20) which is repeated here as (24). For
more examples of indefinite articles, see section 5.8.

(24) John wanted to marry a rich woman …

a. …but she refused him. (specific)
b. …but he couldn’t find any. (nonspecific)

The last type of indefinite articles presented in Figure 3.8 are nonspecific articles. These articles
mark only nonspecific referents and do not occur with specific referents. An example of a non-
specific article in Q’anjobal was presented (32), it is repeated in (25) below (for more examples of
nonspecific articles, see section 5.7).

(25) ay mi junoq yatut thioxh b’ay yich calle?

exist q art:nspec house.of.God at back street
‘Is there a church at the end of the road?’ Q’anjobal (prim. data)

As we saw in this section, the numeral ‘one’ and different types of articles from the indefinite
domain can be distinguished based on the functions and distributions that the markers have, as-
suming the relevant functions or semantic values of such markers to be individuation (quantifi-
cation), the presentational function, specificity, and nonspecificity. Therefore, we do no longer
depend on a catalogue of unrelated properties to define indefinite articles and delimit them from
the numeral.

3.3 Negative polarity items

Negative polarity items (NPIs) are a heterogeneous group of elements such as any, ever, at all,
either, and yet. What all these elements have in common and what motivates their label is that
they are sensitive to polarity and that they typically occur in environments of negation. Some
examples from English are given in (26) to (28).

(26) a. Peter has not read any linguistics paper today.

b. Peter has read *any linguistics paper today.
(27) a. Peter does not want to read about semantics at all.
b. Peter wants to read about semantics *at all.
(28) a. I do not see anyone.
b. I see *anyone.

59
The group of polarity-sensitive elements that is relevant for the purposes of this section can be
narrowed down to the ones that occur in the nominal domain together with a noun, as does any
in (26). Being sensitive to polarity does not restrict the occurrence of NPIs to contexts of negation.
Since NPIs have been a popular topic in semantics during the last 40 years, I cannot mention and
summarize all the aspects of NPIs that previous work has dealt with.8 In this section, I focus on
the aspects that are important with regard to nonspecific articles. The relevant uses of NPIs cover
a wide range of contexts (Haspelmath 1997: 52; Givón 1984: 441-449; Giannakidou 2017), and are
given in Table 3.3.

Table 3.3: Relevant contexts for adnominal NPIs

negation direct
I didn’t see any cake in the fridge.
indirect
I don’t think there was any cake in the fridge.
Tabea wrote a dissertation without having any coffee.
inherent negative predicates
Tomas failed to write any paper.
question Did you drink any coffee today?
conditional If you see any problems with my analysis, please let me know.
irrealis TAM marking, modals
He could have bought any dog.
inherent irrealis predicates
She is looking for any student who can help her with the organization.
comparison Cheetahs can run faster than any human.
free choice9 You can take any cookie.

Without going into detail with respect to the semantic and pragmatic conditions of each type
of contexts illustrated in Table 3.3, we can make use of the concept of veridicality proposed and
argued for in Zwarts (1995) and Giannakidou (1997, 1998), which accounts for these different uses
of adnominal NPIs such as any in English. As Giannakidou (2017: 21) puts it: “[a] veridical context
8 Work on NPIs goes back to Baker (1970), Fauconnier (1975), Horn (1972), Ladusaw (1980) who discuss NPIs in
English focusing on negation contexts. Amongst others, Giannakidou (1997) and Zwarts (1981, 1995) extended the
study of NPIs to other polarity domains outside of negation and free choice contexts, also arguing for the distinction
between different types of NPIs. For recent overviews on the topic and further references, see e.g. Giannakidou
(2017), Hoeksema (2013), Horn & Kato (2000).
9 I will not be concerned with two uses that are usually attributed to NPIs: free choice contexts as you can take any

cookie and comparisons as in cheetahs can run faster than any human. The reason for excluding such contexts is
that even though they are often expressed by NPIs, we find languages that do not use NPIs, but may have a separate
marker. I take these contexts to be functional extensions and not a core function of NPIs.

60
is one that allows the speaker to infer the truth of a sentence; a non-veridical context is one where
truth inference seems to be suspended”. She provides the following example for illustration:

(29) a. Unfortunately, Mary saw a snake. ⇒ Mary saw a snake.

b. Maybe Mary saw a snake. ⇏ Mary saw a snake.
(Giannakidou 2017: 21)

In example (29a), also in the presence of a “factive” adverbial, the truth value of the utterance
(with respect to the discourse universe) can still be assessed which makes the utterance veridical.
The utterance in (29b), on the other hand, lies within the scope of the modal maybe, which blocks
the inference of the truth value of the proposition expressed by that sentence. This is an example
of a nonveridical statement.
Hence, non-veridicality can be taken as the property of propositions that licenses NPIs. Nega-
tion is a special case, as it does not only involve non-veridicality (the truth value of p cannot
be assessed) but antiveridicality, which asserts that the proposition does not hold (p is not). In
addition, Giannakidou (2017: 7) mentions an anti-episodicity constraint: if the proposition makes
reference to a single past event, even e.g. negation cannot license NPIs as, e.g., free choice items:

(30) a. Any cat hunts mice.

b. ?Any cat hunted mice.

In (30a), any makes the referent nonspecific as a free choice marker, entailing that every referent
that belongs to the kind of ‘cat’ can be referred to. Such an interpretation however is only available
because the proposition type allows for reference to multiple nonspecific events. In (30b), on the
other hand, the proposition is tied to a single past event, which blocks the nonspecific free choice
interpretation of any and makes the sentence infelicitous.
As I will show in section 5.7, the contexts for nonspecific articles are exactly the same. They
signal that there is no particular referent that is linked to the expression used, but that each
potential referent of its kind can be linked to the expression.
The distinctive property between adnominal NPIs such as any in English and nonspecific ar-
ticles is their distribution. Languages such as English, where the indefinite article a, depending
on the context, leaves the referent ambiguous between a specific and nonspecific interpretation,
do not need to resolve the ambiguity between a specific and nonspecific referent. For this reason,
we can use the indefinite article a or a bare noun instead of the NPI any in all contexts that were
presented to be relevant to NPIs in Table 3.3:

(31) a. I did not see cake in the fridge. (negation)

b. I don’t think there was cake in the fridge. (neg-raising)

61
 c. Tabea wrote a dissertation without having coffee. (indirect negation)
d. Tomas failed to write a paper. (negative verb)
e. Did you drink coffee today? (question)
f. If you see a problem with my analysis, please let me know. (conditional)
g. He could have bought a dog. (irrealis)
h. She is looking for a student who can help her with the organization. (irrealis verb)

Of course, even though the utterances from Table 3.3 can occur with indefinite articles or bare
nouns instead of NPIs with their truth value being the same (Chierchia 2013: 27), there are dif-
ferences between the interpretations of the sentences in Table 3.3 and in example (31) above. We
find differences of two types. Either, using the article or a bare noun, the referent is ambigu-
ous between a specific and nonspecific interpretation in the contexts in which no other explicit
linguistic clue for nonspecificity is provided. We see this in (31g) and (31h). Or, NPIs can in-
volve pragmatic effects that are absent in the sentences in (31). Such effects have been linked to
exhaustivity (e.g. Chierchia 2013: 27) and alternatives similar to focus (e.g. Chierchia 2013: 34,
Giannakidou 1998: 81). What is important for the purposes of the present section is the fact that
although NPIs generally seem to be available in the languages of the world in the same contexts
as nonspecific articles occur in, we could see for English that NPIs are not necessary in the sense
that there are linguistically simpler alternatives (bare nouns, the indefinite article) that can have
the same semantic interpretation.
To show that we see a similar effect in languages without articles, the following Polish exam-
ples (32) to (37) show that we can equally find corresponding expressions of nonspecific referents
either marked by an NPI (żaden, jakiś, or jakikolwiek) or left as a bare noun:10

(32) Nie widziałam w lodówce (żadnego) ciasta.

neg see.pst.1fsg in fridge.loc (npi.gen.n) cake.gen
‘I didn’t see (any) cake in the fridge.’ Polish (prim. data)11
(33) Tabea powiedziała to bez (żadnej) ironii.
Tabea say.pst.fsg it without (npi:gen.f) irony.gen
‘Tabea said it without (any) irony.’ Polish (prim. data)
(34) Tomas odmówił napisania (jakiegokolwiek) artykułu.12
Tomas decline.pst.3msg writing.gen.n (npi:gen.m) article.gen
‘Tomas declined to write (any) articles.’ Polish (prim. data)

10 Being beyond the scope of the present study, I do not address the functions and distributions of different NPIs and
free choice items in Polish here. For a discussion of NPIs in Polish, cf. Błaszczak (1999, 2003).
11 I thank Joanna Zaleska for providing these Polish examples.
12 The referent that belongs to the word artykuł ‘article’ is ambiguous between a definite and nonspecific interpretation

without the NPI jakiegokolwiek and further context.

62
(35) Czy piłaś dzisiaj kawę?13
q drink.pst.2fsg today coffee.acc
‘Did you have coffee today?’ Polish (prim. data)
(36) Daj mi znać, jeśli zauważysz (jakieś) błędy w mojej analizie.
give.imp me know if notice.fut.2sg (npi:acc.mpl) errors in my analysis
‘Let me know, if you notice (any) errors in my analysis.’ Polish (prim. data)
(37) Szuka (jakiegoś) studenta, który mógłby jej pomóc z organizacją.
look.3sg (npi:gen.m) student.gen which.nom could.3msg her help with organization
‘She is looking for a (any) student who can help her with the organization.’
Polish (prim. data)

After this digression on NPIs and their contexts, we return to the initial question of this section, i.e.
the difference between NPIs and nonspecific articles. As noted earlier, the semantics (referential
functions) of NPIs do not differ from the ones of nonspecific articles. Their difference rather lies
in the additional pragmatic effects that NPIs have and nonspecific articles lack.
The preceding paragraphs showed that although any in English is a dedicated marker of non-
specificity because it occurs in various nonspecific contexts, it is not systematically required to re-
solve the referential vagueness between specific and nonspecific referents. Generally, bare nouns
or nouns with an indefinite article can also be used in nonspecific contexts without resolving the
ambiguity. In languages with no indefinite article, referents expressed by bare nouns can also
receive a nonspecific interpretation, which was exemplified by the Polish examples in (32) to (37).
However, NPIs like any in English or żaden, jakiś, and jakikolwiek in Polish are not necessarily
required, since a simpler expression is available.
On the other hand, nonspecific articles occur systematically to resolve this ambiguity between
a specific and a nonspecific referent.As was already shown in section 2.2.2, Q’anjobal has a non-
specific article. Example (39a) illustrates that the specific indefinite article jun is not acceptable in
nonspecific contexts. Its nonspecific counterpart junoq, on the other hand, is obligatorily used in
contexts in which the referent is interpreted as nonspecific. Moreover, (39b) shows that the use of
a bare noun is not allowed either. This suggests that, indeed, jun is a specific and not an indefinite
article, and that junoq is not an adnominal NPI, but a nonspecific article, being systematically
required to mark the referential value of nonspecificity.
(38) tzeb’ach yul [jun/ *junoq tuktuk]. mayal wawrtej naq tz’umon ch’en.
come.imp in art:spec art:nspec tuktuk already called cl driver
‘Let’s take a tuktuk. I already called the driver.’ Q’anjobal (prim. data)

13 In polar questions, NPIs do not seem acceptable in Polish.

63
(39) a. asi’ yul [junoq/ *jun tuktuk].
go.imp in art:nspec art:spec mototaxi
‘Take a/any tuktuk.’ Q’anjobal (prim. data)
b. *asi’ yul tuktuk.
go.imp in mototaxi
‘Take a/any tuktuk.’ Q’anjobal (prim. data)

Thus, if there is no systematic need to resolve the ambiguity between specific and nonspecific
referents in the noun phrase, markers that can occur to mark the referent as nonspecific are not
treated articles. If referents have to be marked as nonspecific systematically, I treat such markers
as nonspecific articles.

3.4 Summary
In this chapter, I pointed out how certain types of articles can be distinguished from similar ele-
ments. Section 3.1 dealt with the delimitation of anaphoric and definite articles from adnominal
demonstratives. This distinction was shown to be problematic due because of their diachronic
connection. I argued that deictic markers which can also mark anaphora are not articles, but
should be treated as demonstratives. Anaphoric markers that cannot be used to encode deixis, on
the other hand, are treated as anaphoric articles. The third main type of articles in the definite
domain are definite articles, marking anaphora, situational uniqueness, and often deixis as well.
In section 3.2, I defined the criteria necessary to delimit the numeral ‘one’ from exclusive-
specific or indefinite articles. I argued that numerals which can extend to presentational uses
should not be considered as articles, because this functional extension of the numeral ‘one’ seems
to be generally available across languages. If discourse-prominent and nonidentifiable referents
are systematically encoded, however, I treat such markers as presentational articles, also if they
formally correspond to the numeral ‘one’. If a marker is systematically used to encode specificity
independently of the discourse prominence of the referent, it is regarded as an exclusive-specific
article. If it marks referents as nonspecific in addition, it corresponds to an indefinite article.
Section 3.3 discussed the distinction between nonspecific articles and negative polarity items
(NPIs). Negative polarity items are typically licensed under the scope of polarity, but other main
contexts for NPIs include questions, conditionals, and irrealis contexts. Adnominal NPIs have a
similar functions to the markers that I treat as nonspecific articles. I argued that the difference
between nonspecific articles and NPIs thus does not lie in their referential functions, but in the sys-
tematicity of their distribution in nonspecific contexts. Nonspecific articles occur systematically
to encode a referent as nonspecific, while NPIs are not systematically required in such contexts,
and if they occur, they often have additional pragmatic effects that nonspecific articles lack.

64
Chapter 4

Articles and referentiality

4.1 Referentiality and referent types

In this section, I discuss referentiality and the concepts that are relevant to distinguishing between
different article types on the basis of their referential function. Referentiality is understood as a
linguistic category of reference, which concerns the link between a linguistic expression to an
abstract object in the discourse, the discourse referent. I call the values of referentiality, i.e. the
functions that referring expressions have, referent types. In formal semantics, referentiality has
usually been approached from the sentence level. The study of referentiality restricted to the
sentence level is is traditionally located within semantics, also labelled denotational semantics
(cf. Kroeger 2018: 17), while discourse reference has been viewed as belonging to pragmatics (e.g.
Givón 1981, 1984, Hopper & Martin 1987). In this study, I do not engage in the discussion of
distinguishing these two levels of sentence and discourse referentiality and their location within
semantics or pragmatics. In doing so, I follow Hawkins’s agenda: “By looking at the total picture,
instead of arbitrarily excluding either pragmatics or logical considerations, one is led more easily
to the formulation of significant generalisations” (Hawkins 1978: 90).

4.1.1 The discourse model

This section proposes a discourse model that is necessary to discuss referentiality and to distin-
guish between different referential functions (referent types), which in turn is the basis for the
distinction between different types of articles. I start with setting up some basic discourse-related
notions: the discourse situation, the participants (hearer and speaker), and the discourse referents.
To illustrate how the discourse referents relate to the participants and the discourse itself, let us
consider the following example in (1).

65
(1) Context: S and H are chatting at the bus stop, waiting for the bus to come.
A man approaches the bus stop. S is saying to H:
S: That man is tall!

The discourse situation in (1) is described by the context, the discourse participants are the speaker
(S) and the hearer (H). The sentence uttered by the speaker contains the linguistic expression that
man, which is linked to the discourse referent Rman . Figure 4.1 shows how this discourse situation
(as an example of discourse situations in general) can be represented schematically.

discourse universe
hearer’s mental space MH

Rbus.stop Rspeaker R...

Rhearer Rman
R... R...
Rbus

speaker’s mental space MS

x x
utters hears
“That man is tall!”
S H
discourse situation

Figure 4.1: Discourse situation and discourse universe

For reasons of simplicity, I assume that a discourse situation involves two discourse participants,
i.e. the speaker and the hearer. The linguistic expressions (words) that the speaker utters are also
part of the discourse situation, in this case it is the utterance of the sentence That man is tall! In
order to properly distinguish between form and function, referents1 are represented on a different
level than the referring linguistic expressions. Referents are defined as (abstract) elements in the
discourse universe. The discourse universe is understood as an abstract set containing all referents
from the discourse. The referents, even though being elements in the discourse universe, are
linked to the linguistic expression uttered by the speaker. In example (1), the referent Rman is
linked to the expression that man. The discourse universe, as a set of all discourse referents,
can further be partitioned into two overlapping subsets, which I call the speaker’s and hearer’s
mental spaces MS and MH . In section 4.1.3, I show that it is necessary to distinguish between
the referents that are identifiable by the speaker and the hearer in order to define the notion of
definiteness. As is sketched in Figure 4.1, the referents that are identifiable by the speaker and
1I use the shorter expression “referents” in the meaning of “discourse referents”.

66
the hearer are modelled as elements of two intersecting subsets of the discourse universe, the
speaker’s and hearer’s mental spaces, respectively. Referents from the discourse universe that are
identifiable by both the speaker and the hearer are elements in the intersection of the two mental
spaces. Because the two sets of mental spaces only overlap and are not entirely identical, they
allow for referents that are only identifiable by one of the discourse participants.
Since the discourse model that I propose here relates to a number of previous models of refer-
entiality and of the discourse, the next paragraphs discuss a number of important similarities and
differences between my and previous approaches.
The separation of the linguistic expression and the referent is crucial for the discussion of
articles as linguistic expressions that encode referential functions.2 If not distinguished prop-
erly, the crosslinguistic comparison of articles is problematic. For instance, in their influential
paper, Gundel et al. (1993: 275) directly map referential functions3 to linguistic expressions in
English. Therefore, the referential functions rely on linguistic expressions in a single language
and cannot be applied directly to other languages. For this reason, I define referential functions
form-independently as referent types in section 4.2. These referent types can be applied to any
linguistic forms in any language. This is important because the referent types build the basis for
a language-independent definition of different article types (cf. section 4.3).
Modelling discourse referents as abstract objects in an abstract space (here: the discourse
universe) is not a new approach, even though it goes against the tradition of formal semantics.
Going back to Frege (1892), propositions as events about entities (referents) are evaluated as true
or false against the real world in formal semantics and logic. Early views on referentiality and
definiteness based on uniqueness (cf. Russell 1905, Strawson 1950) assumed that referentiality
was tied to the existence of a referent in the real world. This is problematic in contexts such as
the famous sentence in Russell (1905) The king of France is bald. If no king of France exists in
the real world, the truth value of the proposition cannot be assessed. Formal approaches aimed at
resolving this issue with what is called “possible worlds semantics”.4 In possible worlds semantics,
the truth value of a proposition about a referent can be interpreted with respect to any possible
world. The approach that I follow in this study, on the other hand, does not tie denotation to
existence as such and follows a different tradition arguing for referents being interpreted with
respect to a discourse universe instead of the real world or some other possible world. To this
end, Karttunen (1968, 1976) introduced the notion of the “discourse referent”. Thus, referentiality
is always interpreted with respect to the discourse itself, for which Givón (1984) introduces the
2 Krifka & Musan (2012a: 6) made this point on a related note: “ For some reason, this confusion of expression and
meaning occurs particularly often with information structural notions. With notions like subject, predicate or direct
object it does not arise; no one would claim that John the person is the grammatical subject […], it is John the noun
phrase. The imprecision of information structure terms can be endured if one is aware of it.”
3 Gundel et al. (1993) call them giveness statuses.
4 Cf. Carnap (1947), Hintikka (1967), Kaplan (1989), Kripke (1959), Lewis (1970), Montague (1970).

67
notion of discourse universe.5 Givón (1984: 438) argues that representing referents on the level
of the discourse universe does not only make the analysis of referring expressions simpler, but is
actually grounded in language use. To illustrate this, he provides the following examples:

(2) a. The present king of France is not bald.

b. The present queen of England is not bald.
English (Givón 1984: 438)
(3) a. I rode a unicorn yesterday.
b. I rode a horse yesterday.
English (Givón 1984: 438)

We see in (2) and (3) that even though the sentences have a different interpretation with respect
to their truth value in the real world (the referents in (2a) and (3a) do not exist, as opposed to
the ones in (2b) and (3b)), language does not seem to treat them differently. Both referents in (2)
are expressed by the definite article, while both nouns in (3) occur with the indefinite article. As
Givón (1984: 438) puts it:

In human language, it seems, reference–Russell’s denotation–is not a mapping from referring

linguistic expressions to individuals existing in the RW [real world]. Rather, it is a mapping
from linguistic expressions to individuals established verbally in the Universe of Discourse.
(Givón 1984: 438)

Another important concept with respect to referents is their identifiability, which I relate to the
mental spaces of the speaker and the hearer. The concept of “mental space” goes back to Faucon-
nier (1994, 2007) and represents a semantic but cognitively plausible concept. Fauconnier (1997:
11) defines mental spaces as “[…] partial structures that proliferate when we think and talk, al-
lowing a fine-grained partitioning of our discourse and knowledge structures.” In other words,
mental spaces contain our knowledge, beliefs and attitudes of the things we experience and talk
about (Langacker 2008: 41).

4.1.2 The link between linguistic expressions and referent types

Referents are thus elements in the discourse universe, but with a link to linguistic expressions.
Since articles are linguistic expressions that are used to encode certain referent types, the nature
5 In this study, I use this notion of discourse universe.
Other related notions and concepts are “common ground”, used
in Krifka & Musan (e.g. 2012b) and going back to Karttunen (1974), Lewis (1979), Stalnaker (1974). This notion is
closely related to the broader concept of “grounding” in Cognitive Grammar (cf. Clark & Brennan 1987, Langacker
1987: 126-128, Langacker 2001, Langacker 2002, Langacker 2008: chapter 9). Similarly, Hawkins (1978, 1991) makes
use of a set called “P-set” (pragmatic set) that contains all referents known to both speaker and hearer.

68
of this link is important. This section proposes a way to define the link between referents and
referring linguistic expressions. Since it would surpass the purposes of the present study, the goal
of this section is not to offer a cognitively adequate model. It is intended as an explicit proposal
of how referential functions, i.e. referent types, can be linked to linguistic expressions. For the
sake of simplicity, I use the notion of word here in the meaning of referring expressions. I do
not restrict myself to words in the traditional sense; a word can stand for other units of referring
expressions (e.g. phrases, article-noun combinations, larger event-encoding units etc.) as well.
Words are defined as object with certain properties that can be represented in a typed feature
structure.6 Feature structures are used in different linguistic frameworks such as Head-driven
Phrase Structure Grammar or Construction Grammar. Without directly imposing any theoreti-
cal assumptions, they have the advantage of allowing a formal and explicit notation of relations
between abstract types, their features, and their values. Feature structures are recursive and the
values of features can be types again, having their own feature structure. For the purposes of the
present study, words can be defined as having the type shown in Figure 4.2 with referentiality
(ref) as one of its features. At the same time, referentiality takes an object of the type referent.
The latter contains information of the referential function of the word.
 
word
 
phon phon-object 
   
 sem-object 
 
   
sem  lex 1  
 
 … … 
 
 
syn syn-object 
  
 referent 
 
 referent.type referent.type-object 
  
   
  concept-object 
ref  
   
 concept  lex  
}
1
   { 
 
kind K := R1 , R2 , R3 , . . . , Rn

Figure 4.2: Structure of a referring word

An object of type word has the following features: phon, sem, syn, and ref. The feature phon
takes a phonological object as its value and specifies the phonological form of the word. The
feature syn represents syntactic properties, e.g. case, but also the presence of modifiers or articles
in the noun phrase could be included here. The value of sem corresponds to an object of the type
sem-object, specifying the lexical information of the word and other semantic properties. In that
6 For an overview on feature structures, see (Müller 2010: chapter 5).

69
sense, the value for lex corresponds to ungrounded “concepts” as defined in Langacker (2008: 33f).
The feature ref takes an object of type referent as its value, which contains information on the
referential functions of the word. This is how the referents as elements of the discourse universe
as defined in the previous section are built into the properties of a word. Thus, the “link” between
expressions and referents consists of the ref feature of an object of type word which in turn takes
an object of type referent as its value.
A discourse referent of type referent is characterized by the features referent.type and con-
cept. The value of referent.type defines the referent type as a ref.type-object. I define different
referent types in section 4.2; they are important since they serve as the basis for the distinction of
different article types, as is shown in section 4.3. The other feature of the type referent is concept,
taking an object of type concept-object as its value.
An object of type concept-object contains information about the lexical meaning in its feature
lex. The lexical meaning specifies the concept of the referent, understood as ungrounded lexi-
cal meaning in the sense of Langacker (2008: 43ff). The feature kind of the concept-object refers
to all referents that belong to the same kind as the referent of the word in question. In other
words, the value of kind can be defined as a set, namely the kind set K, which contains all ref-
erents that belong to the concept of the word. Figure 4.2 indicates this with the set notation as
K := {R1 , R2 , R3 , . . . , Rn }. This means that each referent, being of type referent, of any referring
expression of type word contains the relation to all (potential) referents that belong to the same
concept. While the implementation of the kind set K into the feature structure of a referring
expression of the type word may seem unnecessary at this point, I show in section 4.2 that the
distinction between different referent types relies on the concept of a kind set K. Therefore, the
latter is included into the properties of the referent, corresponding to an object of type referent in
this feature structure representation.
Thus, this section proposed a way to express what is meant by linguistic expressions having
referential functions, or by the “link” between the referent and the referring expression, e.g. a
noun.

4.1.3 Identifiabilty as a core concept of referentiality

Traditionally, the literature on referentiality has been split between two major approaches based
on uniqueness and familiarity. In the first part of this section, I define the concept of identifiability
which can be related to both uniqueness and familiarity and which is necessary for the distinction
of different referent types, namely definite and specific referents. In the second part of this section,
I briefly introduce uniqueness and familiarity and argue why neither concepts are sufficient for
the purposes of this study.

70
4.1.3.1 Defining identifiability

Identifiability has to do with the knowledge of the speaker and the hearer about certain properties
of the referent. Note that the concept of identifiability has received various labels in the literature,
e.g. “familiarity” (e.g. Christophersen 1939), “givenness” (e.g. Gundel et al. 1993), “accessibility”
(Ariel 1988: e.g.), or “knowledge”. I use “ identifiability” in this study, but it could in principle be
replaced by any of the other labels.
Identifiability, representing knowledge about the referent, can be understood in the sense
of “manifest” as defined in Sperber & Wilson (1986: 39), taken up by Hawkins (1991). Thus, a
definition of identifiability based on (Sperber & Wilson 1986: 39) is presented below:

(4) Identifiability
A fact is identifiable by a discourse participant at a given time if and only if she is capable
at that time of representing it mentally and accepting its representation as true.

Identifiable facts denote online-knowledge and commitment to that knowledge by a discourse

participant. I will use “know” and “knowledge”, as well as “identify” and “identifiability” in that
sense. Combining the notion of identifiability with the discourse model that was presented in
section 4.1.1, we can define identifiable referents as those referents which are part of the mental
spaces of the speaker and/or the hearer.7

(5) Identifiable referent

A referent is identifiable by the speaker if the referent is an element of her mental space
MS , and the referent identifiable by the hearer if the referent is an element of her mental
space MH .

As various authors have pointed out, the discourse universe is not a stable and fixed structure,
but dynamic and constantly updated during the discourse (e.g. Giannakidou 1998, Hawkins 1991,
Krifka & Musan 2012a, Langacker 2008). Hawkins (1991) shows that, even though a referential
expression is linked to a referent that cannot be fully identified by the hearer based on her knowl-
edge at the point of utterance, the hearer is trying to establish identifiability if the referent is
presented as definite. This capacity helps to construct and update the discourse universe and has
been labelled “accommodation” (Karttunen 1974, Krifka & Musan 2012a, Lewis 1979, Stalnaker
1974). It corresponds to a repair strategy to rescue the interpretation of an utterance. Hawkins
(1991: 413) illustrates hearer accommodation using the following example:
7 Thedefinition of an identifiable referent in (5) equally applies to a group of referents of the same kind, which means
that the concept of identifiability can also be applied to the definiteness of plural expressions.

71
(6) Context: Suppose Mary and Peter are looking at a landscape where she has noticed a distant
church.
M: I’ve been inside that church.

Hawkins (1991: 411) points out that the use of the demonstrative in such an utterance is fully
acceptable:

The speaker is simply telling the hearer to extend the relevant P-sets [here: discourse universe]
on these occasions, and expects the hearer to accept these extensions, presuming that they
are at least compatible with what he knows and that he has no reason not to accept them.

It is not required for Mary to ensure that Peter first sees the object ‘church’ in the real world to be
able to identify the referent Rchurch linked to the expression that church. Due to the marking of the
referent Rchurch as definite and thus identifiable, Peter tries to establish identifiability relying on the
fact that Mary marked the referent as definite because he can identify it as well. One consequence
for the nature of the discourse universe is that it needs to be dynamic and updatable. The other
consequence is that identifiability cannot only be marked by certain linguistic expressions, but
that it can also be established.
Another property of the concept of identifiability is that it is context-sensitive, which means
that depending on the context, the speaker’s and hearer’s knowledge of a referent can be sufficient
for full identification or not. It seems to be commonly assumed that deixis (or physical perception
in general) is sufficient for a referent to be identifiable.8 It is in this vein that nouns accompanied
by demonstratives are taken to be definite, since demonstratives typically involve deixis, which
entails knowledge of physical properties based on visibility. For the majority of instances, it seems
to hold true that visibility licenses mutual identifiability by the hearer and the speaker. There are,
however, counter-examples, illustrating that identifiability must be context-sensitive and is not
automatically licensed by visibility. In example (7), the referent Rboy of the expression a boy is
linked to a visible object (‘a boy’) in the discourse situation.

(7) Context: S and H are on a ship in the ocean, looking at the water, when S spots a boy in the
water.
S: Look! There is a boy in the water! (Pirates of the Caribbean 1, movie)

Nevertheless, as the expression referring to Rboy includes the indefinite article a, the referent
Rboy is marked as nonidentifiable, which in turn suggests that visibility is not always sufficient to
license full identifiability.
8 Note that I treat referents and objects in the real world as different elements on different levels. Thus, strictly
speaking, a referent can never be perceived directly, it can only be linked to an object from the world that is visible
(cf. section 4.2).

72
4.1.3.2 Other approaches

The main focus in semantics with respect to referentiality has probably been on definiteness,
mainly divided introduces uniqueness and familiarity as concepts to explain definiteness.
Uniqueness-based accounts of definiteness go back to Russell (1905) and basically require a
definite description to be unique. In Russell (1905), this uniqueness condition is directly built
into the truth condition of a definite description. The meaning of the famous sentence The King
of France is bald equals the conjunction of the following three statements: (i) there is a King of
France and (ii) there is only one single King of France and (iii) the King of France is bald. The
quantificational aspect of the definite article (the fact that uniqueness is part of its denotation, i.e.
meaning) has been subject to criticism, so that a more “moderate” version of the denotation has
been proposed in Frege (1892) and Strawson (1950). There, uniqueness is not part of the denotation
of the definite description any longer, but is a presupposition for the felicitous use of the definite
description (as a precondition for the element that must be met for its successful use). Taking the
example from (7), this means that there has to be one and only one king of France, and if the king
of France is bald, the proposition is true. If he is not bald, the proposition is false. If there is no
unique king of France, the truth value of the sentence is not defined.
Independently of how exactly uniqueness is used to define definiteness, it straightforwardly
accounts for cases where the discourse referent marked as definite somehow stands out as a single
prominent referent. However, two interrelated problems arise with this account. On the one
hand, uniqueness as such does not provide any information on the conditions under which a
certain referent can be unique. And, moreover, referents can be marked as definite although they
are not unique in the entire world, so that uniqueness must be restricted to a certain domain
relevant in a discourse situation. Heim & Kratzer (1998) implement this in a way that a referent
expressed by a noun and a definite article does not have to be unique in the entire world, but
in a contextually salient subset of the world. However, this does not straightforwardly show
what the exact conditions have to be for the referent to be salient. Therefore, also this version
of contextual uniqueness needs an additional mechanism that regulates it with respect to the
discourse situation.
Anaphoric and therefore salient but not necessarily unique referents have received a lot of at-
tention in the semantics literature, which has led to a competing approach to definiteness, namely
the familiarity approach. For instance, Christophersen (1939) points out that instead of unique-
ness, rather familiarity with (or non-ambiguity of) the referent should be the prerequisite for
definiteness. This has been taken up by Karttunen (1968) and the influential model of File Change
semantics (a discourse model that accounts for definite and indefinite uses of referents) in Heim
(1988). What exactly makes a referent familiar differs from account to account; in Heim (1988,

73
2002), familiarity is licensed by previous mentioning in the discourse, i.e. anaphoric uses. This is
also taken to be the distinctive property of definites as opposed to indefinites:

A definite is used to refer to something that is already familiar at the current stage of the
conversation. An indefinite is used to introduce a new referent. (Heim 2002: 223)

Without going into details of the formalization of this framework here, it can be mentioned
that the elaborate formalization offers a precise, cognitively grounded mechanism to account
for anaphoric definites and newly introduced indefinite referents, which uniqueness-based ap-
proaches could not deal with i a straightforward way. On the other hand, there are a number of
contexts with definite referents that this framework cannot deal with either. Examples of these
are briefly given below and discussed in detail in section 5.2.

(8) a. The man drove past in a car. The horn was blaring loudly.
b. The professor that we were just talking about …
c. Pass me the bucket that is over there.
(Hawkins 1991: 410)

Although the discourse referents in (8) are newly introduced to the discourse, they are expressed
by a noun together with the definite article, marking the referents of these nouns as definite. This
shows that also the purely familiarity-based approaches9 such as File Change semantics cannot
capture the variety of contexts that involve definiteness, as they focus on the (dynamic) semantic
status of the discourse referent, but not on the discourse participants, their relation to the referent,
their intentions etc.
Therefore, uniqueness or familiarity alone are not sufficient to account for definite contexts
and to explain the distribution of definite expressions in the discourse. The concept of familiarity
seems to be rather restrictive in that it focuses on anaphoric definites. Therefore, one would need
to add an entirely different mechanism to capture the different non-anaphoric contexts of definites.
Uniqueness, on the other hand, is less restrictive, but not entirely applicable to the naturalistic
use of language, since the domain of uniqueness as restrictive and relativizing factor must be
implemented in some way to make it compatible with different discourse situations. However, a
slightly different interpretation of uniqueness is unambiguous identifiability, which is noted by
Christophersen (1939) and emphasized by Löbner (1998).
9 Roberts (2003) elaborates on the familiarity requirements proposed in Heim (1988) distinguishing between two lev-
els of familiarity. Without going into detail, weak familiarity subsumes discourse referents that are contextually
accessible or given by world / cultural knowledge, while strong familiarity refers to anaphoric uses. This shows
that in order to make the concept of familiarity applicable to more contexts involving definites, a more elaborate
mechanism has to be applied. Although being able to capture phenomena such as bridging, this version can still not
account for those uses of definites that uniqueness can do.

74
 Though the previously acquired knowledge may relate to the very individual meant, yet it
is often only indirectly that one is familiar with what is denoted by the word. It may be
something else that one is familiar with, but between this ‘something’ and the thing denoted
there must be an unambiguous relation. (Christophersen 1939: 72f)

Here it is: the crucial ‘unambiguous’ , i.e. 1-to-1, relationship between the trigger and the
associate, based on general knowledge. (Löbner 1998: 12).

For this reason, uniqueness and identifiability are not necessarily two entirely different concepts;
and unambiguous identifiability is used in section 4.2 to define definite referents.
Indefiniteness, even more than definiteness, has been discussed strictly in combination with a
specific linguistic form, namely the indefinite article a in English in most cases. Heim (1991: 32),
for instance, argued that the indefinite article, in addition to the existence presupposition which
it shares with the definite article, has a non-uniqueness condition. This is based on examples like
in (9), for which she argues that the indefinite article is not felicitous because it is known that
there is can only be single referent.

(9) a. I interviewed the (#a) father of the victim.

b. The (#a) weight of our tent is under 4 lbs. (Heim 1991: 32)

To make the conditions on the indefinite article more restrictive (and distinct from the one of the
definite article), Heim (1991: 32) proposes a non-uniqueness condition for the indefinite article.
Note that the examples in (9) can also be accounted for by a more elaborate distinction of noun
types, as in Löbner (1985, 2011), in which nouns like father or weight are classified as functional,
which means that they have an argument that inherently relates them to a different referent and
makes them contextually unique, or definite, accounting for the incompatibility with the indefinite
article.
Other scholars have argued for a different meaning of the indefinite article. Hawkins (1991:
420ff) shows that the indefinite article can implicate uniqueness in some contexts, whereas in
others, it cannot. Examples for these two types of contexts are shown in (10), respectively.

(10) a. There is a pizza in the fridge, and a cake in the pantry.

b. Pass me a bucket. ~Pass me one of the buckets (before us). (Hawkins 1991: 417ff)

In (10a), although the referents of pizza and cake are marked as nonidentifiable by the indefinite
article. The latter is required by the existential construction and the referents are understood
as contextually unique. Example (10b), on the other hand, provides evidence for nonidentifiable
referents marked by the indefinite article to be contextually non-unique. Therefore, Hawkins
concludes that the meaning of the indefinite article is neutral to uniqueness. In some contexts,

75
this corresponds to the ambiguity of the indefinite expression with respect to uniqueness. In (11),
the referent of a movie is not necessarily contextually unique (Mary could have watched another
movie on that night), but can be interpreted as such (Mary did not see another movie that night).

(11) A movie that Mary was watching last night was really interesting. (Hawkins 1991: 422)

With regard to the purposes of the present study, the problematic aspect of both Heim’s and
Hawkins’s approaches is that they do not account for indefiniteness as a referential function, but
for the meaning(s) of the indefinite article in English. I show in the next section that indefiniteness
can be treated as the union of two referent types, namely specific and nonspecific referents.

4.2 Defining the referent types

This section defines the referent types that are necessary to distinguish between different arti-
cle types. The relevant types are: deictic, recognitional, absolutely unique, anaphoric, bridging,
situationally unique, and establishing referents in the definite domain; specific and nonspecific
referents in the indefinite domain, and generic referents.

4.2.1 Definite referents

Definite referents all have in common that they are identifiable by both the speaker and the hearer,
which I refer to as mutual identifiability. To be more precise, the referents need to be unambigu-
ously identifiable in order to be definite, since their unambiguous identifiability guarantees that
they are the only salient referent of their kind. In the remainder of this study, I use “identifiable”
and “identifiability” in the sense of mutual and unambiguous identifiability.
A deictic referent is identifiable because it is also linked to an object in the real world that is
present in the discourse situation. An examples of a deictic referent is given in (12):

(12) Do you see the house over there?

The presence of the object in the discourse situation makes the referent the most salient referent
of its kind in the discourse universe. Therefore, the referent is also unambiguously identifiable. A
deictic referent can be defined as follows:

(13) Deictic referent Rdei

A referent is a deictic referent Rdei if there is a function f with its inverse function f −1 ,
and an object Odei in the discourse situation such that the following holds:
( f (Rdei ) = Odei ) ∧ ( f −1 (Odei ) = Rdei ).

76
The unambiguous link between the deictic referent and the object in the discourse situation is
implemented by a function f from the referent Rdei to the object Odei , and its inverse function,
mapping the object Odei back to the referent Rdei .
Recognitional referents represent another type of definite referents. They are identifiable by
both the speaker and the hearer because of shared experience or common knowledge. An example
for a noun with a recognitional referent can be seen in (14).

(14) What happened to that dog (we used to have)?

The definition of recognitional referents is given below:

(15) Recognitional referent Rrecog

A referent is a recognitional referent Rrecog if it is unambiguously identifiable by both the
speaker and the hearer based on shared experience or knowledge.

Absolutely unique referents are referents such as ‘sun’ or ‘moon’, which are unique in the default
perception of our world. They are unambiguously identifiable independently from the discourse
situation or even a larger context, as they are the only referents of their kinds. In English, for
instance, the definite article is licensed with the expression of such referents even without any
previous mention:

(16) The moon is bright tonight.

Of course, this is not to say that we could not distinguish between different moons when talking
about astronomy. I only refer here to a handful of concepts that are usually, even though not
necessarily, perceived as absolutely unique. The reason for distinguishing this type of definite
referent is that definite article in the world’s languages are often not compatible with nouns re-
ferring to these concepts; and hence, this type of referents has to be distinguished in order to be
treated separately (cf. section 5.2.4).10
These by idealization absolutely unique referents entail that the kind set K of an absolutely
unique referent Rabs.u only consists of a single referent. Therefore, the existence of another more
10 Amore fine-grained distinction of noun types is proposed in Löbner (e.g. 1985, 2011). He distinguishes between
sortal (e.g. stone, book, water), individual (e.g. moon, weather, Maria), relational (e.g. sister, leg, part), and functional
(father, head, age) nouns. While this classification of nouns accounts in an elegant way for certain effects in German
and English with respect to the use of the definite article, the only relevant effects with respect to the use of articles
in the languages in the world in general seems to be with two subtypes of what Löbner treats as individual nouns,
namely absolutely unique nouns and proper nouns. Therefore, I do not distinguish between the other noun types
here. As for the compatibility of definite articles with unique nouns and proper nouns, I show in sections 5.2.4 and
5.2.5 that the behaviour of definite articles in the world’s languages varies.

77
salient referent of its kind is excluded, which makes the absolutely unique referent unambiguously
identifiable and hence definite.

(17) Absolutely unique referent Rabs.u

A referent is an absolutely unique referent Rabs.u if it is the only element in its kind set K.

Anaphoric definites have been central in works on definiteness and discourse reference from var-
ious theoretical viewpoints. The formal mechanism of File Change semantics (Heim 1988), for
instance, restricts the notion of definiteness to anaphoric use. Another approach to definiteness
and identifiability in which anaphoric uses play a major role is Ariel (1988, 1990, 2001). She ex-
amined the correlation between the linguistic expression of the discourse referent (NP, pronoun,
demonstrative, etc.) and the distance between antecedent and anaphoric referent. This is also a
relevant factor for the distribution of definite articles. Since articles generally occur together with
nouns and not with pronouns, the variation between pronominal and nominal expressions of ref-
erents also restricts the distribution of definite articles in anaphoric contexts. Epstein (2002: 340)
pointed out that neither uniqueness nor familiarity can account for the distribution of the definite
article with respect to other definite descriptions or its infelicitous use in examples such as There’s
a cat in the yard …It (#The cat) is eating a mouse. The Accessibility Hierarchy in Ariel (1988) aims
at explaining such cases. Based on empirical findings, the definite article could be classified as
low accessibility marker, which means that it is not felicitous in contexts in which the referent is
topical (or highly accessible). Since it is not possible to formulate general deterministic rules for
the conditions that trigger the use of definite articles vs. pronominal forms, I will not discuss this
here. According to the previous definite referent types, anaphoric referents Rana are defined in
the following way:

(18) Anaphoric referent Rana

A referent is an anaphoric referent Rana if there is a referent Rante such that the following
holds:
(Rana = Rante ) ∧ (Rante ∈ MS ∩ MH ).

Another type of referents similar to anaphoric ones are bridging referents. “Bridging”11 occurs
when a nominal expression introduces a new referent Rbrid , which is nevertheless licensed as def-
inite by a special relation to a previously mentioned referent Rante . It is not a proper anaphoric
relation because the referents are not identical, but they are nevertheless linked in an unambigu-
ous way. As Schwarz (2013: 536) notes:
11 The term “bridging” goes back to Clark & Haviland (1977), but the phenomenon was already noted in Christophersen

(1939), and was discussed as “associative anaphora” in Hawkins (1978), “inferrables” in Prince (1981), “accommoda-
tion” in Heim (1988), and as “indirect anaphora” in Erkü & Gundel (1987).

78
 One could consider them, for example, to be a special case of the anaphoric use, except that
the antecedent is not the referent of the definite itself, but stands in some salient relationship
to it. (Schwarz 2013: 536)

Hawkins makes the nature of this link more explicit:

[…] the mention of one NP, e.g. a wedding, can conjure up a whole set of associations for
the hearer which permit the bride, the bridesmaids, etc. I shall refer to the first NP as ‘the
trigger’, since it triggers off the associations, and to first-mention definite descriptions which
are dependent on this trigger as ‘the associates’.
(Hawkins 1978: 123)

The examples in (19) show typical bridging contexts. In (19a), the first part of the utterance refers
to a book, establishing the referent of book (Rante ) as a part of the speaker’s and the hearer’s mental
spaces MS and MH . In the following part of the utterance, the referent Rbrid of the word author
is licensed as definite, because it has an unambiguous link to the already established referent of
book (Rante ), which triggers the use of the definite article.

(19) a. John bought a book. The author is French.

b. John was driving down the street. The steering wheel was cold. (Schwarz 2009: 6)

This unambiguous link between the two referents, which is not identity, usually involves a part–
whole relation, product–producer relation, etc. Frame semantics (cf. Fillmore 1975, 1976) is inher-
ently suitable to account for the definite status of a bridging referent:

[…] frames and scenes, in the mind of a person who has learned the associations between
them, activate each other; and that furthermore frames are associated in memory with other
frames by virtue of their shared linguistic material, and that scenes are associated with other
scenes by virtue of the entities or relations or substances in them, or their contexts of co-
occurrence. (Fillmore 1975: 124)

We can apply this to examples like (19a) as follows: when a book is uttered, the frame of its referent
Rante evokes the concept of ‘author’ based on our world knowledge that books are written by
people and these people are authors. By default, lacking more specific context, we can assume
that each book has a single author. Therefore, the semantic frame of the concept ‘book’ includes
one position for the concept ‘author’. This unambiguously links the referent Rbrid of the author to
the referent Rante of a book and requires the noun book to occur with a definite article, marking its
referent as definite, even though it is mentioned for the first time in the discourse. Figure 4.3 is an
illustration of this unambiguous link between the two referents Rante of book and Rbrid of author

79
enabled by the frame of the concept ‘book’ with one ‘author’ role. Imagining the kind set K of
Rante that contains all potential referents of the concept ‘book’, each referent of the kind ‘book’
evokes a different referent of the kind ‘author’:

book1 author1
book2 author2
book3 author3
… …
bookn authorn

Figure 4.3: Link between the referents of the kinds ‘book’ and ‘author’

We can thus define a bridging referent as follows:

(20) Bridging referent Rbrid

A referent is a bridging referent Rbrid if there is a function f with its inverse function f −1
and a referent Rante such that the following holds:
( f (Rante ) = Rbrid ) ∧ ( f −1 (Rbrid ) = Rante ) ∧ (Rante ∈ MS ∩ MH ).

This means that bridging referents Rbrid are unambiguously linked to a preceding referent Rante .
This link is more complex than the identity function between anaphoric referents Rana and their
preceding referents Rante . For bridging referents Rbrid , the link to a preceding referent Rante con-
sists of a function f from Rante to Rbrid and its inverse function f −1 that returns Rante for Rbrid .
Thus, bridging referents Rbrid are identifiable because of their unambiguous link to Rante which
already is an element from MS and MH .
Another type of definite referents is a situationally unique referent, which is very similar to
bridging referents. What I call situationally unique referents in opposition to absolutely unique
referents is what Hawkins (1978: 115-130) discusses as “larger situation use”. Situationally unique
referents are mutually and unambiguously identifiable because they have an unambiguous link
to the discourse situation that the speaker and the hearer are situated in. This link may be very
narrow and concern objects or concepts directly related to the discourse situation, and it can
involve referents that are linked to either the speaker or the hearer directly. Or, the link may be
to a larger context, e.g. the city, or the country the discourse situation is placed in. Two examples
of situationally unique referents are given in (21) and (22) below:

(21) S and H are in the small village H is from.

S: Where is the church in this village?
(22) H is showing S her new flat.

80
 S: Where is the kitchen?

In example (21), the referent Rsit.u of church is expressed by a noun with the definite article. The
latter is used to indicate that the referent is situationally unique. This means it is the only salient
referent of the kind K linked to the concept of ‘church’ within the discourse situation. Because
it is contained in the discourse situation, we can say that the discourse universe also contains a
referent of the concept ‘church’. Hence, situationally unique referents Rsit.u are identifiable due to
their unambiguous link to another referent Rds (ds for discourse situation) that is mutually and un-
ambiguously identifiable based on the context of the discourse situation. Figure 4.4 schematically
represents this link between Rds of the noun village and Rsit.u of church from (21):

village1 church1
village2 church2
village3 church3
… …
villagen churchn

Figure 4.4: Link between the referents of the kinds ‘village’ and ‘church’

Therefore, situationally unique referents Rsit.u are very similar to bridging referent Rbrid in that
the referent is identifiable based on an unambiguous link to another identifiable referent, only
that the preceding referent is not necessarily mentioned in the discourse; it can be implicit. Thus,
situationally unique referents can be defined as follows:

(23) Situationally unique referent Rsit.u

A referent is a situationally unique referent Rsit.u if there is a function f with its inverse
function f −1 and a another referent Rds which is identifiable from the discourse situation
such that the following holds:
( f (Rds ) = Rsit.u ) ∧ ( f −1 (Rsit.u ) = Rds ) ∧ (Rds ∈ MS ∩ MH ).

What the preceding types of definite referents had in common is that their definite status was
licensed by an unambiguous link to a referent that was already identifiable by both the speaker
and the hearer. Therefore, we could say that the referent type licensed the use of the definite
article. Establishing referents are different. In contrast to these other types of definite referents,
the identifiability of the referent needs to be established. Establishing12 referents correspond to
cases in which a referent is marked as definite, although it is not unambiguously identifiable
by the hearer yet. Epstein (2002) argues from a cognitive perspective that, by using a linguistic
12 The label is taken from Hawkins (1978).

81
expression that marks the referent as definite, the hearer can establish or construct it as such,
relying on the fact that she will be able to identify the referent shortly after the utterance.
The pragmatic requirements that allow such a commitment from both the speaker and the
hearer are formulated in Hawkins (1991). They are based on the principles of Quantity and In-
formativeness, which allow the speaker to mark something as definite (i.e. identifiable by the
hearer as well) because the speaker can rely on the hearer’s commitment to accept the referent as
identifiable and to construct identifiability. The hearer, in turn, can rely on the speaker to mark
a referent as definite only if it is unambiguously identifiable for the speaker as well or will be
shortly after the time of utterance. Two examples of establishing referents are shown in (24) and
(25):

(24) London has been buzzing with the rumour that the Prime Minister is going to resign.
(Hawkins 1978: 102)
(25) Have you heard the news? A cabinet minister has just resigned; I didn’t catch which one.
(Lyons 1999: 261)

The referent Rest of rumour in (24) is not yet identifiable by the hearer at the point in time when
the rumour is uttered, but it will be with the utterance of the complement clause that provides
sufficient information on the referent Rest . The referent is marked as definite so that the hearer
can expect and establish identifiability. We can represent the properties of this type of definite
referents as follows:

(26) Establishing referent Rest

A referent is an establishing referent Rest at the point in time t i of the discourse situation,
if Rest is an element of the speaker’s mental space MS in t i and if Rest is an element of the
hearer’s mental space MH at a later point in time t j of the discourse situation.

4.2.2 Specific referents

I define specific referents as single and particular referents from their kind set. In contrast to
definite referents, at least the hearer is not able to identify the referent, and only knows that the
linguistic expression is linked to a particular referent.
This is similar but not identical to what is often understood as specificity, namely that the
speaker is able to identify the referent but that the hearer is not.13 In addition, the concept of
specificity is used in contrast with nonspecificity, defined as a discourse referent that neither the
13 Asvon Heusinger (2002: 247) puts it: “[…] I discuss the widely used description of specific NPs as a subclass of
indefinite NPs as ‘known/identifiable to the speaker’ and show that this description is inadequate.”

82
speaker nor the hearer can identify and that does not have any existence requirement (e.g. Baker
(1973), Fillmore (1967), Karttunen (1976), Hawkins 1978: 203f, Anderson (1985: 179), Lyons (1999:
58)). This leads to the commonly found three-way distinction based on the combination of speaker
and hearer identifiability, displayed in Table 4.1 (von Heusinger 2002: 249).

Table 4.1: Traditional view of specificity and nonspecificity

identifiable by definite indefinite indefinite
specific specific nonspecific
speaker + + –
hearer + – –

However, von Heusinger (2002) argues that this description of specificity is not adequate: he
shows that we find instances of specific referents which are not identifiable by either the speaker
or by the hearer. For the purposes of the present study, I will adapt von Heusinger’s view that the
description of specificity as identifiability by speaker and by the hearer is not sufficient. Consider
the following example:

(27) Sarah wanted to talk to a colleague of mine, but I forgot which one.
(based on Ionin (2006: 182))14

In (27), the referent of a colleague of mine is not unambiguously identifiable by either the speaker or
by the hearer. Nevertheless, we can assume that Sarah knows which colleague she wants to talk to
based on the context. Thus, also the speaker and the hearer know that there is a particular referent
that a colleague of mine is linked to, even though neither of them can identify the referent. Thus,
a specific referent Rspec is a single referent of the kind set K. While neither the speaker nor the
hearer can identify the referent, they can assume that Sarah can identify the referent. Figure 4.5
sketches this scenario:

colleague1
colleague2
colleague3
… Sarah

colleaguen

Figure 4.5: Anchoring of a specific referent

14 The original example in Ionin (2006: 182) is: Sarah wants to talk to a colleague of mine, but I don’t know which one.

83
Figure 4.5 presents the referents of the kind set K of a colleague of mine on the left side, and the
speaker as well as the hearer know that one of these referents is linked to the expression used.
In addition, both know that reference is made to a particular one of them. That it has to be a
particular referent is given by the fact that it is linked to another referent Rds from the discourse
situation (here the referent of Sarah) that the speaker and the hearer can identify. We can hence
define a specific referent Rspec as follows:

(28) Specific referent Rspec

A referent is a specific referent Rspec if there is a function f and a referent Rds such that
the following holds:
( f (Rspec ) = Rds ) ∧ (Rds ∈ MS ∩ MH ).

The link between the specific referent and the other identifiable referent in the discourse situation
is again implemented as a function f from the specific referent Rspec to the identifiable referent
Rds . In contrast to bridging or situationally unique referents, this link is not unambiguous; for
this reason, there is no inverse function f −1 , that would map the identifiable referent Rds back to
the specific referent Rspec . In other words, assuming that we know which colleague in Figure 4.5
Sarah wanted to meet, we could take this referent as input and the function f (the link) would
output Sarah. As is indicated by the dotted lines, this link is not unambiguous and there is no
function f −1 that would take Sarah as an input and output the particular referent of ‘colleague of
mine’ that she wants to meet.

4.2.3 Nonspecific referents

While nonspecific referents are single referents from the kind set as well, they fundamentally
differ from definite and specific referents in that nonspecific referents do not correspond to any
particular referent from the kind set. An example of a nonspecific referent is given in (29):

(29) Lea wants to buy a bike, but she doesn’t know if she’s going to find one.

Often, the existence of a referent is taken as the basis for the definition of nonspecific referents.
Thus, the latter are usually assumed to lack an existence presupposition. In the present study,
I do not base the definition of nonspecificity on the existence of the referent, since nonspecific
referents are referents and therefore automatically elements in the discourse universeA nonspe-
cific referent is a single referent from its kind set K, but no particular referent is linked to the
expression used. With regard to the referent of a bike in (29), this scenario can be represented as
follows:

84
 expression referents
bike1
bike2
a bike bike3
…
biken

Figure 4.6: Nonspecific referent with its concept

Figure 4.6 shows the concept ‘bike’ which is linked to all referents that belong to this concept
(the kind set). Since the expression a bike in (29) has to be linked to one of these referents, but
since there is no information about the identity of the referent from the kind set that is linked,
the referring expression can be linked to any of them, and it is either not important or not known
which one it is. A nonspecific referent can thus be defined as a different type of referent, Rnspec
which corresponds to a placeholder for any referent (R1 to Rn ) in its kind set K.

(30) Nonspecific referent Rnspec

A referent is a nonspecific referent Rnspec if there is a referent Rnspec with its kind set K
such that the following holds: Rnspec ∈ K

4.2.4 Generic referents

The notion of genericity is usually used in two different senses: as reference to kinds (as opposed
to reference to objects) and as the denotation of a general property (Krifka et al. 1995: 2). In this
study, I only treat the reference to kinds, as in the sentences in (31) below:

(31) a. The potato was cultivated in South America.

b. Potatoes were introduced into Ireland by the end of the 17th century.
c. The Irish economy became dependent upon the potato.
(Krifka et al. 1995: 2)

I will take these expressions as referring expressions that are linked to a referent following (Givón
1984: 440):

The facts of natural language, as we shall see below, tend to suggest that generic subject
expressions […] are in fact bona fide referring expressions, but referring to a group or a type
rather than to an individual. (Givón 1984: 440)

85
Generic referents are of a different type than definite, specific, and nonspecific referents. They do
not correspond to a single element of the kind set, but to all referents in this set. Thus, generic
referents Rgen can be treated to correspond to the kind set itself:15

(32) Generic referent Rgen

A referent is a generic referent Rgen if there is a referent Rgen and its kind set such that the
following holds: Rgen = K

4.3 Defining the article types

This section defines the 9 different article types that are distinguished in the present study. I do
not provide examples here; Chapter 5 offers a detailed discussion of examples for each article type
from various languages of the world. These nine article types do not cover all of the four major
referent types defined in the previous section. While we find different types of articles in the
world’s languages that encode definite, specific, and nonspecific referents, articles whose main
function it is to encode generic referents do not seem to be attested. This is why generic referents
are not mentioned in the remainder of this section (for a discussion of generic referents and their
coding by articles, see section 7.2.2).

4.3.1 Article types from the definite domain

4.3.1.1 Definite articles

The main function of definite articles is to mark the referent of the noun that they occur with as
definite. This is not to say that definite articles could not occur in other contexts as well. How-
ever, other uses of definite articles are not relevant for their definition, e.g., the occurrence with
generic referents. Nor are definite articles required to encode all subtypes of definite referents.
I argue in section 5.2 that a definite article must be able to encode anaphoric (Rana ), bridging
(Rbrid ), situational unique (Rsit.u ), and establishing (Rest ) referents. Deictic referents Rdei can be
marked by the definite article in many languages; however, I argue that this is not a necessary
criterion for definite articles because we find articles in the languages of the world which comply
with the other criteria but which show variation with respect to the coding of deictic referents.
Thus, the motivation of excluding deictic referents as criterion for definite articles is based on the
observation of different definite articles. The same holds for absolutely unique referents Rabs.u .
Definite articles in the languages of the world also differ with regard to their compatibility with
15 Thisis very similar to the description of generic referents in Chierchia (1998: 351): “So a kind can be manufactured
out of a property by taking the largest member of its extension (at any given world).”

86
this referent type. Again, since these articles are used to encode different types of definite refer-
ents, I regard them as definite articles, resulting in the exclusion of absolutely unique referents as
criterion (cf. section 5.2.4). What definite articles are required to encode are anaphoric, bridging,
situationally unique, and establishing referents. Since the ability of definite articles to encode
recognitional referents Rrecog is not documented for many languages, I do not include this refer-
ent type as a criterion for definite articles either; note that this is due to practical reasons rather
than linguistic reasons. The same holds for establishing referents Rest . Definite articles are ex-
pected to encode this type of referents; I base this on Himmelmann (1997: 93-101), who argues
that this referent type is central in the development of definite articles. However, since this use
is not well-documented for many definite articles in the sample, I do not include the coding of
establishing referents as a necessary criterion either.

4.3.1.2 Anaphoric articles

Anaphoric articles represent another type of articles from the definite domain. I treat markers
as anaphoric articles if they systematically occur with a noun to encode its referent as anaphoric
(Rana ). I do not treat them as demonstratives, as was argued in section 3.1. This is because, in con-
trast to demonstratives, anaphoric articles have a less basic referential function and are not univer-
sally found in the world’s languages (most languages use the demonstrative to encode anaphoric
referents or the definite article, if available). Markers that encode both deictic and anaphoric ref-
erents are not anaphoric articles, but demonstratives; the extension of deictic demonstratives to
the marking of anaphoric referents is very common across the world’s languages. Examples of
anaphoric articles are discussed in section 5.3.

4.3.1.3 Recognitional articles

Another type of articles in the definite domain is what I call recognitional article. I show in sec-
tion 5.4 that we find markers in the world’s languages whose only function it is to mark recogni-
tional referents Rrecog , i.e. occur systematically with referents that are identifiable based on previ-
ous shared experience or knowledge. Similarly to anaphoric articles, I treat recognitional markers
as articles, since they are referential markers which, unlike demonstratives, are not universal or
crosslinguistically common (most languages probably use demonstratives to mark recognitional
referents).

87
4.3.2 Article types from the indefinite domain
4.3.2.1 Exclusive-specific articles

As I will show in sections 5.5 and 5.6, we can distinguish between two types of specific articles,
which I call exclusive-specific and inclusive-specific articles. Exclusive-specific articles only en-
code specific referents Rspec and cannot be used to mark nonspecific referents Rnspec .

4.3.2.2 Nonspecific articles

The nonspecific article can be viewed as the counterpart of the exclusive-specific article within the
indefinite domain. It marks the referent as nonspecific (Rnspec ); nonspecific articles cannot be used
to encode specific referents Rspec . Examples of nonspecific articles are discussed in section 5.7.

4.3.2.3 Indefinite and presentational articles

Indefinite articles encode specific and nonspecific referents Rspec and Rnspec . Indefinite articles
are semantically vague between coding specific and nonspecific reference, resulting in the ambi-
guity between specificity and nonspecificity in case the context does not resolve this ambiguity
otherwise. Based on the notion of colexification proposed in François & Vanhove (2008), I follow
Hartmann et al. (2014) in using the notion of coexpression to indicate that the article encodes a
combination of referent types. I show examples of indefinite articles in section 5.8.
Presentational articles have the same referential functions as indefinite articles; they encode
specific and nonspecific referents as well. I distinguish between these two article types for the
following reason. Many languages have what may be an indefinite article because it is system-
atically used to encode specific referents and, even though examples are rarer, potentially also
nonspecific referents. Nevertheless, these markers seem to be more restricted in their use in that
they only occur with referents that are prominent in the discourse, so that one may not want to
call them indefinite articles because there is a clear difference in the distribution of indefinite ar-
ticles and what I call presentational articles. On the other hand, presentational articles also differ
from the occasional use of the numeral ‘one’ (or any marker, for that matter) in specific and non-
specific contexts. Since the distribution of presentational articles is systematic, even though they
are more restricted in their distribution than indefinite articles, I include presentational articles
in this study; examples are shown in section 5.9.

88
4.3.3 Domain-crossing article types
4.3.3.1 Inclusive-specific articles

Inclusive-specific articles, as I will show in 5.6, mark both definite and specific referents Rde f
and Rspec , so that their functional domain crosses the division into a definite and an indefinite
domain. As for their marking of different subtypes of definite referents, I apply the same criteria
as for definite articles, i.e. inclusive-specific articles are required to mark anaphoric, bridging, and
situationally unique referents. Again, expressing the different referent types should be understood
as co-expression.

4.3.3.2 Referential articles

Referential articles encode three types of referents: definite Rde f , specific Rspec , and nonspecific
Rnspec referents. Like definite and inclusive-specific articles, referential articles are not require to
encode all subtypes of definite referents. They are expected to be applicable to anaphoric, bridging,
and situationally unique referents, and may or may not be used to encode other types of definite
referents as well. The other two referent types that a referential article necessarily encodes being
specific and nonspecific referents, the referential article corresponds to the the broadest article
type from a semantic point of view. Examples of referential articles are discussed in section 5.10.

4.4 The relation between referent types and article types

This section addresses the question of the relation between referent types and article types in
more detail, comparing the referent types proposed in section 4.2 with other previous hierarchies
and scales concerning referentiality. Figure 8.1 shows the major four referent types, combining
them with the domains of the different article types as defined in the previous section. The article
types are indicated by small caps: definite (def), recognitional (recog), anaphoric (ana), inclusive-
specific (inspec), exclusive-specific (exspec), nonspecific (nspec), indefinite (indef), and referen-
tial (ref).

89
 referent
types
definite Rde f specific Rspec nonspecific Rnspec

def, ana, recog exspec nspec

article
types

indef, pres
inspec
ref

Figure 4.7: Overview of referent types and article types

While the linear order of the referent types as presented in Figure 8.1 seems to be chosen ac-
cording to the functional domain of articles, they are in fact motivated by the properties of the
referent types themselves. All types of definite referents are unambiguously identifiable by both
the speaker and the hearer. Bridging (Rbrid ) and situationally unique (Rsit.u ) referents were defined
by a function mapping these referents to another referent that already is identifiable as well as an
inverse function to map this other referent back to the bridging and situationally unique referent.
Thus, unambiguous identifiability, the property that all definite referents share, is modelled here
by function from the referent to the union of the mental spaces of the speaker and the hearer
and its inverse function. In principle, the unambiguous identifiability of all definite referent types
can be defined by a function mapping them to the mental spaces and mapping the mental spaces
back to the referents by its inverse function. Specific referents, on the other hand, were defined as
referents that are linked by a function to another referent that is part of the union of the speaker’s
and the hearer’s mental spaces, but without an inverse function. This made them be linked, but
not unambiguously identifiable. Therefore, this definition makes specific referents inherently a
supertype of definite referents, or definite referent a special subtype of specific referents. Nonspe-
cific referents, on the other hand, were defined as single elements of their kind sets, but without
further information of their identity. Again, the definition of nonspecific referents only as ele-
ments in the kind set makes them a supertype of specific and definite referents. Only generic
referents are not a supertype of all other referents, since they are not defined as single elements
of the kind set, but as the kind set itself. Figure 4.8 shows this structure, which is entailed by the
definitions of the four major referent types.

90
 referent types

nonspecific Rnspec

specific Rspec

definite Rde f generic Rgen

Figure 4.8: Hierarchy of the major referent types

In a similar vein, we find hierarchies with respect to the structuring of referential functions in
previous works. The three relevant hierarchies are the definiteness hierarchy in Croft (2003: 132)
(based on the discussion in Comrie (1989: 132-35)), the givenness hierarchy in Gundel et al. (1993),
and the reference hierarchy in Dryer (2014) (going back to Givón 1978). What they have in com-
mon is that they structure and distinguish referential functions by mapping linguistic forms onto
these functions. The three hierarchies are presented in (33) to (35).

(33) definite > specific > nonspecific (definiteness hierarchy, Croft 2003)
(34) in focus > activated > familiar > uniquely identifiable > referential > type identifiable
(givenness hierarchy, Gundel et al. 1993)

(35) anaphoric definite >

nonanaphoric definite >
pragmatically specific indefinite >
semantically specific indefinite >
semantically nonspecific indefinite (reference hierarchy, Dryer 2014)

The definiteness hierarchy shown in (33) is based on differential object marking in Spanish, Turk-
ish, and Persian shown in Comrie (1989: 132-135). In Turkish and Persian, accusative marking
is reserved to nouns with definite referents. Spanish marks patients in transitive contexts if they
are animate referents and definite or specific; nonspecific referents are not marked. While the
hierarchy (based on an entirely different coding phenomenon) essentially shows a similar rank-

91
ing of referential functions as the hierarchy in Figure 4.8, although it does not clearly distinguish
between nonspecific and generic referents: the latter are characterized as “those identifiable only
as a type, not as a specific instance or token” (Croft 2003: 132). Nevertheless, this hierarchy shoes
that also other phenomena in the grammar, namely differential object (argument) marking, reflect
this alignment of definite, specific, and nonspecific functions.
The givenness hierarchy in Gundel et al. (1993: 275) structures referential functions in com-
bination with discourse prominence. These functions are called “givenness statuses” and corre-
spond to “six cognitive statuses relevant to the form of referring expressions in natural language
discourse” (Gundel et al. 1993: 275).
The statuses are not identical with the referent types as defined here, but some overlap: Gundel
et al. (1993) distinguish between four types of definite referents (varying in discourse prominence),
specific, and what they call generic referents, which also includes nonspecific referents. Gundel
et al. (1993: 276) mention explicitly that the givenness hierarchy should not be understood as mu-
tually exclusive but as an implicational hierarchy “such that each status entails (and is therefore
included by) all lower statuses, but not vice versa”. The main problem of the givenness hierarchy
is that it cannot separate the organization of referent types and the linguistic expressions; it is
intended as uinversal hierarchy of semantic functions whose definition is based on linguistic ex-
pressions in a single language. Also, the combination of referentiality and discourse prominence
makes it very difficult to properly separate the statuses. Nevertheless, it is a hierarchy of referent
types that cover the entire referential space.
The third relevant hierarchy is the reference hierarchy of Dryer (2014: 235), repeated in (36):

(36) anaphoric definite >

nonanaphoric definite >
pragmatically specific indefinite >
semantically specific indefinite >
semantically nonspecific indefinite (reference hierarchy, Dryer 2014)

This hierarchy is very similar to the one proposed in this study as well. “Pragmatically specific
indefinites” largely correspond to establishing definites. This “hierarchy” is better understood as
a scale and can be read as a semantic map in the traditional sense: the referential functions dis-
tinguished are ordered based on their coding by articles in the languages of the world, predicting
that a single article can co-express different (also more than two) functions, but only as long as
their are neighbouring functions in the structure:

The primary factor defining the ordering of these types on the hierarchy is the following: if a
language has an article that is used for more than one type of noun phrase on the hierarchy,

92
 then the set of types it is used with will be a set that is contiguous on the hierarchy.
(Dryer 2014: 235)

Thus, the motivation for this hierarchy is the organization of referential functions depending on
the combinations of functions that are encoded by articles. Therefore, the hierarchy proposes
more hierarchical distinctions in the definite domain, where I do not impose any hierarchical
structure: Dryer (2014) distinguishes between higher anaphoric and lower nonanaphoric definite
referents; predicting that there are articles that only code anaphoric definites, but also predicting
articles that mark non-anaphoric referents. While anaphoric articles are widely attested, I show
in section 7.2.1 that it is questionable if non-anaphoric articles in the strict sense exist in the
languages of the world.

4.5 Summary
This chapter dealt with the definition of different article types based on their referential functions.
In section 4.1, I introduced the relevant concepts related to referentiality, and defined four major
referent types that correspond to referential functions in section 4.2: definite, specific, nonspecific,
and generic referents. Section 4.3, defined 9 article types based on the combination of referent
types that they expressed: definite, anaphoric, recognitional, inclusive-specific, exclusive-specific,
nonspecific, indefinite/presentational, and referential articles. Section 4.4 discussed the relation
between referent types and article types, proposing a hierarchy of referent types, onto which arti-
cles can be mapped according to their functional domains. Table 4.2 on the next page summarizes
the different referent types from the definite and indefinite domains, indicating the functional
domains of each article type which corresponds to the expression of different combinations of
article types.

93
 Table 4.2: Summary of referent types and article types

Rdei A deictic referent Rdei is identifiable based on its unambiguous

link to an object in the discourse situation.
Do you see the house over there?
recog Rrecog A recognitional referent Rrecog is identifiable based on shared
previous experience or (common) knowledge.
What happened to that dog (we used to have)?
Rabs.u An absolutely unique referent Rabs.u is identifiable because it is
the only referent of its kind.
The Earth is round.

definite domain
Rana An anaphoric referent Rana is identifiable based on shared iden-
def
ana

tity with a previously mentioned referent.

What did you think about the movie last night? Wasn’t it great?
inspec

Rbrid A bridging referent Rbrid is identifiable based on its unambigu-

ous link to another previously mentioned referent.
ref

Eva bought a book. The author is French.

Rsit.u A situationally unique referent Rsit.u is identifiable because it is
the only salient referent of its kind in the discourse situation.
What’s the best way to the center?
Rest An establishing referent Rest is marked as identifiable and thus
constructed as such.
Did you hear the news? They are going to close the museum.
A specific referent Rspec is not identifiable but linked to a partic-

indefinite domain
Rspec
exspec

ular referent of its kind.

indef/pres

I met a strange new neighbour yesterday.

Rnspec A nonspecific referent Rnspec corresponds to a single, but no par-
nspec

ticular (any), referent of its kind.

Do you have a pen? Any pen will do.
Rgen A generic referent Rgen corresponds to the set of all referents of
its kind.
I like cats.

94
Chapter 5

Articles: examples from the world’s

languages

5.1 Introduction
This chapter gives an overview of the different article types as defined in Chapter 4. For each
type, I provide several examples from the world’s languages. Table 5.1 indicates the languages
that are discussed for for each of the article types. In addition to the comprehensive discussion of
the different article types, this chapter makes three important observations.
The first one concerns the referential functions of definite articles. I argue in section 5.2.3 that
the coding of deictic referents is not a necessary criterion for definite articles either. That many
definite articles can occur in deictic contexts can be explained the fact that most definite articles
originate from demonstratives. As I show for Kaqchikel, definite articles with sources other than
demonstratives need not extend their functional domain to the coding of deictic referents. In
addition, I discuss the occurrence with absolutely unique referents (cf. section 5.2.4) and with
proper nouns (cf. section 5.2.5), showing that definite articles can but do not have to occur in
such contexts, without any particular crosslinguistic trend.
The second important observation concerns the source of nonspecific articles. Based on the
data of different languages, I show in section 5.7.4 that nonspecific articles can be traced back
to verbal irrealis markers. The steps include: occurring on nouns in predicate position to en-
code irrealis-related meanings, occurring on nouns in argument position to encode irrealis-related
meanings, and occurring on nouns marking their referents as nonspecific.
The third observation concerns what I call presentational articles. We seem to find many
emerging indefinite articles which are restricted to discourse-prominent referents. I argue in
section 5.9 that they should rather be regarded as a separate article type, i.e. as presentational
articles.

95
 What I do not discuss in this section is the encoding of generic referents by the different types
of articles. As will be argued in section 7.2.2, there are no articles whose main function it is to
code generic referents. Since the purpose of this chapter is to discuss examples of the different
article types and contrast their properties, it would surpass the scope of this chapter to treat the
encoding of generic referents (cf. section 7.2.2 for an overview of the use of articles with generic
referents).

Table 5.1: Examples for different article types

article type Examples discussed in Chapter 5
definite Kaqchikel (Mayan, Guatemala) (section 5.2)
Mokpe (Bantu, Cameroon)
anaphoric Limbum (Bantu, Cameroon) (section 5.3)
recognitional Oksapmin (Oksapmin, Papua New Guinea) (section 5.4)
Gooniyandi (Bunuban, Australia)
Yankuntjatjara (Western Pama-Nyungan, Australia)
exclusive-specific Q’anjobal (Mayan, Guatemala) (section 5.5)
Akan (Kwa, Ghana)
Arawak (Arawak, Suriname)
Palula (Dardic, Pakistan)
inclusive-specific Bemba (Bantu, Democratic Republic of Congo) (section 5.6)
Tongan (Oceanic, Tonga)
nonspecific Q’anjobal (Mayan, Guatemala) (section 5.7)
Tongan (Oceanic, Tonga)
Lakota (Sioux, USA)
indefinite Tz’utujil (Mayan, Guatemala) (section 5.8)
Carib (Cariban, Suriname)
Bonan (Mongolic, China)
presentational Lango (Nilotic, Uganda) (section 5.9)
Logba (Kwa, Ghana)
Aghul (Lezgic, Russland)
Chatino (Zapotecan, Mexiko)
referential Rapa Nui (Oceanic, Chile) (section 5.10)
Halkomelem (Salishan, Canada)

96
5.2 Definite articles
This section deals with definite articles which were defined as articles that mark definite referents
in section 4.3.1. It was noted that definite articles do not necessarily encode all subtypes of definite
referents. I take articles to be definite articles if they systematically mark anaphoric, bridging,
situationally unique, and establishing referents. Especially bridging and establishing contexts are
not always well-documented; therefore, I also included markers in my sample as definite articles if
they systematically encode anaphoric and situationally unique referents. That deictic referents are
not necessarily marked by definite articles is shown for the definite article in Kaqchikel (Mayan,
Guatemala) in section 5.2.1. The compatibility of definite articles with absolutely unique referents
and proper nouns is discussed in sections 5.2.4 and 5.2.5, respectively.

5.2.1 Kaqchikel
Kaqchikel (Mayan, Guatemala) has a marker ri that is usually treated as definite article (Brown
et al. 2006, Chonay Chonay 2006). In this section, I show that ri is systematically used to mark
anaphoric, bridging, situationally unique, and establishing referents. Moreover, it does not occur
with specific or nonspecific referents, for which Kaqchikel uses the indefinite article jun.
In (1), the article ri marks an anaphoric referent: the expression ixoq ‘woman’ refers back to
the referent of jun ixoq in (1).

(1) A: tak’ xinapon chuwe choch xintz’at [jun ixoq] ruk’in kayi aq’ala
when arrive.pst.1sg to.1sg house see.pst.1sg art.indef woman with two child
ech’okol chuchi qaq’
seated close fire
‘When I came home, I saw a woman with two children sitting in my kitchen.’
B: awetaman ruwäch [*(ri) ixoq]?
know.pst.2sg eyes.3sg art:def woman
‘Did you know the woman? Kaqchikel (prim. data)

Two similar anaphoric contexts in which the definite article occurs with the noun are illustrated
in (2) and (3). In B’s reply in (2), B uses the article ri with wäy ‘tortilla’ to refer back to the referent
of the utterance made by A. Example (3) shows that the definite article is also obligatorily used
with the abstract noun ajob’ab’al ‘love’ to mark anaphoric reference.

97
 (2) A: ninwajo ntä wäy
want.1sg eat tortilla
‘I would like to eat tortilla.’
B: [*(ri) wäy] k’o chuchi qaq’
art:def tortilla exist close fire
‘The tortilla is in the kitchen.’
Kaqchikel (prim. data)
(3) ri amistad xok ajob’ab’al. [*(ri) ajob’ab’al] xok etzelanik
art:def friendship turn.pst.3sg love art:def love turn.pst.3sg hatred
‘(The) friendship turned into love. (The) love turned into hatred.’ Kaqchikel (prim. data)
Definite articles are also defined as markers that can encode bridging referents. Similarly to
anaphoric referents, bridging referents are definite because they involve an unambiguous link
between a preceding referent and the bridging referent. Examples (4), (5), and (6) illustrate the
use of the definite article with bridging referents. In (4), we find a part-whole relation between
jay ‘house’ and ruwi ‘roof’: the referent of jay ‘house’ makes the following referent of ruwi ‘roof’
identifiable and thus definite, because the frame of the concept ‘house’ enables us to link a single
referent of the concept ‘roof’. In (5) and (6), the reference to preceding k’ulbik ‘wedding’ and
ch’ich’ ‘car/taxi’ enables the definite marking of xten nik’ule ‘bride’ and chofer ‘driver’ accord-
ingly: the concepts of the preceding referents open a semantic frame with one position for the
bridging referent, linking the latter to the preceding referent and making it definite.1

(4) xk’atz’enkän jun jay tziaqchik chila pa tinamet. [*(ri) ruwi] jmul
see.past.1pl art:indef house old there in village art:def roof completely
manutzachik
non.functioning
‘We saw an old house there in the village. The roof was completely damaged.’
Kaqchikel (prim. data)
(5) xibe pa jun k’ulbik. [*(ri) xten nik’ule] jab’el (ok) ri rutziaq
go.pst.1sg to art:indef wedding art:def lady get.married pretty emph poss:3sg dress
‘ I went to a wedding. The bride had a pretty dress.’ Kaqchikel (prim. data)

(6) k’ochi xinchob jun ch’ich’ chila pa centro. [*(ri) chofer]

have.to.pst.1sg take art:indef car there in centre art:def driver
xup’ij chwa chiri ramajri manaqchik ch’ich’
tell.pst.s:3sg.o:1sg to.1sg comp this.hour is.not car
‘I had to take a taxi to the centre. The driver told me that there were no cars at that hour.’
Kaqchikel (prim. data)
The definite article is further required to mark are situationally unique referent. A situationally
unique referent is identifiable by both the speaker and the hearer because it is linked to another
1I use the notions “definite” and “identifiable” interchangeably.

98
identifiable referent from the discourse situation. Example (7) allows for the unambiguous iden-
tification of the referent of k’amor-b’äl ‘president’ because the concept of ‘president is unam-
biguously linked to the concept of ‘country’, which can be interpreted with the larger discourse
situation being situated in Guatemala.2 Hence, the definite article is used.

(7) [*(ri) k’amor-b’äl] jani rusamaj wakami

art:def guide-way much work.3sg now
‘At the moment, the President (of Guatemala) has a lot of work.’
Kaqchikel (prim. data)

Another example of a situationally unique referent is presented in (8). Accordingly, both the
speaker and the hearer know based on their world knowledge that there is maximally one hospital
in the village of the discourse situation. Therefore, the referent of aq’omanel-jay ‘hospital’ is
situationally unique and hence marked by the definite article as well.

(8) ankuchi k’obi [*(ri) aq’omanel-jay] chin re tinamet re?

where exist art:def cure-house in dem town dem
‘Where is the hospital in this town? Kaqchikel (prim. data)

Furthermore, definite articles are required to mark establishing referents, i.e. mark a referent as
identifiable so that the hearer can establish identifiability. An example of such context is a relative
clause that follows the noun and expresses the referent which is established as definite. This is
shown in examples (9) and (10). Both referents of achik’ ‘dream’ and achin ‘man’ are marked as
identifiable by the use of the definite article ri, and are made identifiable in the following relative
clause, establishing their identity.

(9) tatzijoj chuwe [*(ri) achik’] xab’en rijunkän aq’a

tell.imp to.1sg art:def dream make.pst.2sg last night
‘Tell me about the dream you had last night.’ Kaqchikel (prim. data)
(10) [*(ri) achin] xuk’ul Mario iwir jari xoyompe k’aba
art:def man meet.pst.3sg Mario yesterday be call.pst.3sg recently
‘The man that Mario met yesterday called not long ago.’ Kaqchikel (prim. data)

The previous examples thus showed that the definite article in Kaqchikel systematically occurs
with anaphoric, bridging, situationally unique, and establishing referents. I will now show that
ri is restricted to definite contexts and does not occur with nonidentifiable referents, which are
2 The concept of ‘president’ corresponds to a functional noun according to the classification of nouns in Löbner (1985).

He notes: “For functional nouns, the relation that defines their reference is a function. Functions relate objects
unambiguously (or one-to-one) to others” (Löbner 1985: 293). Thus, situationally unique referents as defined here
and functional nouns are compatible concepts, only that situationally unique referents are referential functions and
independent of the linguistic expression.

99
marked by the indefinite article jun in Kaqchikel. Examples (11), (12), and (13) provide specific
contexts and show that ri cannot be used with specific referents. Specific referents correspond
to particular referents that are linked to the expression used, but they are not identifiable by all
discourse participants. This is the case for the referents of jay ‘house’, lugar ‘place’, and chkop
‘animal’: the hearer cannot identify them, and they have to be marked by the indefinite article
jun.
(11) xk’atz’enkän [jun jay tziaqchik] chila pa tinamet
see.pst.1pl art:indef house old there in village
‘We saw an old house in there in the village.’ Kaqchikel (prim. data)

(12) Marta xutzijoj chuwe chi k’o [jun lugar janila interesante]
Marta tell.pst.s:3sg.o:1sg to.me comp exist art:indef place very interesting
‘Marta told me about a very interesting place.’ Kaqchikel (prim. data)
(13) tatzu xachob’ [jun chkop]
look.imp catch.pst.2sg art:indef animal
‘Look! You caught an animal!’ Kaqchikel (prim. data)

Examples (14), (15), and (16) below show that the same holds for nonspecific referents. Nonspecific
referents correspond to a single but not particular referent of its kind. In (14), for instance, any
referent of concept ‘pen’ can be referred to by the noun atzib’ab’al ‘pen’.

(14) k’o [jun atzib’ab’al]

exist art:indef pen
‘Do you have a pen? (Any pen will do.)’ literally: ‘Is there a pen?’ Kaqchikel (prim. data)
(15) ninwajo [jun ch’ich’]
want.1sg art:indef car
‘I want to have a (any) car.’ Kaqchikel (prim. data)
(16) nk’atzin chuwe [jun b’anoy-qutun]
need.1sg to.me art:indef prepare-food
‘I need a (any) cook.’ Kaqchikel (prim. data)

Thus, we saw that ri in Kaqchikel is a definite article, occurring with different types of definite
referents, and being absent in specific and nonspecific contexts.

5.2.2 Mokpe
Another example of a definite article can be found in Mokpe (Bantu, Cameroon). As is commonly
found in Bantu, Mokpe has an elaborate gender system which is traditionally referred to as noun
class system, since the values of gender and number of the noun occur as a fused marker on the

100
noun, on other elements in the noun phrase and as agreement markers on the verb. Section 6.2.2
discusses the morphological properties of the definite article in Mokpe in more detail. Here, I give
a brief overview of the exponents of the different classes and argue that they have to be exponents
of a single definite article based on their distribution with nouns in definite contexts.
Table 5.2 shows the article exponents for each class. The class distinction is adapted from
Atindogbe (2013); all examples and the analysis of the article system is based on my fieldwork on
Mokpe. Atindogbe (2013: 11) mentions articles in Mokpe as follows: “the indefinite and definite
articles are rendered by the low and the high tone respectively”. I argue that the definite article
indeed surfaces as high tone on the noun. However, a noun that is not marked by this high tone
does not correspond to a noun with an indefinite article. I treat it rather as the citation form which
is used in all contexts except for definite marking definite referents . In addition, Atindogbe (2013:
11) does not mention the segmental marker as an article exponent. Either, it is simply left out in
the description of Atindogbe (2013), or the variety described may differ from the Mokpe variety
spoken by my informants.

Table 5.2: Mokpe article paradigm

class noun article+noun class noun article+noun
singular plural
1 mòtò è mòtò 2 wàtò wátò ‘person’
ŋmánà è ŋmánà wánà wánà ‘child’
3 mòlélí mólélí 4 mèlélí mélélí ‘food’
mólèŋgò mólèŋgò mélèŋgò mélèŋgò ‘sheep’
5 lìkàlà líkàlà 6 màkàlà mákàlà ‘bridge’
líkpà líːkpa mákpá máːkpá ‘bag’
7 ètáŋgùlɛ̀ étáŋgùlɛ̀ 8 βètáŋgùlɛ̀ βétáŋgùlɛ̀ ‘lizard’
élélé élélé βélélé βélélé ‘morning’
9 mbèʒà è mbèʒà 10 í mbèʒà í mbèʒà ‘boy’
mbólì è mbólì í mbólì í mbólì ‘goat’
14 wòʒǒ wóʒǒ ‘face’
/
wɔ́ŋgɔ́ wɔ́ŋgɔ́ ‘fear’
19 ìnɔ̀ ní ínɔ̀ ní ‘bird’
/
dʒùŋgù dʒúŋgù ‘mosquito’

Classes 1-10 correspond to singular-plural pairs. For each class, two nouns are shown with a high
tone and a low tone on their first tone-bearing unit in their singular and plural forms. Nouns of
classes 14 and 19 contain singular nouns; since these nouns have plural forms in (mostly) class 10,
no corresponding plural classes are indicated for class 14 and 19.

101
 For nouns of class 1, 9, and 10, the article exponent is a segmental marker; è for class 1 and
9, and í for nouns of class 10. For nouns of the other classes, the article does not surface as a
segmental marker but as a tonal process on the noun, which can be described as a high tone on
the first tone-bearing unit of the noun. This is a strong argument for the form-independence of
articles. As we will see in this section, the tonal process has the same function as the segmental
marker è in that both markers occur in the same types of contexts with nouns. Therefore, I treat
them as equal exponents of the definite article in Mokpe.
That the definite article is used in anaphoric contexts is shown in (17) and (18). In both ex-
amples, the referents of móláná ‘woman’ and màlúwá ‘water’ are identifiable because they were
previously mentioned in the discourse. The definite article is required in such contexts. In (17),
we see the article exponent è with a noun of class 1. In contrast, the article surfaces as high tone
with the noun màlúlwá ‘water which belongs to class 6.

(17) A: è βóndá na dʒèlí, n-ɛ́ní mólánà à líj-ó kíʒènì nà βánà βá

art:def time I came 1sg-see woman.cl1 cl1 sit-in kitchen with children.cl2 cl2
wàkɛ́
two
‘When I came (home), I saw a woman with two children sitting in the kitchen.’
B: ò mà wíjá [è móláná]?
you pst know art:def woman.cl1
‘Did you know / recognize that woman?’
Mokpe (prim. data)
(18) A: nà βéja è ɲíʒá màlúwá
I feel art:def thirst water.cl6
‘I am thirsty.’
B: málúwá má βél-ǒ kíʒènì
water.cl6.art:def cl6 is-in kitchen
‘(The) water is in the kitchen.’
A: dʒǎ tɛ́, nà ʒɛ́nì málúwá/ *màluwa
come please I see.neg water.cl6.art:def water.cl6
‘Please come, I cannot find the water.’
Mokpe (prim. data)

Also in Mokpe does the definite article occur with the noun in bridging contexts. Examples (19)
and (20) illustrates the use of the Mokpe article with bridging referents. In (19), the referent
of mòtìlèlì ‘writer’ is identifiable because of its unambiguous link to the previously mentioned
referent of kâtì ‘book’. Since the noun mòtìlèlì ‘writer’ belongs to class 1, the article surfaces as è.

102
(19) mòkákè àndí kâtì. [è mòtìlèlì] à βélì.ndí mòtà βɛ́ndì.
Mokake buy.3sg book art:def writer.cl1 cl1 be person.am French
‘Mokake bought a book; the author is French.’ Mokpe (prim. data)

Example (20) shows the Mokpe counterpart of example (4) from Kaqchikel with a part-whole
relation between a house and its roof. Again, we see that the definite article is used together with
the expression of lìkândò ‘roof’. Since this noun belongs to class 5, it takes the high tone exponent.
(20) ì mɛ̀ nɛ́ ŋmánɛ́ βòβá ndáwò ó mbówà. líkândò/ *lìkândò lí mà kúwú
we see.pst.1pl small ugly house in village roof.cl5.art:def roof.cl5 cl5 pst finish
lì βáŋgèjà
inf get.damaged
‘We saw a small ugly house in the village. The roof had been damaged.’
Mokpe (prim. data)
Example (21b) shows a third example for bridging contexts. By mentioning the referent of è
mbówà jèní ‘his village’, the concept of ‘village’ makes it possible to unambiguously link the people
living in that village, and the “typical” landscape surrounding a village (e.g. a river and mountains)
to it. Therefore, the referents of wàtò ‘people’, mòlélí ‘food’, mòʒɔ̂ ‘river’, and βàkó ‘mountain’ are
unambiguously identifiable and thus definite. We see that the referents need to be marked by the
definite article. The nouns wátò, mólélí, and móʒɔ̂ ocur with a high tone on their first tone-bearing
unit, while the noun βàkó takes the segmental article exponent è.

(21) a. mòkákè à lìjà ó lùwálá náŋgɛ̂, ndí ó ŋmèmá ŋméní à lìŋgúní è

Mokake 3sg live in Douala now but in heart.cl3 poss.cl3 3sg love art:def
mbówà jèní …
village.cl9 poss.cl9
b. …wátò/ *wàto mólélí/ *mòlélí móʒɔ̂/
people.cl2.art:def people.cl2 food.cl3.art:def food.cl3 river.cl3.art:def
*mòʒɔ̂ nà *(è) βàkó.
river.cl3 and art:def mountain.cl9
‘Mokake lives in Douala now, but in his heart, he still loves his village. The people,
the food, the river, and the mountain.’ Mokpe (prim. data)

Similarly to bridging referents, a referent that is situationally unique requires the marking with a
definite article. Example (22), the counterpart of (7) for Kaqchikel, illustrates the use of the definite
article in such contexts. Uttered in Cameroon, the referent of ‘president’ is situationally unique3
in the country, which is why the referent of mòkànèlì ‘leader (president)’ is unambiguously identi-
fiable and requires to be marked by the article. Moreover, the utterance in (22) contains the noun
3 In the approach of Löbner (1985, 2011), the functional noun ‘president’ is unambiguously linked to the noun ‘country’.

This corresponds to the definition of situationally unique referents, which are unambiguously linked to another
referent present in the discourse situation, i.e. the referent of ‘Cameroon’ in this case.

103
ndʒùmá ‘fight, strike’ which is equally marked as definite. Again, due to the location of the dis-
course situation being the anglophone part of Cameroon in 2017, both the speaker and the hearer
know about the strikes against the repressions from the francophone government towards the
anglophone population. Therefore, the referent of ndʒùmá ‘fight, strike’ is situationally unique
and unambiguously identifiable for both the speaker and the hearer.

(22) [*(è) mòkànèlì] à ʒóβ-è βóndá jà [*(è) ndʒùmá]

art:def leader.cl1 cl1 neg-have time for art:def fight.cl9
‘The president does not care about the strike.’
literally: ‘The leader does not have time for the fight.’
Mokpe (prim. data)

Examples (23) and (24) below show that the definite article in Mokpe has to be used with estab-
lishing referents as well.

(23) [*(è) mòtò] Ngúdú á mùŋmá wúnùwû à wɛ́lɛ̂

art:def man Ngudu 3sg meet.pst last.night 3sg call
‘The man Ngudu met last night (just) called.’ Mokpe (prim. data)
(24) ŋmɛ́ndì/ *ŋmɛ̀ ndì mó wélì ndí dʒɔ̀ kɛ̌ wúwɛ̀ lɛ̀ lɛ̀ ó mbɛ̀ ngà wàkálà lì
news.cl3.art:def news.cl3 cl3 be emph true leave.3pl loc down white.cl2 inf
dʒǎ lì dʒɛ̌nɛ̀ ánú?
come inf see.1pl here
‘Are the rumors true? Are they coming to visit us from Europe?’ Mokpe (prim. data)

In order to qualify as a definite article, the article in Mokpe additionally needs to be absent with
specific and nonspecific referents. Examples (25) to (27) show specific referents which are ex-
pressed by the bare nouns ndáwò ‘house’, àmà ‘animal’, and lómbá ‘bundle/gift’ in (25), (26), and
(27), respectively. That the definite article is absent with nonspecific referents as well is shown in
(28) to (30).

(25) ì mɛ̀ nɛ́ ŋmánɛ́ βòβá ndáwò ó mbówà

we see.pst.1pl small ugly house in village
‘We saw a small ugly house in the village’ Mokpe (prim. data)
(26) ɔ́ŋgɔ́! è ŋùndèlì è kɔ̌wí ɲàmà
look.imp art:def trap.cl9 cl9 catch animal
‘Look! The trap caught an animal.’ Mokpe (prim. data)
(27) nò óβélì lómbá
1sg.2sg have bundle
‘I have a gift for you.’ Mokpe (prim. data)

104
(28) ówélì èkí já lì lâ èné mbówá
there.is place of inf eat dem:prox village.q
‘Is there a place to eat in this village?’ Mokpe (prim. data)
(29) nà àʒrà mòtò wà mòkáw à wí lííʒrɔ̀ ŋgɔ̀ nómà ná mà wíjê
I need person am hunt 3sg know hunting like I pst know
‘I need a hunter who can hunt like I used to do.’ Mokpe (prim. data)
(30) wúɲá wɔ́kɔ́ nà óβà mòtówà
day one I have car
‘One day I will have a car.’ Mokpe (prim. data)

Thus, we could establish the different markers in Mokpe as exponents of the definite article, used
with nouns in different definite but not indefinite contexts.

5.2.3 Definite articles with deictic referents

Although there are definite articles that can encode deictic referents, I argue in this section that
marking deictic referents is not a necessary criterion for definite articles. The use of a definite
article with deictic referents is probably often due to the development of definite articles from
demonstratives (or other markers that can be used with deictic referents, for that matter). In such
a scenario, the definite articles can retain the function of marking deictic referents as definite. This
is a novel view insofar as deictic referents are often treated as the most typical definite referents,
reflected by the fact that many expressions for definiteness develop from demonstratives (e.g.
Lyons 1999: 322). For instance in Kaqchikel, definite articles do not need to be compatible with
deictic referents. This can be related to the source of the ri otherwise occurs systematically to
mark definite referents, which is why I do not view the marking of deictic referents as necessary
criterion for definite articles. Instead, their compatibility with deictic referents may be often found
due to the frequent development of definite articles from demonstratives.
In Mokpe, the definite article is compatible with deictic referents. Examples (31) and (32) show
its use with deictic referents. Example (33) shows that the demonstrative marker and the definite
article can be used interchangeably to refer to a deictic referent.

(31) Context: S points at the only table in the room.

S: [è téwèlì] è gbèjáná wí wèwámbó
art:def table.cl9 cl9 made of wood
‘The table is made of wood.’ Mokpe (prim. data)
(32) ɔ́ŋgɔ́! [è ŋùndèlì] è kɔ̌wí ɲàmà.
look.imp art:def trap.cl9 cl9 catch animal
‘Look! The trap caught an animal.’ Mokpe (prim. data)

105
(33) [ònó/ è ŋmánà] á má kòkà lìwòtèjà è βóndá ná mà ŋmɛ̀ nê
dem:prox art:def child 3sg res grown since art:def time I pst see.o:3sg
‘This child has grown since the last time I saw him/her.’ Mokpe (prim. data)

The definite article ri in Kaqchikel most probably developed from the anaphoric marker ri …ri.
That the article is not compatible with deictic referents can be accounted by its different origin
which itself was not able to mark deictic referents.
Examples (34a) and (34b) show two utterances with the referent ulew ‘ground’ marked by
a demonstrative and the definite article, respectively. The referent of ulew in these sentences
receives two different interpretations: with the demonstrative in (34a), it is necessarily interpreted
as deictic, as is indicated by the translation. On the other hand, the noun ulew ‘ground’ with the
definite article in (34b) only allows for a non-deictic referent; it cannot be accompanied by a
pointing gesture and, without further context, corresponds to a generic statement.

(34) a. [la ulew] tz’il

dem ground dirty
‘The ground (here) is dirty.’
b. [ri ulew] tz’il
art.def ground dirty
‘The ground (in general) is dirty.’ Kaqchikel (prim. data)

Examples (35) and (36) below illustrate a similar effect for the use of the definite article with
the nouns q’ij ‘sun’ and ik’ ‘moon’. The versions with the demonstrative in (35a) and (36a) are
interpreted deictically, which makes them only felicitous if the objects ‘sun’ and ‘moon’ are visible
to the hearer and the speaker in the moment of utterance. In case the nouns are marked by the
definite article as in (35b) and (36b), the referents of q’ij ‘sun’ and ik’ ‘moon’ can no longer be
deictic. The only felicitous interpretation of the definite referents is as absolutely unique. As a
consequence of this incompatibility of the definite article and a deictic referent, the utterances in
(35b) and (36b) are only felicitous if the objects ‘sun’ and ‘moon’ are not directly visible neither
to the speaker nor to the hearer.

(35) a. man tatzu [la q’ij]

neg look.imp dem sun
‘Don’t look at the sun.’ (the sun being directly visible, e.g. from outside of a house)
b. man tatzu [ri q’ij]
neg look.imp art:def sun
‘Don’t look at the sun.’ (the sun not being visible, e.g. from inside of a house)
Kaqchikel (prim. data)

106
(36) a. [la ik’] jani nqalaj chpam re aq’a re
dem moon very be.visible.3sg in dem night dem
‘The moon is very bright tonight.’ (the moon not being visible, e.g. from inside of a
house)
b. [ri ik’] jani nqalaj chpam re aq’a re
def.art moon very be.visible.3sg in dem night dem
‘The moon is very bright tonight.’ (the moon not being visible, e.g. from inside of a
house)
Kaqchikel (prim. data)

As was already mentioned, it is plausible to assume that the definite article ri in Kaqchikel orig-
inates from an anaphoric marker ri …ri. The system of demonstratives including the anaphoric
marker is given in Table 5.3 below (based on Brown et al. 2006: 26,149ff and Chonay Chonay 2006:
103ff):

Table 5.3: Kaqchikel demonstratives

proximal re noun re
distal la noun (la)
anaphoric ri noun (ri)

For proximal la …la and anaphoric ri …ri, the second postposed part is often omitted (the form la
of the distal demonstrative used in (34a), (35a), and (36a) corresponds to this reduced form). It is
plausible that the anaphoric marker also went through this process of shortening in becoming a
definite article, since its simple preposed form ri corresponds to what I treat as the definite article
ri. Although ri was shown to occur with anaphoric referents in section 5.2.1, contexts that include
anaphoric referents with the anteceding referent in the same sentence and uttered by the same
speaker still require the full anaphoric form ri …ri and do not allow for the short form ri. Two
examples to illustrate this are given in (37) and (38).
(37) a Lu xiroyoj jun b’eychik mambajota yich’o ruk’in [ri vinaq
prop Pedro call.pst.s:3sg.o:1sg one way want.neg.1sg talk with.3sg ana person
*(ri)]
ana
‘Pedro called again. I don’t want to talk to that person.’ Kaqchikel (prim. data)
(38) k’ab’a a Lu xuroyoj ri ru vecino loman chuk’a
recently prop Pedro call.pst.s:3sg.o:3sg art:def poss:3sg neighbour while also
ink’o yin chila’ [ri achin *(ri)] jabey xqa chinwach
be.1sg I there ana man ana nice came in.front
‘Recently Pedro invited his neighbour while I was at his house. I liked that man.’
Kaqchikel (prim. data)

107
To conclude, it can be assumed that the definite article ri originated from the anaphoric marker
ri …ri, and did not undergo the typical first step from deixis to anaphora. Instead, the emerging
article started as an anaphoric marker, extending its function to the marking of bridging, situ-
ationally unique, and establishing referents. The fact that a definite article like ri in Kaqchikel
does not extend its use to the marking of deictic definites can be accounted for by the presence
of other demonstrative markers in the grammar. Demonstratives that cover the coding of deictic
referents may result in the lack of functional pressure on the definite article to be used with deic-
tic referents. This example from Kaqchikel showed two important things. Firstly, definite articles
do not necessarily encode deictic referents, even though many definite articles from the world’s
languages may do so. Therefore, the coding of deictic referents cannot be taken as criterion for
definite articles. Secondly, the fact that many definite articles encode deictic definites rather fol-
lows from their origin as demonstrative (or similar) markers. If the source element of the definite
article is not compatible with deictic referents like the anaphoric marker ri …ri in Kaqchikel, the
emerging definite article does not necessarily extend towards the expression of deictic referents.

5.2.4 Definite articles with absolutely unique referents

Absolutely unique referents are the only referents of their kind. Typical examples are the referents
of ‘sun’ and ‘moon’. This makes the referents automatically identifiable. Some definite articles,
for instance the in English, are able to mark such referents even without any previous mention:

(39) The moon is bright tonight. English (prim. data)

Because this holds for many definite articles found in Europe, the use of definite articles has often
been related to the coding of absolutely unique referents (e.g. Dryer 2013a, 2014, Hawkins 1978,
Löbner 1985) as well, the uniqueness of the referent being the explanation for the presence of
the definite article. In Hawkins (1978: 115), such referents are grouped with situationally unique
referents as “larger situation uses” of definiteness markers and also Löbner (1985) subsumes them
under the noun type of “individual nouns”. This shows that the presence of the definite article
with this type of referents is expected, if not a necessary criterion to treat a given marker as a
definite article. In this section, I argue that although many languages of the world might require
the definite article with such referents, it is not a necessary criterion for definite articles.
One reason to treat absolutely unique referents differently from other types of identifiable
referents is that they consist of only a small (and probably) restricted set. Usually, examples
are restricted to ‘sun’ and ‘moon’, but, depending on cultural and ethnographical settings, we
can extend the set to earth, ocean/sea, rain, God. Also, sometimes the expression for the own
tribe/people is counted in, but even though some more referents might be added to this set, it

108
seems to be confined to a rather small number of members. For the use of the definite article
with nouns expressing these referents, this means that we often deal with conventionalized uses.
Since the set of referents can hardly be extended, we cannot speak of a productive use of the
definite article in such contexts, the number of absolutely unique referents being restricted to a
small number of nouns.
Another reason to treat the expression of absolutely unique referents differently from situ-
ationally unique referents comes from the properties of definite articles across the world’s lan-
guages. Definite articles that systematically encode anaphoric, bridging, situationally unique,
and establishing referents show no apparent trend towards or against the compatibility with ab-
solutely unique referents. Making it a criterion for the definite article would arbitrarily exclude a
number of definite articles that otherwise have the same properties. Therefore, I do not take the
coding of absolutely unique referents as a criterion of definite articles as a crosslinguistic category,
but as a functions they may or may not have.
The crosslinguistic tendency of definite articles to mark absolutely unique referents or not is
difficult to determine at this point, given that many grammars do not provide any information or
examples for this question. Nevertheless, from the cases in my sample that do mention the use of
definite articles in such contexts, no trend is evident. The following examples illustrate both types
of definite articles with respect to their compatibility with absolutely unique referents. Examples
(40) to (42) show definite articles that are used with absolutely unique referents.

(40) ò ʒɔ́ŋgɔ́ ígbě/ *ìgbě

2sg neg.look sun.art:def sun
‘Don’t look into the sun.’ Mokpe (prim. data)

(41) ʔírf-ís múlmíl-ind

moon-art:def round-f
‘The moon is round.’ Dime (Seyoum 2008: 44)

(42) [kui na] ia-re a-e-kiu-ge malav va suni kini

sun.m art:def.m be-hot-nf 3sgm.o-sbd-die-ant people.pl art:def.pl all act
lagi-re lo-v
shelter-nf finish-pl
‘The sun was very hot, and the people all went and sheltered (in the shade).’
Lavukaleve (Terrill 2003: 91)

In (43) and (44), we see examples from languages with definite articles which do not occur with
absolutely unique referents.

109
(43) ɛ-n-hwɛ [awia *(no)]
3sg-neg-look sun art:def
‘Don’t look into the sun!’ Akan (prim. data)

(44) áxxaashe it asíi-ssaa-k

sun yet emerge-neg-decl
‘The sun hasn’t come up yet.’ Crow (Graczyk 2007: 228)

5.2.5 Definite articles with proper nouns

Similarly to the compatibility of definite articles with absolutely unique referents, their occurrence
with proper nouns (including names) is not a criterion for their classification as definite article or
not.
As for definite articles, we may expect them to occur with proper nouns and names because
their referents usually are unambiguously identifiable. On the other hand, one could argue that
such referents are inherently definite and may not need additional marking as such by a definite
article. Crosslinguistically, we find examples supporting both hypotheses, again without evident
trend. In some languages, definite articles occur with names and proper nouns, while their co-
occurrence is blocked in others. We can distinguish between three co-occurrence patterns of
definite articles and names. The article can either be systematically required with names, its use
can be optional, and it can have a different exponent with names and proper nouns in contrast to
contexts with common nouns.
Examples (45b), (46b), and (47b) illustrate the co-occurrence of the definite article with names
or proper nouns in Arawak, Irish, and Tarahumara. The use of the definite article with (certain
types of) proper nouns is generally required in these languages.4

(45) a. li falhetho b. li Wim

art:def white.man art:def Bill
‘the white man’ ‘Bill’
Arawak (Pet 2011: 52-53)

(46) a. an bád b. an Cháisc

art:def boat art:def Easter
‘the boat’ ‘Easter’
Irish (Stenson 2008a: 35-36)

4 Cf. Pet (2011: 86) for Arawak, Stenson (2008a: 36) for Irish, and Cohen (1998: 144) for Tarahumara.

110
(47) a. echi torí b. echi Antonio Loera
art:def rooster art:def Antonio Loera
‘the rooster’ ‘Antonio Loera’
Tarahumara (Cohen 1998: 70, 129)

In addition to definite articles that occur together with names and proper nouns, we also find
languages in which anaphoric articles can be used with such referents. This is shown in (48) for
the anaphoric article in Lango.

(48) òkélò-mɛ́rɛ̂ tɛ̂ rìŋgò

Okelo-art:ana 3sg.then run
’Then the aforementioned Okelo ran’ Lango (Noonan 1992: 251)

In other languages, definite articles can combine with proper nouns and names, but their co-
occurrence is optional, as in German (49) or Papuan Malay (50). Or, definite articles may be
incompatible with proper nouns, which is what is the case in, for instance, English, Sheko (Hel-
lenthal 2010: 142), Biak (van den Heuvel 2006: 221), or Diyari (Austin 2011: 43) do not use definite
articles with proper nouns or names.

(49) (der) Thomas

art:def.m Thomas
‘Thomas’ German (prim. data)

(50) Rahab (de)

Rahab art:def
‘Rahab’ Papuan Malay (Kluge 2017: 355)

In other languages, the definite (or specific or referential) article has different exponents for their
use with common and proper nouns. This is often found in Austronesian languages. Example (51)
illustrates this for the referential article in Rapa Nui: the article exponent occurring with common
nouns that are used in referential contexts is te, the marker being used with proper nouns being
a.

(51) he oho a Hotu ki te hare

ntr go art:ref.prop Hotu to art:ref house
‘Hotu went home.’ Rapa Nui (Kieviet 2017: 102)

111
5.3 Anaphoric articles
As was argued in sections 3.1 and section 4.3.1, anaphoric markers are different from deictic
demonstratives and can be treated as articles. In this section, I first present an example of anaphoric
articles from Limbum (Bantu, Cameroon). The second part of this section provides examples of
anaphoric markers from other languages.

5.3.1 Limbum
According to Fransen (1995: 146), Limbum features a tripartite system of spatial demonstrative
markers: one value expresses closeness to the speaker, the second one closeness to the hearer, and
the third value marks distance from both the speaker and the hearer:

Table 5.4: Limbum demonstratives

closeness S ʃá (cá)5
closeness H ánà
distance S & H ʃà (cà)

In addition to these spatial demonstratives mentioned in Fransen (1995), Limbum has another
marker fɔ̄ which qualifies as an anaphoric article: it systematically occurs in anaphoric contexts,
but it cannot be used to mark other types of definite referents.
That the marker fɔ̄ is obligatory in anaphoric contexts is illustrated in examples (52), (53), and
(54). In example (52), we see that the anaphoric referent expressed by mdzɨ̄p ‘water’ in A’s second
utterance. It requires the presence of the anaphoric article fɔ̄. In addition, the distal demonstrative
ánà is not felicitous in such an anaphoric context.

(52) A: ndɔ̄ŋ à yúʃī mɛ̀

neck is dry 1sg
‘I am thirsty.’
B: à cɔ̄r mdzɨ̄p mɔ̄ʔ ā
you have water some q
‘Do you have some water?’
A: vù mū càr. m̀ cī yɛ̄ [mdzı̄p *(fɔ̄)] kāʔ
come dim small I prog see water art:ana neg
‘Can you come? I do not see the water.’
′
A : vù mū càr. m̀ cī yɛ̄ [mdzı̄p *ánà] kāʔ
come dim small I prog see water dem neg
Limbum (prim. data)
5 The versions of the demonstratives given in brackets correspond to the forms used by my informant.

112
Two other examples illustrating the anaphoric use of fɔ̄ and at the same time the infelicitous use of
the distal spatial demonstrative ánà in such contexts are shown in examples (53) and (54): Again,
we see in (53) that the referent of ndzíŋwɛ̌ ‘woman’ uttered by B, referring back to the same
referent uttered by A has to be marked by fɔ̄, while the demonstrative ánà cannot be used to mark
the referent of ndzíŋwɛ̌ ‘woman’ as anaphoric. The same is shown for the anaphoric referent of
mú ‘child’ in (54).

(53) A: mú mfèʔ mɛ̀ mū bāʔ, mɛ̀ mū yɛ́ ndʒíŋwɛ̀ bá bōō báā ó tʃútèh mí
when time I pst2 arrive I pst2 see woman and children two 3pl sit in
kísı̀n
kitchen
‘When I came home, I found a woman with two children sitting in my kitchen.’
B: wɛ̀ mū rı̄ŋ [ndzíŋwɛ̌ *(fɔ̄)] ā?
2sg pst2 know woman art:ana q
‘Did you know the woman?’
B′ : wɛ̀ mū rı̄ŋ [ndzíŋwɛ̌ *ánà] ā?
2sg pst2 know woman dem q
Limbum (prim. data)
(54) a. àmbò djòʔ à m cī rò bzhı̄ ā tárté ye o mú mɔ̀ ʔ
when elephant s.3sg pst3 prog search food 3sg stumble 3sg with child one
‘When it was looking for food, the elephant stumbled into a child.’
b. …[mú *(fɔ̄)] à kɛ̄ʔ á cí wārī
child art:ana 3sg start inf prog cry
‘The child started to cry.’
c. …[mú ?ánà] à kɛ̄ʔ á cí wārī
child dem 3sg start inf prog cry
Limbum (prim. data)

Although the marker fɔ̄ in Limbum is systematically used to mark anaphoric referents, it is not
obligatory in all anaphoric contexts in the strict sense. Example (55) below shows that with ab-
stract referents, for which a generic reading is available in addition to the situationally grounded
one, the use of the anaphoric article is not obligatory. The “optionality” of the anaphoric article
in this case is most probably due to this alternative generic reading, but it also seems to involve a
stylistic contrast. Since such cases are also attested in other languages with e.g. definite articles
(in fact, the English translation of (55) shows the same effect for definite article in English), we
can still assume that fɔ̄ is an anaphoric article, allowing for exceptions.

(55) ŋkār à bbìtī rkòŋnı̀ [rkòŋnı̀ (fɔ̄)/ *àná] à bìtī rbànì.

friendship has become love love art:ana dem has become hatred
‘Friendship became love. (The) love became hatred.’ Limbum (prim. data)

113
In addition to the previous examples showing that fɔ̄ is used to mark anaphoric referents, it can be
shown that the use of fɔ̄ is not felicitous with other types of definite referents. Example (56) shows
an instance of bridging. The bridging referent, expressed by ŋwɛ̀ rsāŋ ‘author’, is identifiable due
to its unambiguous link to the referent of ŋwāʔ ‘book’. Nevertheless, fɔ̄ cannot be used to mark
the referent of ŋwɛ̀ rsāŋ ‘author’ as definite:

(56) Tāntōh ā mū yū ŋwāʔ mɔ̄. [ŋwɛ̀ rsāŋ (*fɔ̄)] á ŋwɛ̀ -mbòm.

Tantoh has pst2 buy book one person writing art:ana is person-Wimbum
‘Tantoh has bought a book; the author is a Wimbum person.’
Limbum (prim. data)

The same effect can be observed for situationally unique definite referents. The two examples (57)
and (58) below illustrate that the anaphoric article fɔ̄ is not felicitous with this type of referent:

(57) [ntāh (*fɔ̄)] yō fɛ́ à làʔ nâ?

market.cl9 art:ana be.cl9 where in village this?
‘Where is the market in this village?’ Limbum (prim. data)
(58) [tō ntāʔ (*fɔ̄)] tvı́r mfèʔ àmbò bbìr kāʔ.
head chair.am art:ana have time about wars neg
‘The President does not care (does not have time for) about the strikes.’
Limbum (prim. data)

Given its systematic occurrence in anaphoric contexts and absence in other types of definite con-
texts such as deictic, bridging, unique, and situationally unique definites, the marker fɔ̄ in Limbum
is an example for anaphoric articles.

5.3.2 Other examples

Grammars can often only dedicate little space to the discussion of the marking of anaphoric ref-
erents. In many cases, data alone does not provide enough information on the systematicity of
the use of a given marker. Therefore, in many cases, one needs to rely on the descriptions and
judgements of the authors to decide whether the occurrence of the marker in question seems to
be systematic enough to qualify as an article. The examples of anaphoric articles presented in this
section are the cases in which it was necessary to rely on the description of the marker.
The first example comes from Bare in (59)-(60). Bare has two different anaphoric articles, a
proximal and a distal one. The use of the proximal anaphoric article is shown in (59), Rante of
tʃiñu ‘dog’ belonging to the same sentence. Example (60) also refers to the same Rante , but is a
segment occurring a little later in the discourse. Hence, the noun tʃiñu ‘dog’ is marked with the
distal anaphoric article (a)sa instead.

114
(59) tʃiñu yawi mehésa me-wát’uka [da tʃiñu] i-báraka
dog angry 3pl.want 3pl-beat art:ana.prox dog 3sg.nf-run
‘The dog was angry, they wanted to beat the dog, it ran away.’ Bare (Aikhenvald 1995: 24)
(60) nu-yadã nu-tʃána-ka [asa tʃiñu]
1sg-see 1sg-stay-decl art:ana.dist dog
‘I am still seeing that dog.’ Bare (Aikhenvald 1995: 24)

In the following examples from Runyankore (61), Mangarrayi (62), Wardaman (63), and Koasati
(64), only the part containing the anaphoric referent is provided, which is equally expressed by a
noun together with the anaphoric article.

(61) Ku bu-ri-ku-mara ku-sya nya-mu-ishiki ni-a-banza

when cl14-cop-inf-finish inf-get.ready art:ana-cl1-girl pres-1.3sg-first
a-a-ko-za-mu a-ka-kumu a-a-siiga nya-mu-ereere o-mu
1.3sg-pres-dip-appl-in aug-cl13-finger 1.3sg-pres-smear art:ana-cl1-baby aug-in
kanwa
mouth.cl13
‘When it (the porridge) is ready, the (said) girl will first dip her finger in the porridge and
put the finger in the mouth of the (said) baby.’ Runyankore (Asiimwe 2014: 230)

(62) warguj may gi-nara-bayi ḷandi ḷiñ bu-b, marb ṇamdag ŋañ-ṇawu.
pick.up aux.3sg art:ana-dist-foc stick poke aux.3sg-pst tie aux.3sg neck-its
mayawa-ja! wirilmayin marb ṇamdag
now-emph goanna tie aux.3sg
‘(He) picked up that stick, poked him, noosed its nec. Now then! (He) noosed the goanna.’
Mangarrayi (Merlan 1989: 44f)

(63) gumurrinji-marlang bagbag ∅-bu-ndi yiwarli bag yi-bam darang wayana

emu-art:ana break 3-aux-pst arm-abs rump-abs and.all
‘He broke the arm of that emu, broke his head and rump and all.’
Wardaman (Merlan 1994: 96)

(64) má:mo:si-n st-ołá-:ci-n iko:si-:sáya im-alá-k

be.then-ds instr-there-3pl-ds aunt-art:ana poss:3-brother.in.law-s
háki-hco:li-sk hisakmápk-on á:t-ílli-:sáy-on isbakkí
be.drunk-custom-conseq pull.by.hair-ds.foc person-dead-art:ana-o.foc head
st-itta-folomóhli-n6
instr-loc-wave.about-ds

115
 ‘And then, they brought him over there, and the brother-in-law of this aforementioned
aunt, being drunk, grabbing the aforementioned dead man by the hair, was waving his
severed head about it.’ Koasati (Kimball 1985: 345)

A somewhat different type of anaphoric markers is found in Kayardild. Although it does not qual-
ify as an article based on the criteria of systematic distribution, it can be considered as emerging
anaphoric article. In the grammar, Evans (1995: 240) notes with regard to the marker niida ‘the
same’ that “[…] the speaker assumes that the hearer can identify the referent, because it is iden-
tical to something that has just been talked about.” Although this marker does not seem to be a
systematic referential marker, it is similar to an anaphoric marker. One example is shown in (65)
below:

(65) rar-umban-ji dulk-i [niid-a warngiid-a mungkiji kardu] kala-th

south-orig-loc country-loc same-nom one-nom own.nom father.in.law.nom cut-actl
‘In the south land (i.e. on Bentinck Island) the same one true father-in-law performed the
circumcision.’ Kayardild (Evans 1995: 240)

I do not treat niida as an anaphoric article here, since it is not primarily a referential marker and
since it is unclear if the marker is used in a systematic way with anaphoric referents. However, it
is very plausible that anaphoric articles can develop from markers like niida in Kayardild if they
occur more frequently in anaphoric contexts.

5.4 Recognitional articles

Across the world’s languages, we find another type of markers that qualifies as an article. These
markers are usually discussed as “recognitional” demonstrative or “recognitional function” of
demonstratives (e.g. Himmelmann 1997: 61-72, Diessel 1999: 109). What is meant by recogni-
tional function is the identifiability of a referent by both the speaker and and hearer based on
previous (common) knowledge or mutual experience in the in past, which was defined as recog-
nitional referent in section 4.2.1.
In many languages of the world demonstratives are used to mark recognitional referents. This
is illustrated for English in example (66):

(66) Do you remember that cat (we used to have)? English (prim. data)

In (66), the distal demonstrative that does not indicate deixis. Instead, it signals that the referent
of cat is identifiable by the hearer based on past shared experience with the speaker. Demon-
6 The sign “:” in the Koasati example stands for a glottal glide (Kimball 1985: 24).

116
stratives marking recognitional referents are not treated as recognitional articles. I only consider
markers that are exclusively used to encode recognitional referents as recognitional articles. The
motivation for treating them as articles rather than another type of demonstratives follows the
motivation for treating anaphoric markers as articles: marking recognitional referents is an ab-
stract referential function, and dedicated markers are rather rare crosslinguistically.
Oksapmin (Oksapmin, Papua New Guinea) is an example for a language with a recognitional
article. According to Loughnane (2009: 123), “[t]he demonstrative max is usually used when the
referent has not been previously mentioned/activated in the current discourse but is presumed to
be familiar to both the speaker and the addressee”. Two examples of the recognitional article are
given in (67) and (68). Loughnane (2009: 124) notes for example (67): “In the following example,
max is used to refer to the story that the speaker told earlier in the morning. This text had not
been previously mentioned in the current story but all the addressees had been present when he
told the previous story.”

(67) gin i ml-sa jəxe tumbuna paxna sup [stori max] pla gina
now hesit come.up-seq then ancestor hunger illness story art:recog tell.prs.sg now
‘Now, I came up and told that story about famine in the old days. Now …’
Oksapmin (Loughnane 2009: 124)

The context of example (68) is the following: “The text to which the following example belongs
was collected just after New Year’s Day which everybody in the community had known about
and the churches had held special events for” (Loughnane 2009: 124).

(68) [niu jia max] bəten x-t-pel=o li=m xe-ja

new year art:recog pray o-pfv-fut.pl=qot say-seq be-prs.pl
‘They wanted to pray for you, you know, that New Year.’
Oksapmin (Loughnane 2009: 124)

Gooniyandi (Bunuban, Australia) features a recognitional article as well, occurring with definite
referents that are identifiable by shared knowledge between the speaker and the hearer. According
to McGregor (1990: 146), the marker is used for referents from shared previous experiences or
conversations, as is shown in example (69).

(69) ngooddoo-ngga [ginharndi goornboo] wardgilayi-ngangi moolooddja-yidda

that-erg art:recog woman she.had.gone-with.us Mulurrja-all
niyaji-ngga barnnginaddi gaddwaroo
this-erg she.returned.me afternoon
‘The woman who went to Mulurrja with us brought me back yesterday.’
Gooniyandi (McGregor 1990: 146)

117
Yankuntjatjara (Western Pama-Nyungan, Australia) is another language with a recognitional ar-
ticle. Although treated as an anaphoric marker in Goddard (1985: 54), the description of its use
corresponds to the coding of recognitional referents. An example is shown in (70).

panya […] (roughly ‘you know the one’) calls the listener’s attention to the fact that he or she
is already familiar with a referent. It is not usually used about things which are fully topical–
i.e. already being talked about, but rather to re-introduce something into the conversation.
(Goddard 1985: 54)

(70) munu nyaku-payi “waru kampa-nyi [kapi panya-ngka-manti-r]

add see-char fire.nom burn-pres water art:recog-loc-prob-excit
‘And they would see “There’s a fire burning, at that water(hole), you know the water(hole),
most likely, by gee.”’ Yankuntjatjara (Goddard 1985: 54)

Himmelmann (1997: 69) notes that the marker nhenge in Arrernte (Central Pama-Nyungan, Aus-
tralia) is glossed as ‘remember’ in the grammar (Wilkins 1989: 121) and used in recognitional
contexts. In Wilkins (1989: 121), its use is described as follows:

The demonstrative nhenge ‘remember’ indicates that the entity to which the NP refers is some-
thing from before which I (the speaker) think that you (the addressee) should be able to remember.
It often functions as an indicator that something has been mentioned previously in the dis-
course, although it might not have been mentioned recently. However, it can also be used
when something is to be remembered from general context even if there has been no previ-
ous mention of it. (Wilkins 1989: 121)

Wilkins provides the following two examples to illustrate the two uses: example (71) shows
nhenge marking an anaphoric referent (not adnominally in this example), while example (72)
demonstrates that nhenge can also encode a recognitional referent:

(71) …kem-irre-ke thipe kngerrepenhe anteme. kem-irre-me-le ante

get.up-inch-pst.compl bird big-one now get.up-inch-npst.prog-ss and
nhenge alkere-k-irre-ke
art:ana.nom sky-dat-inch-pst.compl
‘…A big bird arose. It arose and the aforementioned (bird) took flight.’
Arrernte (Wilkins 1989: 121)
(72) [inspector nhenge] map-le school nhenhe-rlke inspect-emimle-ke
inspector art:ana pl-erg school dem-too inspect-tr.caus-pst.compl
‘Those inspectors (you remember the ones) inspected this school too.’
Arrernte (Wilkins 1989: 121)

118
Example (71) shows that Arrernte nhenge does not only mark recognitional definites, but would be
better described as an anaphoric marker whose use can extend to recognitional contexts, similarly
to the English demonstrative that in (66). Therefore, I do not treat nhenge as recognitional article
here.
This section presented a number of recognitional articles. They clearly underlie an areal bias
and are mostly attested in Australia and Papunesia. Since little work has been done on this refer-
ential function or on the markers that encode recognitional referents, future research is required
in order to give a more detailed picture of recognitional articles.

5.5 Exclusive-specific articles

This section deals with exclusive-specific articles. As defined in section 4.3.2, exclusive-specific
articles are defined as articles that encode a specific referent. Many of the examples in this section
will consist of contexts that are set in the past, including a referent that is marked as nonidenti-
fiable by the speaker, as in I was attacked by a penguin yesterday. Following e.g. Giannakidou
(1997, 2017), I take contexts that refer to single past events as episodic contexts and involved
nonidentifiable referent as specific.

5.5.1 Q’anjobal
The marker jun in Q’anjobal is an exclusive-specific article.7 I show in this section that it is not
an indefinite article, as it cannot occur with nonspecific referents. The latter are marked by the
nonspecific article junoq in Q’anjobal (see section 5.7.1).
Examples (73) and (74) show the marker jun in specific indefinite contexts, occurring together
with the expressions lugar ‘place’ and -tut ‘house’. The referents of ‘place’ and ‘house’ are specific
because the expressions uttered are linked to particular referents from their kind sets, but they
are not identifiable by the hearer.
(73) wojtaq [jun lugar] baytal chije’ ko-kuywi
know.1sg art:exspec place where can 1du-study
‘I know a (certain) place where we can study.’ Q’anjobal (prim. data)

(74) cham anima may kon tojlaneni ay [jun miman ya-tut]

cl:male person pst comp pay.pst.s:3sg.o:1pl exist art:exspec big 3sg-house
‘The man who paid us has a big house.’ Q’anjobal (prim. data)
7 MateoToledo (2017) and Raymundo Gonzáles et al. (2000) treat jun as an indefinite article in Q’anjobal and they do
not discuss nonspecific referents. In addition, there seems to be a high degree of dialectal variation in Q’anjobal,
also with respect to the exponents of the plural form of the article. The examples presented in this section are based
on the variety spoken in Santa Eulalia (Huehuetenango, Guatemala). Future research is required to show whether
the distinction between a specific-exclusive and a nonspecific article in Q’anjobal is only present in this variety or
also available in others.

119
The specific article jun in Q’anjobal has the plural form juntzan, which is illustrated in (75) and
(76). In (75), its presence marks the referent of watz’il-anima ‘friend’ as specific. In the existential
construction in (76), the article juntzan is also used to mark the referent of qa’jaja ‘mosquito’ as
specific. In this case, the referent is marked as particular but not identifiable because its identifi-
cation is not relevant to the discourse situation.

(75) ay [juntzan heb’ hin-watz’il-anima] max hin ul yulanej yet

exist art:exspec.pl pl 1sg-friend-person pst me come.pst.3pl visit.pst.3pl temp
ma’i aq’ab’alil
today night
‘Some friends (of mine) visited me last night.’ Q’anjobal (prim. data)
(76) yiban te mexha ay [juntzan qa’jaja]
above cl:wood table exist art:exspec.pl mosquito
‘There are (some) mosquitoes (flying) above the table.’ Q’anjobal (prim. data)

In order to be a specific article, jun must be incompatible with nonspecific referents. Examples
(77) and (78) show junin indefinite specific and nonspecific contexts. The referent of yatut thioxh
‘church’ in (77) is specific, as it a particular referent from its kind that is linked to the expression
used. In this case, only the article jun is felicitous. In example (78), on the other hand, the referent
of yatut thioxh ‘church’ is nonspecific because it falls under the scope of the question, so that any
referent from the kind set could be linked to the expression. In such a context, the specific article
jun cannot be used, and the nonspecific article junoq occurs instead.

(77) ay [jun/ *junoq yatut thioxh] b’ay yich calle.

exist art:exspec art:nspec house.3sg God
‘There is a church at the end of the road.’
(78) ay mi [junoq/ *jun yatut
exist q art:nspec art:exspec house.3sg God
‘Is there a church at the end of the road?’
at back street
Q’anjobal (prim. data)
thioxh] b’ay yich calle?
at back street
Q’anjobal (prim. data)

It is very plausible that the nonspecific article junoq is diachronically related to the specific article
jun. Therefore, one could argue that we are dealing with only one marker jun used in indefinite
specific and nonspecific contexts, the latter requiring an additional marker -oq. Even though
the markers are related diachronically and even though the nonspecific marker can formally be
decomposed into the specific article and an additional marker -oq, I treat them as separate articles.
I do so because, synchronically, we see two formally distinct markers with two distinct referential
functions. Neither the diachronic nor the synchronic formal relation between these two markers
is relevant to the question of whether they are articles, as long as we deal with two formally and
functionally distinct markers.

120
 In examples (79) and (80), we see a similar pair of specific and nonspecific referents, marked by
the specific and nonspecific articles, respectively. In (79), the speaker can identify the referent of
tuktuk ‘tuktuk’, as is indicated by the continuation in the second sentence; the referent is therefore
specific. As expected, only jun can be used. In (80) on the other hand, no particular referent of
the kind set of ‘tuktuk’ is linked to the expression tuktuk. Again, we see that jun is not felicitous,
and that junoq is used instead.

(79) tzeb’ach yul [jun/ *junoq tuktuk] mayal wawrtej naq tz’umon ch’en
come.imp in art:exspec art:nspec tuktuk already called cl driver
‘Let’s take a (certain) tuktuk. I already called the driver.’ Q’anjobal (prim. data)
(80) asi’ yul [junoq/ *jun tuktuk]
go.imp in art:nspec art:exspec tuktuk
‘Let’s take a (any) tuktuk.’ Q’anjobal (prim. data)

The following examples illustrate the absence of the specific article with definite referents. In
(81), the referent of itaj ‘herb’ is deictic, and hence definite. The noun itaj occurs together with a
demonstrative and a classifier, the article jun being absent. Example (82) illustrates an anaphoric
referent and its incompatibility with jun. Neither does the article jun occur with an establishing
referent, as can be seen in (83), where the referent is established as identifiable by the absence of
the article.

(81) context: S is pointing at her food.

manxa watz’iloq chiji an itaj ti’
emph delicious eat cl:herb vegetables dem:prox
‘These vegetables are delicious.’ Q’anjobal (prim. data)
(82) A: ayit max inabniok pai in-na maxwuil jun ix
when pst arrive.pst.1sg at 1sg-house see.pst.s:1sg.o:3pl art:exspec woman
yetuq kawan unin tolchotanayeb yul kusina
with two child be.seated in kitchen
‘When I arrived at home, I found a woman with two children sitting in the kitchen.’
B: hojtaqmi pax [ix ix tu’]?
know.2sg aff cl:female woman dem:dist
‘Did you know that woman?’
Q’anjobal (prim. data)
(83) a [xal ix] maktet ichaj’el yuj Lwin yet ewo tolto max awji
foc cl:polit woman who was.met by Pedro temp yesterday just pst call.3sg
‘The woman Pedro met yesterday called not long ago.’ Q’anjobal (prim. data)

121
Thus, this section showed that jun cannot occur in either definite or nonspecific contexts, but has
to occur with specific referents, which makes it an exclusive-specific article.

5.5.2 Other examples

We also find an exclusive-specific article in Akan (Kwa, Ghana).8 In (84), we see that bi is used
with the noun to introduce a new referent to the discourse:

(84) da bi [ɔbea bi] ne ne ba ɔsoɔdenfo bi tena-a

day art:exspec woman art:exspec conj poss child stubborn art:exspec stay-compl
ase
under
‘Once upon a time, there was a woman and her stubborn child.’
Akan (Amfo 2010: 1791)

Examples (86), (87), and (85) show that the exclusive-specific article is used in different contexts
involving specific referents . The article bi occurs together with the expressions akyɛde ‘gift’,
ahwahwadeɛ ‘miracle’, and mayɔnkofoɔ ‘my friends’, which are all linked to a particular referent
that is not identifiable by the hearer. In (85), the referent becomes identifiable by the hearer at a
later point in time, whereas the identifiability of the specific referent in (86) and (87) is not relevant
to the discourse. The context for (86) excludes the hearer from knowing about the ‘gift’.

(85) nanso [ahwahwadeɛ bi] si-iɛ ɔbaabunu no fa-a afuro

but miracle art:exspec happen-pst virgin art:def take-pst stomach
‘But a miracle occurred. The virgin conceived.’ Akan (prim. data)
(86) me-wɔ [akyɛde bi] ma wo
1sg-have gift art:exspec give 2sg
‘I have a gift for you.’ Akan (prim. data)
(87) [m-ayɔnkofoɔ bi] ba-a me-hɔ ɛnora anadwo
poss:1sg-friends art:exspec come-pst poss:1sg-place yesterday night
‘Some friends came over to my place last night.’ Akan (prim. data)

Examples (88) and (89) show that bi cannot be used with nonspecific referents. In (88), the referent
of yere ‘wife’ is nonspecific because it is within the scope of the irrealis reading of the predicate;
(89) shows the same in the context of a question.

(88) me-nya-a anka me-wɔ [yere (*bi)]

1sg-get-pst before 1sg-have wife art:exspec
‘I wish I had a (any) wife.’ Akan (prim. data)
8 The article bi is also treated as a marker of specificity in Arkoh (2011).

122
(89) wo-bɛ-nya [pen (*bi)] a-ma me anaa?
2sg-fut-get pen art:exspec impers-give 1sg q
‘Can you give me a (any) pen?’ Akan (prim. data)

In order for bi to be a specific article of the restrictive type, the last relevant property its incompat-
ibility with identifiable referents. Example (90) below illustrates this with an anaphoric referent
in (90b). As we can see, bi is not felicitous in such cases, and the anaphoric article no occurs
instead. That bi can neither be used to mark a bridging referent is shown in (91b); it can be seen
in (91a) that the bridging referent is marked by a possessive instead.

(90) a. mʊ-tɔ-ɔ ekutu bi

1sg-buy-pst orange art:exspec
‘I bought an orange.’
b. [ekutu no/ *bi] yɛ dɛw dɛɛ
orange art:def art:exspec be nice so
‘The orange is/was so nice.’
Akan (Arkoh & Matthewson 2013: 6)
(91) a. ye-hu-u dan dadaw bi wɔ ekurasi hɔ [ni nkyɛnsıdan]
1pl-see-pst building old art:exspec at village there poss roof
e-hodwow
pfv-worn.out
‘We saw an old building in the village; the roof was worn out.’
b. *ye-hu-u dan dadaw bi wɔ ekurasi hɔ [nkyɛnsıdan bi]
1pl-see-pst building old art:exspec at village there roof art:exspec
e-hodwow
pfv-worn.out
Akan (Arkoh & Matthewson 2013: 14)

As we saw, bi occurs with specific, but not with nonspecific or definite referents. Hence, it repre-
sents another example of an exclusive-specific article.
Another exclusive-specific article is found in Arawak (Arawakan, Suriname). The marker aba
is systematically used to mark specific referents of singular nouns. Pet (2011) does not treat aba as
an article or determiner; he consistently glosses aba as ‘one’, since it originates from the numeral.
However, he makes a distinctions in the translations: aba is always translated as a into English
in contexts with specific referents, while it is translated as one in the discussion on numerals.
Examples (92) to (94) below illustrate the use of aba to mark a referent as specific, meaning
that it indicates the link to a particular referent that is not identifiable by the hearer.
(92) da-fara [aba dodole]
1sg-kill art:exspec peccary
‘I killed a peccary.’ Arawak (Pet 2011: 36)

123
(93) da-malhita [aba bahy]
1sg-make art:exspec house
‘I made a house.’ Arawak (Pet 2011: 63)
(94) [aba wadili] sika khali da-myn miaka
art:exspec man give cassava.bread 1sg-to yesterday
‘A man gave me cassava bread yesterday.’ Arawak (Pet 2011: 62)

While Pet (2011) does not provide glossed examples of contexts with nonspecific referents, a num-
ber of unglossed but translated sentences with nonspecific singular referents show that aba is not
used (e.g. Pet 2011: 139,142,146,158). Therefore, I conclude that aba is restricted to occurring with
specific referents. Neither can it occur with definite referents: definite referents are encoded by
the definite article li in Arawak. An example showing the use of li with an establishing referent
is given in (95).

(95) [li wadili] siki-thi-fa no damyn andy-fa jon

art:def man give-rel-fut it to.me arrive-fut here
‘The man who will give it to me will arrive here.’ Arawak (Pet 2011: 86)

Thus, the article aba is an exclusive-specific article that marks specific referents but that does not
encode definite or nonspecific referents.
Palula (Dardic, Pakistan) also has an exclusive-specific article, as was shown in section 3.2.
Examples (21) to (23), repeated here as (96) to (98), and (99) show that the article áa/ák 9 marks the
referent as specific, i.e. as particular but not identifiable by the hearer.

(96) bhuná [áa ɡíri] heensíl-i

below art:exspec rock stay.pfv-f
‘Down below there was a big rock.’ Palula (Liljegren 2016: 309)
(97) miír thaní [áak míiš] heensíl-u de
Mir qot art:exspec man stay.pfv-m.sg pst
‘There was a man called Mir.’ Palula (Liljegren 2016: 309)
(98) eesé zanɡal-í [áa baṭ-á] ǰhulí harí so kuṇaák
rem forest-obl art:exspec stone-obl on take.away.cv def.nom.m.sg child.nom.m.sg
bheešóol-u
seat.pfv-m.sg
‘In that forest he took the child to a stone and seated him.’ Palula (Liljegren 2016: 283)
(99) [ak táapeṛ-e] ṭék-a
art:exspec hill-gen top-obl
‘on the top of a hill’ Palula (Liljegren 2016: 281)
9 The different forms are used in different varieties of Palula Liljegren (2016: 140).

124
The article áa/ák does not occur with nonspecific referents. Examples (100) to (102) below illus-
trate this. Neither can the article mark definite referents. Definite referents are expressed by the
definite article se in Palula, as can be seen in the anaphoric context in (103):

(100) čúur reet-í ǰheez-íi fláiṭ na bhíl-i hín-i aáǰ bi kansál

four night-pl airplane-gen flight neg become.pfv-f be.prs-f today also cancelled
bhíl-i
become.pfv-f
‘There have been no flights for four days, and also today it was cancelled.’
Palula (Liljegren 2016: 243)
(101) tasíi ba ɡa wása na heensíl-u
3sg.gen top any strength neg stay.pfv-m.sg
‘And he had no strength at all.’ Palula (Liljegren 2016: 310)
(102) muloó díi ɣeér kráam na bháan-u
mullah from without work neg become.prs-m.sg
‘Without a mullah the work is not being done.’ Palula (Liljegren 2016: 332)
(103) théeba [se ṭhaaṭáak-a] bi tas sanɡí khainií široó thíil-u
then art:def monster-obl also 3sg.acc with eat. start do.pfv-m.sg
‘(When the monster came inside, the man was eating.) And the monster started eating
with him.’ Palula (Liljegren 2016: 141)

To conclude, this section showed with examples from Q’anjobal, Akan, Arawak, and Palula that
exclusive-specific articles occur in various areas of the world.

5.6 Inclusive-specific articles

In this section, I present two examples for inclusive-specific articles from Bemba and Tongan.
Inclusive-specific articles encode both definite and specific referents and semantically vague be-
tween these two referent types. This article type is not frequently attested across the world’s
languages; my sample only contains six instances of inclusive-specific articles.

5.6.1 Bemba
Nouns in Bemba (Bantu, Democratic Republic of Congo), like nouns in various other Bantu lan-
guages10 , do not only have prefixes that indicate number and gender (traditionally referred to as
10 E.g. Zulu, Kirundi, Nguni.

125
noun classes). They also feature an additional prefix, often called “pre-prefix” or “augment”.11 An
example is given in (104):

(104) u-mu-ntu aaliishile

art:inspec-cl1-man come.pst.3sg
‘The/a man came.’ Bemba (Givón 1969: 47)

The noun umuntu ‘man’ in (104) consists of the lexical root ntu, a prefix mu- indicating that it
belongs to class 1, and the additional prefix u-. I treat this prefix as an inclusive-specific article
in Bemba. Table 5.5 gives an overview of the noun class prefixes and the corresponding article
exponents based on Givón (1969: 28f). The rightmost column indicates the class prefix as it occurs
on nouns in the absence of the article; the column in the middle presents the forms of the combined
form consisting of the article and the noun class marker.
Classes 1 to 14 are each singular and plural pairs of what could be treated as 7 different genders.
Class 15 indicates abstract or derived nouns and also infinitives. Classes 16 to 18 are somewhat
different in that they are locative markers (pa ‘at’, ku ’at/in’, and mu ‘inside’) which can be added
to nouns of other classes (the original class marker is retained) to express spatial but also other
abstract semantic values. As we see in Table 5.5, the augment or what I treat as specific article
also inflects for the class of the noun.
11 Theform, but also the factors that condition the distribution of the augment, differ significantly across Bantu
languages. In some languages the use of the augment is tied to definiteness or topicality, in others it only occurs in
certain syntactic contexts (e.g. is absent with negation), and in some languages nouns can no longer occur without
the augment. For an overview of the the augment in Bantu languages, see de Blois (1970).

126
 Table 5.5: Bemba noun class markers
noun class article-class class
1 u-mu- mu(u)-
2 a-ba- ba(a)-
3 u-mu- mu(u)-
4 i-mi- mi(i)-
5 (il)i- li(i)-
6 a-ma- ma(a)-
7 i-ci- ci(i)-
8 i-fi- fi(i)-
9 i-N- ni(N)-
10 i-N- ni(N)-
11 u-lu- lu(u)-
12 a-ka- ka(a)-
13 u-tu- tu(u)-
14 u-bu- bu(u)-
15 u-ku- ku(u)-
16 / pa-
17 / ku-
18 / mu-

While Hoch (1964) only lists morphosyntactic factors to account for the absence of the augment,
I follow Givón (1969) in that it is conditioned by the referent type of the noun. Therefore, as the
article systematically occurs with nouns in specific contexts as well as definite contexts, but is
absent with nouns in nonspecific contexts, I treat it as a inclusive-specific article. The contexts
themselves are not provided in Givón (1969). I follow his description as well as the translations
of the examples. If the ambiguity between interpreting the referent as definite or specific is not
resolved by the context, the inclusive-specific article is always translated as the/a into English; this
strongly suggests that the article marks the referent as vague between a definite and an indefinite
interpretation.
Two examples to illustrate this are given in (105) and (106). The inclusive-specific article u- is
obligatory in specific contexts (106a), the lack thereof being ungrammatical (106b):

(105) naalimweene *(a)-ba-ana

see.pst.1sg art:inspec-cl2-child
‘I saw (some/the) children.’ Bemba (Givón 1969: 42)
(106) a. u-mu-ntu aaliishile
art:inspec-cl1-man come.pst-3sg
‘The/a man came.’

127
 b. *muu-ntu aaliishile
cl1-man come.pst-3sg
‘The/a man came.’ Bemba (Givón 1969: 47)

If a nonidentifiable referent occurs within the scope of a negated single past event, it has to be
nonspecific. As examples (107) and (108) show, the article cannot occur in such contexts. The
referent of muana ‘child’ in (107a) contains the article, and is therefore necessarily interpreted as
definite or specific.12 A nonspecific reading is only available in the absence of the augment u- as
in (107b). Example (108) illustrates this for the referent of citabo ‘book’.

(107) a. nshiamwene u-mu-ana

see.pst.neg.1sg art:inspec-cl1-child
‘I didn’t see the child.’
b. nshiamwene mu-ana
see.pst.neg.1sg cl1-child
‘I didn’t see any child.’ Bemba (Givón 1969: 42)
(108) nishiatumine ci-tabo ku-mu-ana
send.pst.neg.1sg cl7-book cl17-cl1-child
‘I didn’t send any book to the child.’ Bemba (Givón 1969: 56)

Although, according to Givón (1969), the article is required in definite as well as specific indefinite
contexts and excluded in nonspecific contexts, we find systematic gaps, i.e. contexts in which the
inclusive-specific article does not occur. For instance, it does not occur together with a prenominal
demonstrative (109), an associative marker (110), or a locative marker consisting of a class prefix
(111).

(109) uyu-(*u)-mu-ntu
dem-art:inspec-cl1-man
‘this man’ Bemba (Givón 1969: 53)
(110) u-mu-ana u-a (*u)-mu-luungi
art:inspec-cl1-child cl1-am art:inspec-cl1-hunter
‘the child of the hunter’ Bemba (Givón 1969: 54)
(111) ali mu-(*u)-mushi
be.3sg cl18-art:inspec-village
‘He’s in the village.’ Bemba (Givón 1969: 55)
12 Eventhough the translation suggests that the referent has to be definite, Givón (1969: 42) marks it as “token”
reference. The distinction between “token” and “type” corresponds to the distinction between specificity vs. non-
specificity/genericity (Givón 1969: 41). In this sense, the referent of ‘child’ in (107a) can most probably also receive
a specific interpretation.

128
That the specific article is blocked by the presence of other prefixes in such contexts and for
referential reasons can be seen in example (112). In (112a), the specific article -u is by the additional
class 17 prefix ku-. In (112b), on the other hand, ku is only present on the first of two coordinated
nouns, namely on muntu ‘man’. The class 17 prefix ku being absent on the noun muana ‘child’,
the latter occurs with the inclusive-specific article u-.

(112) a. aamoneshya i-ci-tabo ku-(*u)-mu-ana

show.pst.3sg art:inspec-cl7-book cl17-art:inspec-cl1-child
‘He showed the book to the child.’
b. naalimoneshya i-ci-tabo ku-mu-ntu na-u-mu-ana
show.pst.1sg art:inspec-cl7-book cl17-cl1-man and-art:inspec-cl1-child
‘I showed the book to the man and the child.’
Bemba (Givón 1969: 55f)

5.6.2 Tongan
In Tongan (Oceanic, Tonga), we find another example of an inclusive-specific article. In contrast to
Bemba, which only has an inclusive-specific article, Tongan also has a nonspecific article. In pre-
vious studies, the Tongan articles were regarded as definite and as indefinite articles (cf. Church-
ward 1985: 23-27, Morton 1962) by analogy with the article systems that are commonly found in
European languages. Although still calling them definite and indefinite articles, Otsuka (2000: 50)
notes that the use of the articles in Tongan differs from their English counterparts. Völkel (2010),
on the other hand, treats the two articles as specific and nonspecific articles. According to Völkel
(2010) and based on their use and distribution, I treat the Tongan articles as inclusive-specific and
nonspecific articles (cf. section 5.7.2 for a discussion of the nonspecific article). The paradigm of
the articles in Tongan (Völkel 2010: 159) is given in Table 5.6 below.

Table 5.6: Tongan article paradigm

neutral affective
specific (h)e13 si’i
nonspecific ha si’a

Both the inclusive-specific and the nonspecific articles have two exponents, one used in a neutral
way, the other one to mark affection. I will only discuss the markers that belong to the neutral
series here.
13 The exponent he of the neutral specific article only occurs after certain prepositions, otherwise the form e is used.

129
 In almost all examples in Völkel (2010) and Otsuka (2000), the referents marked by the inclusive-
specific articles are translated with a definite article in English, suggesting that the referents are
definite.14 However, given the translation in example (113), the article e does not necessarily mark
the referent as definite but can also code it as specific. Also Otsuka (2000: 50) notes that the use
of the article (h)e can mark the referent as definite or specific, as is shown in (114).

(113) ko ’eku lele mai pe ’o fakaa’u ’a [e ’ofa] mei he

pres poss:1sg polit:run to just and polit:present abs art:inspec present abl art
kāinga mei he Lolo ’a Halaevalu ’i teu fakamanatu ’a e
family abl art:indefspec Lolo ’a Halaevalu prepare remind abs art:inspec
valu-ngofulu-ma-nima ta’u ’a e ’Afio na
85th year poss art:inspec polit:you
‘I have come with the relatives from Lolo a Halaevalu to present to your Majesty a gift for
your eighty-fifth birthday.’ Tongan (Völkel 2010: 208)
(114) na’e sio ’a Sione ki [he ta’ahine]
pst see abs Sione to art:inspec girl
‘Sione saw a/the girl.’ Tongan (Otsuka 2000: 50)

In (115) and (116) below, the inclusive-specific article (h)e is shown to occur with definite referents
of type. Example (115) shows the situationally unique referent fakaafe ‘feast’. In (116), the referent
of kā ‘car’ deictic, as the context includes a pointing gesture; it occurs with the inclusive-specific
article as well.

(115) na’a ne talanoa ki [he fakaafe]

pst 3sg talk loc art:inspec feast
‘He talked about the feast.’ Tongan (Völkel 2010: 116)
(116) ’alu ki mui ’i [he kā]
go spat.p.all back loc art:inspec car
‘Go to the back of the car.’ Tongan (Völkel 2010: 135)

Example (128) contains a question with the nonspecific referent of maa ‘bread’. The article (h)e
is absent in such contexts, and the nonspecific article ha is used instead (cf. section 5.7.2 for the
nonspecific article in Tongan).

(117) ’oku ’i ai [ha maa]?

prs in there art:nspec bread
‘Is there some bread?’ Tongan (Otsuka 2000: 50)
14 Morton (1962) and Churchward (1985) hardly provide any examples of full sentences.

130
We saw that the article (h)e can mark referents as definite and specific, but not as nonspecific. It
represent thus another example of an inclusive-specific article.

5.7 Nonspecific articles

In this section I discuss nonspecific articles which mark nonspecific referents. Sections 5.7.1 and
5.7.2 present two examples of nonspecific articles in Q’anjobal and Tongan, which were also
shown to have specific articles. The third example for nonspecific articles comes from Lakota
(Siouan, USA). Except for Tongan, the data from these languages points towards the diachronic
source of nonspecific articles being conditional markers on predicates.

5.7.1 Q’anjobal
Q’anjobal (Mayan, Guatemala) features the marker junoq, which is a good candidate for a nonspe-
cific article. As was already mentioned in section 5.5.1, previous works on Q’anjobal only mention
the marker jun, which is regarded as indefinite article (Mateo Toledo 2017, Raymundo Gonzáles
et al. 2000). Although neither of these grammatical sketches includes contexts with nonspecific
referents similar to the contexts shown here, Mateo Toledo (2017) mentions the irrealis marker
-oq that can be related to the nonspecific article (cf. section 5.7.4). The discussion in this section
is based on my primary data collection in Santa Eulalia (Huehuetenango, Guatemala), featuring
the nonspecific article junoq here.
As was already shown in the previous sections, the nonspecific article junoq occurs in com-
plementary distribution with the specific article. Examples (118a) in opposition to (118b), as well
in (119a) and (119b).

(118) a. ay mi [junoq /*jun yatut thioxh] b’ay yich calle?

exist q art:nspec art:exspec house.3sg God at back street
‘Is there a church at the end of the road?’
b. ay [jun/ *junoq yatut thioxh] b’ay yich calle.
exist art:exspec art:nspec house.3sg God at back street
‘There is a church at the end of the road.’
Q’anjobal (prim. data)
(119) a. tzeb’ach yul [jun/ *junoq tuktuk] mayal wawrtej naq tz’umon ch’en
come.imp in art:exspec art:nspec tuktuk already called cl driver
‘Let’s take a (certain) tuktuk. I already called the driver.’
b. asi’ yul [junoq/ *jun tuktuk]
go.imp in art:nspec art:exspec tuktuk
‘Let’s take a (any) tuktuk.’
Q’anjobal (prim. data)

131
In (120) to (123) below, I show further examples for the use of junoq in other nonspecific contexts.
Note that the possessive marking of the nouns does not interfere with the nonspecific referents
of ha-tz’ib’al ‘(“your”) pen’ and ha-na ‘(“your”) house’.

(120) ma ay [junoq ch’en ha-tz’ib’al]?

neg exist art:nspec cl:stone 2sg-pen
‘Do you have a pen ? (Any pen will do.)’
literally: ‘Don’t you have a (any) pen?’ Q’anjobal (prim. data)
(121) ay [junoq ha-na]?
exist art:nspec 2sg-house
‘Do you have a house?’ Q’anjobal (prim. data)
(122) chin q’an [junoq watz jatnemajna]
1sg need.1sg art:nspec good architect
‘I need a good architect (any architect who is good).’ Q’anjobal (prim. data)
(123) ta’ watz’ max hot mulnaj hab’a ay mi wal [junoq/ *jun
if good pst cond work.2sg 2sg exist for.2sg emph art:nspec art:exspec
ha-mimam na]
2sg-big house
‘If you had worked well, you would have had a big house.’ Q’anjobal (prim. data)

Like the specific article jun, the nonspecific article junoq can also be used to express plural. This
is illustrated in example (124).

(124) a. aq’ [junoq/ *jun hin-xila]

give.imper art:nspec art:exspec 1sg-chair
‘Give me a (any) chair.’
b. aq’ [juntzanoq/ *juntzan hin-dulce]
give.imper art:nspec.pl art:exspec.pl 1sg-candy
‘Give me some candy. (It can be any of the candies that are in the bowl.)’
Q’anjobal (prim. data)

Having established that at least in the Santa Eulalia variety of Q’anjobal, the markers jun and
junoq occur in complementary distribution and can thus be treated as an exclusive-specific and
as a nonspecific article, respectively. Based on Mateo Toledo (2017), I show in section 5.7.4 that
-oq in junoq originates from a verbal irrealis marker.

5.7.2 Tongan
As was mentioned in section 5.6.2, Tongan has a nonspecific article, ha, in addition to the specific
article (h)e. In this section, I present examples that show the distribution of the nonspecific article.

132
Examples (125), (126), and (127) illustrate the use of ha . In (125) and (127), the referents marked
by the nonspecific article ha correspond to the lack of the link between the expression and a par-
ticular referent of the kind set, which is due to the imperative context. Nonidentifiable referents
within the scope of imperatives are probably always nonspecific, since it seems implausible from
a communication point of view to refer to a particular referent which is not identifiable by the
hearer within an imperative. Example (127) also shows that the nonspecific article is compatible
with plural nouns. In (126), the context allows for the referent of vai ‘water’ to be specific or non-
specific. Given that it is presented as nonspecific in the translation and given that it is marked
by ha, we can assume that the nonspecific article marks the referent of vai ‘water’ as nonspecific
here as well.

(125) ha’u mo [ha afo]

come with art:nspec fishing.line
‘Bring a (any) fishing line.’ Tongan (Churchward 1985: 25)
(126) ’oku ou fie ma’u [ha vai mafana]
prs I want receive art:nspec water warm
‘Is there any warm water here?’ literally: ‘I want to receive warm water.’
Tongan (Churchward 1985: 24)
(127) fai [ha ’u tohi]
do.imper art:nspec pl letter
‘Write (some) letters.’ Tongan (Churchward 1985: 24)

That ha is also used to mark nonspecific referents in questions is shown in (128). We also see that
ha cannot be used with specific referents in (129).

(128) ’oku ’i ai [ha maa]?

prs in there art:nspec bread
‘Is there some bread?’ Tongan (Otsuka 2000: 50)
(129) *na’e sio ’a Sione ki [ha ta’ahine].
pst see abs Sione to art:nspec girl
‘Sione saw a girl.’ Tongan (Otsuka 2000: 50)

The distinction between specific and nonspecific markers in Tongan is also part of the possessive
system. The exponents of adnominal possessives can morphologically be decomposed into three
parts: the article, a possessive marker, and a person marker. This also leads to possessive markers
that are sensitive to the specific-nonspecific distinction as well. There are two sets of adnominal
possessive markers, one set (A-set) is used in syntactic S and A positions, the other set (O-set)
is used in O positions (Völkel 2010: 160-162). The paradigms of these two sets are presented in
Table 5.7 and Table 5.8, respectively.

133
 Table 5.7: Paradigm of Tongan adnominal possessive markers (A-set)

specific nonspecific
neutral affective neutral affective
1sg (he)’eku i’eku ha’aku i’aku
2sg ho’o i’o ha’o i’ao
3sg (he)’ene i’ene ha’ane i’ane
1du.ex (he)’ema si’ema ha’ama si’ama
1du.in (he)’eta si’eta ha’ata si’ata
2du ho’oma si’omo ha’amo si’amo
3du (he)’ena si’ena ha’ana si’ana
1pl.ex (he)’emau si’emau ha’amau si’amau
1pl.in (he)’etau si’etau ha’atau si’atau
2pl ho’omou si’omou ha’amou si’amou
3pl (he)’enau si’enau ha’anau si’anau
impers (he)’ete si’ete ha’ate si’ate

Table 5.8: Paradigm of Tongan adnominal possessive markers (O-set)

specific nonspecific
neutral affective neutral affective
1sg hoku si’oku haku si’aku
2sg ho si’o hao si’ao
3sg hono i’ono hano si’ano
1du.ex homa si’oma hama si’ama
1du.in hota si’ota hata si’ata
2du homo si’omo hamo si’amo
3du hona si’ona hana si’ana
1pl.ex homau si’omau hamau si’amau
1pl.in hotau si’otau hatau si’atau
2pl homou si’omou hamou si’amou
3pl honau si’onau hanau si’anau
impers hoto si’oto hato si’ato

An example for the use of possessive forms from these two sets is provided in (130). In both
(130a) and (130b), possessive markers from the O-set are used. In the question in (130a), we see
the nonspecific possessive, while the specific possessive is featured in the answer in (130b).

134
 (130) a. ’oku ’iai [hao tokoua]?
pres exist poss:2sg.nspec sibling
‘Do you have a (any) sibling?’
b. ’oku ’iai [hoku tokoua]
pres exist poss:1.inspec sibling
‘I have a sibling.’
Tongan (Völkel 2010: 162)

5.7.3 Lakota
Another example of a nonspecific article is found in Lakota (Siouan, USA), occurring in comple-
mentary distribution with the specific article in indefinite contexts. This is shown in example
(131). In (131a), the specific referent of igmu’ ‘cat’ is encoded by the exclusive-specific article wã.
In contrast, the marker wãži in (131b) signals that igmu’ has a nonspecific referent.

(131) a. igmu’ [wã wachi]

cat art:exspec want.s:1sg.o:3sg
‘I want a (particular) cat.’
b. igmu’ [wãži wachi]
cat art:nspsec want.s:1sg.o:3sg
‘I want a (any) cat.’ Lakota (Van Valin 1977: 64)

Since this distribution of wã and wãži with respect to specific and nonspecific contexts was re-
ported to be systematic (Rood & Taylor 1996, Van Valin 1977, Williamson 1984), I treat wãži as a
nonspecific article.15 The full paradigm of the nonspecific and specific article is given in Table 5.9
below.16

Table 5.9: Lakota specific and nonspecific article paradigm

specific nonspecific
sg wã wãži
pl eya’ etã

Examples (132) to (135) are further illustrations of the distribution of the two articles in Lakota.
In (132), the referent of makno’xlova ‘cave’ is specific and therefore marked with the specific
article. In (133) and (134), the context does not disambiguate between a specific or a nonspecific
interpretation of the referents thaspã ‘apple’ and šuka ‘dog’. In these examples, it is the nonspecific
15 Even though these two articles encode specific and nonspecific referents, respectively, they are treated as two
indefinite articles in Van Valin (1977: 63), as two indefinite determiners in Williamson (1984: 48), and as two
indefinite markers in Ingham (2003: 90).
16 Based on Van Valin (1977: 64), Williamson (1984: 48), and Rood & Taylor (1996).

135
article that resolves this ambiguity and marks the two referents as nonspecific. Example (135)
illustrates the form etã of the nonspecific article in the plural.

(132) [makno’xlova wã] e’l eũ’papi

cave art:exspec in laid.s:3pl.o:3sg
‘They laid him in a cave.’ Lakota (Van Valin 1977: 66)
(133) [thaspã wãži] wachi
apple art:nspec want.s:1sg.o:3sg
‘I want an (any) apple.’ Lakota (Williamson 1984: 48)
(134) [šuka wãži] ophethu maši
dog art:nspec buy ask.s:3sg.o:1sg
‘He asked me to buy a (any) dog.’ Lakota (Williamson 1984: 49)
(135) [cha etã] aku wa
stick art:nspec.pl bring imp
‘Bring back some (any kind of) sticks!’ Lakota (Williamson 1984: 48)

In (136), the nonspecific referent of thaspã ‘apple’ is under the scope of negation. In this case, the
nonspecific article bears an additional negation marker and occurs as wãži-ni (cf. section 6.2.1).

(136) [thaspã wãži-ni] tebwaye šni

apple art:nspec-neg eat.up.1sg neg
‘I didnt’ eat up an (any) apple.’ Lakota (Williamson 1984: 49)

Rood & Taylor (1996: 36) treat nonspecific articles like in (136) as a separate negative nonspe-
cific article and list additional exponents for combinations with non-human and inanimate plural
nouns. However, examples for nonspecific articles in negation contexts are already rare in the
literature; the only marker that is shown in contexts being wãžini (Ingham 2003, Rood & Taylor
1996, Van Valin 1977, Williamson 1984). Independently of this issue, I do not treat the nonspecific
article exponent wãžini in negation contexts as a separate marker because negation itself does
not result in the coding of a separate referent type.
Van Valin (1977: 64) discusses a further marker cha which he calls an indefinite article. He
notes that cha is usually not regarded as an article but that it “has the same general syntactic
properties of the other articles”. Based on this language-internal argument, he treats it on a par
with the other articles in Lakota. However, cha was described to encode contrast similarly to a
focus marker (Van Valin 1977: 65-67, Williamson 1984: 49f), which is why I do not treat it as an
article here.

136
5.7.4 The source of nonspecific articles
In section 5.7.1, it was mentioned that the nonspecific article can be traced back to a verbal ir-
realis marker. I show in this section, that we find similar evidence for the development of an
irrealis marker to a nominal nonspecific marker (and eventually, article) in a number of Siouan
and Wakashan languages.
The nonspecific article in Q’anjobal junoq was shown to formally consist of the exclusive-
specific article jun and another marker -oq. Even though neither Raymundo Gonzáles et al. (2000)
nor Mateo Toledo (2017) mention this nonspecific article, Mateo Toledo (2017) shows that -oq a
verbal irrealis or infinitive marker -oq for intransitive contexts. It seems highly likely that this
irrealis marker -oq fused withjun and became the nonspecific article. Examples for the verbal uses
of -oq are given in (137) to (139). They show its three uses as irrealis, future, and infinitive marker.

(137) q-q’anjab’ ayach ta q-ach q’anjab’-oq

pot-talk to.2 cond pot-2pl talk-irr
‘X will talk to you, if you talk.’ Q’anjobal (Mateo Toledo 2017: 538)
(138) hoq-ach tyaj-l-oq yekal
pot-2 prayer-intr-irr tomorrow
‘You will pray tomorrow’ Q’anjobal (Mateo Toledo 2017: 538)
(139) x-toj heb’ aw-j-oq b’ay-tu
ipfv-3 3pl voice-intr-inf tomorrow
‘They went to shout there.’ Q’anjobal (Mateo Toledo 2017: 538)

Although Mateo Toledo (2017) does not present examples in which the article and the irrealis
marker co-occur, we find two examples of -oq on nouns, shown in (140) and (141) below. In
example (140), the irrealis marker is on a noun but the noun functions as the predicate of the
sentence. In (141), on the other hand, we see the marker on a noun that functions as arguments
and may have a nonspecific referent, as it is in the scope of the question.

(140) man anima-oq hach

neg person-irr 2sg
‘You are not a person.’ Q’anjobal (Mateo Toledo 2017: 551)
(141) maktxel max h-aq’-kan ko-taynomal-oq
who pfv 2sg-give-dir:stay 1pl-guard-irr
‘Who did you leave as our guard?’ Q’anjobal (Mateo Toledo 2017: 553)

These examples suggest that the irrealis marker is flexible enough to occur on nouns as well, even
though it is not clear why it combines with the article jun and not with the noun, the functionally

137
broad irrealis, future, and infinitive marker -oq can be assumed to be the source of the development
of the nonspecific article junoq in Q’anjobal. As I will show in the remainder of this section, other
unrelated languages show a similar development. However, in all other cases, the irrealis marker
attaches to the noun and not to the article. The only other clear example of an irrealis marker
that occurs with the indefinite article comes from the related language Ch’ol, as is shown in (142).
Note that in Ch’ol, this corresponds to the use of the irrealis marker and not to a nonspecific
article.

(142) ma’an [jum-p’ej=ik vivienda]

exist.neg art:indef-cl=irr residence
‘There isn’t any residence.’ Ch’ol (Vázquez Alvarez 2011: 239)

A development similar to the one in Q’anjobal is found in Hidatsa (Siouan, USA), only that the
development in Hidatsa does not seem to be as advanced as in Q’anjobal. The Hidatsa article sys-
tem consists of the definite article -s and the indefinite article -wa (Park 2012). Also for Hidatsa, it
was noted that the so-called conditional marker -rúg can attach to nouns and trigger a nonspecific
interpretation of the referent (Park 2012: 367). An example of this is shown in (143), contrasted
with the expression of a specific referent in (144):

(143) hiraacá-míà-rúg aru-m-úá-waa-c

Hidatsa-woman-cond irr-1-marry-caus.1-decl
‘I am going to marry a Hidatsa woman.’
literally: ‘If she is a Hidatsa woman I will marry her.’ Hidatsa (Park 2012: 368)
(144) hiraacá-míà-wa m-úá-waa-c
Hidatsa-woman-art:indef 1-marry-caus.1-decl
‘I married a Hidatsa woman. Hidatsa (Park 2012: 368)

As its Q’anjobal counterpart, the Hidatsa conditional marker -rug usually marks irrealis, condi-
tional, and future on both verbs and nouns. Examples (145) to (146) illustrate its use with a verb
and a noun, respectively.

(145) ééhgee-wa-rúg aru-nii-ma-hgi.wéˀ-he

know-1.act-cond irr-2.stat-1.act-tell-emph
If I knew it I would tell you.’ Hidatsa (Park 2012: 228)
(146) áàda-rúg magi-maa-íhgohbi-wihi-ˀo-ˀ
daylight-cond recip-1.act-meet-fut.q.1-pl-q
‘Shall we meet tomorrow?’ Hidatsa (Park 2012: 181)

Also in Crow, another Siouan language, a conditional marker may be related to a marker on
nouns that marks them as indefinite. However, it does not formally correspond to the nonspecific

138
article found in the language. Crow, like Lakota, has an article system with a definite (-sh), spe-
cific (-m) and a nonspecific (-eem) article (cf. example (66) in section 2.2.4). According to Graczyk
(2007: 233), in a certain narrative genre, the two articles from the indefinite domain are replaced
by another marker (-dak). Graczyk (2007: 230f) notes “dak is another indefinite nonspecific de-
terminer; it is homophonous with the conditional and temporal conjunction dak […] Dak occurs
relatively rarely as a determiner. It is best treated as generic irrealis marker that functions both
as a determiner and as a complementizer”. While the development of -dak as indefinite marker
on nouns is not entirely clear, Crow corresponds to another piece of evidence pointing towards
conditional markers from the verbal domain as source for nonspecific articles.
The origin of the nonspecific article wãži in the closely related language Lakota, which for-
mally also corresponds to the specific article wã and another segment -ži, is much less clear. There
is not an obvious marker in the verbal domain that would correspond to -ži (/ʒi/). Ingham (2003:
31-32) mentions a negative marker /ʃni/ (spelled -ṡni) : “The suffix -ṡni negates predicates and
thus only occurs following verbs or nouns when the latter occur as predicates”. However, relat-
ing these two markers at this point would only be speculation.
Two more examples of languages that show this “flexibility” of the conditional marker to
occur on nouns and result in the nonspecific interpretation are the two Wakashan languages
Nuuchahnulth and Makah. Examples (147) and (148) show this for the two languages, respectively.

(147) ʔu-ńaˑḥ-šiƛ=’aƛ=či ƛ’uq-aˑʔa=quː ḿuksýi

so.and.so-look.for-perf=temp=go.imper.2sg broad-on.rocks=cond stone
‘Go find a (any) broad stone!’ Nuuhchahnulth (Davidson 2002: 280)
(148) ʔu-suba=s ɬaʔaˑš=qeyu
so.and.so-need=indic.1sg bag=cond.3sg
‘I need a (any) bag.’ Makah (Davidson 2002: 280)

Thus, this section showed that there is evidence from several languages for the nonspecific article
to develop from a conditional or irrealis marker in the verbal domain, given that this marker is
flexible enough to occur on nouns, or, in the case of Q’anjobal, on the article itself.

5.8 Indefinite articles

Indefinite articles encode specific and nonspecific referents. In this section, I present three ex-
amples of this article type: the indefinite article in Tz’utujil (Mayan, Guatemala), Carib (Cariban,
Suriname), and Bonan (Mongolic, China). Although indefinite articles are commonly found in
European languages, this section shows that they are not restricted to this area of the world.

139
5.8.1 Tz’utujil
The marker jun in Tz’utujil (Mayan, Guatemala) is an example of an indefinite article, and was
treated as such in Dayley (1985) and Tz’utujil Tinaamitaal (2007). In contrast to the cognate
exclusive-specific article jun in Q’anjobal, the article in Tz’utujil is systematically used to mark
specific referents as well as nonspecific referents.
Examples (149), (150), and (151) illustrate the use of jun to mark referents as specific. In (149),
the referent of jay ‘house’ is introduced into the discourse; it is a particular referent from its kind
set which is not (yet) identifiable by the hearer. In (151), the referent of se’ep ‘gift’ is specific and
its identifiability is not relevant to the discourse.

(149) xkatzu’ [jun jay tz’bu’uk’] p’=tinamit

see.pst.s:1pl.o:3sg art:indef house old in=village
‘We saw an old house there in the village.’ Tz’utujil (prim. data)
(150) tetz’ta k’oola [jun ak’aal] chpaam ya’
look.imp exist art:indef boy inside water
‘Look! There is a boy in the water.’ Tz’utujil (prim. data)
(151) xjuwaj qas’utzna rte’et xuya [jun se’ep] tzra xtijjona
dance be.beautiful grandfather give.pst.s:3sg.o:3sg art:indef gift for teacher
kxin nkixjova
of dancers
‘The dance was so beautifully that the grandfather gave a gift to teacher of the dancers.’
Tz’utujil (prim. data)

The indefinite article jun cannot encode a definite referent. It is the absence of the article, as is
shown in example (152b), that leads to a definite interpretation of the referent, in this example of
k’olbek ‘wedding’.

(152) a. ximba p’=[jun k’olbek]

go.1sg to=art:indef wedding
‘I went to a wedding.’
b. ximba p’=k’olbek
go.1sg to=wedding
‘I went to the wedding.’ Tz’utujil (prim. data)

In Tz’utujil, the marking of plural in contrast to unmarked singular is commonly found with
human referents, but not obligatory with most other types of nouns (Dayley 1985: 140f). In (153)
and (154), we see that the indefinite article can be used with plural nouns.

140
(153) a. atrejben [jun xtan] chu jay
wait.s:3sg.o:2sg art:indef woman outside house
‘A lady is waiting for you outside.’
b. atkiyben [julee xtan-ij] chu jay
wait.s:3pl.o:2sg art:indef.pl woman.pl outside house
‘Some ladies are waiting for you outside.’ Tz’utujil (prim. data)
(154) [julee wixb’-ij] xinkeresta’ chaq’a’ iwir
art:indef.pl friend.pl-pl came.see night yesterday
‘Some friends came to visit last night.’ Tz’utujil (prim. data)

The article jun is an indefinite article because it is also used to encode nonspecific referents. Thus,
it does not disambiguate between a specific and nonspecific interpretation of the referent. Exam-
ples (155), (156), and (157) illustrate its use with nonspecific referents.

(155) k’ool [jun atz’ib’ab’el]?

exist art:indef pen
‘Do you have a (any) pen?’ Tz’utujil (prim. data)
(156) nk’atzin chwa’ [jun atz’aqoneel]
need to.1sg art:indef architect
‘I need an (any) architect.’ Tz’utujil (prim. data)
(157) nwajo’ [jun ch’eech’]
want.pres.s:1sg.o.3sg art:indef car
‘I want a (any) car.’ Tz’utujil (prim. data)

Even though examples (155) to (157) show that the indefinite article jun is used to mark nonspecific
indefinite referents, it is not obligatory and even marked in at least certain nonspecific contexts
involving the existential marker k’ool:17

(158) k’ool [(jun) k’olb’al] ab’aar nkat-wawa’ chpaam jawa tinamit

exist art:indef place where eat inside dem:prox village
‘Is there a place where one can eat (a restaurant) in this village?’
Tz’utujul (prim. data)
(159) A: k’ool [(?jun) tiyoxajaay] chpaam jawa tinamit?
exist art:indef church inside dem:prox village
‘Is there a church in this village?’
B: jee’ k’ool jun tiyoxajaay
yes exist art:indef church
‘Yes, there is a church.’
Tz’utujil (prim. data)
17 Neither Dayley (1985) nor Tz’utujil Tinaamitaal (2007) discuss this.

141
Even though there are exceptions to the use of jun with specific and nonspecific referents, I view
its distribution in these contexts as systematic and treat it as an indefinite article.

5.8.2 Carib
Another example of an indefinite article outside of Eurasia is the indefinite article in Carib (Cariban,
Suriname). Example (160) and (161) show that the marker amu is used to introduce new and not
(yet) identifiable referents into the discourse. Also in (162) and (164) below, amu occurs with
specific referents.

(160) dan [amu topu] man kawo

then art:indef rock be.3sg high
‘Now there was a rocky mountain.’ Carib (Courtz 2008: 166)
(161) dan amu ty-wai-je man kurupi
then art:indef 3-become-adj be.3sg kurupi
‘There was a kurupi (a monster), once.’ Carib (Courtz 2008: 150)
(162) [uwapoto-mpo amu] nirompy-i rapa
older.person-dev art:indef die-recpst again
‘It happened again that an old person died.’ Carib (Courtz 2008: 145)
(163) typo roten [akuri amu] si-ene-ja
apart just red.rumped.agouti art:indef 1-see-prs
‘Suddenly, I saw an agouti.’ Carib (Courtz 2008: 182)
(164) moro wara ro rapa [woryi mati amu] si-ene-ja y-jena y-woneto
that like continue again woman black art:indef 1-see-prs 1-in.arms 1-dream
‘I dreamt and saw a black woman in my arms.’ Carib (Courtz 2008: 187)

That amu is an indefinite article is shown by its use with referents that can be nonspecific. Ex-
amples to illustrate this are given in (165), (166), and (167). Here, neither the article amu nor the
context disambiguate between a specific and a nonspecific interpretation of the referent.

(165) erome [tonomy amu] si-epory-taka rapa

now animal art:indef 1-find-fut again
‘Today, I’ll find game again.’ Carib (Courtz 2008: 187)
(166) [akuri teràa amu-kon] si-upi-ja
reed.rumped.agouti already art:indef-pl 1-seek-prs
‘I am going to try and find some agoutis.’ Carib (Courtz 2008: 188)

142
(167) isème te [amu18 po-no amamin-nano kapy-ry] ‘se wa [amu
in.spite.of but art:indef house at-adnom work-nposs make-poss wanting
pyrata-‘me-mpo] enepy-tòme y-‘wa
be.1 art:indef money-little-dev bring-instr 1-to
‘Still, I want to do some work at home to earn some money.’ Carib (Courtz 2008: 198f)

Example (166) shows that the indefinite article is compatible with plural and even bears the only
plural marker in the noun phrase.

5.8.3 Bonan
We also find an indefinite article in Bonan (Mongolic, China). Example (168) shows that the
indefinite article -gə occurs with the noun ahku ‘Buddhist monk’ to introduce the nonidentifiable
referent into the discourse:

(168) hkutə tɕaɕi silaŋ=da o=ku tɕhəχaŋnaŋ atɕaŋ pantɕhe terə

yesterday Jiashi Xining=loc go=ipfv.nmlz time 3sg bus on
ahku=gə tɕolaʁa-tɕə
Buddhist.monk=art:indef meet-pfv
‘Jiashi met a monk on the bus when he went to Xining yesterday.’
Bonan (Fried 2010: 82)

We see the marker -gə with other specific referents in (169) to (172). In (169) and (170), the article
occurs in the object position of a transitive clause. Example (171) shows that the indefinite article
-gə is also used in the subject position.

(169) atɕaŋ dzəjaŋ=gə=tɕə khidzə=gə ap-tɕo

3sg bowl=art:indef=and chopsticks=art:indef buy-ipfv.o
‘He bought a bowl and a pair of chopsticks.’ Bonan (Fried 2010: 167)
(170) atɕaŋ mənda mehtoχ tɕəχaŋ=gə oχ-tɕo
3sg 1sg.loc flower white=art:indef give-ipfv.o
‘He gave me a white flower.’ Bonan (Fried 2010: 146)
(171) deda ŋarwaŋ=gə pətə χeɕaŋ pitəʁaŋ
grandfather powerful=art:indef 1pl.ex.coll Leru small
khər-saŋ=nə thorun hko=nə khərə=tɕə khəl-saŋ
be.required-pfv.nmlz=acc head big=acc be.required=qot.impf say-pos
‘A powerful grandfather said, “as we have to have the small Leru festival, [we] must have
a big well”. Bonan (Fried 2010: 118)
18 Courtz (2008: 139) mentions that amu tends to occur prenominally with count nouns and postnominally with mass
nouns.

143
(172) dedə=la laχɕi tɕo-tɕə sarə=la=nə χamdə hkuraŋ=gə jiχ-sa
grandfather=pl towel wear-ipfv young=pl=acc together circle=art:indef do-cond
thəke-saŋ
permit-pos
‘The grandfathers can wear a towel and (dance) in a circle with the young (men).’
Bonan (Fried 2010: 200)

Fried (2010: 47) notes that definiteness in object positions is signalled by the absence of the ac-
cusative marker -nə, which causes the indefinite article to not be required in this position; a
number of specific and/or nonspecific referents in object positions are indeed expressed as bare
nouns in Fried (2010). An example is given in (173) below.

(173) lapsi=da saŋ χəra-tɕə tɕhəma=nə χara-tɕo

Lapsi=loc offering burn-ipfv Chuma=acc taunt-o.ipfv
‘And at Lapsi (they) burn offerings and taunt Chuma.’ Bonan (Fried 2010: 329)

Nevertheless, the previous sentences in (169) and (170) showed that the indefinite article does
occur in the object position; in (174) below we can see that it even is compatible with the accusative
marker.

(174) nadamada khəl-də=ku tɕəχaŋnaŋ jimaŋ=gə=nə

long.ago speak-imp=ipfv.nmlz time goat=art:indef=acc
sugə-saŋ=tɕə khəl-na
argue-pos=qot.ipfv speak-dur
‘It is said that long ago, when (we) were fighting, (we) were arguing over a goat.’
Bonan (Fried 2010: 330)

As an indefinite article, -gə must also occur with nonspecific indefinite referents. Examples for
contexts with referents whose interpretation is ambiguous between being specific and nonspecific
are given in (175) and (176). Again, we see that the indefinite article -gə occurs with the nouns of
these referents.

(175) dʐoma htɕəχta χənə=gə ap=ku taraŋ ər-tɕo

Droma bicycle new=art:indef take=ipfv-nmlz desire come-ipfv.o
‘Droma wants to buy a new bike.’ Bonan (Fried 2010: 82)
(176) pə dʐoma=da samtəχ=gə oχ=ku taraŋ ər-na
1sg Droma=loc gift=art:indef give=ipfv.nmlz desire come-dur
‘I want to give a gift to Droma.’
literally: ‘The desire that I give a gift to Droma comes.’ Bonan (Fried 2010: 282)

144
That the article -gə does not occur in definite contexts is shown in examples (177), (178), and
(179) below. Definite referents in Bonan are expressed by bare nouns. In (177), ‘the boxes’ are
situationally unique in the context of the story told, the same is the case with ‘the teachers’ in
(179) in the context of a story concerning a certain school. The sentence in (178) occurs in a story
and follows another sentence that refers to a ‘mountain’, making the noun in (178) an anaphoric
referent.

(177) səme=la=nə ula=da dabla-tɕə o khər-na

god.box=pl=acc mountain=loc raise-ipfv go be.required-dur
‘(We) must carry the god boxes to the mountain.’ Bonan (Fried 2010: 330)
(178) de χara-tɕə=ku tɕənsan sowu ula=nə oχ-to=tɕi-saŋ
just taunt-ipfv=nmlz reason important mountain=acc give-pfv=s.qot.ipfv-epist
‘But we consider the main reason for taunting (them) to be that we gave them the moun-
tain.’ Bonan (Fried 2010: 331)

(179) gərgən=da ɕikə sokərgə-tɕə

teacher=loc very respec-pfv
‘Be very respectful toward the teachers.’ Bonan (Fried 2010: 320)

Since it could be established that -gə is systematically used with specific and nonspecific referents,
I treat it as an indefinite article.

5.9 Presentational articles

5.9.1 Introduction
In a number of languages, we find markers which systematically occur with specific referents
in discourse-prominent positions. The presence of such markers exceeds the occasional use of
the numeral ‘one’ in this type of contexts (cf. section 3.2). Nevertheless, this type of article is
somewhat different from indefinite articles: it mainly occurs with referents that are introduced
into the discourse and that represent topics in the sense of first elements of topic chains as well
as with referents that are human or animate. Since these parameters are highly correlated, I
use the cover term “discourse-prominent” here. The most important property of presentational
articles is that they seem to be somewhere in the middle of a scale from the numeral ‘one’ with
an occasional presentational function on one end and with the indefinite article on the other end.
This corresponds to the traditional grammaticalization path for indefinite articles (cf. section 3.2).
Therefore, it seems plausible to regard these presentational markers as some sort of emerging

145
specific articles. In this section, I propose that a presentational article is not necessarily a marker
of such an intermediate stage between the numeral ‘one’ and the indefinite article.
The first reason for treating the presentational article as a separate article type is that its use
is systematic, even though it is restricted to discourse-prominent referents. The second reason
is that their use may not strictly bound to specific referents. Data from Chatino suggests that
presentational articles may also be able to encode nonspecific referents, only that this use can be
expected to be very infrequent given the restriction of the article to discourse-prominent referents.
This means that the presentational article is not necessarily in such a middle position between
the numeral and the indefinite article, and it means that its restriction to discourse-prominent
referents may be, to a certain degree, independent of the restriction to specific referents. The
third reason is that regarding presentational articles as emerging specific or indefinite articles
implies that they necessarily develop further and extend to the coding of less prominent referents.
However, we do not know whether articles that are restricted to discourse-prominent referents
necessarily “unstable” across time; it may also possible that presentational articles do not develop
into specific or indefinite articles.

5.9.2 Examples
5.9.2.1 Lango

One example for what I call presentational articles can be found in Lango. The marker -mɔ́rɔ̂ 19
systematically marks referents as specific, but it only occurs in certain types of contexts. Examples
(180) and (181) below illustrate the use of -mɔ́rɔ̂.

(180) twòl-lɔ́rɔ̂ òkàò àtîn

snake-art:pres bite.perf.3sg child
‘A snake bit the child.’ Lango (Noonan 1992: 162)
(181) pʊ́ nɔ́’-mɔ́rɔ̂ tíê ì ŋè ɔ̀ t
pig-art:pres be.present.hab.3sg in back house
‘There’s a pig behind the house.’ Lango (Noonan 1992: 162)

However, -mɔ́rɔ̂ is not generally required to mark referents as specific, as the following two exam-
ples show. Even though the referents of lócə̀ ‘man’ in (182) and mɛ́ ‘girl’ (183) are not identifiable
by the hearer, they do not require a specific marker.

(182) àŋéô lócə̀ à tɛ̂k à ryɛ̀

know.hab.1sg man attr strong attr wise
‘I know a strong and wise man.’ Lango (Noonan 1992: 164)
19 The initial nasal is phonologically assimilated to the last segment of the noun that the article occurs with.

146
(183) nwàŋ gìn òlárò ɲákô mɛ́ ɲòm
long.ago they compete.pfv.3pl girl for marriage
‘Long ago they were competing to marry a girl.’ Lango (Noonan 1992: 294)

Noonan (1992: 147,162) notes that specific referents, expressed as subjects in existential sentences,
are often accompanied by -mɔ́rɔ̂, and describes the distribution of -mɔ́rɔ̂ in specific indefinite
contexts as follows:

Not all indefinites are marked with -mɔ́rɔ̂ ; in fact, the majority are not […] [It] is almost always
found in subject position. The reason for this is that the Lango sentence is organized roughly
on an old-new continuum. Subjects are always interpreted as definite unless specifically made
indefinite with -mɔ́rɔ̂ (hence the association with subjects in existential constructions). In
other sentence positions, NPs are assigned definite or indefinite interpretations pragmatically,
and do not require an explicit indefinitizer like subjects. (Noonan 1992: 162)

This suggests that the presentational article in Lango is used systematically with exactly those
referents that are introduced to the discourse and that will stay discourse-prominent. This also
correlates with the occurrence of the article in existential constructions. Because the article occurs
systematically in such constructions in the data presented in Noonan (1992), I treat -mɔ́rɔ̂ in Lango
as a presentational article.

5.9.2.2 Logba

Another example of a presentational article is found in Logba (Kwa, Ghana). In example (184), the
referents e-bítsí-klɔyì ‘small child’, a-gbɛ́ ‘dog’, and a-kpákplá ‘frog’ are not yet identifiable by the
hearer, since they are introduced into the discourse. We see that the marker ɔ-kpɛ is used with
the first of the enumerated nouns to mark its referent (and potentially the referents of the other
nouns as well) as specific.

(184) e-bítsí-klɔyì ɔ-kpɛ a-gbɛ́ kp’ɛ a-kpákplá é-nyá20

cl5-child-small cl5-art:pres cl1-dog and cl1-frog s:pl-live
‘There lived a small child, a dog, and a frog.’ Logba (Dorvlo 2008: 62)

Examples (185) and (186) show the use of ɔ-kpɛ in these other contexts. The referents of u-dzɛ
‘woman’ and ɔ-sá ‘man’ are not identifiable by the hearer (and their identification is is not relevant
for the discourse) which is signaled by the use of -ɔkp(i)ɛ.
20 Dorvlo(2008) only glosses class marker morphemes as such, without indicating the class of the noun. He lists
four “singular plural pairings” of noun class markers and shows the agreement morphemes on attributes and verbs
(Dorvlo 2008: 44-45). The class indications in the glosses are mine, based on the overview table of noun classes and
agreement given in (Dorvlo 2008: 47).

147
(185) maŋú [u-dzɛ tsengo ɔ-kpɛ]
1sg-see cl3-woman old cl3-art:pres
‘I saw an old woman.’ Logba (Dorvlo 2008: 82)
(186) [ɔ-sá tsengo ɔ-kpiɛ́] ó-tsi mɔ́
cl7-man old cl7-art:pres s.sg-sit there
‘An old man sat there.’ Logba (Dorvlo 2008: 91)

(187) e-bìtsi-ɛ́ ɔ-kla fɛ́ [ɔ-yɔ́ mango ɔ-kpiɛ] dzáa ko etsi

cl5-child-art:def s:sg-hide into cl7-tree big cl7-art:pres no.addition only under
‘The child went and hid under only a big tree.’ Logba (Dorvlo 2008: 93)

Examples of specific referents that occur as bare nouns are given in (188), (189), and (190). Note
that while the positions in which the presentational article occurred were often subjects, we find
specific referents that are expressed as bare nouns mostly in object positions.

(188) a-wu á-bo-wú

cl9-dress 3sg-stay-2sg
‘You have a dress.’ Logba (Dorvlo 2008: 109)
(189) idze ati-bí-ta ogridi wá anɖa
today 1pl-fut-give story say friend
‘Today, we shall tell a story one another.’ Logba (Dorvlo 2008: 284)
(190) a-gbi=ɛ́ ó-ŋú n-wo u-tsa
cl1-dog=art:def cl1-see cl2-bee cl3-home
‘The dog saw a bee hive.’ Logba (Dorvlo 2008: 45)

5.9.2.3 Aghul

The third example of presentational articles comes from Aghul (Lezgic, Russia). The presentational
article sa is based on the numeral ‘one’. Examples (191), (192), and (193) show that the marker is
used to introduce discourse-prominent referents.

(191) hal č˳ a-s zun qatːaq’.a-s-e za-s ʜabaw.a qatːaq’.u-nde [sa q’isa]
now 2pl-dat I tell.ipfv-inf-cop 1sg-dat granny.erg tell.pfv-ptcp art:pres story
‘Now I am going to tell a story that a granny told me.’ Aghul (Ganenkov et al. 2009: 10)
(192) aχpːa me χinebi qatːq’.a-j-e [sa q’isa] hate tüʜür ʜür.i-ʔ uč.i-l
then dem Khinebi tell.ipfv-cv-cop art:pres story dem Tükhür village-in refl-super
alčarx.u-nde
super-lat.fall.pfv-ptcp
‘Then this Khinebi is telling us a story about an incident, that happened to him in that
village of Tükhür.’ Aghul (Ganenkov et al. 2009: 6)

148
(193) a. x.u-na-a [sa iǯe kas]
become.pfv-res-prs art:pres good man
‘There once was a good man.’
b. mi-l alčarx.a-j-e [sa ʕurčaqan]
dem-super meet.ipfv-cv-cop art:pres hunter
‘He was met by a hunter.’
Aghul (Ganenkov et al. 2009: 10)

Example (194) illustrates the use of the presentational article with the specific referent of idemi
‘man’, which is relevant to the discourse, but whose identifiability is not important. Thus, in con-
trast to the specific referents in (191), (192), and (193), the referent of idemi ‘man’ is not necessarily
going to be identifiable by the hearer in the course of the discourse.

(194) Some young people from our village arrived in Mugan to work.
muʁan.di-ʔ alčarx.a-a [sa jašlu idemi]
Mugan-iness meet-ipfv-prs art:pres old man
‘In Mugan, an old man is meeting them.’ Aghul (Ganenkov et al. 2009: 7)

Furthermore, we see in (195b) and (196) that the presentational article sa can be used with plural
nouns as well:

(195) a. č˳ e χul.a-s [sa qːari] š.u-ne

poss:2pl house-dat art:pres old.woman come-pfv-perf
‘An old woman came to your house.’
b. č˳ e χul.a-s [sa qːari-jar] š.u-ne
poss:2pl house-dat art:pres old.woman-pl come-pfv-perf
‘Some old women came to your house.’
Aghul (Ganenkov et al. 2009: 8)
(196) me ag.u-na-a [sa jiharkːa-jar.i-s]
dem see.pfv-res-prs art:pres milker-pl-dat
‘Some milkers caught sight of her.’ Aghul (Ganenkov et al. 2009: 8)

5.9.2.4 Chatino

In Teotepec Chatino (Zapotecan, Mexico), we find another example of a presentational article.

Referents that are prominent in the discourse can be marked as nonidentifiable in Chatino by the
use of the marker skar . Examples for its occurrence with specific referents that the hearer cannot
identify are provided in (197), (198), and (199) below.21
21 The superscribed numbers in the Chatino examples indicate tone, ‘r’ stands for “relaxed tone” (McIntosh 2011: 60).

149
(197) mn7ą3 mer lo31 [skar kto3 ] ka13
see.compl Carmelo art:pres chicken yesterday
‘Carmelo saw a chicken yesterday.’ Chatino (McIntosh 2011: 106)
(198) xa31 wa734 mdya21 nir la1(+0) kwa13 nir lo713 ntkwa13 [skar nur jn7ą31 ]
when already go.compl 3s.rp place there 3s.rp and sit.hab art:pres nom woman
‘When she arrived, a woman was there.’ Chatino (McIntosh 2011: 134)
(199) nskwa13 [skar na0(34) ] lor msa23
lie.hab art:pres thing on table
‘There is a thing on the table.’ Chatino (McIntosh 2011: 102)

The use of the marker skar is restricted to discourse-prominent referents. With specific referents
that are not relevant to the discourse or whose identifiability does not play a role in the discourse,
we do not find the marker skar to mark them as specific indefinite. Examples for this type of
referents, expressed by bare nouns, are given in (200) and (201).

(200) jykur nda31 nyer nsna23

eat.compl Daniel apple
‘Daniel ate an apple.’ Chatino (McIntosh 2011: 94)
(201) mt7o31 -(n)skwa0(34) 7įr 7ya13 kyee31(r) s7ęr nskwa2 wa734 ni0(34)
go.out.compl-lie.prog to near stone place lie.hab now
‘It came out lying on a stone, where it is to this day.’ Chatino (McIntosh 2011: 117)

Since the use of skar seems really restricted to referents of typical presentational contexts and
other nonidentifiable referents that are prominent within a given discourse sequence, I treat skar
in Chatino as a presentational article. Although being a presentational article, skar is not restricted
to the use with specific indefinite referents. The the two examples in (202) and (203) show that
skar also occurs in contexts that seem less typical for presentational articles. In (202), we see that
skar can also be used with a nonspecific referent. On the other hand, its counterpart in (203) also
features a nonspecific referent which is expressed as the bare noun knya731 .

(202) ja n7įr skar ntę2(+0) re34

neg live.hab art:pres people dem
‘A person doesn’t live here.’ Chatino (McIntosh 2011: 104)
(203) s7i2(+0) knya731 kar nur mn7ą21
neg deer be.prog rel see.compl.2sg
‘You did not see a deer.’ Chatino (McIntosh 2011: 104)

150
5.9.3 The role of presentational articles for article development
The example from Chatino suggested that a presentational article as systematic referential marker
restricted to discourse-prominent referents does not need to be restricted to specific referents, but
can also be used to mark nonspecific referents. This is is evidence for the use of presentational arti-
cles with discourse-prominent referents being independent from specificity, which is an argument
against the seemingly plausible hypothesis of presentational articles being emerging specific arti-
cles. It may be that most discourse-prominent referents are specific, which could account for the
fact that we mostly find presentational articles with specific referents. What Chatino showed is
that presentational articles need not be restricted to specific indefinite contexts. Exclusive-specific
articles, as was shown in 5.5, are strictly restricted to specific contexts. Although it seems tempt-
ing to assume that presentational articles represent an intermediate stage between the numeral
‘one’ and specific articles on the grammaticalization scale, their compatibility with nonspecific
referents shows that this is not necessarily the case.
Another important point shown in this section with the presentational articles in Logba and
Aghul is that they are not necessarily restricted to singular contexts, even though originating from
the numeral ‘one’. Often, the availability of a marker of specificity or indefiniteness is viewed as
argument for or against its classification as an article. The underlying assumption is that some-
where in its development from the numeral ‘one’ (through a presentational and a specific stage)
to an indefinite article, the marker extends its use to plural referents. If a presentational article
really develops into an indefinite article, the fact that presentational articles can already occur in
the plural shows that formal (plural form) and functional (nonspecific contexts) extensions are
independent as well. Again, they are often correlated, the two extensions must not be confused.
This means that formal criteria such as the availability of a plural form cannot be taken as the
basis of classifying a marker as indefinite article or not. Presentational articles show that the
availability in plural contexts or the development of plural forms is not necessarily tied to the
functional extension of a marker to nonspecific contexts, but that it is attested independently.
Hence, the presentational articles in Logba and Aghul underline the importance of functional
criteria as outlined in 3.2 to decide whether a given marker is an article or not.

5.10 Referential articles

This section discusses referential articles. Referential articles encode the three major referent
types of definite, specific, and nonspecific referents. I present two case studies of referential ar-
ticles in more detail and mention a few other examples. The first example of a referential article
comes from Rapa Nui (Oceanic, Chile). The second example of a referential article is found in
Halkomelem (Salishan, Canada). These two referential articles seem very different at first sight:

151
the article in Rapa Nui has one exponent; the article in Halkomelem, on the other hand, corre-
sponds to various exponents. I argue that these are not separate articles or deictic markers, but
exponents of a single referential article that can mark deixis in addition to their referential func-
tion.

5.10.1 Rapa Nui

In Rapa Nui (Oceanic, Chile), the marker te is distributed across different referential contexts; it
is used to mark definite, specific, and nonspecific referents.22 Examples (204) to (206) show the
article in different types of definite contexts. In (204), the referent of nu’u ‘people’ is anaphoric
and deictic. Example (205) shows te with an establishing referent. Examples (206) features a
situationally unique referent.

(204) ki a kōrua [ki te nu’u] hakaroŋo mai ’ī a au he kī atu

to prop 2pl to art:ref people listen hither imm prop 1sg ntr say away
‘To you, to the people listening, I tell you …’ Rapa Nui (Kieviet 2017: 46)
(205) matu ki oho tatou ki Vērene ki u’i i [te me’e] haka ’ite mai
come.on hort go 1pl.in to Bethlehem to look acc art:ref thing caus know hither
ena
med
‘Come, let’s go to Bethlehem, to see the thing announced (to us).’
Rapa Nui (Kieviet 2017: 46)
(206) ’ina he mā’eha mo u’i iŋa i [te kai]
neg pred light for see nmlz acc art:ref food
‘There was no light to see the food.’ Rapa Nui (Kieviet 2017: 93)

That the referential article te also occurs in specific contexts is shown in (207) and (208). In both
examples, the nouns is linked to a particular referent from its kind set, but the referent is not
identifiable by the hearer.

(207) ’i te noho iŋa tuai era ’ā [te taŋata e tahi] te ’īŋoa

at art:ref stay nmlz ancient dist ident art:ref man num one art:ref name
ko Tu’uhakararo
prom Tu’uhakararo
‘In the old times (there was) a man called Tu’uhakararo.’ Rapa Nui (Kieviet 2017: 238)
22 Inthis section, I only discuss the markers te and he which are used for common nouns. Like in many Oceanic
languages, Rapa Nui has a different nominal marker, a, which occurs with proper nouns.

152
(208) ko tu’u ’ana a au ki ruŋa i [te henua e hitu]
perf arrive cont prop 1sg to above at art:ref land num seven
‘(In my dream) I arrived on seven islands.’ Rapa Nui (Kieviet 2017: 238)

In order to be treated as a referential article, te is required to mark nonspecific referents as well.

Examples (209) to (212) show that this is what we find in Rapa Nui. In these examples, the referents
that occur with the article te are nonspecific because they do not correspond to any particular
referent from their kind that is linked to the noun.

(209) e ai rō ’ana hō [te me’e] mo ta’e rova’a e te ’Atua mo aŋa

ipfv exist emph cont dub art:ref thing for neg obtain ag art:ref God for do
‘Would there be any thing that God is not able to do?’ Rapa Nui (Kieviet 2017: 192)
(210) he kī ō’oku ki kō’ku ŋā poki taina era mo oho o mātou mo kimi i
ntr say o.poss:1sg to o.poss:1sg pl child sibling dist for go of 1pl.ex for search acc
[te pipi]
art:ref shell
‘I told my brothers and sisters that we would go to look for (any kind of) shells.’
Rapa Nui (Kieviet 2017: 239)
(211) ko mate atu ’ana ki [te vai] mo unu
perf die away cont to art:ref water for drink
‘I’m dying for (any kind of) water to drink.’ Rapa Nui (Kieviet 2017: 237)
(212) e ai rō ‘ā [te ika] o roto?
ipfv exist emph cont art:ref fish of inside
‘Are there any fish inside?’ Rapa Nui (Kieviet 2017: 241)

Although we saw that the article te occurs with definite, specific, and nonspecific referents, its
status as article is not uncontroversial. From a syntactic point of view, te occurs in complementary
distribution with the marker he. The latter is primarily used to mark nouns in predicate position,
as is shown in (213).23 Moreover, we see in (214) that he is not felicitous in argument positions in
which the article te was shown to occur in.

(213) [he taŋata] tau manu era

pred man dem bird dist
‘That bird was a human being.’ Rapa Nui (Kieviet 2017: 242)
(214) *ko tike’a ’ā a au (i) [he honu]
perf see cont prop 1sg acc pred turtle
intended: ‘I have seen a turtle’ Rapa Nui (Kieviet 2017: 237)
23 Sincea closer look at the nature of the marker he would go beyond the scope of this study, I simply gloss he as
predicate marker following Weber (2003) and Kieviet (2017), although I do not necessarily follow the analysis of he
as a determiner proposed in Kieviet (2017: 244f).

153
Therefore, one could also analyze te and he as syntactic markers, i.e. he as a predicate marker and
te as an argument marker. Or, te could be analyzed as a nominal marker which marks the word
that it occurs with as a noun. The marker he, in turn, would then be a predicate marker, indicating
that the word it occurs with is a verb or a predicate. This is a plausible analysis especially because
Rapa Nui has flexible word classes, meaning that many lexemes can be used as either nouns or
verbs. Example (215) illustrates this for the lexeme poki ‘(be) child’, which is compatible with
nominal and verbal contexts where it can occur as an argument (215a) or as a predicate (215b).
We see the same effect for the lexeme nuinui ‘big(ness)’ in (216a) and (216b).

(215) a. he pōrekoreko [te ŋā poki] ’i Tāhai.

ntr born:red art:ref pl child at Tahai
‘Children were born in Tahai.’ Rapa Nui (Kieviet 2017: 76)
b. mai te hora era ō’oku e poki nō ’ana …
from art:ref time dist poss.3sg.o ipfv child just cont
‘From the time when I was a child …’ Rapa Nui (Kieviet 2017: 76)
(216) a. [te nuinui o Tahiti] e ’āmui atu tāua e ono nuinui nei o Rapa
art:ref big:red of Tahiti ipfv add away 1du.in num six big:red prox of Rapa
Nui
Nui
‘The size of Tahiti altogether is six times the size of Rapa Nui.’
Rapa Nui (Kieviet 2017: 255)
b. mai ki hāpa’o nō tātou i a ia ’ātā ka nuinui rō
hither hort care.for just 1pl.in acc prop 3sg until cntg big:red emph
‘Let us take care of him until he is big.’ Rapa Nui (Kieviet 2017: 330)

While the article te is certainly one of the indicators that poki ‘child’ in (215a) corresponds to a
noun and not to a verb, also the syntactic position and e.g. number marking indicate the syntactic
status of poki. Hence, I follow Kieviet (2017: 238-240) in regarding te as an article. Furthermore,
I show later in this section that an analysis as a referential article accounts for the contexts te
occurs in, and also for those it does not (and in which the marker he is used instead), while a
syntactic only accounts for a subset of contexts in which te does not appear.
Other contexts24 in which te does not occur are the following: the instrumental preposition
hai ‘with’ (217), the comparative preposition pē ‘like’ (218), and appositions (219). While no other
prenominal marker occurs after the preposition hai in (217), examples (218) and (219) show that
he is (can be) used with the noun instead.
24 For other minor constructions that condition the distribution of te and he, cf. Kieviet (2017: 235f).

154
(217) he tunu mā’ea vera haka hopu i te poki [hai vai vera]
ntr cook stone hot caus bathe acc art:ref child instr water hot
‘He cooked (the water) with hot rocks, and bathed the child with hot water.’
Rapa Nui (Kieviet 2017: 264)

(218) ’iti’iti [pē he kio’e] hāpa’o i te hare [pē he paiheŋa] haka ’āriŋa
small:red like pred rat care.for acc art:ref house like pred dog caus face
‘Small like a mouse, guarding the house like an insolent dog.’
Rapa Nui (Kieviet 2017: 244)
(219) he kī [e te matu’a tane era o Te Rau [(he) taŋata pū’oko o te nu’u
ntr say ag art:ref parent male dist of Te Rau pred man head of art:ref people
o Kapiti]] …
of Kapiti
‘The father of Te Rau, the leader of the people of Kapiti, said …’
Rapa Nui (Kieviet 2017: 243)

The absence of te in these contexts can be accounted for by the conventionalization if te is taken
as a marker of referentiality. Especially for nouns that are instruments or that occur in similative
constructions, we can assume that they typically not used to refer but rather to evoke a particular
concept. In appositions, the referent is usually identical to the referent in its anchor (Heringa 2011:
5), which also motivates the absence of the article as a referential marker in such constructions.
If these functionally motivated trends conventionalize, they can become part of the grammar,
resulting in blocking effects for the article. This can account for the incompatibility of the refer-
ential article te in Rapa Nui with the constructions shown in (217) to (219). Moreover, this is in
accordance with the fact that te is not generally blocked in the presence of prepositions. As we
can see in (220), te occurs with the spatial preposition i ‘at’ that probably does not show the same
tendency of the instrumental marker hai ‘with’ of taking nonreferential nouns.

(220) he eke ki ruŋa [i te mā’ea e tahi]

ntr go.up to above at art:ref stone num one
‘He climbed on a stone.’ Rapa Nui (Kieviet 2017: 247)

An explanation based on te being a syntactic marker may also capture these patterns, but it seems
less evident why te cannot occur with certain prepositions, while it is required with others. Hence,
I treat te as a referential article, noting that its referential properties are certainly correlated with
its syntactic behaviour.

155
5.10.2 Halkomelem
Musqueam25 Halkomelem (Salishan, Canada) has a number of markers that frequently occur with
nouns. Although they are treated as separate articles with different deictic functions in Suttles
(2004), I argue in this section that they are better regarded as exponents of a single referential
article.
An overview of the article exponents in Halkomelem based on (Suttles 2004: 339) is given in
Table 5.10.26 In addition, the article has different exponents according to a category of deixis and
visibility. I argue in this section that the marking of deixis is only an additional function of the
referential article. Table 5.10 shows the exponents of the article according to its deictic functions.
Suttles (2004: 340) distinguishes between proximal (he calls it “present”) and visible, proximal
(“nearby”) and nonvisible, and remote.27

Table 5.10: Halkomelem article exponents

article exponent
proximal & visible tə
proximal & invisible kʷθə
remote ḱʷə

The following examples illustrate the deictic functions of the article. In (221) the referent of
sqʷəméý ‘dog’ is marked as proximal and invisible by using kʷθə; Suttles (2004: 342) notes that this
sentence is only felicitous in a scenario in which the speaker is outside of the house but close to it,
the dog being inside the house. Example (222) with the article tə marking visibility corresponds
to a scenario in which the speaker again is outside the house, and in which the dog is coming
outside towards the speaker.

(221) niˀ skʷtéxʷ ˀə tə léləḿ [kʷθə sqʷəméý]

be.there inside obl art:ref.vis house art:ref.nvis dog
‘The dog is in the house.’ Halkomelem (Suttles 2004: 342)
25 Halkomelem can be split into three major groups of dialects: Island varieties, Downriver varieties, and Upriver
varieties. The Musqueam variety of Halkomelem belongs to Downriver varieties of the language. Although I only
refer to the Musqueam variety in this section, the determiner system seems to be similar in Island Halkomelem
(e.g. Gerdts & Hukari 2004). The “determiner system” in Upriver Halkomelem, on the other hand, shows greater
differences, for details see Brown & Thompson (2013), Galloway (1993), Wiltschko (2002).
26 Suttles (2004: 339-340) further distinguishes “feminine” forms of the articles; however, he also notes that these are

used only with human and female referents. For this reason, I do not discuss these forms here.
27 Suttles (2004: 340) lists additional shortened variants of the article exponents. Since these are rare in the examples

that he provides, I do not list them here. In addition, (Suttles 2004: 348) treats another marker as an article. This
marker only occurs with proper nouns in oblique positions. Since its use is restricted to a subset of nouns and
certain syntactic contexts, I do not regard it as a referential marker and hence do not treat it as an article.

156
(222) ˀi ˀəḿí ˀə́ ƛ́qəl [tə sqʷəméý]
be.here come exit art:ref.vis dog
‘The dog came out.’ Halkomelem (Suttles 2004: 342)

Example (223) shows that the marking of deixis and visibility is not always straightforward: the
article tə for visible proximal referents is rather contrasted with the “remote” form ḱʷə instead of
being contrasted with the proximal invisible form kʷθə. Another example for the deictic use of
ḱʷə is given in (224), although Suttles (2004: 344) notes that it additionally implies that the hearer
cannot identify the pasture .

(223) a. niˀ ˀə [tə nə-ˀéθəqən]

be.there obl art:ref.vis my-front
‘in front of me (in a canoe)’ literally: ‘there at my front’
b. niˀ ˀə [ḱʷə nə-słiˀáˀaqʷt]
be.there obl art:ref.rem my-rear
‘behind me (in a canoe)’ literally: ‘there at my rear’
Halkomelem (Suttles 2004: 344)
(224) néḿ cən ˀə [ḱʷə spéłxən]
go I obl art:ref.rem pasture
‘I’m going to the pastures (way off, out of sight.)’ Halkomelem (Suttles 2004: 344)

Although we saw in the preceding examples that the article in Halkomelem is involved in marking
deixis and visibility, they are better accounted for as different exponents of a single referential
article that systematically occurs with definite, specific, and nonspecific referents and is absent in
nonreferential contexts.28
Definite referents that are marked by the article are presented in examples (225) to (227). Ex-
ample (225) shows the first mention of smə́yəθ ‘deer’ in a story about deer hunting. In the context
of the story, the referent is situationally unique and thus definite. An anaphoric referent is shown
in (226). The sentence that it occurs in follows the sentence in (225). Example (227) features a sen-
tence that is uttered with the dog being present in the discourse situation; therefore, the referent
of ˀənsqʷəméý ‘your dog’ is deictic. These examples also show that visibility is coded indepen-
dently from definiteness: in (225) and (226), the form kʷθə is used to mark nonvisibility, while
example (227) shows the form tə, encoding visibility.
28 Gerdts
& Hukari (2004: 159) make a similar remark, noting: “Semantically, articles refer to definite and indefinite
NPs indiscriminately, as is general in Salish languages.”

157
(225) ƛ́á ceˀ kʷə́ x-t [kʷθə smə́ yəθ] wə-mí-əs cəˀ
be.3 fut shoot-tr art:ref.nvis deer when-come-sub:3 fut
wəł-cłáqʷ-θət
already-pass.through-self
‘They will be the ones who will shoot the deer when they start coming through.’
Halkomelem (Suttles 2004: 525)

(226) wə-ḿí-əs wəłwéł-ət-əs [kʷθə sməyəθ]

when-come-sub:3 be.chasing-tr-sub:3 art:ref.nvis deer
‘…when they come chasing the deer.’ Halkomelem (Suttles 2004: 525)
(227) łə́ q́ ˀə q́ íq́ əḱʷəĺs [tə ˀən-sqʷəméý]?
habitually q be.biting art:ref.vis your-dog
‘Does your dog bite?’ Halkomelem (Suttles 2004: 471)

The following examples illustrate the use of the article to mark specific referents. Again, we see
that the deixis parameter of the article exponents is independent of the referential value. Thus,
the form tə, indicating visibility (perhaps due to the reported visibility), is used to express the
specific referent of čičíˀq́ ən ‘mink’ in (228). Another example of a specific referent is provided in
(230); the specific referent is marked by the form kʷθə of the article.

(228) s-niˀ-ct wə-ḱʷə́ c-nəxʷ [tə čičíˀq́ ən]

nmlz-aux-our pfv-see-tr art:ref.vis mink
‘Then we saw a mink.’ Halkomelem (Suttles 2004: 347)
(229) kʷə́ n-ət-əs [tə kʷíˀxʷ]
get-tr-s:3 art:ref.vis pitch
‘She took (some) pitch.’ Halkomelem (Suttles 2004: 348)
(230) wə-ˀəḿí-əs [kʷθə θí qʷłéý]
comp-come-sub:3 art:ref.nvis big log
‘…when a big log comes.’ Halkomelem (Suttles 2004: 348)

Suttles (2004: 345) notes that nonspecific referents are systematically encoded by the “remote”
form of the article ( ḱʷə). Examples of nonspecific referents are presented in (231) to (233). Note
that the article surfaces as shortened ḱʷ in (232). Example (233) shows that also nominalized
predicates, here nə-s-néḿ ‘my going’ can be referential and are marked by the article accordingly.

(231) nə-s-c-ƛ́íˀ [ḱʷə páy]

my-nmlz-do-value art:ref.rem pie
‘I want some (any kind of) pie.’ Halkomelem (Suttles 2004: 345)

158
(232) ˀə́ ý kʷ s-néḿ-ct sə́ wq́ -t [ḱʷ ṕə́ wəý] ˀə tə cáẃcəẃ
good art:ref.vis nmlz-go-our seek-tr art:ref.rem flounder obl art:ref.vis offing
‘We’d better go look for (any kind of) flounders offshore.’ Halkomelem (Suttles 2004: 345)

(233) nə-s-cƛ́íˀ [ḱʷə nə-s-néḿ]

my-nmlz-do-value art:ref.rem my-nmlz-go
‘I want to go.’ literally: ‘What I want is my hypothetical going.’
Halkomelem (Suttles 2004: 345)

Examples (234) to (237) illustrate the lack of the article with non-referring nouns. In (234) to (235),
the nouns are used as predicates and therefore have no referent. In example (237), the expression
təw-cítməxʷ ‘owl’ is used in a similative construction and is non-referring as well.
(234) wəłθíˀθə ḿə θəw ƛ́a
ritualist assert she
‘She is a ritualist.’ Halkomelem (Suttles 2004: 382)

(235) spéˀeθ ćə
bear qot
‘It is said / supposed to be a bear.’ Halkomelem (Suttles 2004: 374)
(236) wə-nét-əs ceˀ nə-sƛ́píwən
my-nom-own fut my-shirt
‘It will be my shirt. Halkomelem (Suttles 2004: 375)
(237) stém stém yəxʷ tθé təw-cítməxʷ waˀ
what what infer dem like-great-horned.owl presump
‘What can that be? It looks like an owl.’ Halkomelem (Suttles 2004: 390)

In contrast to nominal predicates, the article occurs with nouns in equational constructions. This
is shown in (238).

(238) ƛ́a ḿə [kʷθə nəhə́ ĺkʷ]

be.3 assert art:ref.nvis my-be.breaking
‘It is my pocketknife (that he has in his possession, as I noticed just now).’
Halkomelem (Suttles 2004: 382)

Halkomelem (as is commonly attested in Salishan languages (cf. Beck 2013)) has flexible word
classes, which means that many words can occur as nouns or as predicates. I do not discuss the
status of nouns and verbs in the grammar of Halkomelem here; for the purposes of this section, it
suffices to say that words which are typically referring can equally be used as predication and vice
versa. Given that there is no strict distinction between words used for reference or predication

159
in Halkomelem, we would also expect words that are typically predicates to occur as a referring
expression. Examples (239) to (241) show that this is what we find, and that the article occurs
with such expressions to mark them as being referential.

(239) nə́ wə yəxʷ ceˀ xáˀłəm-ət [tə hə́ ýqʷ]

be.you infer fut agree-tr art:ref.vis be.burning
‘You will (honour us to) be the one to look after the fire.’
Halkomelem (Suttles 2004: 376)
(240) stém kʷə [tə kʷəné-t-əxʷ]
what then art:ref.vis hold-tr-you
‘What is it you are holding?’ literally: ‘What is your holding?’
Halkomelem (Suttles 2004: 385)
(241) stém [ḱʷə ˀə-s-cƛ́íˀ]
what art:ref.rem you-nmlz-want
‘What do you want (e.g. of food laid out)?’ Halkomelem (Suttles 2004: 400)

To conclude, this section showed that the three markers tə, kʷθə, and ḱʷə can mark distinctions
of deixis and visibility, but it also showed that they systematically occur with different types of
referents, and that they are systematically absent in nonreferring contexts. Therefore, I treat them
as exponents of a single referential article.

5.10.3 Other examples

Referential articles are also attested in other Oceanic languages, for instance in Niuean (New
Zealand) or Tahitian (French Polynesia, Tahiti)29 . In this section, I briefly present the referential
article in Niuean and in Sabanê (Nambikwaran, Brazil).
The referential article in Niuean has two exponents as is shown in Table 5.11 (based on Gorrie
et al. (2010), Massam et al. (2006)). It occurs as e in the absolutive case, and as he in the ergative
(and also in the genitive and locative case). I addition, we find some prepositions that require the
absolutive form e while others combine with the form he.
29 Cf. Rigo & Vernaudon (2004).

160
 Table 5.11: Niuean article paradigm
absolutive ergative
e he
ma e ‘for’ ke he ‘to’
mo e ‘with’ (comit) mai he ‘from’
aki e ‘with’ (instr)

In example (242), the referent marked by the article is specific; the same referent, now being
anaphoric, is featured in (242b). In both sentences, the article e is used with the noun. That the
article also occurs with nonspecific referents is shown in (243). Although the referent of tau manu
could be specific, it is understood as nonspecific in (243) according to Massam et al. (2006: 11).

(242) a. ne kitia e ia [e taha feke lahi] …

pst see prop.erg 3sg art:ref.abs one octopus big
‘Near the reef edge she saw an octopus …’
b. …kua keli tuai e ia [e feke] ti
perf beat perf prop.erg 3sg art:ref.abs octopus then
‘…She beat the octopus and it died.’
Niuean (Gorrie et al. 2010: 361)
(243) kua fia manaho a lautolu ke [he tau manu]
perf desid want prop.abs 3pl to art:ref.loc pl bird
‘They wanted (any) birds.’ Niuean (Massam et al. 2006: 11)

Outside of Oceanic languages, another example of a referential article is found in Sabanê (Nam-
bikwaran, Brazil).30 Example (244) shows the definite referent of katataliʔ ‘man’ being marked
by the article -mi; the context of this sentence is the question about the best place for the man in
question (Antunes de Araujo 2004: 119).

(244) iawa katataliʔ-mi yeyi-al-i

cl:bark white.man-art:ref stay-pres-assert
‘The white man stays on my boat.’ Sabanê (Antunes de Araujo 2004: 119)

The referential article -mi is also used with specific and nonspecific referents. This is shown in
(245) and (246) for a specific and in (247) for a nonspecific referent.
(245) Manoel anose-mi t-osan-ntal-i
Manoel bowl-art:ref 1.o-give-pst-assert
‘Manoel gave me a bowl.’ Sabanê (Antunes de Araujo 2004: 176)
30 Antunes de Araujo (2004) notes that the related languages Sararé (cf. Borella 2004) and Latundê (cf. Telles 2002)
also feature a similar marker that may be treated as a referential article.

161
(246) deypa-mi pi-imi-tasa-datinan
snake-art:ref 1pl.o-bite-inch-pst:evid
‘A snake almost bit us.’ Sabanê (Antunes de Araujo 2004: 181)
(247) kapune-mata-mi palann-dana
paca-dim-art:ref not.have-pres-evid
‘She dose not have a small paca.’ Sabanê (Antunes de Araujo 2004: 92)

5.11 Summary
This chapter presented and discussed various examples of the nine article types found in the
world’s languages. These are: definite, anaphoric, recognitional, exclusive-specific, inclusive-
specific, nonspecific, indefinite, presentational, and referential articles.
For definite articles, I argued that their occurrence with absolutely unique and deictic definites
may be a common, but not a necessary property. The use of definite articles together with deictic
referents was shown to be dependent on the properties of the source element, a demonstrative in
most cases. This explains that many definite articles can be used to encode deictic referents, but
it also explains how some definite articles are not used in such contexts. In addition to definite
articles, I showed that two other types of articles have to be distinguished in the definite domain:
anaphoric and recognitional articles. Anaphoric articles only encode anaphoric referents; recog-
nitional articles mark referents that are definite based on previous experience shared between the
speaker and the hearer.
In the indefinite domain, we find exclusive-specific, nonspecific, indefinite, and presentational
articles. I argued that we need to distinguish between indefinite and what I called presentational
articles. These two article types do not differ from indefinite articles in their referential functions;
they are restricted to the encoding of discourse-prominent referents. For nonspecific articles, I
show for a number of languages that they originate from verbal irrealis markers that extend to
nominal predicates in irrealis contexts, then to nouns with nonspecific referents in irrealis con-
texts, and eventually these markers can become systematic nonspecific markers, i.e. nonspecific
articles.
In addition to articles within the definite and the indefinite domains, I showed that there
are two article types that mark referent types from either domains. These are inclusive-specific
articles, encoding definite and specific referents, and referential articles, marking definite, specific
and nonspecific referents.

162
Chapter 6

Morphological properties of articles

This chapter discusses the main morphological aspects of articles, focusing on their inflectional
properties and the paradigmatic relations between forms of different articles and article expo-
nents. In section 6.1, I provide a number of examples to illustrate and discuss the inflectional
patterns of articles, and show that we find a crosslinguistic trend against the inflection of articles.
With regard to number, gender, and case marking and presents, two factors are presented that
restrict their marking on the article: the expression of number and case on the noun as well as
the inflectional properties of the source elements of articles. Section 6.2 discusses two paradig-
matic relations within article systems: the form relation between nonspecific and specific articles
and the paradigm of the definite article in Mokpe. The latter consists of a segmental and a tonal
marker which show on several levels that morphological and syntactic properties of two article
exponents can be different and independent of their shared referential function.

6.1 Article inflection

6.1.1 Selected patterns
In the world’s languages, the categories that are marked on the article include number, gender,
case, and deixis. In this section, I present examples of articles inflect for different categories,
focusing on the marking of gender, case, and deixis (for number, cf. section 7.4.2). Since section
6.1.3.1 deals with the correlation of the inflection of the article with the inflection of the noun, the
discussion of article inflection here will focus on its interaction with nominal inflection as well.
Gender marking on the article is common across languages, in fact, gender is the nominal
category with the strongest trend to be marked on the article (cf. section 6.1.2). Before we can
address the question of gender marking on articles, a few preliminary remarks on gender are
required.

163
 I take a language to have gender if we can distinguish different groups of nouns based on
their behaviour with respect to agreement on other elements, e.g., verbs, adjectives, demonstra-
tives, quantifiers, numerals, or the article (cf. Corbett 2003, Corbett & Fedden 2016). I refer to these
different groups as classes of a gender system. I do not distinguish between different types of gen-
der systems; purely sex-based or animacy-based systems are included if they fulfill the agreement
criterion, as well as “noun class” systems and systems with nominal classifiers. I do not further
distinguish between these types of gender systems. This is because a more detailed and accurate
distinction is hardly possible without a lexical analysis for each language, which exceeds the pur-
poses of this study. In other words, I include languages with a distinction between masculine and
feminine “genders” for human referents (e.g. Dime, Konso, Wayuu, Asheninka Perené), “sortal
classifier” systems (e.g. Cubeo, Tepehua, Ch’ol, Rajbanshi), “noun class” systems similar to the
“Bantu type” (e.g. Runyankore, Bemba, Mokpe), and “gender” systems found in e.g. many Indo-
European and Australian languages (e.g. Domari, Palula, Guragone, Wardaman). The same holds
for animacy, treated as gender if animate and inanimate nouns show differences in agreement. In
the sample, only a small number of languages have animacy-based gender (Blackfoot, Ute, and
Biak).
Important is also the article as the target for gender marking: I count an article as inflecting
for gender if it has a different exponent for at least one of the gender classes in the language. Since
I base gender on agreement between different classes of nouns and other elements, I categorically
exclude gender marking on the noun. This is not to say that nouns do not contain information
on the gender class in their phonological form. A case in point is Spanish.1 For Spanish, we
can establish phonology-based rules for gender class assignment that cover a large portion of the
nouns in the language. The following two phonology-based rules can assign a gender class to the
majority of Spanish nouns:

(1) a. -a → feminine
b. -o → masculine

Even though this rule can be applied successfully in most cases, there are two reasons against
treating Spanish nouns as gender-marking, following the argumentation of Harris (1991). Firstly,
there are forms for which the rules predict the wrong gender class, e.g., the noun problema ‘prob-
lem (m)’ or the noun mano ‘hand (f)’. Secondly, there are many nouns in the language that do not
have either the final vowel -a/-o or any suffix that could automatically assign the gender class,
which means that the rules shown in (1) are not applicable to a number of nouns in the language.
Some examples are sol ‘sun (m)’, col ‘cabbage (f)’, tribu ‘tribe (f)’, espíritu ‘spirit (m)’.
1 For a similar discussion of gender marking on nouns in German, cf. Köpcke (e.g. 1982), Köpcke & Zubin (1984).

164
 A well-known example for an article that is inflected for gender is found in German. As
Table 6.1 shows, both the definite and the indefinite article have different exponents according to
the gender class of the noun that they occur with. However, the definite article does not inflect
for gender in the plural. Since this is systematic in that it can be captured as syncretism between
the three gender classes, it is uncontroversial to regard the articles in German as gender-marking.

Table 6.1: German article paradigm

singular plural
masculine feminine neuter
def indef def indef def indef def indef
nom der ein die eine das ein die /
acc den einen die eine das ein die /
dat dem einem der einer dem einem den /
gen des eines der einer des eines der /

A perhaps less intuitive example for gender marking on the articles comes from Ch’ol (Mayan,
Mexico). Ch’ol uses numeral classifiers for nouns that are used together with numerals. The
classifiers, occurring as suffixes on the numerals, are presented in Table 6.2.

Table 6.2: Ch’ol numeral classifiers

classifier semantic properties of the noun
-p’e(j) / (default classifier)
-tyikil person
-kojty animal, chili
-tsijty small, thin, large things
-k’ej flexible, square rectangle or rounded things
-sejl rounded things
-lejch flat things
-tyejk tree

The classifier -p’ej(j) is the default classifier used for the majority of the nouns in Ch’ol. The
indefinite article in Ch’ol originates from the numeral ‘one’; therefore, it obligatorily contains
one of these classifier suffixes according to the noun it occurs with. Based on the data provided in
Vázquez Alvarez (2011), the indefinite article does not occur without being marked by a classifier.
Since the choice of classifier depends on the noun, I regard the presence of the classifier on the
indefinite article as gender marking of the noun for the purposes of the present study. Examples
for the gender marking on the indefinite article are given in (2) to (5) below:

165
(2) ta’=bi i-tyaj-a [jum-p’e juñ]
pfv=qot 3sg-find-tr art:indef-cl:pej paper
‘It is said that he found a document.’ Ch’ol (Vázquez Alvarez 2011: 92)
(3) poj añ lonk-sa’ tyi [jum-p’e bolsaj]
hon exist 1pl.ex-pozol in art:indef-cl:pej plastic.bag
‘We have our pozol in a plastic bag.’ Ch’ol (Vázquez Alvarez 2011: 171)
(4) yi y-il-ä [juñ-tyikil x’ixik li wiñik]
pfv 3sg-see-dt art:indef-cl:tyikil woman
‘The man saw a woman.’ Ch’ol (Vázquez Alvarez 2011: 353)
(5) añ-bi jiñi [juñ-tyikil wiñik]
exist-qot hesit art:indef-cl:tyikil man
‘There, hm, was a man.’ Ch’ol (Vázquez Alvarez 2011: 245)

As was mentioned in section 5.6.1, the inclusive-specific article in Bemba is prefixed to the noun,
preceding the noun class marker that is traditionally regarded as indicator of number and gender.
Table 6.3 below shows the article exponents and the noun class markers again, rearranged into
the gender classes A to H split up into the singular and plural.

Table 6.3: Bemba article and noun class prefix

gender class article noun class prefix
singular plural singular plural
A u- a- -mu- -ba-
B u- i- -mu- -mi-
C (i-) a- -li- -ma-
D i- i- -ci- -fi-
E i- -N-
F u- a- -lu- -ka-
G u- u- -tu- -bu-
H u- -ku-

Comparing the noun class markers and the article exponents, we see that the article exponent
always corresponds a copy of the vowel in the noun class marker. Therefore, the article exponent
may rather be a result of a phonological process which is why I do not treat the inclusive-specific
article in Bemba as gender-marking.
Another interesting case with respect to gender marking on the article and the noun is Supyire
(Gur, Mali). Supyire has a definite article and, according to Carlson (1994), it has the five gender
classes 1-5, shown in Table 6.4. Gender as a category of the noun in Supyire is manifested by
agreement on adjectives and other elements in the noun phrase (Carlson 1994: 75). Nouns of

166
class 1-3 have singular and plural forms, whereas the classes 4 and 5 contain mass nouns which
do not exhibit a number distinction. Carlson (1994: 77) presents two paradigms for the gender
marking of nouns in the language. One corresponds to what he calls “basic noun gender suffixes”,
the other paradigm contains the “definite noun gender suffixes”. I treat the first forms as the
forms of the bare noun, while the other forms correspond to the combination of the noun and the
definite article. The abstract exponents for the definite article in Table 6.4 suggest that the article
inflects for gender, given nouns of the different gender classes are shown to condition a different
exponent of the article.

Table 6.4: Gender marking on Supyire nouns and articles

gender nominal marker definite article
singular plural singular plural
1 -wV -(bi)li -ŋi -pili
2 -gV -yV -ke -yi
3 -lV -gili -ni -kili
4 -rV -te
5 -mV/-bV -pe

Examples of bare nouns and nouns with the definite article are presented in Table 6.5 (based on
Carlson (1994: 79-94)). While the correspondence between the gender class and the nominal suffix
is irregular, the article exponent is consistently conditioned by the gender class of the noun. The
variation in the article exponents within classes that we see for the singular of class 2 and for
class 4 (as well as the tonal effects) can be accounted for by regular phonological processes in the
language. Therefore, I regard the definite article in Supyire as inflecting for gender. The Supyire
patterns contrasts with the one in Bemba in an interesting way; it shows that the ability of the
article to inflect does not depend on how close it is to the noun.

167
 Table 6.5: Examples of gender marking on Supyire articles
gender noun noun-article
1 sg foo foòŋi ‘(the) owner’
nu nùŋi ‘(the) mother’
sò sòŋi ‘(the) duiker’
pl zòmii zòmpíí ‘(the) hearts’
dufáa dufáabíí ‘(the) pockets’
pwùnn pwùunbíí ‘(the) dogs’
2 sg ŋ̀ kéŋɛ̀ ŋ̀ kêŋke ‘(the) branch’
canŋa cànŋke ‘(the) day’
baga bage ‘(the) house’
kafɛɛgɛ kafɛɛ̀ ge ‘(the) wind’
bàhà bàhe ‘(the) poison’
lwɔhɔ lwɔhé ‘(the) water’
pl ciye ciyí ‘(the) trees’
kùùyò kùùyi ‘(the) stools’
canya canyi ‘(the) days’
3 sg sàhàlà sàhàni ‘(the) basket’
cere cení ‘(the) calabash’
fwuu fwuùni ‘(the) yam’
pl sháháŋii sháháŋkíí ‘(the) palm nuts’
jàhii jàhigíí ‘(the) beana’
tugugii tugugíí ‘(the) loads’
4 suro sure ‘(the) mush’
shire shiré ‘(the) feathers’
sɛrɛ sɛɛré ‘(the) honey’
nana nante ‘(the) maleness’
kòònɔ̀ kòònte ‘(the) discussion’
tɔɔnnɔ tɔɔ̀ nnte ‘(the) metal’
5 sìnŋkanma sìnŋkanmpé ‘(the) sourcery’
bɛ̀ ɛ̀nmɛ̀ bɛ̀ ɛ̀nmpe ‘(the) light’
suumɔ suùmpe ‘(the) salt’

Animacy is also regarded as gender if it is reflected in agreement for the purposes of the present
study. However, animacy of the noun is rarely marked on the article in the world’s languages
and is attested only three times in my sample (in Blackfoot, Ute, and Siar Lak). One example for
articles with different exponents marking the animacy of the referent comes from Ute (Numic,
USA). Animacy can be regarded as a gender category in Ute, since it is reflected in the agreement
on e.g. demonstratives and adjectives (Givón 2011: 50,60). The Ute article system consists of

168
a definite article which distinguishes between animate and inanimate referents in addition to
number and case (subject vs. non-subject) of the noun. The paradigm, based on Givón (2011:
163), is given in Table 6.6:

Table 6.6: Ute definite article paradigm

animate inanimate
subject non-subject subject non-subject
singular ’u ’uway
’uru ’uru
plural ’umʉ ’umʉ

Examples to illustrate the animacy distinctions are given in (6) to (8) below, contrasting the ani-
mate (6) and inanimate (7) singular subject markers, and the animate (8) and inanimate (6) object
markers.

(6) kh-’ura [’áapachi ’u] [tʉkuavi ’uru] tʉka-pʉga

then-be boy.an art:def.an meat.inan art:def.inan eat-rempst
‘then the boy ate the meat’ Ute (Givón 2011: 163)
(7) [tʉka’napʉ ’uru] yaqh-kya
table.inan art:def.inan break-ant
‘the table broke’ Ute (Givón 2011: 164)
(8) [tuachi ’uway] [tʉkuavi ’uru] magha-pʉga
child.an art:def.an meat.inan art:def.inan feed-rempst
‘she fed the meat to the child’ Ute (Givón 2011: 164)

Case marking on articles and other determiners may not be unexpected from the Indo-European
perspective. However, in the world’s languages, case marking on articles is relatively rare and it is
more much more often marked on the noun only. As for case marking, I only consider core cases in
order to keep case marking comparable across languages. I take a language to have case marking
if it systematically marks nouns as core arguments of intransitive and transitive constructions.
Thus, if at least one marker consistently occurs with a noun to mark it as the sole S argument
of an intransitive clause, and/or as the A and/or P argument of a transitive clause, I treat it as a
case marker. What may be called case marking but what I do not treat as case marking within
the purposes of the present study is the marking of nouns as oblique arguments, instruments,
benefactives, recipients, genitives/possessives, or spatial relations, etc. (e.g. in Guragone, Irish,
Ao).
One example for core case marking on the article alone is German. Even though German
nouns are traditionally presented in case-number paradigms, the form of nouns generally does

169
not distinguish between its function as S/A argument (nominative) and as P argument (accusative).
Table 6.7 shows that what is partially distinguished are the genitive and dative forms of the noun,
but not the nominative and accusative forms.

Table 6.7: Case marking on nouns and the definite article in German

masculine feminine neuter

singular plural singular plural singular plural
nom der Baum die Bäume die Wiese die Wiesen das Haus die Häuser
acc den Baum die Bäume die Wiese die Wiesen das Haus die Häuser
dat dem Baum den Bäumen der Wiese den Wiesen dem Haus den Häusern
gen des Baums der Bäume der Wiese der Wiesen des Hauses der Häuser

The only marker on the noun that qualifies as core case marker in German is -en, which marks
accusative singular nouns of the weak declension type (Hentschel & Weydt 2003: 139). While it
is still obligatorily used with some nouns, e.g. den Student-en ‘the.acc student-acc’, other nouns
that fall into this inflection class can occur without the marker, e.g. den Bär(-en) ‘the.acc bear-
(acc)’. Neither does the article distinguish case in all relevant contexts. As was shown in Table 6.1,
the exponents of both the definite and the indefinite article are also highly syncretic. Only their
masculine singular forms formally distinguish between the nominative and accusative case. Nev-
ertheless, this distinction is systematic so that I consider both articles in German as case-marking.
Another example of case being marked on the article but not on the noun comes from Albanian.
Usually, Albanian nouns are arranged into a definite and an indefinite paradigm which are marked
for number and case. The so-called definite forms of the noun correspond to the combination of
the noun and the definite article which occurs as a suffix on the noun. On the other hand, the so-
called indefinite forms rather correspond to the default or citation form of the nouns. The definite
article is suffixed to the noun and is fused with the case marker. Table 6.8 below illustrates the
case markers for four different inflection classes, largely corresponding to masculine, feminine,
neuter, and a plural nouns (based on Newmark et al. (1982: 159-162) and Buchholz & Fiedler (1987:
269-271)).2
2 Buchholz& Fiedler (1987: 269) distinguishes between 5 inflection classes; the case syncretisms for singular and
plural nouns are essentially the same in the other classes not presented here.

170
 Table 6.8: Albanian case paradigm
masculine feminine neuter plural
mik ‘friend’ dele ‘sheep’ ballë ‘forehead’ mure ‘walls’
noun noun-art noun noun-art noun noun-art noun noun-art
nom mik-u del-ja
mik dele ballë ball-ët mure mure-t
acc mik-un dele-n
dat/abl miku mik-ut deleje dele-s balli ball-it mureve
gen i miku i mik-ut i deleje i dele-s i balli i ball-it i mureve

As we see in Table 6.8, it is only the definite article with singular nouns that distinguishes between
nominative and accusative forms. Nouns that occur without the definite article do not formally
distinguish these two core cases across the four inflection classes. Therefore, I only treat the
definite article but not the noun as case-marking in Albanian.
As I will show in section 6.1.2, articles that mark case are not common across the languages of
the world, and in the sample, paradigms like the one in German or Albanian are in fact confined
to a handful of languages in Eurasia (German, Albanian, Basque), Australia (Warrwa, Arrernte),
and North America (Ute).
A language that marks core cases both on the article and on the noun is Diyari (Central Pama-
Nyungan, Australia). In addition, the definite article in Diyari shows how different case syn-
cretisms in nouns and articles can come about: the definite article originated from third person
pronouns that distinguish between the nominative, accusative, and ergative, while the majority
of nouns only distinguish between nominative and ergative cases. Table 6.9 shows the core case
forms of the definite article in Diyari (Austin 2011: 68).

Table 6.9: Diyari definite article paradigm (core cases)

singular singular
dual plural
feminine nonfeminine
ergative nhandru nhulu pulali thanali
nominative nhani nhawu pula thana
accusative nhanha nhinha pulanha thananha

We can see in examples (9), (10), and (11) that both the article and the noun are inflected for case.

(9) [thana wilha] wima kirli-yi

art:def.f.pl.nom woman.nom corroboree.nom dance-pres
‘The women are dancing a corroboree.’ Diyari (Austin 2011: 123)

171
(10) nhulu [pulanha mankarra-wurla-nha] kuthi-ipa-yi
3sg.nf.erg art:def.du.acc girl-du-acc hide-caus-pres
‘He hid the two girls.’ Diyari (Austin 2011: 81)
(11) waldrawirti [nhulu kinhala-li] tyukurru nandra-lha wirri-yi
yesterday art:def.sg.nf.erg dog-erg kangaroo.acc hit-fut aux-pres
‘Yesterday, the dog killed a kangaroo.’ Diyari (Austin 2011: 98)

The two examples in (12) and (13) show the combination of an article with a nonfeminine noun
that does not distinguish between nominative and accusative, while the article does.

(12) [nhawu kupa] muka thurara-yi

art:def.nf.nom child.abs sleep.abs lie-pres
‘The child is sleeping.’ Diyari (Austin 2011: 91)
(13) ngathu [nhinha kupa] wapa-lka-yi
1sg.erg art:def.sg.nf.acc child.abs go-appl-pres
‘I take the child for a walk.’ Diyari (Austin 2011: 80)

Core case marking on both the article and the noun being rare in the world’s languages, the
pattern in Diyari with its definite article that formally corresponds to the third person pronouns
and retained its inflectional behaviour is an illustrative example of how such systems can develop,
and it also shows that the case morphology of the article can remain independent of the case value
of the noun.
Articles do not only inflect for nominal categories; it is not uncommon to find articles that
mark some form of deixis, mostly in those cases in which the source element already marked deixis
and the article retained the existing formal opposition. An example of a definite article that can
mark deixis and that developed from a set of demonstrative markers can be found in Macedonian.
Table 6.10 shows the article exponents with the demonstrative markers in brackets.3 The distance-
neutral markers from the first row correspond to the default article exponents which are the least
marked for deixis and also the most frequent forms in language use (Kramer & Mitkovska 2011:
212).

Table 6.10: Macedonian definite article paradigm

deixis singular plural
masculine feminine neuter
neutral ot (toj) ta (taa) to (toa) te (tie)
proximal ov (ovoj) va (ovaa) vo (ova) ve (ovie)
distal on (onoj) na (onaa) no (ona) ne (onie)
3 Based on Lunt (1952: 41), Kramer & Mitkovska (2011: 211f).

172
This relation between three (neutral, proximal, distal) sets, the rows in Table 6.10, of article ex-
ponents that are not base on the “usual” gender, number, and case distinctions relates to a more
general important issue. I treat these forms as exponents of a single definite article in Macedonian,
additionally marking deixis. An alternative view would be to treat these three sets them as three
definite articles based on their deixis value, each being additionally marked for gender and num-
ber. However, since marking distance does not affect their referential function, in cases like the
Macedonian definite article, deixis should be treated as number, gender, or case, corresponding
to a category that is additionally marked on the article (cf. section 6.2.1).
The article exponents from the first row in Table 6.10 formally correspond to the medial demon-
strative markers and represent the default definite article exponents. They can generally be used
in all contexts that require definite articles. The proximal and distal forms are less frequent and
mainly used in certain deictic contexts. However, they are not restricted to deictic functions and
are conditioned by register, style, and creative language usage according to my informant. Exam-
ples (14) and (15) illustrate the use of the proximal and distal forms of the definite article. As (14b)
and (15b) show, the deixis neutral form of the article can equally be used in both cases.

(14) a. ima restoran vo selo-vo?

have.3sg restaurant in village-art:def.prox
‘Is there a restaurant in the / this village?’
b. ima restoran vo selo-to?
have.3sg restaurant in village-art:def
‘Is there a restaurant in the village?’
Macedonian (prim. data)
(15) a. se sekavaš na ideja-na što ja imav?
refl remember.pst.2sg on idea-art:def.dist comp I have.pst.1sg
‘Do you remember the / that idea that I had?
b. se sekavaš na ideja-ta što ja imav?
refl remember.pst.2sg on idea-art:def comp I have.pst.1sg
‘Do you remember the idea that I had?
Macedonian (prim. data)

Other cases of articles marked for deixis are the referential article in Halkomelem (cf. section 5.10.2)
and the anaphoric articles in Koasati and Bare (cf. 5.3.2). For the anaphoric articles, marking
deixis corresponds to marking more or less time between the mentioning of the antecedent and
the anaphoric referent. This was illustrated for Bare in examples (59) and (60), which are repeated
below.
(16) tʃiñu yawi mehésa me-wát’uka [da tʃiñu] i-báraka
dog angry 3pl.want 3pl-beat art:ana.prox dog 3sg.nf-run
‘The dog was angry, they wanted to beat the dog, it ran away.’ Bare (Aikhenvald 1995: 24)

173
(17) nu-yadã nu-tʃána-ka [asa tʃiñu]
1sg-see 1sg-stay-decl art:ana.dist dog
‘I am still seeing that dog.’
Bare (Aikhenvald 1995: 24)

6.1.2 Crosslinguistic trends

As was already mentioned in the previous section, articles inflect most frequently for gender, and
less frequently for number and case. With respect to inflection in general, this section shows that
there is no trend towards inflection for articles in the worlds languages.
What I treat as article inflection here is the systematic marking of number, gender, and case of
the noun that the article occurs with. Gender and case marking is understood here as was defined
in the previous section. I treat number marking accordingly. Number has to be systematically
marked on the noun or the article in order to be counted as number marking. This includes
systematic number marking for a subset of noun, e.g. only on nouns with human or animate
referents (e.g. Oko, Tongan, Arawak, Wubuy). On the other hand, if number marking is described
as “optional” in the nominal domain without further indication of the conditions for number
(mostly plural) marking, I do not treat articles or nouns as number-marking (e.g. Siar Lak, Skou,
Mangarrayi, Jingulu).
Based on the sample of 99 languages with 131 articles in total, Figure 6.1 shows the number
of articles that inflect (infl) and that do not inflect (no.infl), split up into the six macroareas
as distinguished in Dryer & Haspelmath (2013), Hammarström et al. (2018).4 Although there
are minor areal differences in the proportions of inflecting and not inflecting articles, the overall
trend shows more articles without inflection than articles that inflect. The sample of 131 articles
contains 49 inflecting articles and 82 non-inflecting articles; a chi-square test5 confirms that we
can expect this bias against inflection to hold for articles in the languages of the world. 6
4 For details on the properties of single languages and articles see Tables A.2 to A.7 in the appendix.
5 The chi-square test was performed in R (R Core Team 2016), with a p-value = 0.005007.
6 In order to exclude a bias by the languages that have more than one article, I repeatedly selected a random sample

with one article per language; the results did not differ from the sample with 131 articles.

174
 20

art.infl

count
infl
no.infl

10

africa eurasia papunesia australia n.america s.america

Figure 6.1: Article inflection in the six macroareas

Regarding any potential areal biases, a Bayes test of proportion7 confirms that the proportions
between inflecting and non-inflecting articles do not differ statistically across the six macroar-
eas. The Bayes test of proportion gives estimated confidence intervals for the proportions tested.
The proportion values of the whole population from which my sample was drawn (the languages
from the six macroareas) are estimated to lie within these intervals with 95% confidence. If the
intervals overlap, we can take this as a confirmation that there is no statistically significant differ-
ence between the proportions compared; in the case of the different macroareas, the confidence
intervals for the proportions of inflecting and non-inflecting articles all overlap. The numbers for
Figure 6.1 as well as the proportions and confidence intervals are given in Tables A.8 and A.9 in
Appendix A.
Figure 6.2 shows the categories of inflection for the articles in the sample that inflect, again
across the six macroareas. We can see the following areal trends: while articles inflect for number
in all areas, case marking is absent in South America, Papunesia, and Africa. This is a consequence
of case itself being less frequent in these areas. Articles that only mark number or gender are
found in all areas, whereas case marking without inflection for another category is restricted to
Australia.
7 The Bayes test of proportions was performed in R (R Core Team 2016) using the “Bayesian First Aid” package (Bååth

2014). This also applies to the other proportion tests referred to in this section.

175
 12.5

10.0

art.infl.cat
case
7.5
gen
gen+case
count

num

5.0 num+case
num+gen
num+gen+case

2.5

0.0

africa eurasia papunesia australia n.america s.america

Figure 6.2: Categories of article inflection in the six macroareas

As for other factors that would influence the inflection of articles, it can be shown that neither
the position of the article nor the article type seem to have an impact. As is illustrated in Fig-
ure 6.3 below, the position of the article with respect to the noun does not seem to play a role for
the inflectional behaviour of the former. Preceding (pre) and following (post) articles equally
show the overall trend against inflection. A Bayes proportion test confirms that there is no sta-
tistically significant difference between pre- and postposed articles with respect to inflection (cf.
Tables A.10 and A.11 in Appendix A).8

60

art.infl
40
count

infl
no.infl
20

0
post pre

Figure 6.3: Inflectional behaviour of following and preceding articles

8 Theindefinite article of Sri Lanka Malay is not included here, as it appears to be flexible in preceding or following
the noun.

176
In Figure 6.4, we see the proportion of inflecting and non-inflecting articles with respect to the
different article types. Again, there are minor differences; for instance, the indefinite article type
appears to contain a higher proportion of inflected articles than the definite or presentational type.
However, a Bayes proportion test cannot confirm these differences (cf. Tables A.12 and A.13 in
Appendix A).

40

30
art.infl
count

infl

20 no.infl

10

def ana recog exspec nspec indef pres inspec ref

Figure 6.4: Inflectional behaviour of different article types

Thus, so far, we saw that the availability of article inflection is neither influenced by the macroarea,
the position or the type of the article. Considering the categories that different article types inflect
for, we see certain biases. This reflects the correlation between article types, their geographical
distribution, and the nominal categories that are available more or less frequently in certain areas.
Figure 6.5 shows the categories that the 9 article types inflect for. While there is no clear trend
with respect to gender and number marking, we see that case marking (or case together with
number) occurs mainly with definite and anaphoric articles from the definite domain and less
so with articles from the indefinite domain. This reflects the distribution of case as a nominal
category in predominantly Eurasia and Australia. Eurasia mainly contributes to case marking on
definite (and less so indefinite) articles, whereas anaphoric articles are very prominent in Australia
which can explain their high proportion of case marking.

177
 20

15
art.infl.cat
case
gen
gen+case
count

10
num
num+case
num+gen
num+gen+case

def ana recog exspec nspec indef pres inspec ref

Figure 6.5: Inflectional categories of different article types

As was mentioned earlier in this section, 27 languages from the sample have more than one arti-
cle and it seems plausible that different articles within single languages would tend to be similar
regarding their inflectional behaviour. Indeed, in 16 cases, different articles within a single lan-
guage inflect for the same categories (or do not inflect). However, the articles from the other 11
languages do not behave alike with respect to inflection. This shows that there may be a weak
trend towards encoding the same categories on different articles if the language has more than
one, but it certainly is not a strong tendency in the languages of the world.9
Ch’ol is an example of a language with both a definite and an indefinite article which have
different inflectional properties. As was shown in the preceding section, the indefinite article
marks gender. The definite article, on the other hand, does not inflect. Examples (18) and (19)
show this contrast.

(18) a. tyi j-k’ajty-i-b-e [jum-p’e baso]

pfv s:1sg-ask-tr-appl-tr art:indef-cl:p’e glass
‘I ask him for a glass.’
b. tyi j-k’ajty-i-b-e [li baso]
pfv s:1sg-ask-tr-appl-tr art:def glass
‘I ask him for the glass.’
Ch’ol (Vázquez Alvarez 2011: 247)
9 Table A.14 in Appendix A shows the languages with multiple articles in the sample.

178
(19) a. ya’ añ [juñ-tyikil wiñik]
there exist art:indef-cl:tyikil man
‘There is a man.’
b. ts-äch=bi majl-i [li wiñiki]
pfv-aff=qot go-itr.ipfv-o:3 art:def man
‘It is said that the man went’
Ch’ol (Vázquez Alvarez 2011: 247)

To conclude, even though minor areal tendencies of the inflectional behaviour of articles could
be observed, these tendencies cannot explain why articles inflect or do not inflect. I discuss two
factors that have and impact on the inflection of articles in the next section.

6.1.3 Factors that condition the article inflection

6.1.3.1 Inflection of the noun

In this section, I will discuss the inflectional patterns of articles with respect to the inflectional mor-
phology of the noun. As the previous section has shown, article inflection seems to be relatively
independent of the geographical area, the article type. This section shows that the morphological
marking of number and case on nouns is one of the factors that influence the distribution of ar-
ticle inflection that we can observe in the world’s languages. As was mentioned in the previous
sections, articles are regarded as marking number, gender, and case if they do so systematically.
The same applies to nouns, only that nouns are excluded as gender markers. For the purposes of
the present section, I further restrict the marking of number and case on the noun to the contexts
in which it occurs together with the article.
Figure 6.6 presents the distributions for number and case marking on the article, noun, and
on both elements. The corresponding numbers are indicated in Table 6.11. Note that the numbers
correspond to the distributions in the reduced sample of one article per language, containing 99
languages and 99 articles. What Figure 6.6 clearly shows is that both number and case show very
similar distributions: they are predominantly marked on the noun only, while their marking on
the article and on both the article and the noun is disfavored.10
10 For details on the inflectional properties of the articles in the sample, see Tables A.2 to A.7 in Appendix A.

179
 number case

40

30

marked
art

count 20 art+noun
noun

10

art art+noun noun art art+noun noun

Figure 6.6: Number and case marking on the article and the noun

Table 6.11: Numbers for article and noun inflection

number case gender
as category 89 37 44
on article 11 6 22
on article and noun 16 3 /
on noun 43 27 /
on article : as category 11 : 89 6 : 37 22 : 44
proportion 0.12 0.16 0.50
on noun : total 43 : 70 27 : 36 /
proportion 0.61 0.75 /

In addition to to the numbers of Figure 6.6, Table 6.11 shows two other important things. First,
one is the indication of the number of languages that have the categories of number, case, and
gender; this is marked in the first row (“as category”). Second, two proportions are given in the
two last rows. The proportion of “on art : as category” shows the number of articles that mark one
of the three categories with respect to the number of languages that have these categories. The
numbers confirm the expected split between number and case on the one hand, and gender on the
other: crosslinguistically, articles mark gender considerably more frequently than number or case.
A Bayes proportion test confirms that this difference between number and case on the one hand
and gender on the other is statistically significant: the confidence intervals of the proportions

180
(number and case versus gender) do not overlap. The distribution of number and case marking,
however, is similar, also according to the results of the Bayes proportion test (cf. Table A.15 in
Appendix A).
As Figure 6.6 showed, the bias against number and case marking obviously correlates with
the high number of languages in the sample that mark these categories on the noun. Looking at
the proportions of “noun : total” in Table 6.11, we see for number and case that the proportions
of marking on the noun are similar, i.e. the crosslinguistic trend of marking number and case
on the noun versus on both the noun and the article and on the article alone is comparable. The
Bayes test of proportions confirms this similarity (cf. Table A.16 in Appendix A). Given that it
was shown for nouns to equally mark mark number and case, and and given that nouns can
generally occur without articles but not vice-versa, suggests that this is not only a correlation,
but that the marking on the noun influences the marking on the article on this crosslinguistic
level. Since the effect is so similar for both number and case, even though more languages in
the sample mark number than case, it suggests that this influence on the inflection of articles is
a general, category-independent effect. Thus, we can conclude that one factor that explains why
articles generally mark gender much more frequently than number and case is the crosslinguistic
tendency to avoid the marking of these categories on a second element in the noun phrase.
What is most surprising with regard to this effect is that the trend is so strong, given that
neither articles nor nouns necessarily mark the three categories separately, but often combine the
expression of more than one category. As was shown in Figure 6.2, gender marking often occurs
together with number marking, and if case marking occurs on the article, it is almost exclusively
marked in combination with gender and/or number (except for Australia). That the trends are
so robust is another indication for a category-independent trend against repeated marking in the
noun phrase.

6.1.3.2 Inflection of the article’s source element

The other factor that can be shown to influence the inflectional behaviour of articles is the inflec-
tional behaviour of their source element, i.e. the marker that the article has developed from. In
most cases, this is a demonstrative or pronoun (possessive or personal) for articles from the defi-
nite domain (cf. De Mulder & Carlier 2011), and the numeral ‘one’ for articles from the indefinite
domain (cf. Heine 1997).
Three scenarios are possible with respect to the inflection of articles and their source elements:
the article can either retain the inflectional morphology of its source element, simplify it which
results in the loss of inflectional morphology or in the loss of form oppositions; or the article can
10 Fornumber and case, “total” refers to the sum of their marking on the article, on the article and the noun, and on
the noun in the 99-articles sampe, e.g., for number: 11 (on article) + 16 (on article and noun) + 43 (on noun) = 70.

181
develop new inflectional morphology that is not present on its source element. In this section,
I present evidence for the first two scenarios (retaining and losing inflectional morphology) and
show that there is almost no evidence for the emergence of inflectional morphology on the article.
Since diachronic evidence is not available for most of the languages in the sample, I base this
section on the inflectional patterns of the article and its source element in the current stage of
the languages. This is turn is based on the assumption that inflectional morphology is mostly
retained on the source elements. This assumption has two consequences for the results of a syn-
cronic comparison of the inflection on the article and on its source element: on the one hand, the
source element could lose inflection that is retained on the article which I cannot distinguish from
innovative inflection on the article. Since this is attested in only two cases for which I offer a sep-
arate explanation in section 7.4.2, I do not regard potential inflection loss on the source element
without its loss on the article as problematic. On the other hand, I cannot distinguish between the
loss of inflection on the article and the innovation of inflection on its source element. However,
the scenario of inflectional innovations on the source elements (demonstrative, adnominal pos-
sessive and numeral ‘one’) without innovations on the article is not very likely for the following
reason: demonstratives, possessives, or the numeral ‘one’ occur less frequently than their article
counterparts; it is not likely that the grammaticalization leading to new inflectional morphology
would only affect these three elements without affecting the more frequent article.
Tables 6.12 to 6.14 show the inflectional behaviour of articles and different source elements
for those articles in the sample that can be related to their source elements. Table 6.12 lists articles
that originate from personal and possessive pronouns. The third column shows the categories that
the article inflects for, and in the last column we see the inflection of the pronoun in the current
language. Accordingly, Table 6.13 shows this for definite articles that originate from demonstra-
tives, and Table 6.14 for articles from the indefinite domain that developed from the numeral
‘one’.11 In all three cases, we mainly see evidence for two of the three scenarios mentioned above:
articles either keep the inflectional properties of their source elements or they simplify the sys-
tem. Crucially, there are only two examples of a morphological innovation in my sample (the
presentational articles in Basque and Logba).
11 Theforms provided for articles and their source elements in Tables 6.12 to 6.14 do not always correspond to an
exhaustive presentation of all exponents, but serve as an illustration for the similarity between the two elements
or a subset of them.

182
Table 6.12: Inflection of articles and their source elements (personal/possessive pronouns)
language article art inflection pronoun pro inflection
Oko àye̩ (sg), num àye̩ (3sg), num
(def) ábe̩ (pl) ábe̩ (3pl)
Lango ɛ́n / ɛ́n (3sg) num
(ana)
Amharic -w/u (m.sg, pl) num+gen -w/u (poss:m.sg) num+gen
(def) -wa (f.sg) -wa (f.sg)
Pap. Malay de (3sg) num de (3sg) num
(def)
Biak -ya/-i (3sg) num+gen i num+gen
(exspec) (cf. Table 6.15)
Warrwa kinya case kinya (3sg) num+case
(ana)
Diyari nhawu (3nf.sg) num+gen+case nhawu (3nf.sg) num+gen+case
(def) (cf. Table 6.9) (cf. Table 6.9)
Sumu -ni (3sg) / -ni (poss:3sg) num
(ana)
Cupeño pe’ (sg) num pe’ (3sg) num
(def) pem (pl) pem (3pl)
Apinayé ja / ja (3sg) num
(def)

183
 Table 6.13: Inflection of articles and their source elements (demonstratives)
language article art inflection demonstrative dem inflection
Supyire -ke (cl2.sg) num+gen -ŋ̀gé (cl2.sg) num+gen
(def) -pe (cl5) -m̀pé (cl5)
(cf. Table 6.4)
Konso -siʔ (m.sg) num+gen -oosiʔ (m.sg) num+gen
(def) -siniʔ (m.pl) -oosiniʔ (m.pl)
Sheko -ns/nʃ (m.sg) gen yīz (m.sg) num+gen
(def) -in (f.sg) yīnì (f.sg)
Hungarian a / az (nom.sg) num+case
(def)
Armenian -n / -ayn (sg) num+case
(def)
Maori te (sg) num tēnā (sg) num
(def) ngā (pl) ēnā (pl)
Alamblak ɨnd num+gen ɨndar num+gen
(def)
Ch’ol li / ili /
(def)
Ute def ‘u (s.an.sg) num+gen+case ‘uwa (s.an.sg) num+gen+case
(def) ‘umu (s.an.pl) ‘umu (s.an.pl)
(cf. Table 6.6)
Tarahumara echi / echi /
(def)
Arawak li (m.sg), to (nm.sg), num+gen lihi (m.sg), toho (nm.sg), num+gen
(def) na (pl) naha (pl)

184
 Table 6.14: Inflection of articles and their source elements (numeral ‘one’)
language article art inflection numeral num inflection
Logba ɔkpiɛ (cl7.sg) num+gen ikpɛ gen
(pres) ikpiɛ (cl8.pl)
Konso takkain (f) gen takka gen
(pres)
Hungarian indef egy / egy /
(indef)
Albanian një / një /
(indef)
Armenian mi / mek /
(indef)
German ein gen+case eins gen+case
(indef)
Basque pres bat (nom.sg) num+case bat case
(pres) batzuk (nom.pl)
Barwar Aramaic xa (m) gen xa (m) gen
(pres) ða (f) ða (f)
Kharia moɲ / muɖ u (hum) gen
(pres) moɲ (nhum)
Tamil oru / onru /
(indef)
Dameli ek / ek /
(pres)
SL Malay hatthu / satthu /
(indef)
Agul sa / sa /
(pres)
Palula áak/áa / áak/áa /
(exspec)
Sye hai / hai(ten) /
(indef)
Tepehua tam / laqa-tam (cl:gen-one) gen
(indef) puma-tam (cl:hum-one)

185
The data in Tables 6.12 to 6.14 show that the inflectional potential of an article is generally con-
strained by the inflectional morphology of its source element, independently from the type of
source element. The only changes we see correspond to the loss of formal oppositions leading to
the loss of inflection on the article. Thus, the categories that articles inflect for are restricted by
the properties of their source elements to begin with, since only simplification but no innovation
could be observed.

6.2 Article paradigms

This section is concerned with article paradigms. I discuss two types of paradigmatic relations
concerning articles: the first part deals with the form relation of different articles within languages,
and the second part presents the paradigm of the definite article in Mokpe (Bantu, Cameroon). The
latter is an example of a complex article paradigm, showing how different exponents of a single
article can interact.

6.2.1 One article or many?

In the world’s languages, we find article system that have more than one article, with one of
the exponents being the morphological basis for the exponent of the other article. This only
is attested in two combinations of article types: exclusive-specific and definite articles, as well
exclusive-specific and nonspecific articles.
Biak has an exclusive-specific article and a definite article (and also a nonspecific article which
is not relevant here). The specific article has the exponents as shown in Table 6.15 and attaches
to the last element of the noun phrase which is the noun by default (van den Heuvel 2006: 204).

Table 6.15: Biak specific article exponents

article exponent
singular -ya/-i12
dual -su-ya/i
paucal -sko-ya/i
plural -s-ya/i

The exponents of the exclusive-specific article are also used to form the definite article. The latter
formally consists of the marker an and the specific article that attaches to an. In example (20), we
12 van den Heuvel (2006: 68) notes that “-ya can be used in all positions, whereas the use of -i is restricted to prepausal
position”.

186
see the specific article ya attached to a noun and an adjective. In (21), the specific article ya forms
the definite article anya together with the marker an.

(20) ya-fáru man-koko=ya kukru ikák ve=ba=ya

1sg-tell bird-chicken=art:spec with snake rel-big-art:spec
‘I tell about a bird and a big snake.’ Biak (van den Heuvel 2006: 60)
(21) indya fyanu [rofan anya] i-fnovku mankroder=i
so feed.3sg dog art:def 3sg-with frog=art:spec
‘So he took care of the dog, together with a frog.’ Biak (van den Heuvel 2006: 205)

The other cases of article exponents forming the base of other articles are the exclusive-specific
articles in Q’anjobal (cf. section 5.7.1), Lakota (cf. section 5.7.3), and Crow (cf. 2.2.4), whose forms
are repeated in Table 6.16 together with the forms of the nonspecific articles. For Q’anjobal, it
was shown in section 5.7.4 that the nonspecific article originates from an irrealis marker that
combined with the former indefinite article jun and conventionlized into a nonspecific article.
Whether a similar scenario is also responsible for the formal relation of the exclusive-specific and
the nonspecific articles in Crow and Lakota is unclear at this point; it seems plausible though that
nonspecific article developed from the combination of an indefinite article and some other marker
that restricted the combination of markers to nonspecific contexts.

Table 6.16: Specific and nonspecific article exponents

exclusive-specific article nonspecific article
Q’anjobal jun jun-oq
Lakota wã wã-ži
Crow -m -eem
inclusive-specific article nonspecific article
Tongan te ha
Blackfoot animate
(w)a (sg) , -(y)i (pl)
inanimate -i
-iksi (sg) , -itsi (pl)
Siar Lak animate
ep ti
inanimate
a ta

Table 6.16 also shows the specific and nonspecific articles in Tongan (cf. section 5.7.2), Blackfoot
(Frantz 1991: 9-11), and Siar Lak (Rowe 2005: 16). These three article pairs do not show any

187
obvious form relation. This shows that the exponent of the nonspecific article does not necessarily
depend on the exponent of the specific article.
Nevertheless, the formal relation between a number of certain article pairs is not random:
the exclusive-specific and nonspecific article (Q’anjobal, Lakota, Crow) as well as the exclusive-
specific and definite article (Biak). That the exponent of the exclusive-specific article forms the
basis for the definite article in Biak is plausible in a scenario in which the ambiguity between a
former inclusive-specific and definite referents was resolved by an additional marker in definite
contexts, an, that conventionalized and led to the development of a definite article, consisting of
an and the exponent of the later exclusive-specific article.
As was shown for nonspecific articles in section 5.7.4, their development often involves the
combination of a former indefinite article with an irrealis marker (or perhaps another type of
marker) that restricts this indefinite article to nonspecific contexts. Even though the nature of
this additional marker may differ, the fact that we find a number of exclusive-specific and non-
specific articles for which the exponent of the former corresponds to the basis of the exponent
of the latter is evidence that the diachronic development for nonspecific articles proposed in sec-
tion 5.7.4 is plausible. Moreover, it accounts for the different properties of the article pairs in
Tongan, Blackfoot, and Siar Lak, all of which have an article system consisting of an inclusive-
specific article and a nonspecific article. These systems probably did not have an indefinite article
from which the nonspecific article could have developed, which can explain why the exponents
of these two articles are not formally related.

6.2.2 A complex article paradigm in Mokpe

In section 5.2.2, we saw that the definite article in Mokpe (Bantu, Cameroon) has a rather complex
paradigm . In this section, I present the paradigm in more detail and propose a scenario for its
development. Table 5.2, repeated as Table 6.17, illustrates the article exponents for nouns of the
different classes. With regard to the article, this paradigm can be viewed as merging two layers of
article exponents: an older layer that corresponds to the tonal exponent, and a more recent layer
which is the segmental marker.

188
 Table 6.17: Mokpe article paradigm
class noun article+noun class noun article+noun
singular plural
1 mòtò è mòtò 2 wàtò wátò ‘person’
ŋmánà è ŋmánà wánà wánà ‘child’
3 mòlélí mólélí 4 mèlélí mélélí ‘food’
mólèŋgò mólèŋgò mélèŋgò mélèŋgò ‘sheep’
5 lìkàlà líkàlà 6 màkàlà mákàlà ‘bridge’
líkpà líːkpa mákpá máːkpá ‘bag’
7 ètáŋgùlɛ̀ étáŋgùlɛ̀ 8 βètáŋgùlɛ̀ βétáŋgùlɛ̀ ‘lizard’
élélé élélé βélélé βélélé ‘morning’
9 mbèʒà è mbèʒà 10 í mbèʒà í mbèʒà ‘boy’
mbólì è mbólì í mbólì í mbólì ‘goat’
14 wòʒǒ wóʒǒ ‘face’
/
wɔ́ŋgɔ́ wɔ́ŋgɔ́ ‘fear’
19 ìnɔ̀ ní ínɔ̀ ní ‘bird’
/
dʒùŋgù dʒúŋgù ‘mosquito’

The nouns of classes 1 to 10 are singular-plural pairs, while the nouns in classes 14 and 19 are
singular forms with corresponding plural forms mostly in class 10. Each noun class is represented
by a noun with a high and low tone in the first tone-bearing unit,13 which is affected by the tonal
article exponent. What is marked in Table 6.17 are those classes which do not combine with the
high tone exponent. Class 1 and 9 occur with è, while nouns of class 10 combine with the marker í.
I show later in this section that í is not an exponent of the definite article and can occur in both
definite as well as in indefinite contexts.
With nouns of the classes that require the tonal exponent of the definite article, it surfaces
as a high tone on the first tone-bearing unit of the noun if the latter has a lexical low tone in
this position (first row of examples). If the first tone-bearing unit of the noun is already lexically
specified as having a high tone, the tonal article does not affect the tones of the noun (second
row of examples). Only for the nouns in classes 5 and 6, a phonological effect of the tonal article
could be observed with nouns that already have an initial high tone. This effect is represented as
vowel-lengthening here.14
13 My informants and I are not aware of examples with initial high tones for nouns of class 19.
14 Inorder to fully account for the tonal effects of the article in Mokpe, a detailed phonetic and phonological analysis
would be necessary. I present the system in a simplified way here, without discussing contour tones that are affected
in a slightly different way by the tonal article.

189
 As mentioned in the beginning of this section, the tonal article exponent most probably rep-
resents an older article system. It is more pervasive and the only marking for definite referents
mentioned in Atindogbe (2013).
Even though there is no reason that would speak against referential markers being expressed
by tones, tonal article exponents are crosslinguistically rare. To the best of my knowledge, the
other only reported case is found in Ewondo (Lyons 1999: 65f), which is closely related to Mokpe.
Tonal articles probably originate from short(ened) segmental prenominal morphemes that are
phonetically reduced to the nonsegmental material, which then associate to the first tone-bearing
unit of the noun. Other similar processes of phonetic reduction in Mokpe make such a scenario
plausible.15 In Mokpe, this process is pervasive with prepositions and other function words that
often only consist of a single vowel in and that can be reduced to a tone which associates to the
first tone-bearing unit of the following word.
Possessives and demonstratives show an interesting effect with respect to the two article ex-
ponents. The co-occurrence restrictions for the tonal and segmental exponents with possessives
and demonstratives are different in Mokpe. Examples (22) and (23) illustrate that the segmental
article è cannot co-occur with either demonstratives or possessive markers in the noun phrase.

(22) (*è) ònó (*è) mòtò

art:def dem.cl1 art:def child.cl1
‘this child’ Mokpe (prim. data)
(23) (*è) mòtò wǎ-mì
art:def child.cl1 cl1-poss:1sg
‘my child’ Mokpe (prim. data)

The tonal article exponent, on the other hand, can be shown to occur with both demonstratives
and possessives. Demonstratives precede the noun, while possessives follow it. Assuming that
the anchor of the tonal article is the noun phrase and not the noun, one would expect to find it
on the first tone-bearing unit of the first element in the noun phrase. Demonstratives preceding
the noun, we would expect to find that all demonstrative markers of the noun classes that take
the tonal article have an initial high tone, and that the high tone no longer surfaces on the noun.
Possessive pronouns, on the other hand, follow the noun. Accordingly, the high tone is expected
to be realized on the noun. Table 6.18 shows that this is indeed what we find for the tonal article
exponent. On the left, we see the proximal demonstrative in combination with a noun, and on
the right, possessives are illustrated by the possessive form of the first person singular. We see
that all demonstratives of the noun classes taking the tonal article have an initial high tone, while
15 The
phonological assimilation of preposed articles is common across the world’s language, a well-known example
being the definite article le/la in French occurring as shortened l’ before vowel-initial nouns.

190
demonstratives of class 1 and 9 have an initial low tone. Also, the noun is no longer affected by
the tone if the demonstrative is present. In combination with possessives, on the other hand, the
tone is marked on the noun, as it is the first element in the noun phrase. As for the segmental
article é, we see in Table 6.18 that it cannot be applied in the presence of either a demonstrative
or a possessive.

Table 6.18: The definite article in Mokpe with demonstratives and possessives
class this noun noun my
1 (*è) ònó mòtò ‘this man’ (*è) mòtò wǎmì ‘my man’
2 wánù wàtò ‘these women’ wátò wámì ‘my women’
3 mónò mòlélí ‘this food’ mólélí ŋmánì ‘my food’
4 ménè mɛ̀ kɔ̀ ‘these plantains’ mɛ́kɔ̀ méjàmí ‘my plantains’
5 línì lìkàlà ‘this bridge’ líkàlà lámì ‘my bridge’
6 mánù màkàlà ‘these bridges’ mákàlà mámì ‘my bridges’
7 énè èlèlà ‘this duck’ élèlà jámì ‘my duck’
8 wénè wèlèlà ‘these ducks’ wélèlà wéjámì ‘my ducks’
9 (*è) èné mbèʒà ‘this boy’ (*è) mbèʒà dʒǎmì ‘my boy’
10 (*í) ínì mbèʒà ‘these boys’ (*í) mbèʒà dʒámì ‘my boys’
14 wónò wòʒǒ ‘this face’ wóʒǒ wámì ‘my face’
19 ínì ìnɔ̀ ní ‘this bird’ ínɔ̀ ní dʒámì ‘my bird’

This shows that even though the two markers are exponents of a single definite article based on
their distribution in definite contexts (cf. section 5.2.2), there seems to be no requirement for
the two exponents to behave alike with respect to their co-occurrence with demonstratives and
possessives.
Concerning the development of this article paradigm, there is no evidence for the source of the
tonal article. However, it can be assumed to be the older article exponent that, for some reason,
was no longer available for nouns of class 1, 9, and 10 at a certain point in time. This in turn could
have led to the development of the segmental article exponent è. The latter can be traced back
to the agreement marker of class 9, occurring in the preverbal domain and cross-referencing the
subject. An example is given in (24):

(24) è mbúndá è tìmbá ndóló

art:def friendship.cl9 cl9 become love
‘The friendship became love.’ Mokpe (prim. data)

There are two arguments to explain why the agreement marker of class 9 developed into an article
exponent instead of the agreement marker á of class 1. Firstly, class 9 seems to be the most

191
productive noun class in Mokpe; loan words as well as ad-hoc uses of English and French nouns
consistently fall into this class. As a result, class 9 nouns probably also have a high token frequency
in use, which would already make them more accessible for the development into an article than
less frequent forms. Secondly, it is plausible that the referents of class 1 nouns, human referents
in most cases, are interpreted as highly discourse-prominent and as definite by default, making
additional marking as definite less necessary than the marking of other types of referents. Once
the marker è was established as an adnominal definiteness marker with class 9 nouns, the marker
may have been less associated with its earlier function of marking the gender class 9, giving way to
be extended to class 1 nouns. That this is a plausible scenario is reflected in the current paradigm:
the agreement marker of class 1 is not attested as an article exponent.
Nouns of class 10, as shown in table 6.17, neither combine with the tonal nor the segmental
article exponent but are marked as plural nouns with í. Since class 9 and 10 nouns do not have
a noun class prefix that formally distinguishes number, the plural agreement marker of class 10,
otherwise used with verbs, occurs adnominally to mark the nouns as plural forms. Its use as
agreement marker for nouns of class 10 is illustrated in (25).

(25) [í nʒɔ̀ kù] í lìjà ó wàŋgá

cl10 elephant.cl9/10 cl10 live in forest
‘Elephants live in the forest.’ Mokpe (prim. data)

In contrast to its singular counterpart è which occurs adnominally as definite article, the marker
í does not encode definiteness but marks plural. The referent type of the noun is determined by
the context alone. Example (26) shows the noun kémà of class 10 marked by í. The referent of
the noun is interpreted as specific by the context because it is not identifiable by the hearer. In
utterances like (27), the referents of í ŋgɔ̀ndɔ̀ ‘girls’ can only be interpreted as definite according
to my informants. That the plural class 10 marker is also compatible with nonspecific contexts is
shown in (28).

(26) è βóndá na dʒèlí, n-ɛ́ní [í kémà] ó ndáwò

the time I came 1sg-see cl10 monkey.cl10 in house.cl2
‘When I came (home), I saw (some) monkeys in the house.’ Mokpe (prim. data)
(27) [í ŋgɔ̀ ndɔ̀ ] dʒáŋgɔ́ dʒó ámbɛ̀ lɛ̂ ówókà
cl10 girl.cl10 your 2sg wait outside
‘The girls are waiting outside.’ Mokpe (prim. data)
(28) wúɲá wɔ́kɔ́ nà kánɛ̀ lǒβà [í mbúndá]
day one I pray have cl.10 friend.cl10
‘I wish to have friends one day.’ Mokpe (prim. data)

192
Nevertheless, the plural marker í shows some effects related to definiteness. As is shown in
examples (29) and (30), existential constructions that force a non-definite interpretation of the
referent do not allow for the use of í. This makes the noun ambiguous between a singular and
plural interpretation:

(29) ó wélì [(*í) kémà] ó wàŋgá

there be cl10 monkey.cl10 in forest
‘There are / is a monkey(s) in the forest.’ Mokpe (prim. data)
(30) ó wélì [(*í) ŋgɔ̀ ndɔ̀ ] jò ámbɛ̀ lɛ̂ ówókà
there be cl10 girl.cl10 2sg wait outside
‘There are girls / a girl waiting for you outside.’ Mokpe (prim. data)

The Mokpe article paradigm showed two general properties of articles and article paradigms.
Firstly, the two article exponents in Mokpe stress the independence between articles as referential
markers and the forms that these markers have. In addition, Mokpe uses the marker í to indicate
plural (class 10). Even though it is similar to the segmental article exponent é, it does not mark def-
initeness and cannot be considered as an article exponent for nouns of class 10. Secondly, we saw
that the two article exponents showed a distinct behaviour with demonstratives and possessives.
This provides strong evidence against a functional motivation of the co-occurrence restriction
of the definite article and possessives and demonstratives. While it may be less surprising that
we find variation with respect to this parameter across languages, Mokpe shows that different
exponents of the same definite article in a single language can also differ in this respect.

6.3 Summary
In section 6.1, I provide selected examples to illustrate and discuss inflectional patterns of articles
and show that while there are minor areal trends for the inflection of articles, we rather find a
global crosslinguistic trend against the inflection of articles. With regard to number, gender, and
case, I show that two factors can be singled out to restrict their marking on the article: on the one
hand, the expression of number and case on the noun appears to restrict the marking of these two
categories on the article. Thus, the fact that marking gender on articles in the world’s languages is
considerably more frequent than the marking of number or could be due to the general avoidance
of marking the same category on more than one element in the noun phrase, manifested in the
lower number of articles that mark number and case. On the other hand, I showed that articles
mostly inflect for the categories that their source elements inflect for; they either retain inflection
or lose the opposition of forms, but very rarely, if at all, develop new inflection.

193
 Section 6.2 discussed two types of paradigmatic properties of articles. In the first part, I showed
that we find a number of articles that formally correspond to parts of other articles within the lan-
guage. These cases were shown to reflect the diachronic relation between definite and inclusive-
specific articles, as well as exclusive-specific and nonspecific articles, also suggesting that non-
specific articles in system with an inclusive-specific article developed in a different way than
nonspecific articles in systems with an exclusive-specific article. The second part dealt with the
definite article in Mokpe that has a tonal and a segmental exponent conditioned by the noun class.
These two exponents were shown to behave differently with respect to their compatibility with
possessives and demonstratives. This showed that even though the markers are exponents of a
single definite article based on their referential function, there seems to be no pressure on that
would require different exponents of the same article to have similar morphosyntactic properties,
which in turn is an argument for the independence of morphosyntactic properties of articles as
referential markers.

194
Chapter 7

Crosslinguistic variation and universals

In this chapter, I discuss a number of trends and patterns of variation of articles in the world’s
languages, beginning with the areal trends with regard to article types in section 7.1. Section 7.2
presents and discusses two types of articles that are only scarcely attested (non-anaphoric articles)
or entirely absent (generic articles). Section 7.3 proposes a typology of article types and article
systems. I discuss the relation between article types and referent types with regard to the crosslin-
guistic frequency distributions of the major article types. On the bases of these distributions and
the resulting crosslinguistic gaps, I formulate five universals. Section 7.4 deals with articles and
their interaction with three types of constructions: definite articles in superlative constructions,
articles in combination with plural nouns, and articles co-occurring with other determiners. I
show that the distribution of articles in these contexts can mostly be related to the diachrony of
the articles, superlatives being the only case in which semantics seems to play a major role.

7.1 Areal trends

In my sample of 131 articles from 99 languages from around the world, the frequency distribu-
tions of the nine article types are presented in Figure 7.1: definite (def), anaphoric (ana), recog-
nitional (recog), exclusive-specific (exspec), nonspecific (nspec), indefinite (indef), presenta-
tional (pres), inclusive-specific (inspec), and referential (ref) articles. The corresponding num-
bers are given in Table 7.1. We see that definite articles correspond to the most frequent article
type, confirming the trend noted in Dryer (1989, 2014).

195
 40

30

count
20

10

def ana recog exspec nspec indef pres inspec ref

Figure 7.1: Article types

Table 7.1: Distribution of article types in the sample

article type N in sample
definite 47
anaphoric 23
recognitional 5
exclusive-specific 12
nonspecific 6
indefinite 15
inclusive-specific 6
presentational 13
referential 4

However, the more fine-grained distinction of article types also shows that anaphoric articles
are relatively frequent across the world’s languages, as well as indefinite articles, if counted to-
gether with presentational articles. Figure 7.2 shows the distribution of article types across the
six macroareas distinguished in Dryer & Haspelmath (2013) and Hammarström et al. (2018). The
numbers are shown in Table A.1 in Appendix A. We can observe three major areal trends: Africa
and to a lesser degree North America show a relatively high proportion of definite articles. Eura-
sia, on the other hand, features a high number of indefinite (and presentational) articles, and
Australia stands out for its high proportion of anaphoric articles.

196
 type.1
20 def
ana
recog
exspec
count

nspec
indef

10 pres
inspec
ref

africa eurasia papunesia australia n.america s.america

Figure 7.2: Article types in the six macroareas

Since it was already shown that definite articles are crosslinguistically more frequent than other
types of articles (Dryer 1989, 2014), their high frequency in Africa and North America is not
regarded as an areal phenomenon.
The second trend that can be seen in Figure 7.2 is the high proportion of indefinite and presenta-
tional articles in Eurasia. This trend could be due to both internal grammatical properties of the
languages (differential object marking, which appears to be relatively common in Indo-Iranian
languages) or language contact; future research will be required to shed more light on this obser-
vation.
The high number of anaphoric articles in Australia is perhaps most unexpected and may be
related to an interesting property that concerns Australian languages with anaphoric articles, and
perhaps languages from Australia in general. Many of these languages either lack distinct third
person pronouns (and use demonstratives) or the pronouns are described as being rarely used
in speech.1 This is summarized in Table 7.2. Out of the 11 languages with anaphoric articles
in Australia, three languages do not have a third person pronoun that is formally distinct from
demonstratives, and for 4 languages it was reported that third person pronouns either have de-
fective paradigms and/or are rarely used in speech. Also Bhat (2004: 103) notes that “most of the
Australian languages are two-person languages in which [sic] third person pronoun is primarily
a demonstrative.”
1 In
the WALS chapter on third person pronouns and demonstratives (Bhat 2013), the data points for Australia also
show a high number of formally related third person pronouns and demonstratives.

197
 Table 7.2: The relation of third person pronouns and demonstratives in Australian languages
language properties of third person pronouns
no third person pronoun distinct from the demonstrative
Mangarrayi (Merlan 1989: 52)
Wardaman (Merlan 1994: 107f)
Warrwa (McGregor 1994: 20)
third person pronoun formally distinct from the demonstrative
Jingulu pronoun forms only available in acc and gen (Pensalfini 1997: 242)
Guragone pronoun is rarely used (Green 1995: 130)
Bardi pronoun is rare in speech, no anaphoric use (Bowern 2004: 42)
Martuthunira use of the pronoun is restricted (Dench 1994: 100)
Wubuy no reported restrictions (Heath 1984: 241)
Arrernte no reported restrictions (Wilkins 1989: 124)
Duuŋidjawu no reported restrictions (Kite & Wurm 2004: 50)
Yolŋgu no reported restrictions (Wilkinson 1991: 210)

Obviously, this does not mean that the lack of a third person pronoun would directly lead to the
presence of anaphoric articles. However, the development of an anaphoric article (which always
go back to an anaphoric pronominal form)2 could be favored by the absence of a separate third
person pronoun. The latter would make the use of demonstratives and the anaphoric marker
more frequent, which in turn could favor the extension of an anaphoric pronominal marker to an
adnominal marker. This effect seems to be restricted to languages from Australia. In languages
from other areas of the world, there is no evidence for a correlation between the lack of third
person pronouns and the existence of anaphoric articles.
Another interesting property of Australian languages was observed by Bhat (2004: 191):

“They [Australian languages] do not show any distinction between nouns and adjectives and,
correspondingly, they use the same set of proforms either as adnominals or as pronominals.”
(Bhat 2004: 191)

This more general property of the language system may have also facilitated the extension of
a pronominal anaphoric marker to an anaphoric article, and it could be another factor that is
responsible for the high number of anaphoric markers in Australian languages.
2 All
anaphoric in the sample can also occur as pronominal forms. While I am not aware of any languages with an
anaphoric article (or an adnominal anaphoric marker, for that matter) but no pronominal anaphoric marker, the
opposite is commonly found in different areas of the world, e.g. in Chukchi (Dunn 1999: 96), Gamilaraay (Giacon
2014: 192), and Bantawa (Doornenbal 2009: 94).

198
7.2 Unattested and rare article types: two examples
7.2.1 Non-anaphoric definite articles
I proposed three types of articles that belong to the definite domain in this study: definite, anaphoric,
and recognitional articles. Previous studies suggested the distinction of another type of definite
articles, namely non-anaphoric definite articles based on the data of a handful of languages with
markers that seem like definite articles, but that are not used to mark anaphoric referents. In
this section, I discuss the data that has led to the proposal of this article type and argue that the
evidence for non-anaphoric definite articles is not strong, which is why I did not include non-
anaphoric definite articles as a major article type in Chapters 4 and 5.
With his reference hierarchy, Dryer (2014) predicts the existence of non-anaphoric definite
articles, mentioning the language Ma’di as having such an article. Also Schwarz (2009, 2013)
provides a detailed discussion of two distinct types of definite articles co-occurring in individual
languages, one of which is the non-anaphoric definite article. He labels these “weak” definite
articles and opposes them to “strong” definite articles which largely correspond to what I call
anaphoric articles.
Definite articles that cannot encode an anaphoric referent are very rare in the world’s lan-
guages and it seems that there are only two types of scenarios that lead to their development. The
first scenario involves an emerging definite article that does not originate from a demonstrative,
but, e.g., from a possessive. In its first stages, the marker can extend to the functions of a defi-
nite article without necessarily extending its use to anaphoric contexts “immediately”. However,
many languages with a definite article originating from a possessive marker (e.g. Sumu, Amharic)
show that eventually, the article is very likely to extend its use to anaphoric contexts as well. An
example of an emerging definite article that has not yet entirely reached this stage is Indonesian.
In Indonesian, as was mentioned in section 3.1, the third person possessive marker -nya (which
is also used as an object pronoun) can be used to mark referents as definite based on situational
uniqueness, which can be viewed as an extension of the original possessive semantics of the
marker and as an emerging definite article. While it is still usually the demonstrative that is used
to mark anaphoric referents, Rubin (2010: 109) notes that in colloquial speech, -nya can be used
instead. Examples (1) and (2), repeated from section 3.1, show -nya in a uniqueness-based and in
an anaphoric context, respectively.

(1) kalau mau makan nasi-nya di lemari

if want eat rice-def in pantry
‘If you want to eat, the rice is in the pantry.’ Indonesian (Rubin 2010: 107)

199
(2) A: jadi gua ntar ketemu dia langsung di salon
so I soon meet her direct in salon
‘So I’m going to meet her directly at the salon.’
B: salon-nya di deket rumah?
salon-def in near house
‘Is the salon near your house?’ Indonesian (Rubin 2010: 109)

The second scenario that may give rise to a non-anaphoric definite article includes an article
system with a definite article and an additional anaphoric article. If the anaphoric article can
block the use of the broader definite article to encode anaphoric referents, the definite article
corresponds to a non-anaphoric definite article. The languages that were mentioned to have such
a system are Ma’di (Dryer 2014) and Lakota as well as a number of Germanic languages (Schwarz
2013). My sample also contains one language, Jamsay, with both an anaphoric article and a definite
article, making the latter a candidate of a non-anaphoric definite article. I briefly present the
definite articles from these languages, showing that most of them cannot be regarded as proper
non-anaphoric definite article; since in most cases, the presence of the anaphoric articles does not
automatically block the use of the definite article in anaphoric contexts.
In Jamsay and Ma’di, even though they have a dedicated anaphoric article, the latter does not
automatically block the use of the definite article with anaphoric referents. Heath (2008: 164)
notes for Jamsay that the anaphoric article kò is often used together with the definite article kùⁿ.
This is illustrated in (3) below.

(3) tògù pɔ́ːrɔ́ tógó kó bɛ̀ rɛ̂ː bè nûː [kò tôg kùⁿ] úró
shed first shed.building 3sg.nhum in 3pl enter.pfv art:ana shed art:def house
táná-ŋá mèyⁿ nîŋ yɔ́=kɔ̀
become-caus and now exist=nhum
‘The first shed that they built and went into (to live), that (aforementioned) shed having
been transformed into a house, it is still there to this day.’ Jamsay (Heath 2008: 164)

For Ma’di, Blackings & Fabb (2003) distinguish between a definite article rɪ̀ and an anaphoric
article nā. They describe the two markers as follows: “Referents which are to be found in the
discourse context are identified with rì ‘the … in question’ (for more distantly mentioned referents)
and nā ‘the aforementioned’ (for more recently mentioned referents)” (Blackings & Fabb 2003:
17). This description also suggests that even though Ma’di has an anaphoric article, the definite
article is not excluded from marking anaphoric referents. What complicates the picture of the
two articles in Ma’di is the status of the definite article. Throughout the grammar, the so-called
definite article often occurs together with the focus marker ʔɪ̄ and also its translation as “the … in
question” make it questionable whether the marker is a definite article at all. Without discussing
its status here, it can be shown that it is not excluded from occurring with anaphoric referents.

200
Example (4a) shows an utterance referring to ɓá ‘people’. Two sentences later in the conversation,
(4b) is uttered, referring back to the referent of ɓá ‘people’. We see that the noun is marked by
the so-called definite article rɪ̀ and not by the anaphoric article nā.

(4) a. ādrúpì nɪ̄ tʃárágùlé k-ēɓù ānɪ́ nɪ́ ɓá míà ázɪ́á

brother 3sg Caragule 3-call 3s ben people hundred six
‘His brother Caragule should get the service of six hundred people for him.’
b. ìtó ʔà ādrúpɪ̀ ūŋgwē [ɓá rɪ̀] ʔɪ̄
Ito his brother call people art:def foc
‘Hare’s brother called them, the people.’
Ma’di (Blackings & Fabb 2003: 674)

Assuming that rì in Ma’di is a definite article, its relation to the anaphoric article is not essen-
tially different from the relation between definite articles and demonstratives in the languages in
which the definite article can also occur to mark referents as deictic and thus competes with the
demonstrative in deictic and also anaphoric contexts.
Lakota is another language with similar properties in that it was reported to have an anaphoric
and a definite article, which is why Schwarz (2013) claims that the definite article in Lakota is re-
stricted to the marking of only uniqueness-based referents, similarly to what was claimed for
Ma’di by Dryer (2014). While it could be shown for Ma’di and Jamsay that the presence of an
anaphoric article does not automatically block the definite article in anaphoric contexts, the situ-
ation in Lakota is different.
The use of the definite article ki in Lakota being well attested (Curl 1999, Rood & Taylor 1996,
Van Valin 1977, Williamson 1984), the so-called anaphoric article kuŋ is also mentioned in the
works cited above, but not documented by convincing examples in any of these studies, since the
scarce evidence shown always goes back to Buechel (1939) with respect to the anaphoric article
kuŋ. Also in Schwarz’s discussion on the two articles in Lakota, the only source for examples
of kuŋ is Buechel (1939). Moreover, Curl (1999) and Ingham (2003) note that ki is the regular
definite article which can also be used anaphorically, and even though kuŋ often marks anaphoric
referents, its main function is the indication of a topic-shift. Therefore, I conclude that kuŋ in
Lakota cannot be regarded as an anaphoric article.3 Thus, Lakota is not a convincing example of
a language with a definite non-anaphoric article either.
Schwarz (2009) also discusses data from several West Germanic languages, one of which is
Fering (North Frisian, Germany). In Fering, it seems to be the case that that the existence of an
anaphoric article blocks the use of the definite article in anaphoric contexts. Example (5) below
3 Lakota is also part of my sample; I do not treat kuŋ as an article for the reasons mentioned above.

201
shows a minimal pair, contrasting a situationally unique (5a) with an anaphoric (5b) referent. We
see that the two articles, definite a and anaphoric di are mutually exclusive in these contexts.

(5) a. ik skal deel tu [a/ *di] kuupmaan

I must down to art:def art:ana grocer
‘I have to go to the grocer.’
b. oki hee an hingst keeft. [di/ *a hingst] haaltet
Oki has a horse bought art:ana art:def horse limps
‘Oki has bought a horse. The horse limps.’
Fering (Schwarz (2013: 538), cited from Ebert (1971)

The same effect can be found in German with definite articles following certain prepositions. To-
gether with these prepositions, the singular forms of the definite article occur in a contracted
form which can only be used in non-anaphoric definite contexts. At the same time, the use of the
uncontracted combination of the preposition and the definite article is not acceptable in such con-
texts. An example to illustrate this is given in (6). Example (7) shows that a situationally unique
definite referent, on the other hand, requires the contracted form, if available:

(6) In der New Yorker Bibliothek gibt es ein Buch über Topinambur.
in art:def New York library exist 3sg art:indef book about topinambur
Neulich war ich dort und habe [in dem /#im Buch] …
recently was I there and have in art:def in.art:ana book
‘In the New York public library, there is a book about topinambur. Recently, I was there and
(searched) the book (for an answer to the question of whether one can grill topinambur).’
German (Schwarz 2009: 30)
(7) Der Empfang wurde [vom/ #von dem Bürgermeister] eröffnet.
art:def reception was by.art:ana by art:def mayor opened
‘The reception was opened by the mayor.’ German (Schwarz 2009: 40)

I did not consider this distinction between a definite and an anaphoric article in German in the
other parts of this study because it is lexically restricted by the preposition used.
The data from Fering and German are similar to examples (37) and (38) from Kaqchikel showed
in section 5.2.3, repeated in (8) and (9) below.

(8) a Lu xiroyoj jun b’eychik mambajota yich’o ruk’in [ri vinaq

prop Pedro call.pst.s:3sg.o:1sg one way want.neg.1sg talk with.3sg ana person
*(ri)]
ana
‘Pedro called again. I don’t want to talk to that person.’ Kaqchikel (prim. data)

202
(9) k’ab’a a Lu xuroyoj ri ru vecino loman chuk’a
recently prop Pedro call.pst.s:3sg.o:3sg art:def poss:3sg neighbour while also
ink’o yin chila’ [ri achin *(ri)] jabey xqa chinwach
be.1sg I there ana man ana nice came in.front
‘Recently Pedro invited his neighbour while I was at his house. I liked that man.’
Kaqchikel (prim. data)

Although it was shown in section 5.2.1 that the definite article in Kaqchikel occurs in anaphoric
contexts as well, examples like (8) and (9) show that its use can be blocked in certain types of
anaphoric contexts. The most influential factor probably is the distance between the antecedent
and the anaphoric referent. What these examples hint at is that even though the use of the definite
article can be blocked with certain types of anaphoric referents by an anaphoric article, the restric-
tion depends on certain properties of the anaphoric referent, and do not apply to all anaphoric
contexts. Therefore, and also because the number of reported non-anaphoric articles is so low,
it is questionable if non-anaphoric definite articles in the strict sense exist at all in the world’s
languages.
To conclude, non-anaphoric definite articles are extremely rare across the languages of the
world, and non-anaphoric definite articles in the strict sense probably correspond to the idealiza-
tion of a more complex competition pattern between two articles with respect to different types
of anaphoric referents.

7.2.2 Generic articles

A generic referent was defined in section 4.2.4 as a referent that does not correspond to a single
referent of its kind, but to all referents of its kind, i.e. to the kind set.
Articles whose primary function it is to mark generic referents do not seem to be attested in the
languages of the world. This does not mean that articles cannot occur with generic referents; in
many European languages, we find that generic referents are typically expressed by bare nouns in
the plural or nouns in the singular together with a definite article. However, from a crosslinguistic
point of view, we do not find a strong tendency with regard to the use of certain article types with
generic referents.
In general, we find definite, indefinite, inclusive specific, and referential articles that can be
used to express generic referents. Other article types do not seem to be attested in such contexts.
Two examples of definite articles that are used with generic referents come from Armenian and
Bullom So. According to the description in Dum-Tragut (2009), examples (10) to (11) show two
contexts with generic referents in Armenian. We see that the definite article is used together with
a singular noun. Example (12) shows that Bullom So can also express generic referents as a plural
noun together with the definite article.

203
(10) bžišk-ě piti ušadir lin-i
doctor-art:def part careful be-deb.fut.3sg
‘A doctor must be careful.’ Armenian (Dum-Tragut 2009: 111)
(11) mard-u-n kar-oł ē krakel-ov č’-span-es bayc xosk’-ov
person-dat-art:def can-ptcp:prs is shoot.inf-instr neg-kill-fut.2sg but word-instr
span-es
kill-fut.2sg
‘One cannot kill a human with shootings, but with words you kill him.’
Armenian (Dum-Tragut 2009: 162)
(12) [ìmúsmɛ̀ nɛ̀ àcɛ́] ŋà cé-ɛ́n àwòyɛ́n
water.cat art:def 3pl cop-neg frightening
‘Water cats are not scary.’ Bullom So (Childs 2011: 72)

Also inclusive specific articles can be used to express generic referents, e.g. in Basque and Bemba.
Examples (13) and (14) show the use of the inclusive specific article with generic referents in the
singular and plural in Basque, and example (15) shows that Bemba can express a generic with the
inclusive specific article and a plural noun as well. Baure, on the other hand, has a referential
article. The latter can equally occur together with a noun whose referent is generic, which is
shown in (16).

(13) haritz-a zuhaitz saindu-a da

oak-art:inspec tree sacred-art:inspec is
‘The oak is a sacred tree.’ Basque (Hualde & Ortiz de Urbina 2003: 120)
(14) txakurr-a-k ugaztun-a-k dira
dog-art:inspec-pl mammal-art:inspec-pl are
‘Dogs are mammals.’ Basque (Hualde & Ortiz de Urbina 2003: 119)
(15) i-m-bwa nii-nama
art:inspec-cl9-dog cl9-animal
‘Dogs are animals.’ Bemba (Givón 1969: 47)
(16) ni=ipiko-wo [to tiporek]
1sg=be.afraid-cop art:ref chicken
‘I am afraid of chicken.’ Baure (Danielsen 2007: 130)

In other languages, the article types shown in the preceding examples are not compatible with
generic referents. Both Sheko and Oko, for instance, have definite articles. Nouns that express
generic referents, however, cannot be used with the definite article and occur as bare nouns. This
is shown in (17) for Sheko and in (18) for Oko.

204
(17) gárgá íntʃù-rà há=gyá-mə
termite wood-acc 3sg.m=chew-irr
‘A termite eats wood.’ Sheko (Hellenthal 2010: 142)
(18) ógbén e-yíwo
child prog-cry
‘A child is crying / Children are crying’ Oko (Atoyebi 2010: 161)

Bonan has an indefinite article. As is shown in (19) and (20), the indefinite article is not used
with generic referents, which are expressed as bare nouns in the singular. Similarly, Palula has
an exclusive-specific article. As we see in example (21), generic referents are expressed by bare
nouns, while the exclusive-specific article is not used.

(19) ŋantɕhə=da ɬawa zaija bi-saŋ

past=loc shaman strong cop.s-epist
‘In the past, shamans were very powerful.’ Bonan (Fried 2010: 83)
(20) talo naŋda noʁi doχkəsa hkamo wa
tall.building in dog raise difficult cop.o
‘Keeping dogs in an (apartment) building is difficult.’ Bonan (Fried 2010: 83)
(21) amzarái muṛ-u=bhaáu insaán na kha-áan-u
lion die.ptcp-m.sg=adj human.being neg eat-prs-ms.g
‘A lion doesn’t eat a human being that is dead.’ Palula (Liljegren 2016: 398)

These examples show that the use of articles with generic referents fundamentally differs from
their uses in other contexts. Because articles of different types can occur with generic referents,
we can assume that is not their primary referential function that makes them compatible with
generic referents. That it is not the article that marks the referent as generic is supported by
the fact that single languages do not necessarily have one consistent strategy to mark generic
referents but usually use different strategies.
It seems plausible that the article types that are attested with generic referents, i.e. definite,
indefinite, specific, and referential articles, are articles with a relatively high token frequency in
language use to begin with, which probably allowed them to extend their use to contexts with
generic referents. Minor article types4 such as anaphoric articles, recognitional articles, or presen-
tational articles cannot be used. One possible explanation of this distinct behaviour of different
article types with respect to their compatibility with generic referents is based on their functional
domains, meaning that semantically broader articles may have more potential to extend their
functional domain than articles with a more restricted meaning. Another way to account for this
4 Minor
is understood here in the sense of a narrower function of the article, which leads to its occurrence in fewer
contexts and makes it less frequent in language use.

205
difference with respect to the use of articles with generic referents is based on the distribution of
articles in the language. If a language encodes several referent types by articles and if this leads to
bare nouns being infrequent in the language, the system could allow or require the use of articles
with generic referents in order to avoid the occurrence of bare nouns.
Across languages, the high variation that we find with respect to articles and their use with
generic referents in addition to the lack of evidence for articles whose primary function it is
to encode generic referents lead to the conclusion that generic articles do not exist. That generic
referents do not have a marker of their own is plausible from a functional perspective. The generic
referent is the least individuated referent type, and arguably the least prominent referent type in
our communicative needs, which would also make it the least frequent referent type in language
use. Therefore, it is not surprising that languages do not develop a grammatical marker, i.e. an
article, whose main function is the expression of generic referents.

7.3 A typology of articles and article systems

Based on the examples found in the world’s languages, I defined different article types with respect
to the referent types that they mark and illustrated each type in detail in Chapter 5. In this
section, I discuss certain article types lumped together in the following way: definite articles
(including definite, anaphoric, and recognitional articles), inclusive-specific articles, exclusive-
specific articles, indefinite articles (including indefinite and presentational articles), nonspecific
articles, and referential articles.
The intuitive way to distinguish between different article types is to distinguish them accord-
ing to their referential function, i.e. the referent types that they express. This was proposed in
Chapter 4 based on the distinction of definite, specific, nonspecific, and generic referent types.
Another distinction between article types is possible on the basis of the number of referents that
the article encodes. In this vein, Table 7.3 shows a typology of the articles that are attested in the
world’s languages, distinguishing between the following three types: articles of type I encode a
single referent type, articles of type II encode two referent types, and type III articles encode three
referent types.

206
 Table 7.3: A typology of articles
article types
type I type II type III
referent type (encodes 1 ref. type) (encodes 2 ref. types) (encodes 3 ref. types)

definite (75) def/ana/recog

inspec
specific (50) exspec ref
indef/pres
nonspecific (38) nspec

generic (0)

As is shown in Table 7.3, there are three articles types that belong to type I and that mark a sin-
gle referent type: definite (def), anaphoric (ana), and recognitional (recog) articles only encode
definite referents; exclusive-specific articles (exspec) encode specific referents, and nonspecific
articles (nspec) mark nonspecific referents only. We find three article types in the world’s which
combine the marking of two referent types and therefore belong to type II: inclusive-specific arti-
cles (inspec) mark both definite and specific referents, while indefinite (indef) and presentational
(pres) articles encode specific and nonspecific referents. The referential article (ref) corresponds
to type III, marking definite, specific, and nonspecific referents. As was already mentioned in the
previous sections, marking more than one referent type does not mean that the article is poly-
semous. I assume that they are semantically vague with respect to the two functions coded and
coexpress the two/three referent types.
As was argued in the previous section, articles whose primary function it is to encode generic
referents are not attested, which is why no article type is associated with generic referents in
Table 7.3. Related to this, the numbers given in brackets for the four referent types in Table 7.3
represent the sum of articles in the 99-languages sample that encode each of these referent types
(cf. Table 7.1 on p. 196). Thus, the number of 85 for definite referents in Table 7.3 means that the
sample contains 85 instances of articles that encode definite referents, consisting of the 47 definite,
the 23 anaphoric, the 5 recognitional, the 6 inclusive-specific, and the 4 referential articles that
add up to 85 instances of articles encoding a definite referent. The numbers for the other referent
types given in Table 7.3 were calculated accordingly.
What we can see is that in the languages of the world, the referent type that is most frequently
encoded by articles is the definite referent, followed by specific and nonspecific referents. A Bayes
proportion test confirms that the difference between the numbers for encoded definite referents

207
(85) vs. specific (50) and nonspecific (38) referents is statistically significant, while the difference
between the latter two referent types is not (cf. Table A.18 in the appendix). This is similar but not
identical to the findings in Dryer (1989, 2013a,b) who also shows that there are more definite than
indefinite articles in the world’s languages. The numbers in Table 7.3 do not reflect a claim about
the frequency of article types in the world’s languages per se. They show how often each referent
type is marked by an article in a crosslinguistic sample. What the numbers in Table 7.3 suggest is
that finding more definite articles than, for instance, indefinite articles in the world’s languages
is a consequence of the general trend towards marking definite referents in the grammar. As I
show in the following paragraphs, a typology of article systems allows us to find the same trend
reflected in the frequency of certain types of article systems in the languages of the world.
Similarly to the typology of articles, article system of individual languages can be distin-
guished on the basis of the number of referents that the article system encodes. Tables 7.4 to
7.6 show the major types of attested article systems in the world’s languages, encoding one, two,
and three referents types, respectively. The last row in Tables 7.4 to 7.6 provides the frequency
of the article system in the sample. Note that definite, anaphoric, and recognitional articles are
lumped together under the label of definite articles, as well as indefinite and presentation articles
are both counted in as indefinite articles. Table A.17 in Appendix A lists all the languages for each
article system type.5 Article systems encoding one referent type also consist of a single article.6
The only article types that are attested as single articles in the language are definite (including
definite, anaphoric, and recognitional) and exclusive specific articles. Nonspecific articles do not
seem to occur without other articles in the language.

Table 7.4: Article systems encoding 1 referent type

referent type system I system II

definite def

specific exspec

nonspecific

N systems 50 2
5 Since the sample includes 99 languages, the total number of the article systems in the sample should also add up
to 99; however, Basque was excluded here for having a “hybrid” article system with an inclusive-specific and an
indefinite article. Therefore, the total number of article systems considered here is 98.
6 The only exception are article systems that only code definite referents but have a distinct definite and anaphoric or

recognitional article.

208
Article systems that encode two referent types are shown in Table 7.5. These can either consist of
a single article that co-expresses two referent types, namely the inclusive-specific article (system
IV) and the indefinite (including presentational) article (system VI). Or, the system consists of two
separate articles for each referent type: system III, consisting of a definite (including anaphoric
and recognitional) article, and system V with an exclusive-specific article and a nonspecific article.
The combination of an exclusive specific article within a nonspecific article is not attested in my
sample. However, as I showed in Chapter 5, Q’anjobal may be considered as an example of a
language with a specific and nonspecific article and with no definite article.

Table 7.5: Article systems encoding 2 referent types

referent type system III system IV system V system VI

definite def
inspec
specific exspec exspec
indef
nonspecific nspec

N systems 7 2 / 15

The last group of article systems is shown in Table 7.6 and consists of article systems that mark
all three referents. The two combinations with two markers for the three referents that we find
are the following: system VII with a definite and an indefinite article and system VIII with an
inclusive-specific article and a nonspecific article. Another attested system marking all three
referent types is system IX with each referent type being coded by a separate article, namely
a definite, an exclusive-specific, and a nonspecific article. The last type (X) of article systems
attested in the world’s languages consists of a single referential article, coexpressing all three
referent types.

209
 Table 7.6: Article systems encoding 3 referent types
referent type system VII system VIII system IX system X

definite def def

inspec
specific exspec ref
indef
nonspecific nspec nspec

N systems 12 3 3 4

Out of the possible combinations of different article types, a number of systems are unattested.
These systems are shown in Table 7.7. In general, we can say that they all involve restrictions
imposed on the marking of nonspecific referents. As for systems with one marker, it was already
mentioned earlier in this section that nonspecific articles are not attested without other articles
existing in the language, which correspond to system *I in Table 7.7. Systems marking two referent
types also show a gap involving nonspecific referents; there is no article system with one article
that that would coexpress definite and nonspecific referents (system *II). The same gap is found
with systems that express all three referent types with two markers, one coding specific referents,
and another single article for definite and nonspecific referents (system *IV). Neither do we find
article systems that consist of a definite and a nonspecific article only (system *III).

Table 7.7: Unattested article systems

referent type system *I system *II system *III system *IV

definite def-nspec def def-nspec

specific exspec

nonspecific nspec def-nspec nspec def-nspec

As was mentioned in the beginning of this section, the crosslinguistic trend towards certain article
systems provides additional evidence for the general preference of coding definite referents. With
respect to article systems that mark only one referent type, this trend translates into articles
marking a definite referent (system I, N=50) being more frequent than articles which only mark

210
a specific (system II, N=2) referent (cf. Table 7.4). Moreover, article systems that consist of a
nonspecific article only (system *I) are not attested (cf. Table 7.7).
The trend towards the coding of definite referents can also be observed for articles systems
that mark two referent types, but in a slightly different way. As was shown in Table 7.5 for sys-
tems that mark two referent types, there are two systems with a single article, namely system
IV with inclusive-specific articles (encoding definite and specific referents) and system VI with
indefinite articles (encoding specific and nonspecific referents). That the difference in frequency
between inclusive-specific (N=2) and indefinite (N=15) articles is statistically significant is con-
firmed by a Bayes proportion test (cf. Table A.19 in Appendix A). The important finding here is
that nonspecific referents are more likely to be coexpressed with specific referents (as indefinite
articles), than specific with definite referents (as inclusive specific articles). Hence, we see a trend
against the coexpression of definite referents with other referent types. As Table 7.7 shows, the
coexpression of definite and nonspecific articles (system *II) is not attested.
This last observation extends to article systems marking all three referent types (cf. Table 7.6).
Systems marking the definite referent by a definite article in opposition to an indefinite article
co-expressing specific and nonspecific referents (system VII, N=12) are attested in 12 out of 22
languages with an article system marking 3 referents. On the other hand, the article system that
has a nonspecific article as opposed to an inclusive-specific article, expressing both definite and
specific referents (system VIII, N=3), is attested but rare across languages. A Bayes proportion
test confirms that this difference is statistically significant as well (cf. Table A.20 in Appendix A).
Accordingly, a system with a single article marking both definite and nonspecific referents and
with another article expressing specific referents (system *IV) is not attested.
What has not been mentioned so far is the possibility of article systems to have two articles
with overlapping functional domains, e.g. a definite article together with an inclusive specific
article, or an exclusive specific article with an indefinite article. Except for the co-occurrence of
definite and anaphoric (cf. section 7.2.1) or recognitional articles, it seems to be the case that
such systems do generally not occur across the world’s languages. The only other example of
functionally overlapping article types within a language in the sample is Basque. Basque has an
inclusive-specific article that encodes definite and specific referents, and a presentational article
that also marks discourse-prominent specific referents. While this may seem redundant and in-
efficient from a functional point of view, such a system can in fact be motivated functionally in
that it helps to reduce ambiguity in specific contexts: the inclusive specific article co-expressing
a definite and a specific referent, a noun marked with this article needs be interpreted as having a
definite or specific referent according to other contextual cues, if available. However, in contexts
in which a noun that would be usually associated with a definite referent, but which is intended

211
to have a specific referent, the presentational article can be used to resolve this mismatch of usual
and intended contextual interpretation.
The trends presented in this section allow for five main generalizations, presented as Univer-
sals 1 to 5 below.

Universal 1 The absence of generic articles

There are no articles whose main referential function is the coding of generic referents.

Universal 2 The co-expression of definite and nonspecific referents

Articles can only coexpress definite and nonspecific referents if they also encode specific referents.

Universal 3 Nonspecific articles depend on specific articles

Nonspecific articles only occur in article systems with either inclusive-specific or exclusive-specific
articles.

Universal 4 Systems of articles with overlapping functions are avoided

In the world’s languages, article systems that include articles with overlapping functions are rare.
Exceptions to this are systems with a definite and an anaphoric/recognitional article.

Universal 5 The split into the definite and the indefinite domain
In the world’s languages, articles show a preference to encode definite versus specific and non-
specific referents. This is manifested in the overall higher number of articles that encode definite
referents, but also in the lower number of articles that coexpress definite and specific referents
versus articles that coexpress specific and nonspecific referents.

Universals 1 and 2 concern article types; they are expected to hold within single languages. Uni-
versal 2, excluding articles that encode definite and nonspecific referents, reflects the gaps cor-
responding to systems *II and *IV in Table 7.7. It also confirms earlier scalar arrangements of
definite, specific and nonspecific referents (Croft 2003, Dryer 2014, Gundel et al. 1993), as was
discussed in section 4.4. Dryer (2014) explicitly notes that his reference hierarchy (see p.91) is
motivated by the functional domains that articles can have, allowing only those articles which
express neighbouring functions on this scale. Universals 3, 4, and 5 apply to article systems. Uni-
versal 3 is implicational and restricts the existence of a nonspecific article to the existence of an
inclusive-specific or exclusive-specific article in the system. It reflects the crosslinguistic gaps
of systems *I and *III. The last universal corresponds to the crosslinguistic tendency that article
systems tend to encode definite referents in contrast to both specific and nonspecific referents,
which motivates the split of referential functions into a definite and an indefinite domain. The

212
trend is not only manifested in the number of definite versus specific and nonspecific articles, but
also in the higher number of article systems with an indefinite article (VI, VII) against a lower
number of systems with an inclusive-specific article (VI, VIII), showing that there also is a trend
against articles that coexpress definite and specific referents.

7.4 Interactions between articles and selected constructions

This section discusses three types constructions in which articles can but do not need to occur, re-
sulting in crosslinguistic variation. I show that the variation can be accounted for by two types of
interactions: on the one hand, the semantic properties of the article can make it compatible with
the construction and license its use. This is the case for definite articles in superlative construc-
tions (section 7.4.1). On the other hand, I argue that in most cases, the compatibility of an article
with constructions in which it is not required referentially can be explained by the properties of
the article’s diachronic source element. I show this discussing the incompatibility of indefinite
articles with the plural (section 7.4.2) as well as the co-occurrence of definite articles with demon-
stratives and possessives (section 7.4.3). In the last part, I also show examples of different types
of co-occurring articles within languages.

7.4.1 Definite articles in superlative constructions

It is known that we usually find definite articles in superlative constructions as well. For instance,
Lyons (1999: 246) notes: “It is a general fact that languages which have definiteness marking
use it with superlatives”, and Hawkins (1978: 233-241) discusses definite (and indefinite) articles
with respect to superlatives in English in great detail, concluding that the definite article is oblig-
atory while the indefinite article cannot be used with superlatives (except for certain non-default
interpretations).
The aim of this section is not a detailed discussion of definite articles and their use in superla-
tive contexts, but to show that this is a case of the interaction between the semantic properties of
the article and the superlative, but also that it is only a strong correlation that can be outranked
by other requirements of the grammar.
Definite articles are inherently compatible with superlatives, because both involve the expres-
sion of a single salient referent (in a given context). Therefore, we could assume that in a language
with a definite article, the article should be present in superlative contexts, provided the language
has a distinct superlative construction as well, and disregarding non-default interpretations of the
latter. This also holds for attributes such as ‘first’, ‘last’, ‘next’, since they equally entail that there

213
is only a single referent with the property that they denote.7 In the languages from the sample
with a definite article and a superlative construction, no clear counter-example is attested; most
languages however do not have superlative constructions.
Examples (22) and (23) illustrate the obligatory use of the definite article in these contexts in
Mokpe. The superlative in Mokpe is expressed by a periphrastic construction that schematically
corresponds to ‘art:def noun more adj than all noun’.

(22) jɔ́ndɔ̀ / *jɔ̀ndɔ̀ já wòʒǒ

journey.cl7.art:def journey.cl7 cl7 front
‘the first journey’ Mokpe (prim. data)
(23) míːʒɔ̀ / *míʒɔ̀ má lɔ́ lǎkà mííʒɔ̀ má-ʒɛ̌
eyes.cl6.art:def eyes.cl6 cl6 nice more eyes.cl6.def cl6-all
‘the most beautiful eyes’ Mokpe (prim. data)

In the periphrastic expression in the examples of Mokpe, the definite article is not the only marker
of the superlative. However, we find examples of the article being the only marker of the superla-
tive in the world’s languages; Romance languages being a well-known example. In my sample,
the only languages with the definite article as the only formal marker of a superlative are Irish and
Zoque (Mixe-Zoque, Mexico), while the superlative construction in Zoque is an obvious calque
from Spanish. Examples (24) and (25) illustrate this. The comparative forms are given in (24a)
and (25a), and the definite article marking the superlative together with the adjective and the
comparative marker are shown in (24b) and (25b).

(24) a. ta Áine níos óige ná Seán

be Áine compar young.compar than Seán
‘Áine is younger than Seán.’
b. an gasúr is óige8
art:def child cop young.compar
‘the youngest child’
Irish (Stenson 2008b: 80-82)

(25) a. tehi-ʔawə mas wəhə ke miš

there-is compar good that you
‘There is one better than you.’
7 This general assumption is probably based on the fact that definite articles in European languages generally occur
with superlatives. Otherwise, one may also expect that the definite article is absent in superlative constructions
because it is not semantically required.
8 The superlative construction in Irish probably originates from a periphrastic construction and still obligatorily con-

tains the copula is, even though it is used attributively (Stenson 2008b: 82).

214
 b. [yoya bi mas šaša] ʔəy-kəš-pa mas mok
pig art:def compar fat 3-eat-ipfv more corn
‘The fattest pig eats the most corn.’
Zoque (Johnson 2000: 263)

We can see in example (25b) that the definite article may not be used with superlatives consistently
(Johnson (2000) does not provide more information on superlatives), as it does not occur with the
phrase mas mok ‘the most corn’. That this pattern is rare in the world’s languages is supported
by the fact that also Bobaljik (2012: 52-55) only provides a handful of examples of this pattern in
his detailed study on comparatives and superlatives, namely from Modern Greek, Maltese, Mid-
dle Armenian, and Livonian.9 This tells us something important about the use of the definite
article with superlatives. The fact that we find many languages with definite articles but without
superlatives, combined with the observation that the definite article is rarely the only marker of
the superlative shows that definite articles are only compatible with superlatives. This means,
they may be universally able to co-occur with superlatives, but they are by no means universally
sufficient to mark superlatives.
In this vein, Bobaljik (2012: 55) shows for Manx (Celtic, United Kingdom) that even when no
separate superlative marker is available, the combination of the comparative and a definite article
does not automatically lead to a superlative interpretation:

(26) yn boayl share

art:def place good.compar
‘the better/best place’ Manx (Bobaljik (2012: 53), cited from Phillips (2004: 28))

The opposite effect is shown for German in (27). Although the referent marked by meisten ‘most’
is not unambiguously identifiable and not necessarily definite, the definite article has to be used.
In this case, its use is no longer conditioned by the semantic compatibility with the construction,
but by a syntactic requirement of meisten as a superlative form in this construction:

(27) [Die meisten Kinder] sind infiziert.

art:def most children are infected
‘Most children are infected.’ German (Löbner 2011: 297)

Moreover, even though the definite article is used with superlatives in predicative contexts in
many European languages, it is not clear to what extent they are used in such contexts in the
languages of world.
9 Mentioned somewhat implicitly in Bobaljik (2012), the Livonian example involves a demonstrative and not a definite

article.

215
 To conclude, it seems that the use of definite articles in (attributive) superlative constructions
is universal, given that the language has superlatives. This can be related to the semantic prop-
erties of the definite article, and it is perhaps the clearest example for such a regular behaviour
of articles outside of contexts in which they are required referentially. However, we also saw
that the definite article is neither sufficient to mark superlatives (Manx), and we saw that it can
occur with superlative forms in construction with which it is no longer semantically compatible
(German).

7.4.2 Articles and plurals

Already Heine (1995: 68) notes that indefinite articles are “usually confined to determining the
singular of count nouns” and he finds that only a small number of indefinite articles is compat-
ible with nouns in the plural. Considering the compatibility of the articles in my sample, we
see a clear difference in the behaviour of articles from the definite domain vs. articles from the
indefinite domain with respect to plural nouns.10 Table 7.8 shows the numbers for each article
type in the sample that can and cannot occur with nouns in the plural. I group together definite,
anaphoric, and recognitional articles as articles from the definite domain, and indefinite, presenta-
tional, exclusive specific, and nonspecific articles as articles from the indefinite domain. In order
to distinguish between the definite and the indefinite domains, I exclude inclusive-specific and
referential articles, since their functional domains overlap with either domains. We see in Ta-
ble 7.8 that from the total number of articles which are available in the plural, 65 articles belong
to the definite domain, and only 19 to the indefinite domain. The single languages are provided in
Tables A.21 to A.24 in the appendix.11 A Bayes proportion test (cf. Table A.25 in Appendix A) con-
firms a statistically significant difference between articles from these two domains with respect
to their behaviour with plural nouns.
10 Here, I only consider singular and plural, ignoring that some languages have articles that inflect for more than these

two number values.

11 The
numbers in Table 7.8 indicating how many articles are compatible with plural nouns may differ from the
numbers of languages with plural as a nominal category: compatibility here is not restricted to systematic plural
marking, but is fulfilled as long as the article can occur with plural nouns.

216
 Table 7.8: The compatibility of articles with plural nouns
incompatible compatible marking
article type with pl with pl number
definite 0 45 18
definite
domain
anaphoric 0 16 3
recognitional 0 4 1
indefinite 12 3 1
indefinite
domain

presentational 6 6 5
exclusive-specific 4 7 3
nonspecific 2 3 2
Σ definite domain 0 65 22 (=0.34)
Σ indefinite domain 24 19 11 (=0.58)

In addition, the last column of Table 7.8 indicates how many of the articles inflect for number. In
contrast to the compatibility of the articles with plural nouns, articles from the indefinite domain
show a higher proportion of plural marking than articles from the definite domain. A Bayes
proportion test does not confirm this difference (cf. Table A.26). However, it is an interesting
observation that from 65 articles in the definite domain that are available in the plural, 22 inflect
for number, while we find 11 out of 19 articles from the indefinite domain with plural marking.
Moreover, it was noted in section 6.1.3.2 that the only contexts in which articles seem to have
innovative inflection with respect to their source element are indefinite articles that are inflected
for number unlike the numeral.
As was indicated in Table 7.8, the sample contains 9 articles from the indefinite domain that are
morphologically marked for number. This mostly corresponds to an unmarked singular form and
an additional plural marker in plural contexts. Three examples are given in (28), (29), and (30)
for Carib (Arawakan, Suriname), Basque, and Oko (Benue-Congo, Nigeria). Examples (28a), (29a),
and (30a) show their singular, and (28b), (29b), and (30b) their plural forms, respectively.

(28) a. erome [tonomy amu] si-epory-taka rapa

now animal art:indef 1-find-fut again
‘Today, I’ll find game again.’ Carib (Courtz 2008: 187)
b. [akuri teràa amu-kon] si-upi-ja
reed.rumped.agouti already art:indef-pl 1-seek-prs
‘I am going to try and find some agoutis.’ Carib (Courtz 2008: 188)
(29) a. Bilbon [etxe bat] erosi behar dut
Bilbao.loc house art:pres buy need aux
‘I need to buy a house in Bilbao.’ Basque (Hualde & Ortiz de Urbina 2003: 694)

217
 b. zergatik ez erregutu hari [bitxi bat-zuk] utz ziezazkion?
why not beg him jewel art:pres-pl lend aux.s.comp
‘Why not beg her to lend her some jewels?’
Basque (Hualde & Ortiz de Urbina 2003: 503)
(30) a. [ógbén ó̩bè̩n] pè̩n ó̩te̩le̩ àye
child art:spec.sg break pot art:def
‘A child broke the pot.’
b. [égbén é̩bè̩n] pè̩n ó̩te̩le̩ àye
children art:spec.pl break pot art:def
‘Some children broke the pot.’
Oko (Atoyebi 2010: 218)

A closer look at the 8 articles from the indefinite domain that inflect for number reveals that either
their source element is already morphologically marked for number, or, a plural marker is used
that also occurs with other nominal elements. The singular and plural forms12 forms of these
articles are given in Table 7.9.
Except for Konso, the plural markers of the articles displayed in Table 7.9 can be related to
plural markers on other elements in the languages. Moreover, in some cases, the plural marker
corresponds to a clitic, and it simply occurs on the article if the latter is present in order to occur
at the edge of the noun phrase.
The plural forms of the articles in Basque, Carib, Oko, and Logba use a marker that is found
on other nominal elements as well. In all three cases, these markers cover the majority of nouns,
and we do not find many other exponents for plural marking of nouns, which means that these
markers can be assumed to be relatively frequent in language use.
12 The specific article in Biak also has distinct dual and trial forms.

218
 Table 7.9: Singular and plural forms of indefinite articles
language article
Konso pres sg tokka (m) / takka (f)
pl takka-n (Orkaydo 2013: 94)
Basque pres sg bat
pl bat-zuk (Hualde & Ortiz de Urbina 2003: 122)
Carib indef sg amu
pl amu-kon (Courtz 2008: 110)
Oko exspec sg ó̩bè̩n
pl é̩bè̩n (Atoyebi 2010: 68)
Logba pres sg ɔ-kpiɛ
pl i-kp(i)ɛ (Dorvlo 2008: 89,333)
Lavukaleve pres sg roa (m) / ro (f) / roge (n)
pl rovo (Terrill 2003: 53)
Hausa pres sg wani (m) / wata (f)
pl wa(ɗan)su (Newman 2000: 153)
Biak exspec sg i / ya
pl sya (an) / na (inan) (van den Heuvel 2006: 72)
Biak nspec sg o
pl no (van den Heuvel 2006: 72)

In Basque, the segment -k generally occurs with nouns and adjectives to mark plural in the ab-
solutive and ergative (Hualde & Ortiz de Urbina 2003: 173). Also the inclusive specific article
a in Basque is followed by -k in these contexts. In addition, the plural marker only occurs on
the right edge of the noun phrase (Hualde & Ortiz de Urbina 2003: 113). Thus, the fact that the
presentational article occurs with the plural marker is not due to innovative morphology, but a
consequence of an independent property of the plural marker in the language. Also, the incom-
patibility with the numeral ‘one’ and the article probably is not a morphosyntactic but a semantic
one.
Also in Carib, -kon is used to mark plural on nouns, only a group of nouns with animate
referents uses a different marker -jan (Courtz 2008: 59). Similarly to Basque and as was shown in
example (28b), the plural marker only occurs on one element in the noun phrase and seems to be
flexible with regard to the element it occurs on (Courtz 2008: 143).
The third language where we see that the article from the indefinite domain has a plural marker
corresponding to the main plural marker for nouns is Oko. Oko has [ATR] harmony, the exclusive-
specific article has the value of [-ATR]. The alternation between ó̩- and é̩- on the specific article

219
is also found on both nouns and adjectives with the value [-ATR] to indicate singular vs. plural.
Nouns and adjectives that are [+ATR] have the singular marker ó- and take either é- or -í as plural
markers (Atoyebi 2010: 66-69). Although somewhat different from Basque and Carib, also Oko
shows that the number marking of the exclusive-specific article follows a more general number
marking pattern in the nominal domain. The same holds for the presentational article in Logba;
it expresses number with the regular noun class markers. These are prominent in the nominal
domain in that they are used by nouns and demonstratives.
The remaining articles from Table 7.9 have a plural marker (or distinct singular and plural
forms) that does not necessarily correspond to a pervasive number marking strategy in the nomi-
nal domain, but that can be traced back to the source elements of the articles. In Hausa and Biak,
the source elements of the articles are pronouns. In Lavukaleve, it was probably an adjective-like
element with gender and number marking that gave rise to the article.
In Lavukaleve, the presentational article has the paradigm shown in Table 7.10 (Terrill 2003:
53). Its paradigm is similar to the “irregular” paradigm of the adjective ru ‘big’ (Terrill 2003: 53).
Both share some forms with the “regular” paradigm for gender and number marking that is used
by predicates including verbs, but also attributive adjectives (Terrill 2003: 252). In contrast to
the presentational article, the recognitional article in Lavukaleve inflects according to this regular
paradigm (Terrill 2003: 83), given in Table 7.11. The definite article in Lavukaleve, on the other
hand, has yet another paradigm shown in Table 7.12 (Terrill 2003: 91); nevertheless, the plural
form va is similar to the plural forms of the other paradigms. With respect to the plural marker
of the presentational article in Lavukaleve, we can thus note that it has its own paradigm which
most probably goes back to a more conservative paradigm of predicates which it shares with the
adjective ‘big’. The paradigm for predicates in the current language is given in Table 7.13. It is
similar but not identical to the paradigm of the presentational article and of the adjective ‘big’.
What this means for the number marking on the presentational article in Lavukaleve is that was
most probably already a part of the paradigm of the article’s source element.

Table 7.10: Lavukaleve presentational article Table 7.11: Lavukaleve ru ‘big’

sg du pl sg du pl
m roa roal m rua rual
f ro rol rovo f ru rul ruvalev
n roge rogel n rugi rugil

220
 Table 7.12: Lavukaleve definite article Table 7.13: Lavukaleve predicate
sg du pl sg du pl
m na nala m -m -mal
f la va f -a -aol -v
n ga gala n -∅ -gel

The presentational and specific articles in Hausa and Biak, and the nonspecific article in Biak go
back to pronouns and have preserved their number marking as well. In Hausa, the presentational
article has the following forms: wani (m.sg), wata (f.sg), and wa(ɗan)su (pl). The distinctive sec-
ond segment of these three forms consists of the personal pronoun forms, ta ‘her’, a reconstructed
form *ni ‘him’, and su ‘them’. Also in Biak, both the exclusive-specific and the nonspecific articles
originate from pronouns that distinguish number. The specific article developed from personal
pronouns (van den Heuvel 2006: 72) of the forms iya (3sg), siya (3pl.an), and naya (3pl.inan),
giving way to the article exponents (i/ya in the singlar and sya / na in the plural). Also the non-
specific article o with its plural form no originates from a pronoun set translated as ‘some’, with
the singular form oso and the plural form (n)ono (van den Heuvel 2006: 72).
To conclude, this section showed two important points. Firstly, it provided more evidence for
the findings in section 6.1.3.2, showing that the availability of plural forms of articles from the
indefinite domain is mainly conditioned by the properties of their source elements. The same holds
for the incompatibility with articles from the indefinite domain with plural nouns. We saw that
articles from the indefinite domain are significantly less compatible with plural nouns than articles
from the definite domain in the world’s languages. This effect can also be accounted for by the
source elements of the articles and not the properties of the articles themselves: articles from the
indefinite domain often originate from the numeral ‘one’ which is often said to be incompatible
with plural nouns. The second related finding of this section is that cases which appear to be
innovations of plural marking on the article could be shown to go back to pervasive number
markers in the nominal domain with little competing other number markers that can occur on
various markers and not only on the noun. This suggests that the form properties of articles are
indeed heavily restricted by their source elements, and only in combination with these types of
broad markers, the articles can have innovative forms.

221
7.4.3 Definite articles co-occurring with demonstratives and possessives
In this section, I show that definite articles from the world’s languages do not show a strong
overall tendency towards or against co-occurring with demonstratives and possessives. Only for
article demonstrative pairs that occur on the same side of the noun do we find this effect.
Going back to Bloomfield (1933), many syntactic approaches include articles in a syntactic
category of determiners that also contains demonstratives and pronouns. Since Abney (1987),
determiners are often viewed as heads of determiner phrases and as taking noun phrases as com-
plements. This syntactic analysis of articles and demonstratives and adnominal possessives entails
that they are expected to be mutually exclusive, since they all belong to the same syntactic cat-
egory of determiners. Given the patterns in especially Germanic languages, this is intuitive in
that it captures the mutual exclusivity of articles, demonstratives, and adnominal possessives13 in
many languages. Nevertheless, in many other languages, these markers can co-occur.
Haspelmath (1999) proposes a functional account to explain the crosslinguistic tendency of
absent definite articles with possessives based on form-frequency correspondence. This approach
captures the co-occurrence as well, since it allows other factors to override coding efficiency.
As for demonstratives, Haspelmath (1999: 227) notes that “an economic-motivation account is
straightforward in the case of demonstratives, which always make an NP definite”, but that this is
not necessarily the case with possessives, since possessives do not entail definiteness, but strongly
correlate with definite referents.
With both demonstratives and possessives, definite articles can either systematically co-occur
or be blocked, and in a small number of languages their co-occurrence is allowed but not nec-
essary in all contexts. Examples (31) to (36) show these three occurrence patterns for article-
demonstrative pairs: co-occurring, optionally co-occurring, and not co-occurring. These three
patterns are given for markers that occur on different sides of the noun in (31) to (33), and on
the same side of the noun in (34) to (36). What is treated as optionally co-occurring follows the
grammatical descriptions; it may not correspond to optionality but variation conditioned by other
factors.

• co-occurring article and demonstrative on different sides of the noun

(31) na maku man
art:def house dem:dist
‘that house’ Guazacapán Xinkan (Rogers 2010: 209)
• optionally co-occurring article and demonstrative on different sides of the noun
(32) a. hinne baleiichiweé
dem:prox story
‘this story’
13 I only consider pronominal adnominal possessives, excluding genitive noun phrases.

222
 b. hinne iisáaskshee-sh
dem:prox young.man-art:def
‘this young man’
Crow (Graczyk 2007: 70)
• not co-occurring article and demonstrative on different sides of the noun
(33) a. fa rea kaaroo
art:def sg car
‘the car’ Yapese (Jensen 1977: 161)
b. (*fa) rea booch neey
art:def sg boat dem:prox
‘this boat’ Yapese (Jensen 1977: 152-154)
• co-occurring article and demonstrative on the same side of the noun
(34) pɔ̄ bì *(l’ɛ̀)
thing dem:prox art:def
‘this thing’ Beng (Paperno 2014: 98)
• optionally co-occurring article and demonstrative on the same side of the noun
(35) a. tanta dia itu
aunt art:def dem:dist
‘that aunt’ Papuan Malay (Kluge 2017: 281)
b. setang itu
evil.spirit dem:dist
‘that evil spirit’ Papuan Malay (Kluge 2017: 296)
• not co-occurring article and demonstrative on the same side of the noun
(36) a. tēnei mahi
dem:prox work
‘this task/work’ Maori (Harlow 2012: 217)
b. te mahi
art:def work
‘the work’ Maori (Harlow 2012: 138)

Accordingly, examples (37) to (42) show the same co-occurrence patterns of definite articles and
possessives.

• co-occurring article and possessive on different sides of the noun

(37) ú cí-ge
3sg.m tree-art:def
‘his tree’ Supyire (Carlson 1994: 66)
• optionally co-occurring article and possessive on different sides of the noun
(38) a. ʔádo-sá lí
2sg-gen house
‘your house’

223
 b. kasu-sá lí-re
1sg-gen house-art:def
‘my house’
Lepcha (Plaisir 2006: 58)
• not co-occurring article and possessive on different sides of the noun
(39) a. kani ‘uru
house art:def
‘the house’ Ute (Givón 2011: 283)
b. ‘uwayas kani
3sg.gen house
‘her/his house’ Ute (Givón 2011: 205)
• co-occurring article and possessive on the same side of the noun
(40) juu 7ix-7ukxtin
art:def 3sg-boss
‘his boss’ Huehuetla Tepehua (Kung 2007: 358)
• optionally co-occurring article and demonstrative on the same side of the noun
(41) a. ʔəm-nəhiʔ
2sg-name
‘your name’ Zoque (Johnson 2000: 383)
b. bi ʔəm-cek
art:def 2sg-belly
‘your belly’ Zoque (Johnson 2000: 290)
• not co-occurring article and demonstrative on the same side of the noun
(42) a. naké=inga
dog=art:def
‘the dog’
b. naké-ké=ke
dog-3sg.nf.gen=3sg.nf.dat
‘his dog’ Skou (Donohue 2004: 150)

Figure 7.3 below shows the co-occurrence patterns for the definite articles from the sample with
demonstratives (on the left) and possessives (on the right).14 For both markers, we see the num-
ber of languages in which the definite article co-occurs, can co-occur, and cannot co-occur with
demonstratives and possessives.15
14 For details on the languages and the co-occurrence patterns with demonstratives and possessives, see Tables A.35
to A.38 in Appendix A.
15 I did not include all languages with definite articles here; in some cases, e.g. because in the position of the demon-

strative or possessive may vary or the markers may consist of two elements on both sides of the noun. For details
on the counts and languages, see Tables A.27, A.29, and A.31 in the appendix.

224
 30

20

20 15
co.dem co.poss
not.co−occur

count
not.co−occur
count

can.co−occur can.co−occur
10
co−occur co−occur
10

0 0

other.side same.side other.side same.side

def.art + demonstrative def.art + possessive

Figure 7.3: Co-occurrence of definite articles with demonstratives and possessives

We first turn to the distributions of definite articles with demonstratives. Figure 7.3 shows two
important things: firstly, we see a higher number of articles and demonstratives that occur on the
same side of the noun than pairs on different sides. This follows a larger general trend, shown
in Dryer (to appear) for the demonstrative, the numeral, and the adjective with respect to their
possible orders with the noun. He shows that although we find a variety of patterns, languages
generally tend to have all determiners occur on the same side of the noun (called “intercategorical
harmony”).
Secondly, a we find a clear difference in co-occurrence trends between article-demonstrative
pairs on the same side and on different sides of the noun. If the two elements occur on different
sides of the noun, they do not show any tendency with respect to their occurrence if we treat
obligatory and optional co-occurrence as a single strategy as opposed to not co-occurring. For
article-demonstrative pairs that occur on the same side of the noun, on the other hand, we see
a trend against co-occurring definite articles and demonstratives. Based on a Bayes proportion
test, this difference between co-occurring and not co-occurring article-demonstrative pairs on
the same side is not statistically significant (cf. Table A.30). However, this does not mean that
the relative position of the article and the demonstrative does not have any influence at all. The
overlap of the estimated confidence interval is very small; a larger sample and a more detailed
distinction of the order of the demonstrative and the article might be able to consolidate the trend
found in the present data.
The co-occurrence patterns of the definite article with adnominal possessives differ in a crucial
way. On the one hand, we do not see a trend for definite articles and adnominal possessives to

225
occur on either the same side or different sides of the noun. On the other hand, there is no apparent
difference between the preferences for either of the three occurrence patterns based on the relative
sides that the two elements occur on. It seems that in either case, there is a slightly higher number
of articles that do not co-occur with possessives than articles that co-occur with them. However, if
we treat co-occurrence and optional co-occurrence as a single strategy, then there is no statistically
significant trend for either strategies, i.e. co-occurrence vs. no co-occurrence (cf. Table A.32).
These findings show that the co-occurrence restrictions of definite articles with demonstra-
tives and with possessives differ. For demonstratives, we saw their co-occurrence with definite
articles was disfavored if they occurred on the same side of the noun. Possessives, on the other
hand, showed no trend with respect to their co-occurrence with definite articles. It was argued
that there should be a stronger co-occurrence restriction for definite articles with demonstratives
than for definite articles with possessives because demonstratives automatically make the referent
of the noun definite. Possessive markers, on the other hand, often mark the referent as definite
as well, but can also be combined with an indefinite referent (e.g. a friend of mine) and strongly
correlate with but do not imply the definiteness of the referent (cf. Haspelmath 1999). To account
for variation, Haspelmath also notes that economy is not the only motivation that can lead to
the conventionalization in grammar. Explicitness, for instance, can outrank economy, resulting
in patterns with co-occurring definite articles and possessives. Haspelmath (1999: 236-239) pro-
poses two potential scenarios that can lead to patterns with definite articles and possessives that
do not co-occur, arguing that the only attested scenario is what he calls the “inhibition scenario”.
In this scenario, the language has an older possessive construction; an emerging definite article
extends its use to various definite contexts but does “not become established in possessed NPs,
however, because definiteness marking is to some extent redundant in this environment” (Haspel-
math 1999: 237). This leads to the complementary distribution of definite articles and possessives.
Haspelmath also argues that languages with a prenominal possessive are more likely to avoid
its co-occurrence with definite articles for reasons of economy. Thus, this would mean that in
contrast to the relative positions of the article and the possessive which did not show any effect
for their co-occurrence, the absolute position of the possessive (independently of the article posi-
tion) could influence the co-occurrence of definite article with possessives. Figure 7.4 shows the
distribution patterns for definite articles with demonstratives (on the left) and possessives (on the
right) depending on the absolute position of the demonstrative and the possessive, i.e. whether
these occur in a prenominal or postnominal position (note that the position of the definite article
is not taken into account here).16
16 For the languages considered, see Tables A.35 to A.38 in Appendix A.

226
 30

20

15 20
co.dem co.poss
not.co−occur

count
not.co−occur
count

can.co−occur can.co−occur
10
co−occur co−occur
10

0 0

pre post pre post

def.art + demonstrative def.art + possessive

Figure 7.4: Position of demonstratives and possessives and co-occurrence with definite articles

For demonstratives, we can see that the absolute position does not have an impact on their co-
occurrence with definite articles. Possessives, on the other hand, show a tendency towards occur-
ring prenominally. As for the co-occurrence with definite articles, the data in Figure 7.4 suggests
that the proportion of co-occurring possessives and definite articles is higher with prenominal pos-
sessives, which is the opposite of the claim in Haspelmath (1999). However, a Bayes proportion
test does not confirm a statistically significant difference between the proportions of co-occurring
articles and possessives for prenominal vs. postnominal possessives (cf. Tables A.33 and A.34 in
Appendix A). This does not exclude an effect of the absolute position for possessives and their
co-occurrence with definite articles; it only shows that this effect may not be large enough to be
observable due to other more influential factors.
Returning to the observed difference in the co-occurrence restrictions of definite articles with
demonstratives and possessives shown in Figure 7.3, I propose that this difference is not due to
a semantic effect of the article. Rather, the difference in behaviour can be accounted for by the
diachronic relation between the article and the demonstrative. It has was argued that demonstra-
tives represent the most common source for definite articles (e.g. Greenberg 1978b, Diessel 1999,
De Mulder & Carlier 2011). Hence, definite articles that originate from demonstratives should
be more frequent crosslinguistically than definite articles that develop from possessives. This
diachronic relation, and not a synchronically motivated syntactic constraint, can explain why def-
inite articles and demonstratives are more frequently attested on the same side of the noun in the
first place. Given that the demonstrative and the emerging definite article can most probably not

227
be distinguished formally in the initial stages17 , speakers are likely to interpret them as the same
element. As the demonstrative and the emerging definite article formally and functionally over-
lap to a high degree in the contexts in which demonstratives are used, their co-occurrence can
be assumed to very rare.18 Since possessives present a less frequent source for definite articles,
the simultaneous overlap of form and function (in this case, in possessive-like constructions) is
expected to be much more negligeable across languages. This corresponds to the findings shown
in Table 7.3, namely no trend based on the relative positions of the possessives and the definite
article across languages.
That this co-occurrence restriction of the definite article and the demonstrative is not semanti-
cally19 motivated is supported by the overall numbers of co-occurring vs. complementary article-
demonstrative (19 co-occurring vs. 23 complementary) and article-possessive (20 co-occurring vs.
21 complementary) pairs. These numbers does not differ considerably. This means that the trend
against co-occurrence with definite articles cannot be due to a inherent property of the definite
article. Languages with co-occurring definite articles and possessives but complementary articles
and demonstratives show evidence in support of this diachronic explanation. One such case can
be found in Ch’ol. Example (43) shows that both the demonstrative and the diachronically related
definite article precede the noun; the definite article does not co-occur with the demonstrative in
Ch’ol. On the other hand, (44) shows that the definite article can occur together with possessive
marking on the noun. As we see in example (45), also its source element, the proximal demon-
strative, is attested together with adnominal possessive markers. Thus, the diachronic relation of
a demonstrative and a definite article does not only restrict the co-occurrence of the demonstra-
tive and the definite article, but it can also license the use of the definite article with possessive
markers in the language.

(43) a. ili xkaläl-ob

dem:prox girl-pl
‘these girls’ Ch’ol (Vázquez Alvarez 2011: 85)
b. li xk’aläl
art:def girl
‘the girl’ Ch’ol (Vázquez Alvarez 2011: 355)
17 This is not mean to imply that the definite article necessarily becomes formally distinct from its demonstrative
source; this should rather be viewed as a correlation with processes of phonetic reduction that are commonly
treated together as grammaticalization.
18 Future research on the diachronic relation and development of definite articles is needed to show how unlikely

or even excluded cases of form-identical (emerging) definite articles and demonstratives in the same position are;
Hungarian being a case in point. The distal demonstrative az co-occurs with the definite article a; with vowel initial
nouns, the form of the article is identical to the demonstrative resulting in the demonstrative-article string of az az.
19 Indirectly, semantic reasons may be responsible for trend that demonstratives develop into definite articles much

more frequently than possessives.

228
(44) li k-alo’bil
art:def poss:1-son
‘my son’ Ch’ol (Vázquez Alvarez 2011: 327)
(45) ili la=k-tyaty=i
dem:prox pl.in-poss:1-lord-pm
‘(this) our Lord’ Ch’ol (Vázquez Alvarez 2011: 341)

Summing up, I showed in this section that the co-occurrence of demonstratives and definite arti-
cles is only disfavored if they occur on the same side of the noun. The co-occurrence of definite
articles and possessives, on the other hand, was shown to be independent from their relative and
absolute positions in the noun phrase. Again, diachrony instead of synchronic functional mo-
tivations could be shown to explain this difference in the compatibility of definite articles with
demonstratives and possessives.

7.4.4 Articles co-occurring with other articles

Another common assumption based on articles in many European languages is that articles cannot
co-occur with other articles within languages, as in English:

(46) a. *the a boy

b. *a the boy

Given that articles are defined as primarily referential markers, it does not seem surprising from a
functional perspective that different articles do not co-occur. However, the use of articles is rarely
exclusively conditioned by their referential function, and we find co-occurring articles in many
languages for a number of reasons.
The combination of two articles can be semantically motivated. (Heine 1997: 70-77) reports
that partitives are often expressed by the combination of definite and indefinite articles. Most
examples provided, however, involve a partitive marker and not an indefinite article that combines
with the definite article, or they showed both articles that co-occurred in contexts without nouns.
A less controversial example of partitives expressed by the combination of two articles is found in
Akan (Amfo 2010). Example (47) below shows the noun nkɔrɔfo ‘people’ with both an anaphoric
and an exclusive-specific article, and we can see in the translation that this combination has a
partitive interpretation.

(47) [[nkɔrɔfo no] bí] ka-a sɛ wɔ.re-m-pene

people art:ana art:exspec say-compl comp 3pl-prog-neg-agree
‘Some of the people said that they will not agree.’ Akan (Amfo 2010: 1795)

229
The two articles in Akan can also be combined in the reverse order and scope relation; in that case,
the noun is indicated as having a non-identifiable specific referent, but the additional anaphoric
article indicates that the referent has been mentioned before in the discourse. Example (48) illus-
trates this:

(48) [[maame bí] no] a ɔ-bɛ-hwehwɛ-ɛ wo no

woman art:exspec art:ana rel 3sg-mp-look.red-compl you cd
‘That (certain) woman who came looking for you …’ Akan (Amfo 2010: 1797)

Another type of of article combinations is the combination of two articles for other, nonsemantic
motivations. That means that only one of the articles is interpreted semantically, while the one no
longer contributes to the interpretation of the referent. We find such a combination of a definite
and an indefinite article in Albanian, involving “conflicting” requirements with respect to the use
of articles that result in the presence of both articles. As was shown in section 6.1.1, the definite
article in Albanian is suffixed to the noun, while the indefinite article consists of the free form
një preceding the noun. In most contexts, the indefinite article cannot combine with a noun that
is at the same time marked with the definite article. Only if an indefinite article combines with
an adnominal possessive, the noun must also contain the definite article because the possessive
marker formally requires the latter to be present. Example (49) shows a noun with a definite
referent being marked by a possessive. The noun libri ‘the book’ contains the definite article.
As example (50) shows, even if the indefinite article is used to mark the referent of the noun as
non-identifiable, the definite article remains obligatory.

(49) ky është libr-i im

this is book-art:def poss:1sg
‘This is my book.’ Albanian (Zymberi 2004: 58)
(50) një mik-*(u) im
art.indef friend-art:def poss:1sg
‘a friend of mine’ Albanian (Newmark et al. 1982: 150)

Although it does not correspond to a combination of different articles, the specific and indefinite
articles in Q’anjobal and Awakateko show a similar effect with demonstratives. In both languages,
articles from the indefinite domain occur in contexts in which they are not required semantically.
As was mentioned in section 2.2.4, demonstratives in Q’anjobal follow the noun and usually occur
with a classifier that precedes the noun. In case a classifier is not available for semantic reasons,
for instance, because there is no classifier it can combine with, the specific article is used instead,
as is shown in examples (51) to (53) below. The same combination of an indefinite article with a
demonstrative is attested in Awakateko, as in (54) and (55).

230
(51) jun ora tu’
art:indef hour dem:dist
‘at that hour’ Q’anjobal (prim. data)
(52) jun b’eqan ti’
art:indef moment dem:prox
‘at this moment’ Q’anjobal (prim. data)
(53) jun tz’uy ti’
art:indef bag dem:prox
‘this bag’ Q’anjobal (prim. data)
(54) jun tq’uj-ä
art:indef moment-dem:prox
‘this moment’ Awakateko (prim. data)
(55) jun p’uch-ä
art:indef flower-dem:prox
‘this flower’ Awakateko (prim. data)

What examples (51) to (55) from Q’anjobal and Awakateko show is that the article can occur in
contexts that are in conflict with its referential functions, similarly to the example of the German
definite article occurring together with a non-definite superlative form shown in example (27).
Thus, syntactic factors can influence the distribution of articles in a significant way, overriding
their referential functions.

7.5 Summary
This chapter was concerned with crosslinguistic trends and variation with respect to article types
and certain properties of articles.
One important finding in section 7.1 concerned the areal bias on article types. Anaphoric arti-
cles were shown to be very frequent in the languages of Australia, probably due to the lack of third
person pronouns that are formally distinct from demonstratives in many Australian languages.
Section 7.2 discussed two article types that are hardly attested. For non-anaphoric articles,
i.e. definite articles that cannot encode anaphoric referents, I proposed two scenarios that lead
to their development. However, it could be shown that most of these articles are not entirely
blocked in anaphoric contexts, but rather compete with another article, being used under certain
circumstances. Therefore, it is questionable whether non-anaphoric definite articles exist in the
strict sense. The other article type that was argued to be absent in the world’s languages was the
generic article. By showing that various other article types are used to mark generic referents
in the world’s languages, I argued that the marking of generic referents cannot be a primary

231
function of these articles. In addition, due to the lack of any evidence that would point towards
the existence of dedicated generic articles, I concluded that generic articles do not exist in the
world’s languages.
In section 7.3, I proposed a typology of article types and article systems on the basis of the
numbers of referent types that the articles and systems encode. The crosslinguistic frequency
distributions of different article types showed a strong trend towards the encoding of definite
referents, independently of single article types. This trend was reflected in the frequency distribu-
tions of article systems: we found more systems that singled out definite referents compared to
both specific and nonspecific referents. The distributions of article systems also pointed towards
crosslinguistic gaps, all involving nonspecific articles. Nonspecific articles do not occur without a
specific article within the same language, and articles cannot coexpress nonspecific with definite
referents unless they also encode specific referents.
The last section of this chapter, section 7.4, dealt with the interaction of certain article types
and three selected constructions in which articles occur but are not required as referential markers.
For definite articles in superlative constructions, it could be shown that their semantic properties
generally allow and require the use of the article. However, I argued that this should only be
taken as a strong correlation and not as an inherent property of definite articles, since the article
alone is not generally able to function as a superlative marker, for instance, if the language does
not have a separate superlative marker. The properties of the article’s source element, however,
were shown to be a much more pervasive factor to influence the distribution of articles. I showed
this for the compatibility of articles from the indefinite domain with plural nouns. A similar
kind of explanation could be applied to the co-occurrence of definite articles with demonstratives
and possessives. The diachronic relation between definite articles and demonstratives accounted
for the observation that their co-occurrence is not generally restricted, and only disfavored if
both the article and the demonstratives occurred on the same side of the noun. I argued that
this is not necessarily due to a synchronically motivated syntactic restriction, but points towards
demonstratives as sources of the articles, blocking the use of the latter if there was a high degree of
form and function overlap at an earlier stage of the definite article. Since definite articles originate
from possessives much less frequently, this explanation could capture that no effect with respect
to the co-occurrence of articles and possessives was observed. These three patterns showed that
even though the distribution of articles in contexts which do not require them referentially may
be correlated to their semantic properties, the properties of their source elements are a much more
influential factor.

232
Chapter 8

Conclusions

8.1 Main findings

In this dissertation, I defined articles as a crosslinguistic category and discussed various properties
of in articles on the basis of a 99-language sample.
The first important step towards comparing articles across languages is a language-independent
definition of articles, their delimitation from other similar categories, and the distinction of dif-
ferent article types. A detailed solution to these three issues was provided in Chapters 2 to 4,
respectively. In Chapter 2, I defined articles as referential markers that occur systematically with
nouns to indicate their referential function. Chapter 3 proposed criteria to delimit articles in the
definite domain from demonstratives, articles in the indefinite domain from the numeral ‘one’,
and nonspecific articles from NPIs. It is difficult to draw a line between demonstratives and arti-
cles from the definite domain because of their frequent diachronic relation. The same holds for
the numeral ‘one’ and articles from the indefinite domain. However, I showed how we can define
cut-off points on function scales to distinguish between articles and these categories. Chapter 4
established referential functions as different referent types; the major types being definite refer-
ents Rde f , specific referents Rspec , nonspecific referents Rnspec , and generic referents Rgen . Based
on the referential properties of articles that we find in the world’s languages, I defined 9 article
types according to their coding of different combinations of referent types. The article types are:
definite (def), anaphoric (ana), recognitional (recog), exclusive-specific (exspec), nonspecific
(nspec), indefinite (indef), presentational (pres), inclusive-specific (inspec), and referential (ref).
Figure 8.1 sketches the relation between the main referent types and the 9 article types. Anaphoric
and recognitional articles are subtypes of definite articles since they encode subtypes of definite
referents, namely anaphoric and recognitional ones, respectively. Indefinite and presentational
articles do not differ in their referential function, but the occurrence of presentational articles

233
is more restricted by the discourse prominence of the referent than the occurrence of indefinite
articles.

referent
types definite Rde f specific Rspec nonspecific Rnspec

def, ana, recog exspec nspec

article
types

indef, pres
inspec
ref

Figure 8.1: Overview of referent types and article types

In Chapter 5, I illustrated the use of these article types in detail by discussing examples from
various languages of the world. Concerning definite articles, I showed that there are markers
which fulfill all criteria except for the coding of deictic and/or absolutely unique referents. Since
their distribution was otherwise shown to be similar to other definite articles that can occur with
these referent types, I argued that the coding of deictic and absolutely unique referents is not a
necessary criterion for definite articles. Using the example of the definite article in Kaqchikel, I
traced back the incompatibility with deictic referents to the source element of the article. On the
basis of data from a number of Mayan, Salishan, and Wakashan languages, I also showed that
nonspecifc articles can develop from irrealis markers in the verbal domain. I showed that the
irrealis marker was first applied to nominal predicates and then to nouns in argument positions
in the scope of an irrealis predicate scoping over the referent. In some languages, this led to the
emergence of a nonspecific article.
In Chapter 6, I discussed the morphological properties of articles focusing on the inflection of
articles for gender, number, and case. I showed a global trend against inflection in the world’s
languages. Two main factors were discussed that restrict the inflection of articles: on the one
hand, the frequent marking of number and case on nouns that was shown to restrict the marking
of these categories on the article, reflected by a crosslinguistic tendency of articles against the
marking of number and case as opposed to gender. On the other hand, the inflectional behaviour
of articles largely depends on the properties of their source elements. For articles which originate
from demonstratives, pronouns, and the numeral ‘one’, I showed that they either retained the in-
flectional distinctions present in their source elements or lost certain formal oppositions resulting
in the loss of inflection. Inflectional innovations on the articles, however, proved to be extremely
rare. The only 4 examples of innovative inflectional markers were plural markers on articles from

234
the indefinite domain. However, it was shown in Chapter 7 that these examples are not clear
cases of innovative inflection either. The diachrony of the article could also be shown to play a
role for certain syntactic properties of articles, namely the co-occurrence restrictions of definite
articles with demonstratives, discussed in section 7.4.3. I argued that the frequent diachronic re-
lation between definite articles and demonstratives in contrast to possessive markers can explain
why we find a positional co-occurrence restrictions for definite articles and demonstratives, but
not for definite articles and possessives.
Section 7.3 dealt with article systems, i.e. the attested and unattested combinations of articles
that we find in single languages. On the basis of the possible article combinations encoding defi-
nite, specific, and nonspecific referents, I proposed a typology of article systems, summarized in
Tables 8.1 to 8.4.

Table 8.1: Systems encoding 1 referent type Table 8.2: Systems encoding 2 referent types
referent type I II referent type III IV V VI

definite def definite def

inspec
specific exspec specific exspec spec
indef
nonspecific nonspecific nspec

N systems 50 2 N systems 7 2 (1) 15

Table 8.3: Systems encoding 3 referent types Table 8.4: Unattested systems
referent type VII VIII IX X referent type *I *II *III *IV

definite def def definite def.nspec def def.nspec

inspec
specific exspec ref specific exspec
indef
nonspecific nspec nspec nonspecific nspec def.nspec nspec def.nspec

N systems 12 3 3 4

Tables 8.1 to 8.3 show the article systems that encode 1, 2, and 3 referent types, respectively.
The numbers in the last rows indicate how many instances of such systems were found in the
99-languages sample. The number given for the system of type V in Table 8.2 corresponds to
Q’anjobal, which was not part of the sample but which has such an article system. In Table 8.4,
we see the unattested article systems.
This typology of article systems, the frequency distributions of the systems in the sample as
well as the unattested systems allowed for the following universals with respect to articles and ar-
ticle systems in the languages of the world: Firstly, articles whose main referential function it is to

235
code generic referents are not attested. Secondly, articles that coexpress definite and nonspecific
referents without encoding specific referents are not attested either. Thirdly, nonspecific articles
only occur in article systems with another article encoding specific referents (either inclusive-
specific or exclusive-specific). Thirdly, articles show a preference to encode definite versus specific
and nonspecific referents. This is manifested in the overall higher number of articles that encode
definite referents, but also in the lower number of articles that co-express definite and specific
referents (inclusive-specific) versus articles that coexpress specific and nonspecific referents (in-
definite). Fourthly, there are generally no article systems with more than one article encoding
the same referent type. Exceptions to this were shown to be systems with definite and anaphoric
articles, (e.g. Jamsay), systems with definite and recognitional articles, (e.g. Lavukaleve), and the
Basque system with an inclusive-specific as well as a presentational article.

8.2 The path forward

There are still many more things that could be said about articles. A question that was only
touched upon but which deserves a closer and more systematic look concerns article systems
and to what extent the presence of one article can influence the emergence, stability, or loss of
a another article in the same language. What this study showed is that systems of functionally
overlapping articles are generally not attested. The other interaction of certain article types within
single article systems discussed was the nonspecific article that depended on the presence of a
specific article in the language. These are synchronic observations; but one could also look at
such interactions from a diachronic perspective: does the presence or absence of a certain article
type favor or disfavor the emergence or loss of another article type?
This study made a few observations concerning the areal distributions of certain article types:
a high number of indefinite and presentational articles in Eurasia, a high number of definite arti-
cles in Africa, and a strong bias towards anaphoric articles in Australia. These areal biases make
it plausible that articles in different languages did not develop independently. It is left for further
research to determine whether language contact or family-internal developments due to shared
grammatical properties may have led to a preference for or against certain types of articles. How-
ever, these three areas in which major trends towards a certain article type was observed all have
one very dominant language family in terms of number of languages: Indo-Aryan, Niger-Congo,
and Pama-Nyungan. It may thus be possible that the development of certain article types mainly
occurred in the languages of these large families because of certain shared grammatical properties,
but could also influence unrelated languages by language contact. This may be related to the find-
ings of Leiss (2000), who showed for a number of Indo-European languages that the emergence
of articles was caused by the loss of aspectual distinctions in the verbal system. However, on a

236
broader crosslinguistic scale, this correlation between articles and aspectual distinctions does not
seem to be very prominent, which also points towards a common development within a language
family, probably facilitated by language contact.
I showed for two article types that certain grammatical properties can favor their emergence.
The languages of Australia suggested that the lack of formally distinct third person pronouns,
probably paired with a complex demonstrative system, favors the development of anaphoric arti-
cles, while a number of languages in North America suggested that nonspecific articles can emerge
if the language has a semantically broad irrealis marker and a less strict distinction between verbs
and nouns, allowing the marker to occur on nouns as well.
Thus, there are still many unanswered questions especially with respect to the factors that
condition the presence or absence of articles. However, this study gave an answer to some of
the questions. I presented the article types that are attested in the world’s languages, provided a
detailed overview of their referential and morphosyntactic properties, and showed how different
articles can combine within languages, proposing a typology of article systems.

237
Language Index

Aghul, 148–149 Finnish, 31

Akan, 109, 122–123 German, 21, 28, 35, 38, 53, 111, 165, 169–170
Albanian, 170–171, 230 Gooniyandi, 117
Arawak, 110, 123–124 Halkomelem, 156–160
Armenian, 33, 203 Hausa, 37, 221
Arrernte, 118 Hebrew, 57
Awakateko, 30, 230 Hidatsa, 138
Ayutla Mixe, 54 Indonesian, 51, 199
Bare, 28, 114 Irish, 110
Basque, 204, 217 Jamsay, 200
Baure, 204 Kaqchikel, 97–100, 106–108
Bemba, 27, 29, 125–129, 166, 204 Kayardild, 116
Beng, 223 Khwarshi, 35
Biak, 186–187, 221 Koasati, 32, 115
Blackfoot, 187 Konso, 40
Bonan, 143–145, 205 Lakota, 135–136, 187, 201
Bullom So, 33, 203 Lango, 111, 146–147
Carib, 142–143, 217 Lavukaleve, 37, 109, 220
Ch’ol, 138, 165–166 Lepcha, 223
Chatino, 149–150 Limbum, 49, 112–114
Chinese, 21 Logba, 147, 148
Coptic, 41 Ma’di, 200
Crow, 29, 40, 110, 138–222 Macedonian, 52
Dime, 109 Makah, 139
Diyari, 171–172 Mangarrayi, 115
Ejagham, 50 Manx, 215
English, 26, 40, 59 Maori, 223
Fering, 202 Mokpe, 34, 39, 100–106, 109, 214

238
Niuean, 160 Sri Lanka Malay, 34
Nivaclé, 22 Sumu, 33
Nuuhchahnulth, 139 Supyire, 166–168, 223
Nyaturu, 35 Tarahumara, 110
Oko, 204, 217 Tepehua, 33, 224
Palula, 58, 124, 125, 205 Tikuna, 48, 57
Papuan Malay, 111, 223
Tongan, 129–135, 187
Polish, 62
Turkish, 33
Q’anjobal, 30, 42, 59, 119–122, 137–138, 187,
Tz’utujil, 43, 140–142
230
Udmurt, 21
Rapa Nui, 30, 152–155
Ughele, 31
Runyankore, 115
Sabanê, 161 Ute, 33, 168–169, 224
Sheko, 39, 204 Wardaman, 115
Siar Lak, 187 Xinkan, 222
Skou, 224 Yankuntjatjara, 118
Slovenian, 36 Yapese, 223
Spanish, 164 Zoque, 214, 224

239
Appendix A

Article properties (99-languages sample)

Table A.1: Distribution of article types in the 6 macroareas

article type Africa Eurasia Papunesia Australia North America South America
definite 15 8 10 1 10 3
anaphoric 6 1 / 11 2 3
recognitional / / 3 2 / /
exclusive-specific 3 2 3 / 2 2
nonspecific / / 3 / 3 /
indefinite / 9 3 / 2 1
presentational 3 6 1 / 1 2
inclusive-specific 1 1 2 / 2 /
referential / / 1 / 1 2

240
 Table A.2: Article properties (Africa)
language article position art.infl nom.cat nom.infl number gender case
Akan ana post / num+gen num noun / /
exspec post / num+gen num noun / /
Amharic def post num+gen num+gen num art+noun art /
Bemba inspec pre / num+gen num noun / /
Beng def post / num / / / /
Bullom So def post num+gen num+gen num art+noun art /
Dime def post / num+gen+case num+case noun / noun
Gaahmg def post / num+case num+case noun / noun
Hausa pres pre num+gen num+gen num art+noun art /
ana post / num+gen num noun / /
Jamsay def post / num+gen num noun / /
ana pre / num+gen num noun / /
Konso def post num num+gen num art+noun / /

241
pres post num+gen num+gen num art+noun art /
Koyra Chiini def post / num num noun / /
Lango ana post / num num noun / /
exspec post / num num noun / /
Limbum ana post / num+gen num noun / /
Logba def post / num+gen num noun / /
pres post num+gen num+gen num art+noun art /
Mokpe def pre num+gen num+gen num art+noun art /
Moloko def post / num num noun / /
Oko def post num num num art+noun / /
exspec post num num num art+noun / /
Runyankore ana pre / num+gen num noun / /
Sandawe def post num+gen num+gen / art art /
Sheko def post gen num+gen / / art /
Supyire def post num+gen num+gen / art art /
 Table A.3: Article properties (Eurasia)
language article position art.infl nom.cat nom.infl number gender case
Aghul pres pre / num+case num+case noun / noun
Albanian def post num+case num+gen+case / art / art
indef pre / num+gen+case case / / noun
Ao ana post / / / / / /
Armenian def post / num+case num+case noun / noun
indef pre / num+case case / / noun
Basque inspec post num+case num+case / art / art
pres post num+case num+case / art / art
Bonan indef post / num+case case / / noun
Dameli pres pre / num+case case / / noun
Domari exspec post / num+gen+case / / / /
German def pre num+gen+case num+gen+case num art+noun art art
indef pre gen+case num+gen+case num noun art art

242
Hebrew def pre / num+gen num noun / /
Hungarian def pre / num+case num+case noun / noun
indef pre / num+case case / / noun
Irish def pre / num+gen num noun / /
Kharia pres pre / num+case case / / noun
Lepcha def post / num num noun / /
Neo-Aramaic pres pre gen num+gen / / art /
Palula def pre / num+gen+case num+case noun / noun
exspec pre / num+gen+case case / noun
Persian indef post / num+case num+case noun / noun
Rajbanshi pres pre gen num+gen+case case / art noun
Tamil indef pre / num+case case / / noun
Turkish indef pre / num+case case / / noun
Sri Lanka Malay indef flexible / num+case case / / noun
 Table A.4: Article properties (Papunesia)
language article position art.infl nom.cat nom.infl number gender case
Abui recog post / num num noun / /
exspec post / num num noun / /
Agta def pre / num num noun / /
exspec post / num / / / /
Alamblak def pre / num num noun / /
Biak exspec post num+gen num+gen / art art /
def post num+gen num+gen / art art /
nspec post num num+gen / art / /
Lavukaleve def post num+gen num+gen num art+noun art /
pres post num+gen num+gen num art+noun art /
recog post num+gen num+gen num art+noun art /
Maori def pre num num+case case art / noun
Menya def post / num / / / /
Oksapmin indef post / num / / / /

243
recog post / num / / / /
def post / num / / / /
Papuan Malay def post num num / art / /
Rapa Nui ref pre / num+case num+case noun / noun
Siar Lak inspec pre num+gen num+gen / art art /
nspec pre gen num+gen / / art /
Skou def post / gen / / / /
Sye indef pre / num / / / /
Tongan inspec pre / num+case num+case noun / noun
nspec pre / num+case num+case noun / noun
Yapese def pre / num num noun / /
indef pre / num / / / /
 Table A.5: Article properties (Australia)
language article position art.infl nom.cat nom.infl number gender case
Arrernte ana post case case / / / art
Bardi ana pre / case case / / noun
Diyari def pre num+gen+case num+gen+case case art art art+noun
Duunidjawu ana pre / case case / / noun
Gooniyandi recog pre / num+case case / / noun
Guragone ana pre gen gen / / art /
Jingulu ana pre num+gen+case num+gen+case case art art art+noun
Mangarrayi ana pre / num+gen+case case / / noun
Martuthunira ana pre case num+case num+case noun / art+noun
Wardaman ana post / num+gen+case num+case noun / noun
Warrwa ana pre case num+case / / / art
Wubuy ana post num+gen num+gen num art+noun art /

244
Yankuntjatjara recog post / num+case case / / noun
Yolŋu ana pre num num+gen+case num+case art+noun / noun
 Table A.6: Article properties (North America)
language article position art.infl nom.cat nom.infl number gender case
Blackfoot inspec post num+gen num+gen / art art /
nspec post / num+gen / / / /
Chatino pres pre / / / / / /
Chimariko def post / / / / / /
Ch’ol def pre / num+gen num noun / /
indef pre gen num+gen / / art /
Crow def post / num / / / /
exspec post / num / / / /
nspec post / num / / / /
Cupeño def pre num num+case num+case art+noun / noun
Halkomelem ref pre gen num+gen num noun art /
Koasati ana post / num+case num+case noun / case
Lakota def post / num / / / /

245
exspec post num num / art / /
nspec post num num / art / /
Nuuchahnulth inspec post / num num noun / /
Sumu (Mayangna) ana post / num num noun / /
Tarahumara def pre / num / / / /
Tepehua def pre / num / / / /
indef pre / num / / / /
Ute def post num+gen+case num+gen+case num art+noun art art
Xinkan def pre / num+gen num noun / /
Zoque def pre / num num noun / /
 Table A.7: Article properties (South America)
language article position art.infl nom.cat nom.infl number gender case
Apinayé def post / num num noun / /
Arawak def pre num+gen num+gen num art+noun art /
exspec pre / num+gen / / / /
Asheninka Perené pres pre / num / / / /
Baré (Barawana) ana pre gen num+gen / / art /
Baure ref pre gen num+gen num noun art /
Carib indef pre num num / art / /
Cubeo ana pre / num+gen num noun / /
Epeena Pedee ana pre / num+case num+case noun / noun
Karajá exspec post / / / / / /
Parecis pres pre / num num noun / /
Sabanê ref post / / / / / /

246
Wayuu def post num+gen num+gen num art+noun art /
 Table A.8: Article inflection by area
area inflected total
Africa 12 28
Eurasia 7 27
Australia 8 14
Papunesia 10 26
North America 7 23
South America 5 13

Table A.9: Bayes proportion test for article inflection by area

proportion estimated
area inflected : total confidence interval
Africa 0.43 [0.26, 0.61]
Eurasia 0.26 [0.12, 0.44]
Australia 0.57 [0.32, 0.79]
Papunesia 0.38 [0.22, 0.57]
North America 0.30 [0.15, 0.50]
South America 0.38 [0.18, 0.63]

Table A.10: Article inflection by article position

position inflected total
preposed 22 62
postposed 27 68

Table A.11: Bayes proportion test for article inflection by article position
proportion estimated
position inflected : total confidence interval
preceding 0.40 [0.29, 0.51]
following 0.35 [0.25, 0.48]

247
 Table A.12: Article inflection by article type
article type inflected total
anaphoric 8 23
recognitional 1 5
definite 19 47
presentational 7 13
inclusive-specific 3 6
exclusive-specific 3 12
nonspecific 3 6
indefinite 3 15
referential 2 4

Table A.13: Bayes proportion test for article inflection by article type
proportion estimated
article type inflected : total confidence interval
anaphoric 0.35 [0.19, 0.55]
recognitional 0.20 [0.018, 0.59]
definite 0.40 [0.28, 0.55]
presentational 0.54 [0.29, 0.77]
inclusive-specific 0.50 [0.18, 0.82]
exclusive-specific 0.25 [0.072, 0.51]
nonspecific 0.50 [0.19, 0.81]
indefinite 0.20 [0.059, 0.44]
referential 0.50 [0.16, 0.86]

248
 Table A.14: Inflectional behaviour of different articles within single languages

language article type article inflection language article type article inflection
Logba def / Abui recog /
pres num+gen exspec /
Lango ana / Tongan inspec /
exspec / nspec /
Hausa pres num+gen Yapese def /
ana / indef /
Konso def num Biak exspec num+gen
pres num+gen def num+gen
nspec num
Oko def num
exspec num Agta def /
exspec /
Jamsay def /
ana / Siar Lak inspec num+gen
nspec gen
Akan ana /
exspec / Oksapmin indef /
recog /
Hungarian def /
def /
indef /
Blackfoot inspec num+gen
Albanian def num+case
nspec /
indef /
Crow def /
Armenian def /
exspec /
indef /
nspec /
German def num+gen+case
Ch’ol def /
indef gen+case
indef gen
Palula def /
Tepehua def /
exspec /
indef /
Basque inspec num+case
Lakota def /
pres num+case
exspec num
Lavukaleve def num+gen nspec num
pres num+gen
Arawak def num+gen
recog num+gen
exspec /

249
 Table A.15: Bayes proportion test for article inflection
proportion estimated
category on article : as category confidence interval
number 0.13 [0.067, 0.20]
case 0.16 [0.070, 0.31]
gender 0.50 [0.34, 0.62]

Table A.16: Bayes proportion test for noun inflection

proportion estimated
category on noun : total confidence interval
number 0.61 [0.50, 0.72]
case 0.75 [0.59, 0.86]

250
 Table A.17: Distribution of article systems
system I (def)
Alamblak, Amharic, Ao, Apinayé, Arrente, Bardi, Bare, Beng, Bullom So, Chi-
mariko, Cubeo, Cupeño, Dime, Diyari, Duunidjawu, Epena Pedee, Gaahmg,
Gooniyandi, Guragone, Hebrew, Irish, Jamsay, Jingulu, Koasati, Koyra Chiini,
Lepcha, Limbum, Mangarrayi, Maori, Martuthunira, Menya, Mokpe, Moloko,
Papuan Malay, Runyankore, Sandawe, Sheko, Skou, Sumu, Supyire, Tarahu-
mara, Ute, Wardaman, Warrwa, Wayuu, Wubuy, Xinkan, Yankuntjatjara,
Yolŋu, Zoque
N = 50
system II (exspec)
Domari, Karaja
N=2
system III (def+exspec)
Abui, Agta, Akan, Arawak, Lango, Oko, Palula
N=7
system IV (inspec)
Bemba, Nuuchahnulth
N=2
system V (exspec+nspec)
/
system VI (indef)
Aghul, Asheninka Perené, Bonan, Carib, Chatino, Dameli, Kharia, Neo-
Aramaic, Parecis, Persian, Rajbanshi, Sri Lanka Malay, Sye, Tamil, Turkish
N = 15
system VII (def+indef)
Albanian, Armenian, Ch’ol, German, Hausa, Hungarian, Konso, Lavukaleve,
Logba, Oksapmin, Tepehua, Yapese
N = 12
system VIII (inspec+nspec)
Blackfoot, Siar Lak, Tongan
N=3
system IX (def+exspec+nspec)
Biak, Crow, Lakota
N=3
system X (ref)
Baure, Halkomelem, Rapa Nui, Sabanê
N=4

251
Table A.18: Bayes proportion test for different referent types expressed by articles
proportion estimated
type expressed type expressed : total expressed confidence interval
definite 0.49 (85 : 173) [0.42, 0.57]
specific 0.29 (50 : 173) [0.22, 0.36]
nonspecific 0.22 (38 : 173) [0.16, 0.29]
generic / (0 : 173) [1.4e-07, 0.017]

Table A.19: Bayes proportion test for 2-referent-types systems

proportion estimated
article type article type : total confidence interval
inclusive specific 0.08 (2 : 24) [0.015, 0.25]
indefinite 0.63 (15 : 24) [0.41, 0.78]

Table A.20: Bayes proportion test for 3-referent-types systems

proportion estimated
article type article type : total confidence interval
inclusive specific 0.14 (3 : 22) [0.042, 0.31]
indefinite 0.55 (12 : 22) [0.35, 0.74]

252
 Table A.21: Definite articles and plural nouns
language with pl marking plural language with pl marking plural
Supyire YES YES Lavukaleve YES YES
Logba YES no Maori YES YES
Dime YES no Yapese YES no
Gaahmg YES no Papuan Malay YES YES
Konso YES YES Biak YES YES
Oko YES YES Agta YES no
Sandawe YES YES Oksapmin YES no
Amharic YES YES Skou YES no
Moloko YES no Menya YES no
Sheko YES no Diyari YES YES
Jamsay YES no Crow YES no
Beng YES no Ute YES YES
Mokpe YES YES Tarahumara YES no
Bullom So YES YES Ch’ol YES no
Koyra Chiini YES no Zoque YES no
Hungarian YES no Tepehua YES no
Albanian YES YES Cupeño YES YES
Irish YES no Xinkan YES no
Armenian YES no Lakota YES no
German YES YES Arawak YES YES
Hebrew YES no Wayuu YES YES
Palula YES no Apinayé YES no
Lepcha YES no

253
 Table A.22: Anaphoric and recognitional articles and plural nouns
language article type compatible with plural marking plural
Akan anaphoric YES no
Lango anaphoric YES no
Hausa anaphoric YES no
Runyankore anaphoric YES no
Jamsay anaphoric YES no
Limbum anaphoric YES no
Mangarrayi anaphoric YES no
Wardaman anaphoric YES no
Warrwa anaphoric YES no
Jingulu anaphoric YES YES
Wubuy anaphoric YES YES
Yolŋu anaphoric YES YES
Koasati anaphoric YES no
Sumu anaphoric YES no
Bare anaphoric YES no
Cubeo anaphoric YES no
Lavukaleve recognitional YES YES
Oksapmin recognitional YES no
Gooniyandi recognitional YES no
Yankuntjatjara recognitional YES no

254
 Table A.23: Indefinite and presentational articles with plural nouns
language article type compatible with plural marking plural
Hungarian indefinite no no
Albanian indefinite no no
Armenian indefinite no no
German indefinite no no
Persian indefinite YES no
Turkish indefinite no no
Sri Lanka Malay indefinite no no
Bonan indefinite no no
Tamil indefinite no no
Sye indefinite no no
Yapese indefinite no no
Oksapmin indefinite YES no
Ch’ol indefinite no no
Tepehua indefinite no no
Carib indefinite YES YES
Logba presentational YES YES
Hausa presentational YES YES
Konso presentational YES YES
Basque presentational YES YES
Neo-Aramaic presentational no no
Kharia presentational no no
Dameli presentational no no
Rajbanshi presentational no no
Aghul presentational YES no
Lavukaleve presentational YES YES
Chatino presentational no no
Asheninka Perené presentational no no

255
Table A.24: Exclusive specific and nonspecific articles with plural nouns
language article type compatible with plural marking plural
Lango exclusive-specific YES no
Oko exclusive-specific YES YES
Akan exclusive-specific YES no
Palula exclusive-specific no no
Domari exclusive-specific no no
Biak exclusive-specific YES YES
Agta exclusive-specific YES no
Crow exclusive-specific YES no
Lakota exclusive-specific YES YES
Arawak exclusive-specific no no
Karajá exclusive-specific no no
Tongan nonspecific YES no
Biak nonspecific YES YES
Siar Lak nonspecific no no
Blackfoot nonspecific no no
Lakota nonspecific YES YES

Table A.25: Bayes proportion test for compatibility with plural

proportion estimated
article domain with plural : total confidence interval
definite 0 [2.2e-06, 0.044]
indefinite 0.56 [0.41, 0.70]

Table A.26: Bayes proportion test for plural marking

proportion estimated
article domain marking plural : with plural confidence interval
definite 0.34 [0.23, 0.46]
indefinite 0.58 [0.37, 0.78]

256
 Table A.27: Co-occurrence of definite articles and demonstratives
language co-occurrence position language co-occurrence position
Supyire YES other side Lepcha no same side
Logba no same side Lavukaleve no same side
Dime YES other side Maori no same side
Gaahmg YES same side Alamblak no same side
Konso no same side Yapese no other side
Oko no same side Papuan Malay possible same side
Sandawe possible other side Biak no same side
Amharic no other side Oksapmin no same side
Moloko no same side Skou YES same side
Sheko no same side Diyari no same side
Jamsay no same side Crow possible other side
Beng YES same side Ute no other side
Mokpe YES same side Tarahumara no same side
Bullom So YES same side Ch’ol no same side
Koyra Chiini YES same side Zoque no same side
Hungarian YES same side Tepehua YES same side
Albanian possible other side Chimariko YES other side
Irish YES other side Xinkan possible other side
Armenian YES other side Arawak no same side
German no same side Wayuu YES other side
Palula no same side Apinayé no same side

257
 Table A.28: Co-occurrence of definite articles and possessives
language co-occurrence position language co-occurrence position
Supyire YES other side Lavukaleve no other side
Logba no other side Maori YES same side
Dime possible other side Alamblak no same side
Konso no same side Yapese no other side
Oko no other side Agta no other side
Sandawe YES other side Oksapmin possible other side
Amharic no same side Skou no same side
Moloko no same side Menya possible other side
Sheko possible other side Diyari no same side
Jamsay no other side Crow no other side
Beng no other side Ute no other side
Mokpe YES other side Tarahumara YES same side
Bullom So no same side Ch’ol possible same side
Koyra Chiini YES other side Zoque possible same side
Albanian YES other side Tepehua YES same side
Irish no same side Cupeño possible same side
Armenian YES other side Chimariko no same side
German no same side Lakota YES same side
Hebrew YES other side Arawak YES same side
Palula no same side Apinayé no other side
Lepcha possible other side

Table A.29: Co-occurring definite articles and demonstratives

co-occurrence other side same side

co-occur 6 8
can co-occur 4 1
cannot co-occur 3 20
(can) co-occur 10 9
cannot co-occur 3 20

Table A.30: Bayes proportion test for definite articles and demonstratives
proportion estimated
co-occurrence other side : total confidence interval
(can) co-occur 0.52 [0.32, 0.72]
cannot co-occur 0.13 [0.040, 0.34]

258
 Table A.31: Co-occurring definite articles and possessives
co-occurrence other side same side
co-occur 7 5
can co-occur 5 3
cannot co-occur 10 11
(can) co-occur 12 8
cannot co-occur 10 11

Table A.32: Bayes proportion test for definite articles and possessives
proportion estimated
co-occurrence other side : total confidence interval
(can) co-occur 0.60 [0.40, 0.79]
cannot co-occur 0.48 [0.28, 0.67]

Table A.33: Co-occurring definite articles with pre-/postnominal possessives

co-occurrence (can) co-occur cannot co-occur
prenominal 17 13
postnominal 3 8

Table A.34: Bayes proportion test for definite articles and pre-/postnominal possessives
proportion estimated
position co-occur : total confidence interval
prenominal 0.57 [0.40, 0.74]
postnominal 0.27 [0.086, 0.55]

259
Table A.35: Co-occurrence of the definite article and the demonstrative (i)
on the same side of the noun
language co-occur position of demonstrative
Logba no POST
Gaahmg YES POST
Konso no POST
Oko no POST
Moloko no POST
Sheko no POST
Jamsay no POST
Beng YES POST
Mokpe YES pre
Bulom So YES POST
Koyra Chiini YES POST
Hungarian YES pre
German no pre
Palula no pre
Lepcha no POST
Lavukaleve no POST
Maori no pre
Alamblak no pre
Papuan Malay possible POST
Biak no POST
Oksapmin no POST
Skou YES POST
Diyari no pre
Tarahumara no pre
Chol no pre
Zoque no pre
Tepehua YES pre
Arawak no pre
Apinayé no POST

260
Table A.36: Co-occurrence of the definite article and the demonstrative (ii)
on different sides of the noun
language co-occur position of demonstrative
Supyire YES pre
Dime YES pre
Sandawe possible pre
Amharic no pre
Albanian possible pre
Irish YES POST
Armenian YES pre
Yapese no POST
Crow possible pre
Ute no pre
Chimariko YES pre
Xinkan possible POST
Wayuu YES pre

Table A.37: Co-occurrence of the definite article and the possessive (i)
on the same side of the noun
language co-occur position of possessive
Cupeño possible pre
Zoque possible pre
Arawak YES pre
Chol YES pre
Lakota YES POST
Maori YES pre
Tarahumara YES pre
Tepehua YES pre
Alamblak no pre
Amharic no POST
Bulom So no POST
Chimariko no POST
Diyari no pre
German no pre
Irish no pre
Konso no POST
Moloko no POST
Palula no pre
Skou no POST

261
Table A.38: Co-occurrence of the definite article and the possessive (ii)
on different sides of the noun
language co-occur position of possessive
Supyire YES pre
Dime possible pre
Lepcha possible pre
Menya possible pre
Oksapmin possible pre
Sheko possible pre
Albanian YES pre
Armenian YES pre
Hebrew YES POST
Koyra Chiini YES pre
Mokpe YES POST
Sandawe YES pre
Agta no POST
Apinayé no pre
Beng no pre
Crow no pre
Jamsay no pre
Lavukaleve no pre
Logba no pre
Oko no pre
Ute no pre
Yapese no POST

262
Appendix B

Questionnaire

Compatibility of articles and different nouns types

(1) person (2) other animate (3) place

a. child a. dog a. shop
b. man b. cat b. house
c. woman c. sheep c. market
d. neighbor d. goat d. church
e. teacher e. school
f. kitchen

(4) abstract concept (5) mass concept

a. dream a. salt
b. love b. sand
c. idea c. rice
d. mistake d. food
e. hunger e. water
f. promise f. blood
g. agreement g. oil
h. happiness
i. time
j. miracle

Anaphoric contexts

(6) A: When I arrived, I saw a woman with two children sitting in the kitchen.
B: Did you know the woman?

263
(7) Two friends are sitting in the house, it is already time for lunch.
A: I would like some rice.
B: The rice is in the kitchen.
(8) Friendship became love. Love became hatred.
(9) A: I am thirsty. Do you have some water?
B: Water is in the kitchen.
a. A goes to the kitchen.
A: Could you come? I can’t find the water.
(10) A and B are eating fufu that A prepared.
A: Do you like the fufu?
B: Normally, I don’t like fufu a lot, but your fufu is really good!

Bridging contexts

(11) X bought a book. The author is from Z.

(12) They have just arrived from Z. The plane/bus was five hours late.
(13) I had to get a taxi from the city center. On the way the driver told me buses were not
operating at that time.
(14) I’ve just been to a wedding. The bride was wearing a beautiful dress.
(15) In his/her heart, X is still in love with his/her hometown/village. The people, the food, the
rivers, the mountains, …
(16) X recently invited his/her neighbor when I was there. I really like that guy/man.
(17) X called again. I don’t want to talk to this idiot.
(18) We saw an old building in the village; the roof was totally damaged.
(19) The dance was so beautiful that the chief gave the trainer a gift.

Cataphoric contexts

(20) The company that hired us does not make enough money.
(21) Te man X met last night just called.
(22) the dream you had last night
(23) the love her mother gave to her

264
(24) the idea I just had
(25) the meat you bought yesterday
(26) the sand you find on the beach in Z

Situationally unique

(27) Let’s go to the market.

(28) Where is the church/school in this village?
(29) In country X
The president does p.

Absolutely unique

(30) The moon was very bright last night.

(31) Don’t look at the sun!
(32) The ground is dirty.

Establishing definites

(33) Are the rumors true? Are they really coming to visit us from abroad?
(34) X just told me about this great place. Let’s go and check it out / see.
(35) I know how to find your secrets. I know the tricks.

Deictic and identifiable

(Contexts involve visibility and pointing)

(36) Please, can you hold this bag for a second?

(37) This boy has grown so much since I saw him last.
(38) Look at that flower, it is so beautiful.
(39) These mosquitoes/flies are so annoying!

Deictic and nonidentifiable

(Contexts involve visibility and pointing)

265
(40) A and B are in a boat on the river
A: Look! There is a boy in the water!
(41) A and B are checking their mailbox/traps. B usually never gets letters / catches animals. A:
Look, you just got a letter / caught an animal!
(42) There are some girls waiting for you outside.

Specific

(43) A friend returns from his trip and brings back presents.
I have a present for you.
(44) I know a place where we can study.
(45) Some friends came over to my place last night.
(46) The car belonged to a man I met in Z.

Predicative use

(47) a. This is a beautiful morning.

b. This was a delicious meal.
c. This is a good friend.
(48) He is a farmer/cook/…
(49) They are teachers.
(50) She has become a real woman.
He has become a real man.
(51) It’s a miracle that we found each other in this city.
(52) They are the only family he has.

Identificational use

(53) A and B are in school and see a man they have not met; it could be the new teacher.
A: Do you know that guy?
B: I think he is the new teacher.
(54) This is the truth.
(55) Her new husband is the problem/matter/issue.
(56) He will be the next chief/…

266
Generic use

(57) X come’s to his friend’s house. The friend asks whether there is a special occasion for his visit.
X: Does a friend/brother / Do friends need a reason/purpose/announcement to come to
visit?
(58) The prince is talking to the king
P: I thought a king / kings could do whatever he wants?
(59) X belongs to group XX. X is late.
X: An XX man / person is never late.
(60) A teacher is never wrong.

(61) A craftsman/cook receives a compliment for his work.

C: A craftman is always pleased to hear his work is appreciated.
C: A cook is always pleased to hear his food was good.
(62) Pain is in the mind.
(63) I do not like cars/bikes/trains.
I hate dogs.
(64) Visitors / Strangers are always welcome in this house.

Nonspecific contexts

(65) Do you have a (any) pen / Can you give me a (any) pen?
(66) I need to find an (any) architect/cook who is as good as I was.
(67) I wish I had a (any) car.

Associative constructions

(68) the President of Country X

(69) the chief of the tribe
(70) the mistake of the teacher
(71) I want to know the name of your employer.
(72) And you shall never see the light of another day.
(73) But the power of the ring alone could not be undone.

267
Demonstratives and contrastive use

Contexts include visibility and pointing

(74) This house is big; that house is smaller.

(75) I like being in this village. The people in that village are less friendly.
(76) Who did you ask? That girl or this boy?

Instruments

(77) by bike/car/bus/train
(78) with force
(79) by hand

Other conventionalized expressions

(80) be in bed
(81) be made of X
(82) by chance
(83) out of control

Existential construction

(84) There are so many mosquitoes.

(85) Is there a place to eat in this town/village?
(86) There is a church at the end of the road.

268
Bibliography

Abney, Steven Paul. 1987. The English noun phrase in its sentential aspect. Cambridge, MA: MIT
dissertation.
Aikhenvald, Alexandra Y. 1995. Bare. München: Lincom Europa.
Aikhenvald, Alexandra Y. 2008. Classifiers: a typology of noun categorization devices. Oxford: Ox-
ford University Press.
Álvarez, José. 2017. Compendio de la gramática de la lengua wayuu.
Amfo, Nana Aba Appiah. 2010. Indefiniteness Marking and Akan Bi. Journal of Pragmatics 42(7).
1786–1798.
Anderson, Stephen R. 1985. Inflectional Morphology. In, vol. 3 (Language Typology and Syntactic
Description), 151–201. Cambridge: Cambridge University Press.
Anderson, Stephen R. & Edward L. Keenan. 1985. Deixis. In Language Typology and Syntactic.
Description III: Grammatical categories and the lexicon, 259–308. Cambridge: Cambridge Uni-
versity Press.
Andonian, Hagop. 2005. Beginner’s Armenian. New York: Hippocrene Books.
Antunes de Araujo, Gabriel. 2004. A grammar of Sabanê. A Nambikwaran language. Utrecht: LOT.
Ariel, Mira. 1988. Referring and Accessibility. Journal of Linguistics 24(1). 65–87.
Ariel, Mira. 1990. Accessing Noun-Phrase Antecedents. London: Routledge.
Ariel, Mira. 2001. Accessibility Theory: An Overview. In Ted J.M. Sanders, Joost Schilperoord &
Wilbert Spooren (eds.), Text Representation: Linguistic and Psycholinguistic Aspects, 29–87. Am-
sterdam: Benjamins.
Arkhangelskiy, Timofey & Maria Usacheva. 2016. Functions of the 3sg Possessive in Beserman Ud-
murt: Corpus Analysis. St Petersburg.
Arkoh, Ruby. 2011. Semantics of Akan bi and nu. Vancouver: University of British Columbia dis-
sertation.
Arkoh, Ruby & Lisa Matthewson. 2013. A familiar definite article in Akan. Lingua 123. 1–30.
Asiimwe, Allen. 2014. Definiteness and specificity in Runyankore-Rukiga. Stellenbosch: Stellen-
bosch University dissertation.

269
Atindogbe, Gratien. 2013. A grammatical sketch of Mòkpè (Bakweri) (Bantu A20). Kyoto: The Center
for African Area Studies of Kyoto University.
Atoyebi, Joseph Dele. 2010. A reference grammar of Oko. Köln: Köppe.
Austin, Peter. 2011. A Grammar of Diyari. 2nd edn. Cambridge: Cambridge University Press.
Bååth, Rasmus. 2014. Bayesian First Aid: A package that implements Bayesian alternatives to the
classical *.test functions in R. In UseR! 2014 - the International R User Conference.
Baker, Carl Lee. 1970. Double negatives. Linguistic Inquiry 1(2). 169–186.
Baker, Carl Lee. 1973. Definiteness and Indefiniteness in English. Bloomington, IN: Indiana Univer-
sity Linguistics Club.
Barros Brandão, Ana Paula. 2014. A reference grammar of Paresi-Haliti (Arawak). Austin, TX: Uni-
versity of Texas dissertation.
Beck, David. 2013. Unidirectional flexibility and the noun–verb distinction in Lushootseed. In Jan
Rijkhoff & Eva Helena van Lier (eds.), Flexible word classes typological studies of underspecified
parts of speech, 185–220. Oxford: Oxford University Press.
Bhat, D.N.S. 2004. Pronouns. Oxford: Oxford University Press.
Bhat, D.N.S. 2013. Third Person Pronouns and Demonstratives. In Matthew S. Dryer & Martin
Haspelmath (eds.), The World Atlas of Language Structures Online. Leipzig: Max Planck Institute
for Evolutionary Anthropology.
Bisle-Müller, Hansjörg. 1991. Artikelwörter im Deutschen. Tübingen: Niemeyer.
Blackings, Mairi & Nigel Fabb. 2003. A Grammar of Ma’di. Berlin: De Gruyter Mouton.
Błaszczak, Joanna. 1999. -kolwiek Pronouns in Polish: Negative Polarity Items or Free Choice
Items or Both? In Beiträge der Europäischen Slavistischen Linguistik (POLYSLAV) 2. Vol. 4 (Die
Welt Der Slaven. Sammelbände), 51–62. München: Sagner.
Błaszczak, Joanna. 2003. (Negative) polarity items in Polish: Towards a uniform account. In (A)Symmetrien
- (A)Symmetries. Beiträge zu Ehren von Ewald Lang - Papers in Honor of Ewald Lang, 25–56.
Tübingen: Stauffenburg.
Bloomfield, Leonard. 1933. Language. New York: Holt.
Bobaljik, Jonathan David. 2012. Universals in Comparative Morphology. Cambridge, MA: MIT Press.
Borella, Cristina. 2004. Gramática da lı́ngua Sararé/Katithaulhu. Amsterdam.
Bowern, Claire Louise. 2004. Bardi verb morphology in historical perspective. Cambridge, MA: Har-
vard University dissertation.
Brown, Jason & James J. Thompson. 2013. The emergence of determiner clisis in Upriver Halkomelem.
Northwest Journal of Linguistics 7(1.1). 1–13.
Brown, R. McKenna et al. 2006. ¿La ütz awäch? Introduction to Kaqchikel Maya language. Austin,
TX: University of Texas Press.

270
Bruce, Leslie P. 1979. A grammar of Alamblak (Papua New Guinea). Canberra: Australian National
University dissertation.
Brugmann, Karl. 1904. Die Demonstrativpronomina der indogermanischen Sprachen: eine bedeu-
tungsgeschichtliche Untersuchung. Leipzig: Teubner.
Buchholz, Oda & Wilfried Fiedler. 1987. Albanische Grammatik. Leipzig: Vlg Enzyklopädie.
Buechel, Eugen. 1939. A grammar of Lakhota. The language of the Teon Sioux Indians. Rosebud:
Rosebud Educational Society.
Bühler, Karl. 1934. Sprachtheorie: die Darstellungsfunktion der Sprache. Jena: Fischer.
Butterworth, George. 1998. What is special about pointing in babies? In Francesca Simion &
George Butterworth (eds.), The development of sensory, motor and cognitive capacities in early
infancy. From perception to cognition, 171–190. Hove: Psychology Press.
Campbell, Lyle. 2012. Typological characteristics of South American indigenous languages. In
Lyle Campbell & Verónica Grondona (eds.), The Indigenous Languages of South America. A
Comprehensive Guide, vol. 2 (The World of Linguistics), 259–330. Berlin: De Gruyter Mouton.
Carlson, Robert. 1994. A Grammar of Supyire. Berlin: De Gruyter Mouton.
Carnap, Rudolf. 1947. Meaning and necessity: a study in semantics and modal logic. Chicago: Uni-
versity of Chicago Press.
Chacon, Thiago Costa. 2012. The phonology and morphology of Kubeo: The documentation, theory,
and description of an Amazonian language. Manoa: University of Hawaii dissertation.
Chen, Ping. 2003. Indefinite determiner introducing definite referent: a special use of ‘yi ‘one’ +
classifier’ in Chinese. Lingua 113. 1169–1184.
Chierchia, Gennaro. 1998. Reference to Kinds across Languages. Natural Language Semantics 6(4).
339–405.
Chierchia, Gennaro. 2013. Logic in grammar: polarity, free choice, and intervention. Oxford: Oxford
University Press.
Childs, George Tucker. 2011. A grammar of Mani. Berlin: De Gruyter Mouton.
Chonay Chonay, Ana Marı́a. 2006. Runuk’ulem pa rub’eyal rutz’ib’axik ri kaqchikel ch’ab’äl [Gramática
normativa del idioma maya Kaqchikel]. Guatemala: Academia de Lenguas Mayas de Guatemala.
Christophersen, Paul. 1939. The Articles: A Study of their Theory and Use in English. Copenhagen:
Einar Munksgaard.
Churchward, C. Maxwell. 1985. Tongan grammar. Tonga: Vava’u Press.
Clark, Herbert H. 1996. Using Language. Cambridge: Cambridge University Press.
Clark, Herbert H. & Susan E. Brennan. 1987. Grounding in communication. In Lauren B. Bresnick,
John M. Levine & Stephanie D. Teasley (eds.), Perspectives on socially shared cognition, 127–149.
Washington: American Psychological Association.

271
Clark, Herbert H. & Susan E. Haviland. 1977. Comprehension and the given-new contract. In Roy
O. Freedle (ed.), Discourse Production and Comprehension, 1–40. Norwood, NJ: Ablex.
Coffin, Edna Amir & Shmuel Bolozky. 2005. A reference grammar of modern Hebrew. Cambridge:
Cambridge University Press.
Cohen, Diana. 1998. A grammatical description of Tarahumara. Arlington: The University of Texas
at Arlington dissertation.
Comrie, Bernard. 1989. Language universals and linguistic typology. 2nd edn. Oxford: Blackwell.
Corbett, Greville G. 2003. Gender. Cambridge: Cambridge University Press.
Corbett, Greville G. & Sebastian Fedden. 2016. Canonical Gender. Journal of Linguistics 52(3). 495–
531.
Coupe, Alec R. 2007. A grammar of Mongsen Ao. Berlin: De Gruyter Mouton.
Courtz, Hendrik. 2008. A Carib grammar and dictionary. Toronto: Magoria Books.
Cristofaro, Sonia. 2008. Grammatical categories and relations: Universality vs. language-specificity
and construction-specificity. Language and Linguistics Compass 3(1). 441–479.
Croft, William. 1991. Syntactic categories and grammatical relations: the cognitive organization of
information. Chicago: University of Chicago Press.
Croft, William. 2003. Typology and universals. 2nd edn. Cambridge: Cambridge University Press.
Crowley, Terry. 1998. An Erromangan (Sye) grammar. Honolulu: University of Hawaii Press.
Csúcs, Sándor. 1998. Udmurt. In Daniel Abondolo (ed.), The Uralic languages, 276–304. London:
Routledge.
Csúcs, Sándor. 2003. A votják tárgyról. Nyelvtudományi Közlemények 100. 126–136.
Culicover, Peter W. 1999. Syntactic nuts: hard cases, syntactic theory, and language acquisition. Ox-
ford: Oxford University Press.
Cunha de Oliveira, Christiane. 2005. The language of the Apinajé people of Central Brazil. Eugene,
OR: University of Oregon dissertation.
Curl, Tracy S. 1999. The Lakhota definite articles and topic marking. Kansas Working Papers in
Linguistics 24(2). 1–16.
Danielsen, Swintha. 2007. Baure: An Arawak language of Bolivia. Leiden: CNWS Publications.
Davidson, Matthew. 2002. Studies in Southern Wakashan (Nootkan) grammar. Buffalo, NY: Univer-
sity at Buffalo dissertation.
Dayley, Jon P. 1985. Tzutujil grammar. Vol. 107 (Publications in linguistics). Berkeley, CA: Univer-
sity of California Press.
de Blois, Kornelis Frans. 1970. The augment in the Bantu languages. Africana Linguistica 4. 85–
165.

272
De Mulder, Walter & Anne Carlier. 2011. The Grammaticalization of Definite Articles. In Bernd
Heine & Heiko Narrog (eds.), The Oxford Handbook of Grammaticalization, 1st edn. Oxford:
Oxford University Press.
de Rijk, Rudolf P. G. 2008. Standard Basque: a progressive grammar. Cambridge, MA: MIT Press.
Dench, Alan Charles. 1994. Martuthunira. A language of the Pilbara region of Western Australia.
Canberra: Australian National University.
Diessel, Holger. 1999. Demonstratives: Form, function and grammaticalization. Vol. 42 (Typological
Studies in Language). Amsterdam: Benjamins.
Diessel, Holger. 2013a. Distance contrasts in demonstratives. In Matthew S. Dryer & Martin Haspel-
math (eds.), The World Atlas of Language Structures Online. Leipzig: Max Planck Institute for
Evolutionary Anthropology.
Diessel, Holger. 2013b. Where does language come from? Some reflections on the role of deictic
gesture and demonstratives in the evolution of language. Language and Cognition 5(2-3). 239–
249.
Dixon, Robert Malcolm Ward. 1980. The languages of Australia. Cambridge: Cambridge University
Press.
Donohue, Mark. 2004. A grammar of the Skou language of New Guinea. Singapore: Singapore Uni-
versity.
Doornenbal, Marius. 2009. A grammar of Bantawa. Utrecht: LOT.
Dorvlo, Kofi. 2008. A Grammar of Logba (Ikpana). Utrecht: LOT.
Dryer, Matthew S. 1989. Article-noun order. Chicago Linguistic Society 25. 83–97.
Dryer, Matthew S. 1997. Are grammatical relations universal? In Joan Bybee, John Haiman &
Sandra A. Thompson (eds.), Essays on language function and language type: Dedicated to T.
Givón, 115–143. Amsterdam: Benjamins.
Dryer, Matthew S. 2013a. Definite Articles. In Matthew S. Dryer & Martin Haspelmath (eds.), The
World Atlas of Language Structures Online. Leipzig: Max Planck Institute for Evolutionary An-
thropology.
Dryer, Matthew S. 2013b. Indefinite Articles. In Matthew S. Dryer & Martin Haspelmath (eds.),
The World Atlas of Language Structures Online. Leipzig: Max Planck Institute for Evolutionary
Anthropology.
Dryer, Matthew S. 2014. Competing Methods for Uncovering Linguistic Diversity: The Case of
Definite and Indefinite Articles (Commentary on Davis, Gillon, and Matthewson). Language
Language 90(4). 232–249.
Dryer, Matthew S. to appear. On the order of demonstrative, numeral, adjective and noun. Lan-
guage.

273
Dryer, Matthew S. & Martin Haspelmath (eds.). 2013. WALS Online. Leipzig: Max Planck Institute
for Evolutionary Anthropology.
Dum-Tragut, Jasmine. 2009. Armenian. Amsterdam: Benjamins.
Dunn, Michael John. 1999. A grammar of Chukchi. Canberra: Australian National University dis-
sertation.
É. Kiss, Katalin & Orsolya Tánczos. 2017. From possessor agreement to object marking: the gram-
maticalization path of the Udmurt -(j)ez suffix. Budapest.
Ebert, Karen. 1971. Zwei Formen des bestimmten Artikels. In Dieter Wunderlich (ed.), Probleme
und Fortschritte der Transformationsgrammatik, 159–174. München: Hueber.
Edygarova, Svetlana. 2009. Attributive Possession in Udmurt. Linguistica Uralica 45. 101–118.
Eilan, Naomi et al. (eds.). 2005. Joint Attention: Communication and Other Minds: Issues in Philoso-
phy and Psychology. Oxford: Oxford University Press.
Engel, Ulrich. 1988. Deutsche Grammatik. Heidelberg: Groos.
Epstein, Richard. 2002. The Definite Article, Accessibility, and the Construction of Discourse Ref-
erents. Cognitive Linguistics 12(4). 333–378.
Eriksson, Mats. 2009. Referring as interaction: On the interplay between linguistic and bodily
practices. Journal of Pragmatics 41(2). 240–262.
Erkü, Feride & Jeanette K. Gundel. 1987. The pragmatics of indirect anaphors. In The Pragmatic
Perspective. Selected Papers from the 1985 International Pragmatics Conference, 533–545. Ams-
terdam: Benjamins.
Evans, Nicholas. 1995. A grammar of Kayardild: With historical-comparative notes on Tangkic.
Vol. 15 (Mouton Grammar Library). Berlin: De Gruyter Mouton.
Evans, Nicholas & Stephen C. Levinson. 2009. The myth of language universals: Language diversity
and its importance for cognitive science. 32(5). 429–448.
Everett, Daniel Leonard. 1992. A lingua pirahã e a teoria da sintaxe: descrição, perspectivas e teoria.
Campinas (SP): Editora da UNICAMP.
Everett, Daniel Leonard. 2005. Cultural constraints on grammar and cognition in Pirahã. Current
Anthropology 46(4). 621–646.
Everett, Daniel Leonard. 2009. Pirahã culture and gammar: a sesponse to some criticisms. Lan-
guage 85(2). 405–442.
Fauconnier, Gilles. 1975. Pragmatic scales and logical structure. Linguistic Inquiry 6(3). 353–375.
Fauconnier, Gilles. 1994. Mental Spaces. Cambridge, MA: Cambridge University Press.
Fauconnier, Gilles. 1997. Mappings in thought and language. Cambridge: Cambridge University
Press.
Fauconnier, Gilles. 2007. Mental Spaces. In Dirk Geeraerts & Hubert Cuyckens (eds.), The Oxford
Handbook of Cognitive Linguistics, 351–376. Oxford: Oxford University Press.

274
Fillmore, Charles J. 1967. On the syntax of preverbs. Glossa 1. 91–125.
Fillmore, Charles J. 1975. An Alternative to Checklist Theories of Meaning. In Proceedings of the
First Annual Meeting of the Berkeley Linguistics Society, 123–131. Berkeley, CA: BLS.
Fillmore, Charles J. 1976. Frame Semantics and the Nature of Language. Annals of the New York
Academy of Sciences 280(1). 20–32.
François, Alexandre & Martine Vanhove. 2008. Semantic maps and the typology of colexification:
Intertwining polysemous networks across languages. In From polysemy to semantic change:
Towards a typology of lexical semantic associations, 163–215. Amsterdam: Benjamins.
Frank, Michael C. et al. 2008. Number as a cognitive technology: Evidence from Pirahã language
and cognition. Cognition 108(3). 819–824.
Fransen, Margo Astrid Eleonora. 1995. A grammar of Limbum, a Grassfields Bantu language spoken
in the north-west province of Cameroon. Amsterdam: Vrije Universiteit te Amsterdam disserta-
tion.
Frantz, Donald G. 1991. Blackfoot Grammar. Toronto: University of Toronto Press.
Fraurud, Kari. 2001. Possessives with extensive use. A source of definite articles? In Irene Baron,
Michael Herslund & Finn Sørensen (eds.), Dimensions of possession, 243–267. Amsterdam: Ben-
jamins.
Frege, Gottlob. 1892. Über Sinn Und Bedeutung. Zeitschrift für Philosophie und philosophische
Kritik, Neue Folge 100. 25–50.
Fried, Robert Wayne. 2010. A grammar of Bao’an Tu, a Mongolic Language of Northwest China.
Buffalo, NY: University of Buffalo, State University of New York dissertation.
Friesen, Dianne. 2017. A grammar of Moloko. Berlin: Language Science Press.
Frostad, Benedicte Haraldstad. 2012. A grammar of Ughele. Utrecht: LOT.
Galloway, Brent. 1993. A grammar of Upriver Halkomelem. Berkeley, CA: University of California
Press.
Ganenkov, Dmitry, Timur Maisak & Solmaz Merdanova. 2009. Diskursivnaja anafora v agul’skom
jazyke. In NeFestšrift: Stat’i v podarok (k jubileju A. E. Kibrika).
Gerdts, Donna B. & Thomas E. Hukari. 2004. Determiners and transitivity in Halkomelem texts.
In Donna B. Gerdts & Lisa Matthewson (eds.), Studies in Salish linguistics in honor of M. Dale
Kinkade, 151–171. Missoula, MT: University of Montana.
Gerland, Doris. 2014. Definitely not possessed? Possessive suffixes with definiteness marking func-
tion. In Frames and Concept Types (Studies in Linguistics and Philosophy), 269–292. Cham:
Springer.
Ghomeshi, Jila, Ileana Paul & Martina Wiltschko (eds.). 2009. Determiners: universals and variation.
OCLC: 717784871. Amsterdam: Benjamins.

275
Giacon, John. 2014. A grammar of Yuwaalaraay and Gamilaraay. Canberra: Australian National
University dissertation.
Giannakidou, Anastasia. 1997. The landscape of polarity items. Groningen: University of Groningen
dissertation.
Giannakidou, Anastasia. 1998. Polarity sensitivity as (non)veridical dependency. Vol. 23 (Linguistik
Aktuell / Linguistics Today). Amsterdam: Benjamins.
Giannakidou, Anastasia. 2017. Polarity in the semantics of natural language. In Oxford Research
Encyclopedia of Linguistics. Oxford: Oxford University Press.
Givón, Talmy. 1969. Studies in Chibemba and Bantu grammar. Los Angeles, CA: University of
California dissertation.
Givón, Talmy. 1978. Definiteness and Referentiality. In Joseph Harold Greenberg, Charles Albert
Ferguson & Edith A. Moravcsik (eds.), Universals of Human Language: Syntax, 292–330. Stan-
ford, CA: Stanford University Press.
Givón, Talmy. 1980. The Binding Hierarchy and the Typology of Complements. Studies in Lan-
guage 4. 333–377.
Givón, Talmy. 1981. On the development of the numeral ’one’ as an indefinite marker. Folia Lin-
guistica Historica 15(2.1). 35–54.
Givón, Talmy. 1984. Syntax: a functional-typological introduction. Volume 1. Amsterdam: Benjamins.
Givón, Talmy. 2011. Ute reference grammar. Amsterdam: Benjamins.
Glinert, Lewis. 2005. Modern Hebrew: An essential grammar. London: Routledge.
Goddard, Cliff. 1985. A grammar of Yankunytjatjara. Alice Springs: Inst. for Aboriginal Develop-
ment.
Göksel, Aslı & Celia Kerslake. 2005. Turkish: a comprehensive grammar. London: Routledge.
Gordon, Peter. 2004. Numerical cognition without words: evidence from Amazonia. Science 306(5695).
496–499.
Gorrie, Colin, Alexandra Kellner & Diane Massam. 2010. Determiners in Niuean. Australian Jour-
nal of Linguistics 30(3). OCLC: 4637709385, 349–365.
Graczyk, Randolph. 2007. A grammar of Crow (Studies in the Native Language of the Americas).
Bloomington, IN: University of Nebraska Press.
Green, Rebecca. 1995. A Grammar of Gurr-goni (North Central Amhem Land). Canberra: Australian
National University dissertation.
Greenberg, Joseph Harold. 1978a. Generalizations about numeral systems. In Joseph Harold Green-
berg (ed.), Universals of human language. Vol. 3: Word structure, 249–95. Stanford: Stanford
University Press.
Greenberg, Joseph Harold. 1978b. On language: selected writings of Joseph H. Greenberg. Suzanne
Kemmer & Keith Denning (eds.). Stanford, CA: Stanford University Press.

276
Gundel, Jeanette K., Nancy Hedberg & Ron Zacharski. 1993. Cognitive Status and the Form of
Referring Expressions in Discourse. Language 69(2). 274–307.
Hammarström, Harald & Mark Donohue. 2014. Some Principles on the Use of Macro-Areas in
Typological Comparison. Language Dynamics and Change 4(1). 167–187.
Hammarström, Harald et al. 2018. Glottolog 3.2. Jena: Max Planck Institute for the Science of Hu-
man History.
Harlow, Ray. 2012. Maori: a linguistic introduction. Cambridge: Cambridge University Press.
Harms, Phillip Lee. 1994. Epena Pedee syntax. Dallas, TX: Summer Institute of Linguistics.
Harris, James W. 1991. The Exponence of Gender in Spanish. Linguistic Inquiry 22(1). 27–62.
Hartmann, Iren, Martin Haspelmath & Michael Cysouw. 2014. Identifying semantic role clus-
ters and alignment types via microrole coexpression tendencies. In Seppo Kittilä & Fernando
Zúñiga (eds.), Advances in research on semantic roles, 463–484. Amsterdam: John Benjamins.
Haspelmath, Martin. 1997. Indefinite Pronouns. Oxford: Oxford University Press.
Haspelmath, Martin. 1999. Explaining Article-Possessor Complementarity: Economic Motivation
in Noun Phrase Syntax. Language 75(2). 227–43.
Haspelmath, Martin. 2008. Parametric versus functional explanations of syntactic universals. In
Theresa Biberauer (ed.), The limits of syntactic variation, vol. 132 (Linguistik Aktuell/Linguistics
Today), 75–107. Amsterdam: Benjamins.
Haspelmath, Martin. 2010. Comparative concepts and descriptive categories in crosslinguistic
studies. Language 86(3). 663–687.
Hawkins, John A. 1978. Definiteness and Indefiniteness: A Study in Reference and Grammaticality
Prediction. London: Croom Helm Humanities Press.
Hawkins, John A. 1991. On (In)Definite Articles: Implicatures and (Un)Grammaticality Prediction.
Journal of Linguistics 27(2). 405–42.
Heath, Jeffrey. 1984. Functional grammar of Nunggubuyu. Canberra: Australian Institute of Abo-
riginal Studies.
Heath, Jeffrey. 1999. A Grammar of Koyra Chiini: The Songhay of Timbuktu. Berlin: De Gruyter
Mouton.
Heath, Jeffrey. 2008. A grammar of Jamsay. Berlin: De Gruyter Mouton.
Heim, Irene. 1988. The Semantics of Definite and Indefinite Noun Phrases. New York: Garland.
Heim, Irene. 1991. Artikel und Definitheit. In Arnim von Stechow & Dieter Wunderlich (eds.),
Semantics: An International Handbook of Contemporary Research, 487–535. Berlin: De Gruyter
Mouton.
Heim, Irene. 2002. File Change Semantics and the Familiarity Theory of Definiteness. In Paul
Portner & Barbara H. Partee (eds.), Formal Semantics: The Essential Readings, 223–248. Oxford:
Blackwell.

277
Heim, Irene & Angelika Kratzer. 1998. Semantics in Generative Grammar. Malden, MA: Blackwell.
Heine, Bernd. 1995. Conceptual grammaticalization and prediction. In John R. Taylor & Robert E.
MacLaury (eds.), Language and the cognitive construal of the world, vol. 82 (Trends in Linguis-
tics Studies and Monographs), 119–136. Berlin: De Gruyter Mouton.
Heine, Bernd. 1997. Cognitive foundations of grammar. Oxford: Oxford University Press.
Heine, Bernd & Tania Kuteva. 2002. World Lexicon of Grammaticalization. Cambridge: Cambridge
University Press.
Heine, Bernd et al. 1995. On Indefinite Articles. University of New Mexico, Albuquerque.
Hellenthal, Anne-Christie. 2010. A grammar of Sheko. Utrecht: LOT.
Hentschel, Elke & Harald Weydt. 2003. Handbuch der deutschen Grammatik. Berlin: De Gruyter
Mouton.
Heringa, Herman. 2011. Appositional constructions. Groningen: Rijksuniversiteit Groningen dis-
sertation.
Hill, Jane H. 2006. A grammar of Cupeño. Berkeley: University of California Press.
Himmelmann, Nikolaus P. 1997. Deiktikon, Artikel, Nominalphrase: Zur Emergenz Syntaktischer
Struktur. Tübingen: Niemeyer.
Himmelmann, Nikolaus P. 2001. Articles. In Martin Haspelmath et al. (eds.), Language Typology
and Language Universals. An International Handbook, vol. 1, 831–841. Berlin: De Gruyter Mou-
ton.
Hintikka, Jaakko. 1967. Individuals, possible worlds, and epistemic logic. Noûs 1(1). 33–62.
Hoch, Ernst. 1964. A Bemba grammar with exercises. Language Centre.
Hoeksema, Jack. 2013. On the natural history of Negative Polarity Items. Linguistic Analysis 38(1-
2). 3–33.
Hoffmann, Ludger. 2009. Handbuch der deutschen Wortarten. Berlin: De Gruyter Mouton.
Hopper, Paul J. & Janice Martin. 1987. Structuralism and diachrony: the development of the indef-
inite article in English. In Anna Giacalone Ramat, Onofrio Carruba & Giuliano Bernini (eds.),
Papers from the 7th International Conference on Historical Linguistics, vol. 48 (Current Issues in
Linguistic Theory), 295–304. Amsterdam: Benjamins.
Horn, Laurence Robert. 1972. On the semantic properties of logical operators in English. Los Angeles,
CA: UCLA dissertation.
Horn, Laurence Robert & Yasuhiko Kato. 2000. Negation and polarity: syntactic and semantic per-
spectives. Oxford: Oxford University Press.
Hualde, José. 1989. Double object constructions in Kirimi. In Robert Botne & Paul Newman (eds.),
Current approaches to African linguistics, 179–189. Dordrecht: Foris.
Hualde, José Ignacio & Jon Ortiz de Urbina. 2003. A Grammar of Basque. Berlin: De Gruyter Mou-
ton.

278
Ingham, Bruce. 2003. Lakota. München: Lincom Europa.
Ionin, Tania. 2006. This Is Definitely Specific: Specificity and Definiteness in Article Systems. Nat-
ural Language Semantics 14(2). 175–234.
Jany, Carmen. 2009. Chimariko grammar. Areal and typological perspective. Berkeley, CA: Univer-
sity of California Press.
Jensen, John Thayer. 1977. Yapese reference grammar. Honolulu: University Press of Hawaii.
Johnson, Heidi Anna. 2000. A Grammar of San Miguel Chimalapa Zoque. Austin, TX: The Univer-
sity of Texas at Austin dissertation.
Kaplan, David. 1989. Demonstratives. In Joseph Almog, John Perry & Howard Wettstein (eds.),
Themes From Kaplan, 481–563. Oxford: Oxford University Press.
Karlsson, Fred. 1999. Finnish. An essential grammar. London: Routledge.
Karttunen, Lauri. 1968. What Makes Definite Noun Phrases Definite? Tech. rep. Report P3871. Santa
Monica, CA: The RAND Corporation.
Karttunen, Lauri. 1974. Presuppositions and linguistic context. Theoretical Linguistics 1. 181–194.
Karttunen, Lauri. 1976. Discourse referents. In Syntax and semantics, vol. 7: Notes from the Linguis-
tic Underground, 363–385. New York: Academic Press.
Kemmerer, David. 1999. ”Near”and ”far”in language and perception. Cognition 73(1). 35–63.
Kenesei, Istvan, Robert M. Vago & Anna Fenyvesi. 1998. Hungarian. London: Routledge.
Khalilova, Zaira. 2009. A grammar of Khwarshi. Utrecht: LOT.
Khan, Geoffrey. 2008. The Neo-Aramaic dialect of Barwar. Volume 1: Grammar. Leiden: Brill.
Kieviet, Paulus. 2017. A grammar of Rapa Nui. Berlin: Language Science Press.
Kimball, Geoffrey David. 1985. A descriptive grammar of Koasati. New Orleans, LA: Tulane Uni-
versity dissertation.
Kita, Sotaro. 2003. Pointing. A foundational building block of human communication. In Sotaro
Kita (ed.), Pointing: Where language, culture, and cognition meet, 1–8. Mahwah: Lawrence Erl-
baum.
Kite, Suzanne & Stephen Adolphe Wurm. 2004. The Duuŋidjawu language of southeast Queensland:
grammar, texts and vocabulary. Canberra: Australian National University.
Kluge, Andrea. 2017. A grammar of Papuan Malay. Berlin: Language Science Press.
Köpcke, Klaus-Michael. 1982. Untersuchungen zum Genussystem der deutschen Gegenwartssprache.
Tübingen: Max Niemeyer.
Köpcke, Klaus-Michael & David A. Zubin. 1984. Sechs Prinzipien für die Genuszuweisung im
Deutschen: Ein Beitrag zur natürlichen Klassifikation. Linguistische Berichte 93. 26–50.
Kramer, Christina Elizabeth & Liljana Mitkovska. 2011. Macedonian: a course for beginning and
intermediate students. Madison, WI: University of Wisconsin Press.
Kratochvı́l, František. 2007. A grammar of Abui. A Papuan language of Alor. Utrecht: LOT.

279
Krifka, Manfred & Renate Musan. 2012a. Information structure: overview and linguistic issues. In
The expression of information structure, 1–43. Berlin: De Gruyter Mouton.
Krifka, Manfred & Renate Musan. 2012b. The expression of information structure. Berlin: De Gruyter
Mouton.
Krifka, Manfred et al. 1995. Genericity: An introduction. In Gregory N. Carlson & Francis Jeffry
Pelletier (eds.), The Generic Book, 1–124. Chicago: University of Chicago Press.
Kripke, Saul. 1959. A completeness theorem in modal logic. The Journal of Symbolic Logic 24(1).
1–14.
Kroeger, Paul. 2018. Analyzing meaning: An introduction to semantics and pragmatics. Vol. 5 (Text-
books in Language Sciences). Berlin: Language Science Press.
Kung, Susan Smythe. 2007. A descriptive grammar of Huehuetla Tepehua. Austin, TX: University
of Texas dissertation.
Ladusaw, William Allen. 1980. Polarity sensitivity as inherent scope relations. New York: Garland.
Langacker, Ronald W. 1987. Foundations of cognitive grammar. Stanford: Stanford University Press.
Langacker, Ronald W. 2001. Discourse in Cognitive Grammar. Cognitive Linguistics 12(2). 143–188.
Langacker, Ronald W. 2002. Remarks on the English grounding system. In Frank Brisard (ed.),
Grounding: the epistemic footing of deixis and reference, vol. 21 (Cognitive Linguistics Research).
Berlin: De Gruyter Mouton.
Langacker, Ronald W. 2008. Cognitive grammar: a basic introduction. Oxford: Oxford University
Press.
Lazard, Gilbert. 2005. What are we typologists doing? In Zygmunt Frajzyngier, Adam Hodges &
David S. Rood (eds.), Linguistic diversity and language theories, vol. 75 (Studies in Language
Companion Series), 1–23. Amsterdam: Benjamins.
Lehmann, Thomas. 1993. A grammar of modern Tamil. Pondicherry: Pondicherry Institute of Lin-
guistics and Culture.
Leiss, Elisabeth. 2000. Artikel Und Aspekt, Die Grammatischen Muster von Definitheit. Berlin: De
Gruyter Mouton.
Leslau, Wolf. 1995. Reference grammar of Amharic. Wiesbaden: Harrassowitz.
Levinson, Stephen C. 2004. Deixis and pragmatic. In Laurence Horn & Gregory Ward (eds.), The
handbook of pragmatics, 97–121. Oxford: Blackwell.
Lewis, David. 1970. General semantics. Synthese 22(1/2). 18–67.
Lewis, David. 1979. Scorekeeping in a language game. Journal of Philosophical Logic 8. 339–359.
Liljegren, Henrik. 2016. A grammar of Palula. Vol. 8 (Studies in diversity linguistics). Berlin: Lan-
guage Science Press.
Löbner, Sebastian. 1985. Definites. Journal of Semantics 4. 279–326.

280
Löbner, Sebastian. 1998. Definite Associative Anaphora. In Approaches to discourse anaphora. Pro-
ceedings of DAARC96 - Discourse Anaphora and Resolution Colloquium. Lancaster: Lancaster
University.
Löbner, Sebastian. 2011. Concept Types and Determination. Journal of Semantics 28(3). 279–333.
Loughnane, Robyn. 2009. A Grammar of Oksapmin. Melbourne: The University of Melbourne dis-
sertation.
Lunt, Horace Gray. 1952. A grammar of the Macedonian literary language. Skopje: Državno knigoiz-
datelstvo.
Lyons, Christopher. 1999. Definiteness. Cambridge: Cambridge University Press.
Mace, John. 2003. Persian grammar. For reference and revision. London: Routledge.
Marušič, Franc & Rok Žaucer. 2014. A definite article in the AP - Evidence from Colloquial Slove-
nian. In The nominal structure in Slavic and beyond, vol. 116 (Studies in Generative Grammar),
183–208. Berlin: De Gruyter Mouton.
Massam, Diane, Colin Gorrie & Alexandra Kellner. 2006. Niuean determiners: everywhere and
nowhere. In Proceedings of the 2006 annual conference of the Canadian Linguistic Association.
Toronto: Canadian Linguistic Association.
Mateo Toledo, Eladio. 2017. Q’anjob’al. In Judith Aissen, Nora C. England & Roberto Zavala Mal-
donado (eds.), The Mayan languages. London: Routledge.
Matras, Yaron. 2012. A grammar of Domari. Berlin: De Gruyter Mouton.
McGregor, William. 1990. A functional grammar of Gooniyandi. Amsterdam: Benjamins.
McGregor, William. 1994. Warrwa. München: Lincom Europa.
McIntosh, Justin Daniel. 2011. Grammatical sketch of Teotepec Chatino. Austin, TX: The University
of Texas at Austin dissertation.
Merlan, Francesca. 1989. Mangarayi. London: Routledge.
Merlan, Francesca. 1994. A Grammar of Wardaman: a language of the Northern Territory of Aus-
tralia. Berlin: De Gruyter Mouton.
Mihas, Elena. 2010. Essentials of Asheninka Perené grammar. Milwaukee, WI: University of Wisconsin-
Milwaukee dissertation.
Montague, Richard. 1970. Pragmatics and intensional logic. Dialectica 24(4). 277–302.
Moravcsik, Edith A. 1969. Determination. In Working papers on language universals 1, 64–98. Stan-
ford, CA: Stanford University.
Moroz, George. 2017. Lingtypology: easy mapping for Linguistic Typology.
Morton, Ermel Joseph. 1962. A descriptive grammar of Tongan (Polynesian). Bloomington, IN: In-
diana University dissertation.
Müller, Stefan. 2010. Grammatiktheorie. 1st edn. (Stauffenburg Einführungen 20). Tübingen: Stauf-
fenburg.

281
Newman, Paul. 2000. The Hausa language: an encyclopedic reference grammar. New Haven: Yale
University Press.
Newmark, Leonard, Peter Prifti & Philip Hubbard. 1982. Standard Albanian: a reference grammar
for students. Stanford: Stanford University Press.
Newmeyer, Frederick J. 2007. Linguistic typology requires crosslinguistic formal categories. Lin-
guistic Typology 11(1). 133–157.
Nikolaeva, Irina. 2003. Possessive affixes in the pragmatic structuring of the utterance: Evidence
from Uralic. In Pirkko Suihkonen & Bernard Comrie (eds.), Proceedings of the International
symposium on deictic systems and quantification in languages spoken in Europe and North and
Central Asia. Collection of papers, 130–145. Leipzig / Izhevsk: MPI-EVA Leipzig, Udmurt State
University.
Noonan, Michael. 1992. A grammar of Lango (Mouton Grammar Library 7). Berlin: De Gruyter
Mouton.
Nordhoff, Sebastian. 2009. A grammar of Upcountry Sri Lanka Malay. Utrecht: LOT.
Norwood, Susan. 1997. Gramática de la lengua Sumu. Managua: CIDCA.
Orkaydo, Ongaye Oda. 2013. A grammar of Konso. Utrecht: LOT.
Otsuka, Yuko. 2000. Ergativity in Tongan. Oxford: University of Oxford dissertation.
Paperno, Denis. 2014. Grammatical sketch of Beng. Vol. 51 (Mandekan). Paris: INALCO.
Park, Indrek. 2012. A grammar of Hidatsa. Bloomington, IN: Indiana University dissertation.
Pensalfini, Robert J. 1997. Jingulu grammar, dictionary, and texts. Massachusetts, MA: MIT disser-
tation.
Perder, Emil. 2013. A grammatical description of Dameli. OCLC: 940717110. Stockholm: Depart-
ment of Linguistics, Stockholm University.
Pet, Willem J.A. 2011. A grammar sketch and lexicon of Arawak (Lokono Dian). SIL International.
Peterson, John M. 2011. A grammar of Kharia: a South Munda language. Leiden: Brill.
Phillips, John D. 2004. Manx. München: Lincom Europa.
Plaisir, Heleen. 2006. A Grammar of Lepcha. Leiden: Universiteit Leiden dissertation.
Plumley, John Martin. 1948. An introductory Coptic grammar (Sahidic dialect). London: Home &
Van Thal.
Prince, Ellen F. 1981. On the Inferencing of Indefinite this NPs. In Aravind Joshi, Bonnie Webber
& Ivan A. Sag (eds.), In Elements of Discourse Understanding (1981), pp. 231–250. Cambridge:
Cambridge University Press.
Quine, Willard van Orman. 1953. Reference and modality. In From a Logical Point of View, 137–138.
Cambridge, MA: Harvard University Press.
R Core Team. 2016. R: A language and environment for statistical computing. Vienna, Austria.

282
Raymundo Gonzáles, Sonia et al. 2000. Sk’exkixhtaqil yallay koq’anej. Variación dialectal en Q’anjobal.
Cholsamaj: OKMA.
Ribeiro, Eduardo Rivail. 2012. A grammar of Karajá. Chicago: University of Chicago dissertation.
Rigo, Bernard & Jacques Vernaudon. 2004. De la translation substantivante à la quantifi- cation:
Vers une caractérisation sémantique de l’article te en tahitien. Bulletin de la Société de linguis-
tique de Paris 94(1). 457–480.
Roberts, Craige. 2003. Uniqueness in Definite Noun Phrases. Linguistics and Philosophy Linguistics
and Philosophy 26(3). 287–350.
Robinson, Laura C. 2008. Dupaningan Agta: Grammar, vocabulary, and texts. Honolulu: University
of Hawaii dissertation.
Rogers, Christopher. 2010. A comparative grammar of Xinkan. Salt Lake City, UT: University of
Utah dissertation.
Romero-Méndez, Rodrigo. 2008. A reference grammar of Ayutla Mixe (Tukyo’m Ayuujk). Buffalo,
NY: University of Buffalo, State University of New York dissertation.
Rood, David S. & Allan Ross Taylor. 1996. Sketch of Lakhota, a Siouan Language. In Ives Goddard
& William C. Sturtevant (eds.), Handbook of North American Indians (languages), vol. 17, 440–
482. Washington: Smithsonian Institution.
Rowe, Karen. 2005. Siar-Lak grammar essential. Vol. 50 (Data papers on Papua New Guinea lan-
guages). Ukarumpa: SIL.
Rubin, Aaron. 2010. The development of the Amharic definite article and an Indonesian parallel.
Journal of Semitic Studies 55(1). 103–114.
Russell, Bertrand. 1905. On Denoting. Mind 14. 479–493.
Schroeder, Christoph. 2006. Articles and Article Systems in Some Areas of Europe. In Giuliano
Bernini & Marcia L. Schwartz (eds.), Pragmatic Organization of Discourse in the Languages
of Europe, vol. 8 (Empirical Approaches to Language Typology), 545–611. Berlin: De Gruyter
Mouton.
Schwarz, Florian. 2009. Two Types of Definites in Natural Language. Amherst, MA: University of
Massachusetts Amherst dissertation.
Schwarz, Florian. 2013. Two Kinds of Definites Cross-Linguistically. Language and Linguistics
Compass. 534–559.
Seyoum, Mulugeta. 2008. A grammar of Dime. Utrecht: LOT.
Slobin, Dan Isaac. 1997. The origin of grammaticizible notions: beyond the individual mind. In
Dan Isaac Slobin (ed.), The crosslinguistic study of language acquisition. Vol. 5. Expanding the
contexts, 265–323. Mahwah, NJ: Lawrence Erlbaum.

283
Slobin, Dan Isaac. 2001. Form function relations: How do children find out what they are? In
Melissa Bowerman & Stephen C. Levinson (eds.), Language acquisition and conceptual devel-
opment, 406–449. Cambridge: Cambridge University Press.
Song, Jae Jung. 2001. Linguistic typology: morphology and syntax. Harlow: Pearson Education.
Sperber, Dan & Deirdre Wilson. 1986. Relevance: communication and cognition. Oxford: Blackwell.
Stalnaker, Robert. 1974. Pragmatic presuppositions. In Milton K. Munitz & Peter K. Unger (eds.),
Semantics and Philosophy, 197–214. New York: New York University Press.
Stampe, David. 1976. Cardinal numeral systems. Chicago Linguistic Society 12. 594–609.
Stassen, Leon. 1985. Comparison and universal grammar. Oxford: Blackwell.
Steeman, Sander. 2011. A grammar of Sandawe. A Khoisan language of Tanzania. Utrecht: LOT.
Stenson, Nancy. 2008a. Basic Irish. A grammar and workbook. London: Routledge.
Stenson, Nancy. 2008b. Intermediate Irish: A grammar and workbook. London: Routledge.
Stirtz, Timothy M. 2011. A grammar of Gaahmg. A Nilo-Saharan language of Sudan. Utrecht: LOT.
Strawson, P. F. 1950. On Referring. Mind 59(235). 320–344.
Suttles, Wayne. 2004. Musqueam reference grammar. Vancouver: UBC Press.
Taylor, Allan Ross. 1969. A grammar of Blackfoot. Berkeley, CA: University of California disserta-
tion.
Telles, Stella. 2002. Fonologia e Gramática Latundê/Lakondê. Amsterdam: Vrije Universiteit te Am-
sterdam dissertation.
Terrill, Angela. 2003. A Grammar of Lavukaleve. Vol. 30 (Mouton Grammar Library). Berlin: De
Gruyter Mouton.
Tomasello, Michael. 1999. Cultural Origins of Human Cognition. Cambridge: Harvard University
Press.
Tz’utujil Tinaamitaal. 2007. Rub’eeyaal rutz’ijb’axiik ja tzijob’al Tz’utujil. Gramática normativa
Tz’utujil. Ciudad de Guatemala: Academia de las Lenguas Mayas de Guatemala.
van den Heuvel, Wilco. 2006. Biak. Utrecht: LOT.
Van Valin, Robert D. 1977. Aspects of Lakhota syntax. Berkeley, CA: University of California dis-
sertation.
Vázquez Alvarez, Juan Jesús. 2011. A grammar of Chol, a Mayan language. Austin, TX: University
of Texas dissertation.
Völkel, Svenja. 2010. Social structure, space and possession in Tongan culture and language: an eth-
nolinguistic study. Amsterdam: Benjamins.
von Heusinger, Klaus. 2002. Specificity and Definiteness in Sentence and Discourse Structure.
Journal of Semantics 19. 245–274.
Weber, Robert. 2003. Morfologı́a verbal del rapa nui, lengua polinésica de Isla de Pascua, y su función
en la narración. Valparaı́so: Ediciones Universitarias de Valparaı́so.

284
Whitehead, Carl R. 2006. A reference grammar of Menya, an Angan language of Papua New Guinea.
Winnipeg: University of Manitoba dissertation.
Wierzbicka, Anna. 1998. Anchoring linguistic typology in universal semantic primes. Linguistic
Typology 2. 141–194.
Wilde, Christopher P. 2008. A sketch of the phonology and grammar of Rājbanshi. Helsinki: Uni-
versity of Helsinki dissertation.
Wilkins, David P. 1989. Mparntwe Arrernte (Aranda) : studies in the structure and semantics of
grammar. Canberra: Australian National University dissertation.
Wilkinson, Melanie P. 1991. Djambarrpuyŋu. A Yolŋu variety of Northern Australia. Sydney: Uni-
versity of Sydney dissertation.
Williamson, Janis Shirley. 1984. Studies in Lakhota grammar. San Diego, CA: University of Cali-
fornia dissertation.
Wiltschko, Martina. 2002. The syntax of pronouns: Evidence from Halkomelem Salish. Natural
Language & Linguistic Theory 20(1). 157–195.
Zimmermann, Malte. 2008. Quantification in Hausa. In Quantification: A cross-linguistic perspec-
tive, vol. 64 (North-Holland linguistic series), 415–475. Bingley: Emerald.
Zwarts, Frans. 1981. Negatief polaire uitdrukkingen I. Glot International 4. 35–132.
Zwarts, Frans. 1995. Nonverdical contexts. Linguistic Analysis 25. 286–312.
Zymberi, Isa. 2004. Colloquial Albanian. London: Routledge.

285
Selbstständigkeitserklärung und Erklärung über frühere Promotionsversuche

Hiermit versichere ich, dass ich die vorliegende Arbeit ohne unzulässige Hilfe Dritter und ohne
Benutzung anderer als der angegebenen Hilfsmittel angefertigt habe; die aus fremden Quellen
direkt oder indirekt übernommenen Gedanken sind als solche kenntlich gemacht.

Ich habe nicht die Hilfe eines Promotionsberaters in Anspruch genommen. Die Arbeit wurde zu-
vor weder im Inland noch im Ausland in gleicher oder ähnlicher Form einer an deren Institution
vorgelegt. Sie ist weder Bestandteil eines ruhenden Verfahrens noch wurde sie in einem gleichar-
tigen Promotionsverfahren als endgültig nicht bestanden erklärt. Die Arbeit ist vorher auch noch
nicht veröffentlicht worden.

Ich erkläre, dass ich bisher noch keine Promotionsversuche unternommen habe.

Ort, Datum

Unterschrift

Name in Druckbuchstaben