Gender and its interaction with number and evaluative morphology Francesca Di Garbo

Gender and its interaction with number and evaluative morphology Francesca Di Garbo
Gender and its interaction with
number and evaluative morphology
An intra- and intergenealogical typological survey of Africa
Francesca Di Garbo
Doctoral Thesis in Linguistics at Stockholm University, Sweden 2014
Gender and its interaction with number and
evaluative morphology
An intra- and intergenealogical typological survey of Africa
Francesca Di Garbo
Gender and its interaction with
number and evaluative morphology
An intra- and intergenealogical typological survey of Africa
Francesca Di Garbo
c
Francesca
Di Garbo, Stockholm 2014
ISBN 978-91-7447-952-2
Printed in Sweden by US-AB, Stockholm 2014
Distributor: Department of Linguistics, Stockholm University
Cover design: Paolo Cerva
Map design: Ljuba Veselinova
Calipso: Tu che hai visto l’Oceano, i mostri e l’Eliso,
potrai ancora riconoscere le case, le tue case?
Odisseo: Tu stessa hai detto che porto l’isola in me.
Cesare Pavese, I dialoghi con Leucò
Abstract
This dissertation investigates the interactions between gender and number and gender and evaluative morphology in a sample of 100 African languages, and provides a method for assessing
the role that these interactions play in the grammatical complexity of gender systems. Within
the sampling space of one continental area, namely, Africa, in-depth studies of genealogically
related languages are combined with large-scale comparison across unrelated genealogical units.
The dissertation is organised around three main research foci.
First, the dissertation provides a systematic overview of patterns of interaction between gender
and number along the following dimensions: exponence, syncretism, indexation, correlations in
type of marking, and semantic interactions through gender assignment. Several of these patterns
are identified, and the dissertation provides examples of their distribution across the languages
of the sample. The study provides evidence for the possibility that, similar to the verbal domain,
nominal features are also organised in a relevance hierarchy and that this can be studied by
looking at nominal features as encoded both on nouns and on indexing targets. In addition,
the study shows that, in languages with sex-based gender, gender and number may compete
through indexation patterns and that animacy and/or lexical plurality play a crucial role in the
expansion of the domains of use of special patterns of plural indexation. The study also shows
that the development of pervasive indexation systems in the languages of Africa tends to always
involve both gender and number. Finally, the study shows how gender assignment can be used
as a means for encoding variation in the countability properties of nouns and noun phrases.
Second, the dissertation provides a systematic overview of patterns of interaction between gender and evaluative morphology in the languages of the sample. Two major types of interactions
are found, and their occurrence depends on whether or not languages have dedicated diminutive
and augmentative genders. The study shows that the distribution of the two types depends on
three factors: (1) the type of gender system (sex-based vs. non-sex-based), (2) the number of
gender distinctions and (3) the possibility of assigning the same noun to more than one gender.
Third, the dissertation investigates the role that interactions of gender and number and gender and evaluative morphology play in the absolute complexity of gender systems. The study
proposes a metric for gender complexity and uses this metric to compute complexity scores for
each of the gendered languages of the sample. The results suggest that the gender systems of
the languages of the sample lean toward high complexity scores, that languages from the same
genealogical units tend to have the same or similar complexity scores, and that the distribution
of the outliers can often be understood as the result of language contact.
Finally, this dissertation provides an insight into the advantages of looking at nominal features
through their morphosyntactic and semantic interactions rather than as isolated domains of
grammar.
Contents
List of Figures
ix
List of Tables
xi
Abbreviations
xiii
Acknowledgements
xvii
1 Introduction
2 Background: gender, number, evaluative morphology and relevant interactions
2.1 Gender . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
2.1.1 Gender assignment . . . . . . . . . . . . . . . . . . . . . . . . . . .
2.1.2 The morphosyntax of gender: indexation . . . . . . . . . . . . . .
2.1.2.1 Beyond agreement: the notion of indexation and what
we gain with it . . . . . . . . . . . . . . . . . . . . . . . .
2.1.2.2 Patterns of gender indexation . . . . . . . . . . . . . . . .
2.1.3 The function, distribution and diachrony of gender systems . . . .
2.1.3.1 The function of gender . . . . . . . . . . . . . . . . . . .
2.1.3.2 The distribution and stability of gender systems . . . . .
2.1.3.3 The diachrony of gender . . . . . . . . . . . . . . . . . . .
2.2 Number . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
2.2.1 Countability properties of nouns and noun phrases . . . . . . . . .
2.2.2 Types of number systems and number values . . . . . . . . . . . .
2.2.2.1 General number . . . . . . . . . . . . . . . . . . . . . . .
2.2.2.2 Obligatory number marking and possible number values .
2.2.3 Number marking and the Animacy Hierarchy . . . . . . . . . . . .
2.2.4 The morphosyntax of number: indexation . . . . . . . . . . . . . .
2.2.5 Distribution and diachrony of number systems . . . . . . . . . . .
2.2.5.1 The distribution of number systems . . . . . . . . . . . .
2.2.5.2 Sources of number . . . . . . . . . . . . . . . . . . . . . .
2.3 Evaluative morphology . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
2.3.1 The semantics, pragmatics and functions of evaluative markers . .
2.3.2 Types of marking in evaluative morphology . . . . . . . . . . . . .
2.3.2.1 Phonetic iconicity in evaluative morphology: universal
or language-specific? . . . . . . . . . . . . . . . . . . . . .
1
3
3
4
7
7
9
14
14
15
15
18
18
20
20
22
22
23
25
25
25
26
27
29
29
iii
Contents
2.4
2.5
2.6
2.7
2.8
2.9
2.3.2.2 Morphosyntactic encoding of evaluative markers . . . .
2.3.3 Distribution and diachrony of evaluative markers . . . . . . . . .
2.3.3.1 The distribution of evaluative markers . . . . . . . . . .
2.3.3.2 The diachrony of evaluative markers . . . . . . . . . . .
Gender and its interaction with number and evaluative morphology . . .
The interaction between gender and number: state of the art . . . . . .
2.5.1 Exponence . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
2.5.2 Syncretism . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
2.5.3 Cumulative exponence, syncretism and relevance hierarchies . . .
2.5.4 The interaction between gender and number in the languages of
Africa . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
The interaction between gender and evaluative morphology: state of the
art . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Grammatical complexity of gender systems . . . . . . . . . . . . . . . .
Research questions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Summary of the chapter . . . . . . . . . . . . . . . . . . . . . . . . . . .
3 Method: language sampling, data collection and organization
3.1 Sampling methodology . . . . . . . . . . . . . . . . . . . . .
3.2 African languages: genealogical classification . . . . . . . .
3.3 African languages: large-scale language contact . . . . . . .
3.4 Sampling procedure . . . . . . . . . . . . . . . . . . . . . .
3.5 Data collection and organization . . . . . . . . . . . . . . .
3.6 Summary of the chapter . . . . . . . . . . . . . . . . . . . .
.
.
.
.
.
.
4 Gender, number and evaluative morphology in the languages of
an overview
4.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . .
4.2 Gender . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
4.2.1 Criteria of classification of gender systems: summary .
4.2.2 Sex-based gender . . . . . . . . . . . . . . . . . . . . .
4.2.3 Non-sex-based gender . . . . . . . . . . . . . . . . . .
4.2.3.1 Noun Classes in Bantu . . . . . . . . . . . .
4.2.3.2 Noun classes in Atlantic . . . . . . . . . . . .
4.3 Number . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
4.3.1 General number . . . . . . . . . . . . . . . . . . . . .
4.3.2 Tripartite number systems . . . . . . . . . . . . . . . .
4.4 Evaluative morphology . . . . . . . . . . . . . . . . . . . . . .
4.5 Summary of the chapter . . . . . . . . . . . . . . . . . . . . .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
30
31
31
32
32
33
34
35
37
. 38
.
.
.
.
39
40
43
44
.
.
.
.
.
.
45
45
47
52
53
58
59
the sample:
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
61
61
61
69
71
74
76
78
81
87
89
91
93
5 Gender and number
95
5.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 95
iv
Contents
5.2
5.3
5.4
5.5
5.6
5.7
5.8
Cumulation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
5.2.1 Cumulation between gender and number: results . . . . . . . . . .
5.2.1.1 Cumulative encoding of gender and number with all the
indexes . . . . . . . . . . . . . . . . . . . . . . . . . . . .
5.2.1.2 Cumulative encoding of gender and number with some
indexes . . . . . . . . . . . . . . . . . . . . . . . . . . . .
5.2.1.3 Noncumulative encoding of gender and number on the
indexing targets . . . . . . . . . . . . . . . . . . . . . . .
5.2.1.4 Cumulation between gender and number on nouns . . . .
5.2.1.5 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . .
Syncretism . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
5.3.1 Syncretism of gender in the context of number: results . . . . . . .
5.3.2 Semanticization of gender distinctions under syncretism . . . . . .
5.3.3 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Cumulation and syncretism: summary and discussion . . . . . . . . . . .
5.4.1 Cumulation, syncretism, and grammatical complexity . . . . . . .
5.4.2 Syncretism and relevance hierarchy for nominal features . . . . . .
Split plural indexation systems . . . . . . . . . . . . . . . . . . . . . . . .
5.5.1 Animacy-based uses of Dedicated Plural Indexation . . . . . . . .
5.5.2 Dedicated Plural Indexation in Cushitic . . . . . . . . . . . . . . .
5.5.2.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . .
5.5.2.2 The distribution of DPI in the Cushitic sample . . . . . .
5.5.2.3 Two peculiar languages within the Cushitic sample: Kambaata and Daasanach . . . . . . . . . . . . . . . . . . . .
5.5.3 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Presence of gender and type of number marking . . . . . . . . . . . . . . .
Gender shifts and semantic interactions between gender and number . . .
5.7.1 Gender shifts, number/countability in Bantu and North-Central
Atlantic . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
5.7.2 Gender shifts, number/countability in Berber, Semitic and Eastern
Nilotic . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
5.7.3 Gender shifts and semantic interactions between gender and number: summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Summary of the chapter . . . . . . . . . . . . . . . . . . . . . . . . . . . .
6 Gender and evaluative morphology
6.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
6.2 Gender and evaluative morphology: overview of results . . . . . . . . . .
6.3 Type 1: Diminutive and augmentative genders . . . . . . . . . . . . . .
6.3.1 Distribution of diminutive and augmentative genders and general
characteristics . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
6.3.2 Multiple or additive class marking with diminutives and augmentatives . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
97
98
100
101
102
105
109
109
110
113
116
117
117
117
118
119
121
121
122
128
128
130
134
136
139
141
142
145
. 145
. 145
. 147
. 148
. 152
v
Contents
6.3.3
6.4
6.5
6.6
6.7
The renewal of evaluative morphology in the southeastern Bantu
languages . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 154
6.3.4 Diminutive marking in SElEE . . . . . . . . . . . . . . . . . . . . . 158
6.3.5 Type 1: summary . . . . . . . . . . . . . . . . . . . . . . . . . . . 160
Type 2: Diminutives and augmentatives in languages with sex-based gender161
6.4.1 Examples of Type 2 languages . . . . . . . . . . . . . . . . . . . . 163
6.4.2 The distribution of Type 2 languages . . . . . . . . . . . . . . . . . 167
6.4.3 Type 2: summary . . . . . . . . . . . . . . . . . . . . . . . . . . . 169
Evaluative morphology and biological gender in languages without (sexbased) gender . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 170
6.5.1 Bantu . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 170
6.5.2 Western Nilotic . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 172
6.5.3 Akan . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 173
6.5.4 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 174
Absence of interaction between gender and evaluative morphology . . . . 174
Summary of the chapter . . . . . . . . . . . . . . . . . . . . . . . . . . . . 175
7 Gender and grammatical complexity
177
7.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 177
7.2 Background . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 177
7.3 How to integrate interactions of gender into a metric for gender complexity178
7.4 How to elaborate maximally local complexity measures . . . . . . . . . . . 180
7.5 A metric for grammatical gender complexity . . . . . . . . . . . . . . . . . 181
7.6 Method . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 184
7.7 Results and discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 186
7.7.1 Genealogical and areal biases in the distribution of GCSs . . . . . 190
7.7.2 Same complexity score does not mean same type of gender system 193
7.7.3 Implicational relationships between the features in the metric . . . 193
7.7.4 Some features may be stronger predictors of gender complexity
than others . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 194
7.8 Summary of the chapter . . . . . . . . . . . . . . . . . . . . . . . . . . . . 198
8 Summary of the main results, prospects for future research and concluding
remarks
199
8.1 Summary and assessment of results . . . . . . . . . . . . . . . . . . . . . . 199
8.1.1 Gender and number . . . . . . . . . . . . . . . . . . . . . . . . . . 200
8.1.2 Gender and evaluative morphology . . . . . . . . . . . . . . . . . . 201
8.1.3 Interactions of gender and grammatical complexity . . . . . . . . . 202
8.2 Assessment of the sampling methodology . . . . . . . . . . . . . . . . . . 202
8.3 Prospects for future research . . . . . . . . . . . . . . . . . . . . . . . . . 204
8.3.1 Nominal relevance hierarchy . . . . . . . . . . . . . . . . . . . . . . 204
8.3.2 Split plural indexation systems and Dedicated Plural Indexation . 205
8.3.3 Gender assignment and its manipulation . . . . . . . . . . . . . . . 205
8.3.4 Absolute complexity of gender systems . . . . . . . . . . . . . . . . 206
vi
Contents
8.4
Concluding remarks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 207
A The language sample
209
A.1 Genealogical units and internal composition of the subsamples . . . . . . 209
A.2 The language sample: alphabetical index . . . . . . . . . . . . . . . . . . . 215
B Database coding sheet
B.1 Language data . . . .
B.2 Gender systems . . . .
B.3 Number systems . . .
B.4 Evaluative morphology
B.5 Gender complexity . .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
219
. 219
. 219
. 221
. 221
. 223
C Gender- and number-indexing targets in the languages of the sample
225
D Examples of lexical plurals in four Cushitic languages
227
E Singular and plural suffixes in Mabaan, Dinka and Luwo
229
F Complexity scores for the individual features in the metric
231
Sammanfattning på svenska
235
Bibliography
243
vii
List of Figures
2.1
2.2
2.3
2.4
2.5
np-internal and np-external indexation . . . . . . . . . . . . .
The gender system of Romanian . . . . . . . . . . . . . . . .
The grammaticalization of gender from demonstratives . . . .
Universal semantic and pragmatics properties of diminutives .
Possible patterns of syncretism . . . . . . . . . . . . . . . . .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
3.1
3.2
3.3
Greenberg’s (1963a) genealogical classification of the African languages . . 48
African language families and isolates . . . . . . . . . . . . . . . . . . . . 51
Linguistic macro-areas within Africa . . . . . . . . . . . . . . . . . . . . . 54
4.1
4.2
4.3
Sex-based and non-sex-based gender systems in the languages of the sample 72
General and obligatory number in the languages of the sample . . . . . . 86
Types of evaluative morphology systems in the languages of the sample . 94
5.1
5.2
5.3
5.4
5.5
5.6
Cumulation and syncretism . . . . . . . . .
The noun class system of Wamey . . . . . .
Types of gender syncretism across number .
Triggers of DPI in the Cushitic sample . . .
DPI in the Cushitic sample . . . . . . . . .
Gender- and number-indexing targets in the
7.1
7.2
7.3
7.4
Distribution of the GCSs . . . . . . . . . . . . . . . . . .
Geographical distribution of the GCSs . . . . . . . . . . .
Correlation coefficients between the features of the metric
GCSs (Average) stratified according to feature values . . .
. . . . . . . . . . . . . .
. . . . . . . . . . . . . .
. . . . . . . . . . . . . .
. . . . . . . . . . . . . .
. . . . . . . . . . . . . .
languages of the sample
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
9
12
16
27
36
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
96
104
110
126
127
133
.
.
.
.
.
.
.
.
.
.
.
.
188
189
195
197
ix
List of Tables
3.1
3.2
Genealogical relationships within Khoisan . . . . . . . . . . . . . . . . . . 49
Genealogical units in the sample . . . . . . . . . . . . . . . . . . . . . . . 57
4.1
4.2
4.3
4.4
4.5
4.6
4.7
4.8
4.9
4.10
4.11
4.12
4.13
Sex-based and non-sex-based gender systems in the language sample .
Number of genders in the language sample . . . . . . . . . . . . . . . .
Systems of gender assignment in the languages of the sample . . . . .
Number of gender-indexing targets in the language sample . . . . . . .
Overt coding of gender on nouns in the languages of the sample . . . .
Noun classes and indexation patterns in Kirundi . . . . . . . . . . . .
Noun classes and indexation patterns in Bandial . . . . . . . . . . . .
Nominal number systems in the languages of the sample . . . . . . . .
Number values in the languages of the sample . . . . . . . . . . . . . .
Number of number-indexing targets in the languages of the sample . .
Singular marking in Kambaata, Borana-Arsi-Guji Oromo and Tsamai
Plural marking in Kambaata, Borana-Arsi-Guji Oromo and Tsamai . .
Evaluative morphology systems in the languages of the sample . . . .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
63
65
67
68
70
77
79
82
83
85
88
88
92
5.1
5.2
5.3
5.4
5.5
5.6
.
.
.
.
.
99
101
102
103
106
.
.
.
.
.
.
.
111
112
114
114
124
125
131
5.14
5.15
Cumulation between gender and number on the indexing targets . . . .
The semantics of the noun class system of Ju|’hoan . . . . . . . . . . . .
Independent Personal Pronouns in Kabyle . . . . . . . . . . . . . . . . .
The Definite Article in Beja . . . . . . . . . . . . . . . . . . . . . . . . .
Cumulation between gender and number on nouns . . . . . . . . . . . .
Gender syncretism and its relationship with cumulation in the language
sample . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Gender syncretism in Turkana restrictive markers . . . . . . . . . . . . .
Gender and number distinctions in Gola . . . . . . . . . . . . . . . . . .
Gender and number distinctions in Nuclear Wolof . . . . . . . . . . . . .
Morphologically plural nouns taking plural indexation in Baiso . . . . .
Lexical plurals in Baiso . . . . . . . . . . . . . . . . . . . . . . . . . . .
Number-indexing targets in Creissels et al.’s (2008) type (a) languages .
Distribution of the use of gender shifts for the encoding of variation in
the countability properties of nouns . . . . . . . . . . . . . . . . . . . .
Noun classes and singular-plural pairs of Swahili . . . . . . . . . . . . .
Noun classes and singular-plural pairs of Bandial . . . . . . . . . . . . .
6.1
6.2
Interactions between grammatical gender and evaluative morphology . . . 146
Distribution of diminutive and augmentative genders in Type 1 languages 149
5.7
5.8
5.9
5.10
5.11
5.12
5.13
. 135
. 137
. 137
xi
List of Tables
6.3
6.4
6.5
The diminutive suffix in some southeastern Bantu languages . . . . . . . . 155
Number of gender distinctions and type of attested gender shifts in Type
2 languages . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 162
The meanings and function of -hadi in Sotho . . . . . . . . . . . . . . . . 171
7.1
7.2
7.3
Cues for assessing grammatical complexity of gender systems . . . . . . . 183
Gender complexity metric . . . . . . . . . . . . . . . . . . . . . . . . . . . 185
GCSs of the languages of the sample . . . . . . . . . . . . . . . . . . . . . 187
A.1 The language sample . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 210
A.2 The language sample: alphabetical index . . . . . . . . . . . . . . . . . . . 215
C.1 Number of gender- and number-indexing targets in the languages of the
sample . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 225
D.1
D.2
D.3
D.4
Lexical
Lexical
Lexical
Lexical
plurals
plurals
plurals
plurals
in
in
in
in
Iraqw . . . . . .
Borana-Arsi-Guji
Rendille . . . . .
Baiso . . . . . .
. . . . .
Oromo
. . . . .
. . . . .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
227
228
228
228
E.1 Singular markers in Mabaan, Dinka and Luwo . . . . . . . . . . . . . . . . 229
E.2 Plural markers in Mabaan, Dinka and Luwo . . . . . . . . . . . . . . . . . 230
F.1 Complexity scores . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 231
xii
Abbreviations
The glossing of the examples conforms to the Leipzig Glossing Rules: http://www.
eva.mpg.de/lingua/resources/glossing-rules.php. In case of conflict between the
Leipzig Glossing Rules and the abbreviations used in my sources, nonstandard abbreviations were substituted with the corresponding standard abbreviations in the Leipzig
Glossing Rules. The glossing of the examples was adapted accordingly. The orthography
used in the examples is the same as the one used in the individual sources. Tones are
thus only marked if marked in my sources.
2
second person
3
third person
abl
ablative
abs
absolutive
acc
accusative
adj
adjective
aug
augmentative
c
common gender
cl
noun class
cop
copula
crd
coordination
dat
dative
decl
declarative
def
definite
dem
demonstrative
det
determiner
dim
diminutive
xiii
Abbreviations
ds
different subject
dup
reduplication
f
feminine
gen
genitive
general
general number
icp
instrumental-comitative-perlative
indf
indefinite
m
masculine
n
neuter
neg
negation, negative
nom
nominative
obj
object
pass
passive
pauc
paucal
pco
perfective converb
perm
permissive
pfv
perfective
pl
plural
poss
possessive
prf
perfect
prs
present
pst
past
recp
reciprocal
red
reduplication
rel
relative
xiv
Abbreviations
rp
recent past
sbj
subject
sg
singular
subord
subordination marker
unm
unmarked
voc
vocative
xv
Acknowledgements
The path towards the completion of this dissertation is populated by people, places,
random extraordinary events and ordinary daily routines that have inspired me in various
ways and contributed to my growth as a linguist and as a human being. It is to this
beautiful crowd of individuals and life experiences that I would like to say “Thank you!”
today.
First of all, my deepest gratitude goes to the Department of Linguistics at Stockholm
University for welcoming me as a PhD student and for giving me the opportunity to
become part of the most friendly and open-minded environment that a researcher could
wish to live in.
My supervisors, Maria Koptjevskaja Tamm and Ljuba Veselinova, provided me with
rigorous guidance and friendly support from the first day to the last. I thank them for
letting me work independently throughout my doctoral studies, while, at the same time,
always being there to express the most thorough and thought-provoking criticism as
well as the most sincere enthusiasm towards the progress of my work. Our relationship
during these four years has been based on a perfect combination of mutual trust, respect,
empathy and a good dose of humour, all of which have contributed to create an ideal
atmosphere for me to work through during the ups and downs of my life as a PhD
student. Thank you also, Masha and Ljuba, for being such a great source of inspiration
on how to combine professional fulfillment with the challenges of being a woman and a
mother.
A preliminary draft of this dissertation was read and examined by Östen Dahl in
the spring of 2014. His sharp comments, suggestions and insights have been precious
for improving and finalising my work. Matti Miestamo, Micke Parkvall and Bernhard
Wälchli contributed invaluably to my chapter on complexity by helping me designing the
study and reading the first drafts of the chapter. Raphaël Domange, Thomas Hörberg,
Robert Östling, and Bernhard Wälchli patiently helped me with the statistical analysis
of my data on complexity.
I am very grateful to Eva Lindström for translating the summary of the dissertation
into Swedish and to Lamont Antieau for his meticulous proofreading. Sofia Gustafson
Capková patiently helped me find my way through the six-month procedure that precedes
the doctoral defence. Ljuba Veselinova managed to find some extra time to design the
beautiful maps that accompany the thesis text. Paolo Cerva designed the perfect cover
for this dissertation.
Many people helped me with data collection and answered endless questions on their
languages or theoretical domains of expertise. First of all, my deepest gratitude goes to
the members of the French research institute LLACAN (Langage, Langues et Cultures
d’Afrique Noire) that hosted me for two weeks in the spring of 2012. I am particularly
xvii
Acknowledgements
grateful to Dmtry Idiatov, Maria Khachaturyan (Mande), Stefano Manfredi (Standard
Arabic), Amina Mettouchi (Berber), Konstantin Pozdniakov (Atlantic), Nicolas Quint
(Kordofonian), Guillaume Segerer (Atlantic), Yvonne Treis (Cushitic), Mark Van de
Velde (Bantu) and Martine Vanhove (Beja) for giving me the opportunity to discuss
my work with them and for sharing with me their knowledge, resources and fieldwork
data. I am also very grateful to Clement Appah (Akan), Nana Aba Amfo (Akan),
Tom Güldemann (Khoisan), Bernd Heine (Khoisan), Serge Sagna (Bandial), Graziano
Savà (Cushitic and Omotic), Anne Storch (Western Nilotic), and Suzanne van de Meer
(Cushitic) for their generosity in sharing data on their languages of expertise. Desalegn
Hagos Asfawwesen answered a long list of questions on Amharic, his mother tongue,
and Yvonne Agbetsoamedo helped me with Ewe, her mother tongue, and SElEE. Thank
you also, Yvonne, for the great honour of working together on temperature terms and
evaluative morphology in SElEE. The pioneering work on evaluative morphology conducted by Nicola Grandi and Livia Körtvélyessy greatly inspired my own research. I
thank them for the always interesting discussions we had as well as for inviting Yvonne
(Agbetsoamedo) and me to participate in their project on the Handbook of Evaluative
Morphology. I am also very grateful to Jenny Audring for her inspirational work on
gender as a complex feature of grammar and for the numerous email exchanges we had.
Thanks to Harald Hammarström, Suzanne van de Meer and Hedvig Skirgård for incorporating and adapting some of the features of my database into their questionnaire for
the Nijmegen Typological Survey.
This dissertation could have never been accomplished without the support and generosity of the people mentioned above. Shortcomings and mistakes in the analysis and
interpretation of the data are, of course, entirely mine.
I am immensely grateful to all my colleagues within the General Linguistics session
at the Department of Linguistics at Stockholm University, for an exceptionally inspiring
working environment, and for the always cheerful and intellectually stimulating chats
during our måndagsfika, when I would try to practise my Swedish while listening to the
most interesting stories about Swedish society, history and culture. I would not have
made it through these four years without the solidarity of the other doctoral students
in linguistics within and outside the department: Andrea (Kiso), Benjamin (Brosig),
Calle (Börstell), Clelia (La Monica), Desalegn (Hagos Asfawwesen), Emil (Perder),
Franco (Pauletto), Geraldine (Quartararo), Ghazaleh (Vafaeian), Hatice (Zora), Kerstin
(Lindmark), Lena (Renner), Natalia (Perkova), Pernilla (Hallonsten Halling), Robert
(Östling), Susanne (Vejdemo), Thomas (Hörberg), Yvonne (Agbetsoamedo) and all the
others. Thank you guys for our countless lunch breaks in the gloomy darkeness and
the bright sunshine of the Swedish winters and summers, for the many adventures we
experienced together, the joyful moments and the stressful ones. Robert, Yvonne, Kerstin, and, although very briefly, Miina (Norvik) and Gintaré (Grigonyté), you have been
the perfect office mates during these years. Thank you for tolerating my mess and for
brightening up all my days at work. Finally, I would like to thank Nada Djokic, Linda
Habermann, Cilla Nilsson and Rickard Franzén for their patience, efficiency and kindness and, most importantly, because, no matter what they were doing and how busy
they might have been, they always welcomed me with a smile.
xviii
Acknowledgements
The story that brought me to Stockholm began between the spring and summer of
2009, in Albuquerque (New Mexico) where I studied for five months under the supervision of Bill Croft. Thank you, Bill, for giving me the first hint that interactions of
gender with other nominal features could be an interesting phenomenon to look at, as
well as for letting me work on your collection of reference grammars during the sunny
mornings of the high desert summer.
Luisa Brucale, already more than ten years ago you made me fall in love with linguistics. Giorgio Iemmolo, from the old good days in the courtyard of the Department
of Classical Philology at the University of Palermo, you have been the best companion
of linguistic adventures. Shelece Easterday, our deeply felt friendship goes beyond the
ocean and the many time zones that separate us.
It is not easy to move to a new country as an adult and build a life there. I was lucky,
because in Stockholm I met some of the most precious people of my life. Adeile, Fabio,
Franco, Geraldine, you have been and are my beloved “Little Italy” in Stockholm. Eti
and Chris, my weeks are brighter since you came back to Stockholm and brought Elsie to
this world. Ljuba and Vera-Katlheen, the sun always shines in your company. Andrea,
you are the first friend that I made in Sweden, and have become one of the closest.
Stockholm is much more empty since you and Gladstone left. Marjatta, I arrived at
your place on a cold Sunday afternoon, you welcomed me with kanelbullar and a cup
of strong Swedish coffee, and I immediately felt at home. Your beautiful house soon
became the most secure place that I know of within Sweden. A couple of days later,
on the day of my 27th birthday, Yvonne, you also arrived in Stockholm. It didn’t take
long to realise that you are my third sister, born on the other side of the world and from
different parents, but from whom I will never be apart again. Benjamin, the world would
be much happier if everyone could have a friend like you, your generosity and altruism
trascend obviousness, and this is what makes them so precious.
Much of who I am has come to be thanks to the friends I grew up with. Together
we explored our hometown, our island and many places outside of it. Together we also
learned to be apart, to make the best out of the always little time available, and to find
our way through one of the most brutal political and economical crises that Italy has
ever experienced. Anna, Carla, Dario, Diana, Francesca, Graziana, Germana Co. and
Germana Civ., Giorgio, Lollo, Marco, Nadia, Paolo, La Seppia (Peppe Filippi), Rosi and
Sara, thank you for all the years together, for your visits to Stockholm and for making
Berlin, Milan, Paris, Reggio Emilia, Tunisi or Zurich “another” Palermo to visit and in
which to find myself.
My sisters, Chiara and Agnese, and my brother-in-law, Cristiano, are fundamental
pillars of my daily existence. I am very grateful to them for being there when I cannot
be there, and for visiting me here every time they could. Ever since I was born, my
parents, Vera and Enzo, have never left me alone. Their support has been unconditional
without ever being oppressive. They taught me to always look at what’s around me with
humility and to never stop learning. Raphaël gives a sense of direction to all my days.
He is home and future. From my grandmother, Lilla, I learned how to cook, sew, knit
and water plants, but, most importantly, she taught me how to stand my ground and
to never give up. Unfortunately, she passed away in December 2013 and could not see
xix
Acknowledgements
the end of this work. I dedicate this dissertation to her and the memory of her beautiful
rebellious eyes that always saw beyond appearance.
xx
1 Introduction
Gender is one of the most extensively researched domains of grammar. In the typological literature on gender systems, a great number of studies have focussed on identifying
and classifying the types of gender systems attested in the world’s languages, mapping
their distribution across genealogical units and geographical areas, and, to a lesser extent, attempting to model their origins and functions in grammar and discourse. This
dissertation focusses on a less-explored aspect of the nature of gender systems: their
interactions with other domains of grammar.
Gender is often claimed to be closely linked to other grammatical domains such as
argument marking, definiteness, number or evaluative morphology. These connections
can be both morphosyntactic and semantic in nature. For instance, gender distinctions
can be realised together with case, definiteness or number distinctions; in Italian, e.g.,
gender values are always realized together with number values. Or, individual genders
in a language can be associated with the encoding of diminutive and augmentative
meanings; e.g., in Serbian-Croatian, the Neuter Gender is associated with diminutive
meanings (Grandi 2001).1 Accounting for these interactions can be very relevant to
understanding how the gender system of a language functions synchronically, or how
it developed diachronically. Even though phenomena of this type are likely to be at
least mentioned in reference grammars, a systematic survey of the kinds of interactions
attested between gender and other domains is still missing in the typological literature.
This dissertation investigates the morphosyntactic and semantic interactions of gender
systems with number and evaluative morphology. This is done by combining crosslinguistic research and in-depth studies of genealogically related languages within one continental area, Africa. Diachronic and synchronic aspects of the interactions among the
above-mentioned grammatical domains are considered. The notion of interaction is used
in this dissertation in a rather loose way. In fact, in African languages in particular,
gender and number, on the one hand, and gender and evaluative morphology, on the
other, can be so closely intertwined with each other that they do not only interact but
exhibit indications of actually being merged.
The dissertation is built on three main research foci:
(1) Interactions between gender and number
1
In this dissertation, I use capital letters to refer to language-specific categories (e.g., the Neuter Gender
in Serbian-Croatian) and lowercase letters to refer either to a specific marker within a language
(e.g, the neuter marker in Serbian-Croatian) or to grammatical domains as objects of crosslinguistic
comparison (e.g., the neuter genders in the languages of the sample). See Haspelmath (2007) for a
discussion on this topic.
1
1 Introduction
– Do gender and number have the same encoding? How are gender distinctions
mapped across number values?
– Do gender and number compete through indexation patterns?
– Do gender and number have the same relevance to nouns?
– Do gender and number interact semantically?
(2) Interactions between gender and evaluative morphology:
– Can size be a criterion for gender assignment?
– How do diminutive and augmentative genders interact with the other gender
distinctions of a language?
(3) Interactions of gender and grammatical complexity:
– Is it possible to measure the grammatical complexity of gender systems?
– Can interactions of gender with other domains of grammar be accounted for
by such a complexity metric?
– What is the role of these interactions in the overall complexity of a gender
system?
The dissertation is organised as follows. In chapter 2, I introduce the three grammatical domains under investigation by looking at their functional and semantic properties,
their structural encodings and typological distributions. The chapter also provides an
overview of previous literature on the interaction between gender and number and gender
and evaluative morphology as well as on grammatical complexity. The research questions addressed in this study, and already partially introduced here, are outlined at the
end of chapter 2. In chapter 3, I describe the sampling methodology established for this
study as well as the protocol for data collection and organization. Chapter 4 provides
an overview of the gender, number and evaluative morphology systems attested in the
languages of the sample. The results of the investigation are presented and discussed in
chapter 5 (gender and number) and chapter 6 (gender and evaluative morphology). In
chapter 7, I propose a metric for measuring the grammatical complexity of the gender
systems attested in the language sample and show how morphosyntactic and semantic
interactions with number and evaluative morphology can be accounted for by this metric. A summary of the results, prospects for future research and concluding remarks are
presented in chapter 8.
2
2 Background: gender, number, evaluative
morphology and relevant interactions
In this chapter, I define the grammatical domains investigated in the dissertation –
gender, number and evaluative morphology. I also provide an overview of previous
literature on the interactions between gender and number and gender and evaluative
morphology, on the one hand, and the grammatical complexity of gender systems, on
the other. The chapter thus consists of two parts. In the first part, gender, number and
evaluative morphology are defined along four dimensions:
(1) Semantics and functions
(2) Morphosyntax
(3) Typological distributions
(4) Diachrony
The three domains investigated in the dissertation are each allotted a separate section,
starting with gender (§2.1), continuing with number (§2.2) and concluding with evaluative morphology (§2.3).
Previous studies on the patterns of interaction between gender and number and gender
and evaluative morphology as well as on the grammatical complexity of gender systems
are discussed in the second part of the chapter (§§2.5, 2.6 and 2.7, respectively). In
§2.8, an outline of the research questions addressed in the dissertation is provided. A
summary of the chapter is presented in §2.9.
2.1 Gender
Gender is a type of noun classification strategy. The label noun classification strategy
is conventionally used to refer to heterogeneous sets of grammatical constructions that
are specialised in the categorisation of nouns (Aikhenvald 2003). Different types of noun
classification strategies can be determined according to the following criteria: meaning,
number of distinctions within the system, locus of marking, historical development and
degree of grammaticalization. Following Corbett (1991, 2013a,b), I define gender as the
particular type of noun classification strategy which must be reflected beyond the nouns
themselves, that is, through indexation patterns or, following Corbett’s terminology,
“agreement” (see §2.1.2.1 for a critical reappraisal of the notion of agreement).
3
2 Background: gender, number, evaluative morphology and relevant interactions
The notion of gender adopted in this dissertation conforms to that of the mainstream
typological literature, in which gender and noun classes are viewed as the same grammatical phenomenon and the term gender is used as a hyperonym of the two. When
describing language-specific types of gender systems, I use only the label gender for systems that are built on a limited number of distinctions (from two to three, e.g., animate
vs. inanimate, or masculine vs. femine vs. neuter). Conversely, for richer systems (that
make four or more distinctions) I use the term noun class to refer to individual singular
and plural markers and the term gender to refer to the system as a whole and, particularly, to the pairings between singular and plural class markers (see also Aikhenvald
2003; Corbett 1991, 2013b).
In §2.1.1, I discuss gender assignment rules, that is, the dynamics according to which,
in a given language, nouns are distributed across different genders. The morphosyntax
of gender and patterns of gender indexation are discussed in §2.1.2. Finally, the function
of gender, the distribution of gender systems across the world’s languages and aspects
of their historical development are discussed in §2.1.3.
2.1.1 Gender assignment
Nouns can be assigned to a given gender on the basis of a combination of semantic
and formal (morphological and/or phonological) properties. All gender systems are in a
sense mixed since “there is always a semantic core [...] but this is never the entire story”
(Aikhenvald 2003: 22).2
The semantic underpinnings of the gender system of a language can be very blurred
in actual use. Nonetheless, as shown by Corbett (1991) (see also Aikhenvald 2003; Dahl
2000a,b), it is always possible to recall the fundamental semantic notion(s) on which
the categorisation is based, even if only for a restricted portion of the nominal lexicon
of a language. Semantic gender assignment generally involves nouns denoting animate
entities, with the cut-off points within the domain of animacy being highly languagespecific: “between humans and animals, between higher and lower animals, or between
animals and inanimates” (Dahl 2000a: 101). In many languages, for example, only nouns
denoting sex-differentiable entities – that is, humans and animals – can be assigned to a
gender on a semantic basis. The rest of the nominal lexicon is distributed in one gender
or the other on the basis of semantically more arbitrary morphophonological patterns or
mechanisms of semantic extension such as metaphorical or metonymic associations.
Animacy, sex, shape and size are the most common semantic notions upon which
a gender system is based.3 Their distribution across language-specific gender systems
is not equal: sex-based systems are more frequent than other types of systems, which
are generally built upon some notion of animacy (Corbett 2013c). According to the
2
In Uduk (Koman), semantics seems to play no role in the allocation of nouns to one gender or the
other (Don Killian, personal communication). The language is currently being described, but, if this
preliminary analysis holds, the gender system of Uduk represents an interesting challenge for the
current understanding of grammatical gender crosslinguistically.
3
See Croft (1994) for a discussion of the significance of the notion of shape in the semantics of noun
class systems.
4
2.1 Gender
data in WALS, in almost all geographic areas where gender is found, sex-based gender
systems are also found: 84 out of the 112 gendered languages in Corbett’s (2013c)
sample have sex-based gender.4 Moreover, on a general basis, animacy and sex are more
frequent criteria of gender assignment than physical properties (shape and size). In many
languages, “physical properties are only rarely employed to assign genders to animates”;
their relevance as a criterion of gender assignment is restricted to inanimate entities
(Aikhenvald 2003: 278). Finally, among the four most typical semantic underpinnings
of gender, size seems to be the least likely to occur as an independent criterion for
classification (see also §2.6).
In morphological gender assignment, individual word-formation strategies and/or inflectional classes tend to be associated with a particular gender. The nature of these
associations is often problematic, and it is not obvious how or if one dimension (gender or morphological class) derives from the other. The regularity of these associations
is largely language-specific, and, within languages, it can vary according to individual
morphological patterns. In Italian, for example, the suffix for action nominalizations,
-zione, is used to derive feminine nouns. Conversely, when the morphological criterion
and the gender value that would be assigned to a noun according to its semantics are
in conflict, morphology can be overridden by semantics. For example, the Italian noun
soprano is morphologically equivalent to the noun vaso ‘vase.’ These nouns are grammatically masculine. However, since soprano typically denotes female singers, speakers
tend to treat it as feminine.
The notion of phonological gender assignment is a tricky one. In most cases, it is
in fact very difficult to tease it apart from that of morphological assignment. In order
to distinguish between the two, Corbett (1991: 51) proposes the following rule: if the
gender of a noun can be established by taking into consideration more than one form, we
are dealing with morphological assignment; if the gender of a noun can be established
based on one form only, we are dealing with an assignment rule that is also a phonological
rule. The most frequently quoted language with phonological gender assignment is Qafar
(Afro-Asiatic, Cushitic, East Cushitic) (Corbett 1991: 51-52). In Qafar, nouns whose
citation form ends in an accented vowel are assigned to the Feminine Gender. The overlap
between semantic and phonological properties of nouns is generally systematic. With
a few animate nouns, phonology and semantics are in conflict; in such cases, semantics
wins. Thus, the word for ‘father’ abbà, despite its phonology (it ends in an accented
vowel), is assigned to the Masculine Gender.
Dahl (2000a) reassesses the dichotomy between semantic and formal dynamics of gender assignment by introducing a distinction between lexical and referential gender. In
principle, both can be viewed as instances of semantic gender, even though in different
ways and to different extents. Lexical gender refers to the denotation of nouns, that is,
to semantic properties of nouns as lexical items; it may also be conditioned by formal
properties that are characteristic of a noun as a lexeme belonging to a certain class with
certain morphological properties. On the contrary, referential gender is concerned with
4
The total number of languages in Corbett’s sample is 257; out of this sample, 145 languages do not
have gender.
5
2 Background: gender, number, evaluative morphology and relevant interactions
semantic properties of the noun phrase (henceforth np) referents in the extralinguistic
world and is, as such, always semantic. The Russian word for ‘judge’ sud’ja is presented
by Dahl (2000a) as an illustration of this polarity. Russian grammars treat the noun
as masculine. This is in line with the general (semantic) pattern according to which
underived nouns denoting professions are always masculine, with the exception being
those nouns denoting professions that are stereotypically associated with women (medsestra, ‘(medical) nurse,’ from Dahl 2000a: 109). Conversely, sud’ja ends in -a, which
is typically used as a feminine suffix. In actual spoken language, and when referring to
a female judge, Russian speakers are conflicted between treating the noun as feminine
strogaja sud’ja ‘strict (fem.) judge’– thus based on referential semantics – or as masculine strogij sud’ja ‘strict (masc.) judge’ – thus based on the lexicalized, stereotypical
association between the profession and the Masculine Gender. Given that the noun ends
in -a, a prototypical feminine suffix, the use of masculine indexing forms associated with
a female referent cannot be regarded as an instance of formal assignment. Following
Dahl’s analysis, this case can be rather viewed as a conflict between assignment criteria
that are both semantic, even though in different ways, that is, lexically and referentially.
One last phenomenon to be mentioned in relationship with gender assignment is the
existence of non-rigid gender assignment rules, whereby nouns are assigned to different
genders depending on the context in which they occur. In his survey of gender systems of
African languages, Heine (1982) introduces the distinction between free and fixed gender.
Free gender systems are those that allow nouns “to change the gender affiliation” (Heine
1982: 198), while fixed gender systems are those in which nouns are rigidly and invariably
assigned to a gender. Heine suggests that in free gender systems, ideally, any noun could
be allocated to any gender. In reality, he notices that in the African languages with
relatively free gender systems, semantic or morphological gender assignment preferences
exist at least for some nouns. Corbett (2013a) refers to the possibility of multiple gender
affiliation as recategorization, whereby speakers modify the construal of the np referent
by manipulating its gender assignment. The notion of recategorization was previously
used by Corbett (2000, 2012) in relationship to nominal number to describe those cases
in which speakers manipulate the countability properties of nps by treating mass nouns
as counts and counts as masses (see also Lyons 1968; Quirk et al. 1985). According to
Corbett, typical instances of recategorization are the English utterances I’d like three
coffees, please, and There was dog all over the road (Corbett 2000: 85). In the former
case, the mass noun coffee is treated as countable and thus receives plural marking.
In the latter case, the countable noun dog is construed as mass and thus cannot be
pluralized.
Corbett (2013a) suggests that, when referring to gender systems, the most obvious
type of recategorization is based on sex-differentiability. He mentions the case of Kupto,
a West Chadic language spoken in Nigeria, in which animal nouns are usually treated as
feminine. However, in storytelling, and if personified, certain animal nouns – e.g., the
word for ‘hyena,’ ‘elephant’ or ‘mouse’ – can be treated as masculine. In languages with
sex-based gender, inanimate nouns tend to be assigned to the feminine or the masculine
gender depending on the way the np referent is construed in the discourse (Corbett
2013a: 123). According to Corbett, recategorization through gender can also be based on
6
2.1 Gender
the notions of size and value. This happens in languages in which nouns can change their
gender when speakers want to express diminutive and augmentative meanings – both in
terms of size variation (small vs. big) and value (good vs. bad). To date, very little
has been done to explore the crosslinguistic distribution of this phenomenon. Corbett
(2013a) mentions a handful of languages where size and value-related recategorization is
found – e.g., Lavukaleve (Lavukaleve), Maung (Iwaidjic), Savosavo (Savosavo), Walman
(Nuclear Torricelli), Yawuru (Nyulnyulan). Similar phenomena are also investigated by
Aikhenvald (2012) with a special focus on the languages of Papua New Guinea.
Recategorization phenomena related to gender are more commonly referred to in the
literature as gender shifts. The latter label is adopted in this dissertation. I use the label
manipulable gender assignment to refer to the possibility of assigning nouns to multiple
genders (“free gender” according to Heine 1982) and the label rigid gender assignment
(“fixed gender” according to Heine 1982) to refer to those languages where nouns are
invariably assigned to one gender. Gender shifts and manipulation of gender assignment
are explored in detail in §5.7, as well as in chapters 6 and 7.
2.1.2 The morphosyntax of gender: indexation
Hockett (1958: 231) defines gender as a grammatical category that is “reflected in the
behavior of associated words.” As a consequence, for the gender system of a language to
be considered productive, gender needs to be cross-referenced by those elements in the
utterance that entertain some kind of morphosyntactic and/or semantic relation with the
noun itself or the referent of the np (typically, demonstratives, determiners, pronouns,
relative pronouns, adjectives, verbs, but also adpositions, complementizers and phrase
markers). Cross-reference patterns are traditionally treated in the literature as agreement
phenomena. However, following an already established tradition in typological literature
(see, among others Croft 2001, 2003, 2013; Iemmolo 2011), in this dissertation the term
indexation is used rather than agreement. The reasons behind this choice are briefly
explained in §2.1.2.1.
2.1.2.1 Beyond agreement: the notion of indexation and what we gain with it
Typically, the term agreement refers to a particular type of asymmetric syntactic relationship between two entities within an utterance. This syntactic relation is such that
one of the two entities takes an inflectional form that is determined by certain semantic
or morphosyntactic properties of the other entity (typically gender, number, person). In
his model of agreement, Corbett (1991, 2000, 2006) refers to the former entity as target
and to the latter as controller. According to Corbett, the most typical or, to use his
terminology, “canonical”, instances of agreement need to comply with, among others,
the following conditions (for a complete list of the properties of canonical agreement, see
Corbett 2006: 9):
(a) The controller is nominal and overtly expressed in the discourse.
(b) The syntactic dependence between the controller and the target is local in the
7
2 Background: gender, number, evaluative morphology and relevant interactions
sense that, within a given utterance, the “structural distance” (Corbett 2006: 21)
between controller and target is minimal.
Thus, the notion of agreement implies that “there is a phrase in the utterance that
is ‘agreed with’ (the controller ) that is necessarily present” (Croft 2003: 34). In fact,
crosslinguistic evidence suggests that this is far from being the most common state of
facts and that the whole notion of agreement can be very misleading. For instance, contrary to what is stated in (a), in many languages both pronouns and np-internal targets
do not necessarily presuppose the presence of a syntactic antecedent or a controller (for
a useful discussion see Croft 2013). In addition, contrary to what is stated in (b), in the
case of pronominal targets, even when an antecedent is overtly coded in the discourse,
the anaphoric pronoun and the syntactic antecedent are usually not structurally close to
each other (for a more detailed discussion see §2.1.2.2). Finally, to give an example that
is even more closely related to the grammatical phenomena investigated in this thesis,
in many languages in which number marking on nouns is not obligatory, information
about the number construal of the np referent can be marked throughout the discourse
(e.g., on verbs or pronouns) despite there being no overt linguistic entity that can be
interpreted as being agreed with.
In light of the examples discussed so far, I use the term indexation to refer to the
grammatical strategies that languages use to signal (1) lexical and grammatical properties of nouns or (2) semantic properties of np referents, independently of the presence of
an overtly expressed syntactic antecedent/head noun (for a similar understanding of the
notion of indexation, see also Croft 2001, 2003, 2013). More specifically, this dissertation
investigates indexation patterns that are used to signal gender and number. Accordingly,
I use the terms indexing target or, alternatively, index to refer to entities whose inflectional morphology is used to signal gender and number throughout the discourse (e.g.,
adjectives, determiners, verbs, pronouns, etc.). The term syntactic antecedent is used
in cases of np-external indexation to refer to the np indexed by the pronominal target.
Finally, the labels trigger or indexation trigger are used to refer to the entities that
activate the use of a certain indexation pattern (i.e., pronouns and nouns) in a given
discourse domain.
Indexation patterns can differ depending on the np-internal or np-external nature of
the indexing targets. Figure 2.1 (courtesy of Östen Dahl) proposes one way of interpreting some major differences between np-internal and np-external indexing targets. As
shown in figure 2.1, in np-internal and np-external indexation referents can be signalled
by the indexing targets with or without the intermediation of a head noun or a syntactic
antecedent, respectively. In addition, in np-internal indexation the relationship between
indexing targets and referent is mediated by the internal structure of the np. The indexing target can signal properties of the referent either as lexically/grammatically realised
on the head noun or directly, that is, without the intermediation of the head noun. Finally, the information to be indexed may flow from the np to the head noun (as, for
instance, in the case of number) and vice versa, from the head noun to the np (as, for
instance, in the case of gender). Similarly, in np-external indexation, the relationship
between referent and indexing target can be direct, if the syntactic antecedent is missing,
8
2.1 Gender
or indirect, if it is overtly expressed in the discourse.
Figure 2.1: np-Internal and np-External indexation
In the literature on indexation and agreement phenomena, differences between npinternal and external indexation have been modelled in the form of a typological hierarchy known as the Agreement Hierarchy. As Croft (2013: 112) puts it, the hierarchy
attempts to describe “the degree of syntactic closeness or integration” between different
indexing targets and the syntactic antecedent or the head noun. This is discussed in
detail in §2.1.2.2.
2.1.2.2 Patterns of gender indexation
One – very often debated – problem in the literature on gender is how to account for those
languages, such as English, in which the only evidence for gender distinctions appears on
pronouns. In the literature on agreement, pronouns are often defined as non-prototypical
agreement targets insofar as they “violate the expectation that agreement targets should
share a local domain with their antecedent, preferably the phrase” (Audring 2009). However, in spite of their non-prototypical status, in the literature on (gender) agreement,
pronouns are considered to be possible agreement targets (Audring 2009; Corbett 1991,
2006, 2012, 2013a). Within the indexation model introduced in §2.1.2.1, pronominal and
np-internal indexes are also part of one and the same functional domain in the sense
that they all function as strategies for signalling reference through the discourse (on the
9
2 Background: gender, number, evaluative morphology and relevant interactions
functional continuum between np-internal and np-external indexing strategies, see also
Barlow 1992; Corbett 2006; Croft 2013; Siewierska 1999, 2004).
Based on these assumptions, which I also hold in this dissertation, languages like
English are considered to be gendered languages, despite their gender system being less
pervasive in discourse than gender systems in languages with richer indexation are. In
this thesis, gender systems of the English type are singled out through the use of the
label pronominal gender systems. Pronominal gender systems are crosslinguistically very
rare5 (Audring 2009; Corbett 2013b), and, as shown in the typological survey carried
out by Audring (2009), they tend to pattern with strictly semantic principles of gender
assignment. Applying Dahl’s (2000a) dichotomy between lexical and referential gender,
one could think of gender systems of the English type as being referential in nature. In
languages with pronominal gender systems, gender indexation signals salient properties
of the np referents, e.g. male vs. female vs. sexually undifferentiated entities, rather
than aspects of the lexical semantics of nouns.
Examining gender-related indexation patterns within a language is crucial for establishing the number of genders in that language. On the basis of the observation of
independent patterns of indexation, indexation classes (or, in Corbett’s terms, “agreement classes”) can be established. These are defined as sets of nouns whose indexing
targets have the same morphological realisations “under all conditions, that is, if we
hold constant the values of other features such as case and number” (Corbett 2013a).
Examples (2.1) and (2.2) illustrate the indexation patterns associated with two sets of
Italian nouns, one set belonging to the Masculine Gender and the other to the Feminine.
(2.1) The masculine indexation class in Italian (Indo-European, Romance)
(constructed example)
(a) un-o
zio
fantastic-o
indf-m.sg uncle.m.sg fantastic-m.sg
‘a fantastic uncle’
(b) un
dolce
delizios-o
indf.m.sg dessertm.sg delicious-m.sg
‘a delicious dessert’
(2.2) The feminine indexation class in Italian (Indo-European, Romance) (constructed
example)
(a) un-a
zia
fantastic-a
indf-f.sg aunt.f.sg fantastic-f.sg
‘a fantastic aunt’
(b) un-a
nuov-a
chiave
indf-f.sg new-f.sg key.f.sg
‘a new key’
5
The typological distribution of pronominal gender systems is rather striking. In principle, gender
systems of the English type “would appear to be the most natural and well-motivated ones,” and,
also, the easiest to learn (Östen Dahl, personal communication).
10
2.1 Gender
Both nouns in example (2.1) trigger the same indexation pattern, a pattern that
differs from that of the nouns in (2.2). The two noun sets belong to two independent
indexation classes and thus to two different grammatical genders, the Masculine (because
many nouns denoting males are assigned to this gender) and the Feminine (because many
nouns denoting females are assigned to this gender). Italian is often described in the
literature as a language with overt gender, that is, a language in which the gender of
a noun can be inferred from the morphological appearance of the noun itself. Even
though gender assignment in Italian can be often predicted on the basis of the final
vowels of nouns, this is true only to a certain extent. The nouns dolce and chiave in
examples (2.1b) and (2.2b), for instance, have the same final vowel and share the same
type of number marking – that is, they both end in -e when singular and -i when plural.
However, since they trigger different indexation patterns, they do not belong to the
same gender. This is not an exceptional pattern in the language: the -e/-i class is a
very frequent noun class in Italian, and the nouns belonging to this class can be both
feminine and masculine (see Gudmundson 2012 for token frequency counts based on the
corpus of spoken Italian lip, Lessico di frequenza dell’italiano parlato (De Mauro et al.
1993)). In sum, examples (2.1b) and (2.2b) show that the covert/overt dichotomy is not
a very powerful criterion by which to classify gender systems and that the genders of a
language can be identified only by identifying indexation classes.
A less clear-cut, and very much debated, case in the literature on gender indexation
is Romanian (Indo-European, Romance). There are three indexation classes in the language, traditionally labelled as Masculine, Feminine and Neuter. These are illustrated
in example (2.3), where, for each gender, both singular and plural indexation patterns
are given. Notice that the examples show gender indexation on adjectives only. The
definite clitics also index the gender of the noun, but, for the sake of simplicity, this is
not accounted for in the examples. The indexation patterns illustrated in the examples
are summarized in figure 2.2.
(2.3) Gender indexation in Romanian (Indo-European, Romance) (adapted from
Corbett 1991: 150)
(a) Masculine Singular
bărbatul e bun
man.the is good.m.sg
‘The man is good’
(b) Masculine Plural
bărbat»ii sînt bun-i
men.the are good-m.pl
‘The men are good’
(c) Feminine Singular
fata
e bun-ă
girl.the is good-f.sg
‘The girl is good’
11
2 Background: gender, number, evaluative morphology and relevant interactions
(d) Feminine Plural
fetele
sînt bun-e
girls.the are good-f.pl
‘The girls are good’
(e) Neuter Singular
scaunul e bun
chair.the is good.n.sg
‘The chair is good’
(f) Neuter Plural
scaunele sînt bune
chairs.the are good-n.pl
‘The chairs are good’
Figure 2.2: The gender system of Romanian (adapted from Corbett 1991: 152)
As shown in example (2.3) and figure 2.2, in Romanian, nouns belonging to the Neuter
Gender do not have indexation forms that are uniquely associated with them. On
the contrary, they share one of their indexation patterns with nouns assigned to the
Masculine Gender and the other with nouns assigned to the Feminine Gender. Since
the language has three distinct indexation classes (ø/-i, -ă/-e and ø/-e), it has three
genders. For a more detailed account of the gender system of Romanian, see Corbett
(1991: 150-154).
12
2.1 Gender
In his detailed overview of patterns of gender indexation, Corbett (1991) discusses
the case of the hybrid nouns, that is, nouns that tend to take indexation for more than
one gender. The choice among different genders is constrained by the type of target
the nouns are indexed by and follows the predictions made by the Agreement Hierarchy
(Corbett 1991, 2006):
attributive > predicate > relative pronoun > personal pronoun
According to Corbett, the configuration of the hierarchy is such that the more rightwards one moves along the hierarchy, the more likely it is for semantic indexation (he
uses the term agreement) to override syntactic indexation. Thus non-prototypical indexing targets, such as the personal pronouns, are those that are more likely to trigger
indexation based on semantic properties of nouns (see discussion at the beginning of the
section). Indexation is defined by Corbett as “semantic” when it is consistent with the
biological gender of the np referent or other semantic properties of the noun; on the
other hand, indexation is defined as “syntactic” when it is consistent with the gender as
assigned by virtue of morphological and/or phonological rules. One of the most quoted
examples of hybrid nouns is the German noun Mädchen ‘girl’, which is formally assigned
to the Neuter Gender insofar as it is morphologically marked as diminutive (in German,
nouns marked by the diminutive suffix are assigned to the Neuter Gender). The example
below shows how targets that rank differently on the hierarchy manifest different types
of indexation according to the above-mentioned predictions.
(2.4) German (Indo-European, Germanic) (Corbett 1991: 228)
Schau dir dieses Mädchen an, wie gut
sie/es Tennis spielt
look
you this.n girl
at, how good she/it tennis plays
‘Look at this girl, see how well she plays tennis’
Gender indexation with the demonstrative can only be syntactic – dieses – whereas with
the personal pronoun, speakers can choose between semantic and syntactic indexation –
sie or es.
Following Dahl’s (2000a) distinction between lexical and referential gender (see §2.1.1),
conflicts between “semantic” and “syntactic” indexation can instead be seen as conflicts
between referential and lexical gender. Gender indexation is referential when it indexes
extralinguistic properties of the np referent, and it is lexical when it indexes aspects
of the denotations of nouns viewed as lexical items. These notions, however we label
them, play a crucial role in the understanding of some of the grammatical phenomena
investigated in this dissertation (see chapter 5 in particular). When accounting for
conflicts between possible indexation patterns in this study, the terminology proposed
by Dahl (2000a) will be used.
I shall come back to the Agreement Hierarchy in §2.2.4, where similar phenomena are
discussed in connection with number.
13
2 Background: gender, number, evaluative morphology and relevant interactions
2.1.3 The function, distribution and diachrony of gender systems
Gender is a highly grammaticalized phenomenon in language: it presupposes rather long
evolutionary chains and is generally very stable in the history of language families (Dahl
2004). Given these attributes, at least three questions concerning gender are worthy of
discussion:
1. What is the function of gender in grammar and discourse?
2. How are gender systems distributed among the world’s languages?
3. How do gender systems arise?
The rest of this section is devoted to the discussion of how answers to these questions
have been sought in previous research.
2.1.3.1 The function of gender
When discussing the function of gender in language, scholars usually align themselves
with one of the following assumptions:
1. Gender is a useless feature in human communication, and its appearance in grammar is the result of chance (due to grammaticalization) rather than functional
necessity (McWhorter 2001; Trudgill 1999).
2. Gender systems are functional systems; their functions are mostly connected with
reference tracking and disambiguation (Foley & Van Valin 1984; Greenberg 1978).
A new insight into the understanding of the function of gender has recently been
provided by Dahl (2004). According to this view, the idea that gender is devoid of any
communicative function does not mesh with its diachronic stability and the frequency
of its use in gendered languages. Conversely, viewing gender only as a grammatical
device for reference tracking and disambiguation is, according to Dahl, not completely
satisfactory either, since this explanation does not take into account the whole range of
occurrences of gender in a language (e.g., gender indexation internal to the np). Rather,
reference tracking and disambiguation “can be probably seen as one of several functions
of grammatical gender that are all part of a larger scheme of redundancy management”
(2004: 202). According to Dahl, gender systems synchronically function similarly to
checksum digit systems,6 that is, as devices for error checking in the online process of
lexical item selection by speakers. In a language such as French, in which grammatical
gender is marked on the article, speakers “know that a masculine article has to go with
a masculine noun,” and any other combination is perceived as a signal that “something
has gone wrong” (Dahl 2004: 202).
6
Checksum digits are added to bank account and card numbers as a security control device. They are
calculated on the basis of various algorithms. For instance, the checksum digit of a code could be
the last digit of the sum of its individual digits.
14
2.1 Gender
Dahl’s hypothesis of gender as a grammatical device for online error checking provides
– I believe – an insightful synchronic explanation for the function of gender in language.
His hypothesis applies to manifestations of gender in grammar not only at the level of
reference tracking and disambiguation but also in terms of indexation patterns internal
to the np. However, this still leaves the puzzle of the function of gender at the level
of diachronic explanation unresolved at this point. The issue is discussed in detail in
§2.1.3.3.
2.1.3.2 The distribution and stability of gender systems
According to the data in WALS, out of 256 languages, 112 languages, or less than half of
the language sample used by Corbett (2013c), have gender. Gendered languages tend to
cluster both genealogically and areally – and the occurrence of gender goes hand in hand
with the presence of rich morphology. Nichols (1992) describes the genealogically and
areally skewed distribution of gender systems as one of their most striking characteristics
worldwide. In this respect, she introduces the notion of class (i.e., gender) hotbeds and
class outliers. Hotbeds represent “areas in which most languages have classes, classes are
found in languages of more than one family, and the formal implementation of classes
[...] takes more than one form” (Nichols 1992: 130). Conversely, outliers are those
languages with grammatical gender that are located outside of hotbeds. An example of
a class/gender hotbed would be Africa, and an example of class/gender outlier would be
the isolate Burushaski (Nichols 1992: 130-131).
Gender is also one of the longest standing features in the history of language families.
For instance, within Indo-European, Armenian is the only primary branch in which the
gender system of the protolanguage is not reflected by any of the member languages
(Dahl 2004: 199). The diachronic stability of gender is directly proportional to areal
and genealogical entrenchment. Gender systems tend to be unceasingly inherited but
only rarely borrowed. The development of gender is the result of long evolutionary
chains whose stages can only be partially detected based on empirical data (see §2.1.3.3).
In languages with gender, gender distinctions are very much entrenched in the native
speaker’s usage. On the other hand, reduction and loss of gender systems by languages
appear to be correlated with degree of contact and of adult second language learning
(Dahl 2004; Trudgill 1999). This will be made clearer in chapter 7, where the impact of
language contact on the complexity of gender systems is discussed more at length.
2.1.3.3 The diachrony of gender
Given that grammatical gender generally presupposes long evolutionary chains, in most
cases, it is very difficult to establish how the gender system of a particular language
or language family originated. A typical example is the debate over the diachrony of
gender in the Indo-European languages: while many hypotheses and theories have been
formulated over the years, the issue still remains unresolved.7 With respect to systems
7
In this respect, see Matasovič (2004) and Luraghi (2011), two recent contributions that combine
comparative reconstruction with crosslinguistic evidence in formulating hypotheses on the origin of
15
2 Background: gender, number, evaluative morphology and relevant interactions
that have emerged relatively recently, more can be said about the development of gender.
One of the most relevant and quoted pieces of research on the origin of gender markers
is Greenberg (1978). In his paper, Greenberg analyses patterns of grammaticalization
involving gender in different branches of the Niger-Congo family and elsewhere in the
world as part of a three-stage process. The entire grammaticalization chain is illustrated
in figure 2.3.
Figure 2.3: The grammaticalization of gender from demonstratives, according to Greenberg
(1978)
As shown in the figure, demonstratives are interpreted by Greenberg as the ultimate
diachronic source of grammatical gender in the languages of his sample. Demonstratives
develop as definite articles when they begin being used to refer to entities that were
previously mentioned in the discourse and become obligatory in order to mark a referent
as having been identified (Stage 1). The next step in the grammaticalization chain
(Stage 2) is taken when the article begins being used with specific but non-identified
referents as well. At Stage 2, the absence of the article is restricted to generic contexts
only (as in the English sentence I like flowers as opposed to I like the flowers). An
additional step forward is made when one of the two options – usually the co-occurrence
of articles and nouns – is generalized to all contexts. At this stage (Stage 3), the article
has become a general marker of nominality and the grammaticalization chain has reached
its final step of development. The entire process is led, according to Greenberg, by the
need for “referential devices of identification” (1978: 78) or, in other words, by the need
to track np referents throughout a given stretch of discourse. Demonstratives play a
very central role within this scenario since they constantly foster indexation (Greenberg
1978: 77).8 A similar pattern of grammaticalization has been proposed by Heine (1997)
and Kilian-Hatz & Heine (2010) for the gender system of the Khoe languages. In this
case, the grammaticalization chain starts with third person pronouns which evolve as
markers of definiteness and eventually as markers of nominality, that is, as bona fide
gender markers.
Yet, a crucial question remains unanswered: how do markers of gender become attached to demonstratives (and/or third person pronouns) in the first place? The suggestion made in the literature (Corbett 1991; Dahl 2004; Greenberg 1978) is that certain
8
gender in the Indo-European language family.
For an overview of demonstratives as a source of grammaticalization of a wide range of grammatical
phenomena besides gender, see Diessel (1999a,b).
16
2.1 Gender
nouns such as ‘person,’ ‘thing,’ ‘female,’ and ‘male’ may acquire a classificatory function
and then start being regularly used with demonstratives, to the point of fusing with them
and thus result in gender marking. This can eventually spread further, for example, to
anaphoric pronouns and generate indexation (Dahl 2004: 199). Nouns, then, seem to be
the ultimate source of gender systems of this type.
An interesting case of a gender system originating from classificatory nouns is the
Australian language Ngang’ityemeri (Australian, Daly, Southern Daly), which has been
extensively investigated by Reid (1990, 1997). In Ngan’gytyemeri, multiple strategies of
gender marking are available:
• (generic noun)9 + specific noun + (generic noun) modifiers
• proclitic=Noun + proclitic=Modifiers
• prefix-Noun + proclitic=Modifiers
Diachronically, this system seems to have generated from the use of combinations of
generic and specific nouns as a frequent structure of nps and from the use of generic
nouns with noun modifiers as a strategy for reference tracking. Over time, increased
frequency of use resulted in morphological reduction and phonological attrition to the
extent that certain generic nouns became proclitic words and, in some cases, grammaticalized as prefixes. Synchronically, the distribution of the different strategies of class
marking is skewed according to semantics. For example, noun class marking with nouns
denoting males, females and human groups is always performed by means of obligatory
proclitic words marked on head nouns and noun modifiers. The individual stages of
the grammaticalization path from classifier nouns to gender markers are synchronically
instantiated by the multiple strategies of class marking listed above. Interestingly, there
seems to be a synchronic correlation between the size of a noun class and the degree of
grammaticalization of its marker: the larger the number of nouns assigned to a class,
the more likely it is for the marker of that class to be a bound morpheme and to trigger
indexation on noun modifiers (Reid 1997).
Contrary to Greenberg and Corbett’s hypothesis, whereby classificatory nouns are
viewed as the ultimate sources of gender, Nichols (1992: 139-142) proposes that in the
emergence of gender systems, indexation precedes actual noun categorization processes.
Indexation arises as a result of the grammaticalization of hidden – in the sense of not
morphosyntactically coded – animacy distinctions or other types of cognitively salient
distinctions. Once indexation surfaces, noun classes follow.
Whichever evolutionary pattern is hypothesised for individual gender systems, reconstructing the very early stages of development of gender is in many cases very critical.
The notion of exaptation (Lass 1990) or, to use Croft’s (2000) terminology, hypoanalysis,10 becomes thus useful for explaining the diachronic transitions from whatever functions the sources of gender had at the beginning of the grammaticalization chain - or at
9
10
The parentheses indicate optionality.
Both terms refer to processes of functional reanalysis of grammatical constructions. These occur when
contextual properties of a construction are interpreted by language users as its inherent syntactic
properties, and the unit as a whole acquires new meanings and functions (Croft 2000: 127).
17
2 Background: gender, number, evaluative morphology and relevant interactions
any stage along it - to the functions that full-fledged gender systems display synchronically. In sum, synchronic and diachronic explanations might be hard to relate to each
other in the case of longstanding grammaticalization chains as those presupposed by
gender systems.
For an overview of paths of grammaticalization of gender and noun class systems
across languages of the world, see Aikhenvald (2003: 352-412), in which the emergence
of gender in grammar is analysed within a more general discussion of the diachrony of
noun classification devices.
2.2 Number
Nominal number is one of the strategies available in human languages to quantify entities (linguistically encoded as nouns or pronouns). I use the label nominal number in
opposition to verbal number. The latter term is used to refer to strategies available in
the verbal domain in order to quantify events in terms of the number of times they occur
or the number of participants that are involved in them (Veselinova 2013).
Nominal number distinctions align with the following possible semantic cores: one,
more than one, two, three, four, a few, many. Conventionally, the grammatical correspondents of such semantic notions are: singular, plural, dual, trial, quadral, paucal and
greater plural, respectively. Nominal number distinctions can be encoded directly on the
nouns by means of affixation, via indexation or by means of plural words. The marking
of number on nouns or via indexation is discussed in details in §2.2.3 and §2.2.4. Plural
words are especially common among languages that lack inflectional number marking
(e.g., in the languages of Southeast Asia). The categorial status of plural words varies
substantially across languages: they can be articles or numerals, but they can also constitute a word class of their own or pattern together with other modifying words, most
often used to express size and evaluation. To date, the most relevant crosslinguistic
survey of plural words is that of Dryer (1989b).
Various types of number systems are found in the languages of the world, and these
systems depend on: (1) the interaction between the lexicon and grammar in the encoding
of number distinctions, (2) whether or not number marking is obligatory for all nouns and
(3) the internal distribution of number values within a language-specific number system.
In the next sections, I discuss each of these points in detail. In §2.2.1, I begin with
an overview of the countability properties of nouns and noun phrases. Major types of
number systems and number values in the languages of the world are discussed in §2.2.2.
In §2.2.3, I discuss the interaction between number and animacy; this is followed by an
overview of the morphosyntax of nominal number in §2.2.4. Finally, the distribution and
patterns of grammaticalization of nominal number are examined in §2.2.5.
2.2.1 Countability properties of nouns and noun phrases
Information about the quantificational properties of nouns can be expressed both grammatically and lexically. The quantificational properties of nouns at the level of lexical
semantics are referred to in the literature as countability properties and are generally
18
2.2 Number
treated as a manifestation of the polarity between count and mass (see, among others, Corbett 2000; Koptjevskaja-Tamm 2004). Nouns can be count or mass depending
on whether they denote “discrete entities with a well-defined shape and precise limits” or “homogeneous undifferentiated stuff without any certain shape or precise limit”
(Koptjevskaja-Tamm 2004: 1067). The same noun can be used both countably or uncountably depending on the context of its occurrence. This is illustrated in example
(2.5); the English noun table is uncountable in (2.5a) and countable in (2.5b).
(2.5) Allan (1980: 547)
(a) There’s not enough table for everyone to sit at.
(b) We need a bigger table.
Based on this evidence, Allan (1980) claims that countability properties are a characteristic of nps rather than of nouns as individual lexical items (see also the discussion in
Corbett 2000: 84-85.). Yet, he notes that, for instance, in English the countability properties of nouns can be empirically determined on the basis of the relative frequencies of
their occurrence as mass or count. Countability preferences can thus be viewed as part
of the lexical semantics of nouns that can then be manipulated at the level of construal
depending on the discourse context. Corbett (2000) uses the term recategorization to
refer to the manipulation of the countability properties of a noun at the level of construal
(see §2.1.1 for the application of the notion of recategorization to the gender domain).
An important contribution to the understanding of the lexical semantics of nouns
in relation to their countability properties is the notion of boundedness, which has been
discussed in the literature largely within the framework of cognitive grammar (Langacker
1987a,b; Talmy 2000a,b) and conceptual semantics (Jackendoff 1991). Count nouns are
bounded by default in the sense that “one cannot divide [their referent] up and still get
something named by the same count noun,” e.g., apple (Jackendoff 1991: 18). Conversely,
mass nouns are unbounded in the sense that “one can divide [their referent] up and
still get something” named by the same mass noun, e.g., water (Jackendoff 1991: 18).
According to the literature mentioned above, the notion of boundedness can be applied to
the lexical semantics of both nouns and verbs,11 thus connecting countability properties
of nouns and types of event structure. In Jackendoff’s (1991) analysis, a second semantic
feature is necessary for the characterization of types of nouns in terms of countability
properties: internal structure. Both boundedness and internal structure work as binary
values. The combination of the possible values with respect to the two parameters gives
the following possible types of nouns:
+ bounded, – internal structure = individual entities e.g., a cat
+ bounded, + internal structure = groups e.g., a jury
– bounded, – internal structure = substances e.g., milk
11
In the verbal domain, the opposition bounded vs. unbounded would be instantiated by the opposition
perfective vs. imperfective (Langacker 1987a).
19
2 Background: gender, number, evaluative morphology and relevant interactions
– bounded, + internal structure = aggregates (plurals and collectives) e.g., cats, cattle.
The opposition between mass and count nouns can be reflected in various ways at the
level of morphosyntax, for example:
(1) Count nouns may combine quite well with numerals and measure words. On the
other hand, in order to count or measure a mass noun, many languages use more
elaborated constructions, as in English three cars as opposed to three glasses of
water. The items that are added as a unit of measure for mass nouns are referred
to in the literature as mensural classifiers (Gil 2013).12
(2) Mass nouns often participate in number distinctions in a noncanonical way. In
languages whose number system is based on the opposition between singular vs.
plural, for example, mass nouns are often either only singular or only plural.
The countability properties of nouns interact with number marking, but the dynamics
of such interactions are highly language-specific.
2.2.2 Types of number systems and number values
Not all languages need to obligatorily express nominal number values. In this section,
I discuss types of number systems according to the presence or absence of obligatory
number marking.
2.2.2.1 General number
Languages that do not encode number distinctions obligatorily have been defined in the
literature as having transnumeral number, unit reference or general number. In this dissertation, I use the latter label, general number, which was introduced by Andrzejewski
(1960) in a study of number marking in Borana-Arsi-Guji Oromo (Afro-Asiatic, Cushitic),
and was later adopted by Corbett (2000) in his typological investigation of the category
of number. Nouns with general number are by definition outside the system of number
distinctions (and are therefore morphologically zero-marked for number); their translation equivalent in languages with obligatory number could be either singular or plural.
Normally, in languages with general number, there are additional means used to specify
the number value of the noun as singular or plural, if needed. Moreover, in languages
with general number, singularity and plurality are generally obligatorily coded on personal pronouns. The frequency of use and the distribution of nouns with general number as opposed to number-marked nouns vary a great deal across languages, as do the
strategies of formal marking. It is generally quite rare to have three distinct markers for
differentiating among general, singular and plural number with all nouns. What is more
12
As Gil (2013) puts it, “most or all languages” have mensural classifiers. Overall less frequent are
sortal classifiers. Sortal classifiers – or sortal numeral classifiers – are used when nouns occur in
combination with numerals and classify nouns according to inherent properties of the np referent
such as animacy and shape. Sortal numeral classifiers can be both optional and obligatory, and are
especially common among the languages of East and Southeast Asia (for an overview, see Gil 2013).
20
2.2 Number
common, instead, is that the general number form patterns, at least for some nouns,
either with the singular or with the plural. In Cushitic (Afro-Asiatic) languages, for
instance, number unmarked nouns can be: 1) nouns with general number, 2) inherently
singular nouns and 3) inherently plural nouns. Plurality with inherently singular nouns
is marked by means of plural markers, and singularity with inherently plural nouns is
marked by singular markers. In principle, nouns with general number can be marked
either as singular or plural, if needed. However, Cushitic languages vary a great deal as
to how much overt number marking is allowed with nouns with general number. Number
marking in the Cushitic language Kambaata is illustrated in example (2.6).
(2.6) Kambaata (Afro-Asiatic, Cushitic, East Cushitic) (adapted from Treis 2008)
(a) General number noun
haqq
‘tree(s)’
(b) Inherently singular noun
meseleta
‘girl’
(c) Plural of an inherently singular noun
masal-áakkaáta
girl-pl.f
‘girls’
(d) Inherently plural noun
meentú
‘women’
(e) Singular of an inherently plural noun
meent-iccúta
woman-sg.f
‘woman’
In Kambaata, nouns with general number can sometimes be marked by the singular
markers, but the pragmatic trigger of such uses is not very clear (Treis 2008: 140). In
Treis’ (2008) reference grammar of Kambaata, and more generally, in the literature on
Cushitic languages, the singular and plural markers are referred to as pluratives and
singulatives, respectively. This special terminology is used to signal the fact that overt
number markers in Cushitic are usually derivational in nature. A more detailed account
of the system of number marking in Cushitic languages will be given in chapters 4 and
5.
21
2 Background: gender, number, evaluative morphology and relevant interactions
2.2.2.2 Obligatory number marking and possible number values
Languages like English do not have general number, and speakers need to choose the
number value of a noun when they use it (Corbett 2000). As mentioned earlier, the number values that are found in the languages of the world are: singular, plural, dual, trial,
quadral, paucal and greater plural. Their distribution is constrained by implicational
tendencies. A Number Hierarchy has been formulated on the basis of the observation of
such tendencies:
singular > plural > dual > trial
The hierarchy can be seen as equivalent to Greenberg’s Universal 34, which he elaborated
on the basis of a sample of 36 languages:
No language has a trial number unless it has a dual. No language has a dual
unless it has a plural (1963b: 94).
The Number Hierarchy is structured in such a way that the values to the right imply those
to the left. Three values are missing both from the Number Hierarchy and Greenberg’s
Universal 34: paucals, greater plurals and quadrals. The paucal marks a small number
of entities in the real world. It is more common in systems with singular, dual and
plural number values than in systems with singular and plural values only. The paucal
number is widespread among Oceanic languages (Corbett 2000: 24). The greater plural
marks very large numbers of entities in the real world. Fula (Atlantic-Congo, NorthCentral Atlantic) is an instance of a language with a greater plural (Corbett 2000: 31).
The existence of quadral markers has been suggested for a number of Austranesian
languages. However, Corbett (2000: 30) argues that for all such cases, there is no clear
evidence to claim that quadrals, defined as markers “for referring to four distinct real
world entities in the way that trials refer to three” actually exist. One problem with the
formulation of the Number Hierarchy then is when and how the the paucal and greater
plural should be incorporated. This is discussed in detail by Corbett (2000: 39-42), who
ultimately proposes to think about paucal and greater plural as types of number values
that differ in nature from the ones appearing in the Number Hierarchy.
2.2.3 Number marking and the Animacy Hierarchy
Typological investigations of nominal number (Corbett 2000; Smith-Stark 1974) have
shown that the distribution of plural marking in a language is conditioned by the animacy
of np referents. This is ultimately connected to the countability properties of nouns,
whereby animate nouns tend to also be countable nouns. The version of the Animacy
Hierarchy that is canonically used to account for typical and atypical strategies of number
marking on nouns and pronouns within and across languages is the following:
speaker > addressee > 3rd person > kin > human > animate > inanimate
Haspelmath (2013) elaborates a new version of the Animacy Hierarchy, which he uses
to describe the distribution of plural marking on nouns only:
22
2.2 Number
kin > other humans > “higher animals” > “lower animals” > discrete inanimates >
nondiscrete inanimates
In contrast to the version used by Smith-Stark (1974) and Corbett (2000), the Animacy
Hierarchy proposed by Haspelmath (2013) does not take into account pronouns – because
they often display peculiar encodings of plurality – and is based on a more fine-grained
classification of noun types in between kinship terms and inanimate nouns.
Both in its original and revised version, the Animacy Hierarchy is concerned with the
way in which overt plural marking is distributed according to the semantic properties
of nouns. Generally speaking, the likelihood of obligatory plural marking increases the
more one moves toward the leftmost end of the hierarchy. In many languages, only
human or animate nouns (together with personal pronouns) undergo obligatory plural
marking. At the opposite end of the Animacy Hierarchy, inanimate entities and nondiscrete inanimate entities, in particular, may be insensitive to number distinctions. The
role of animacy in number marking has been explained as the result of cognitive salience:
“the distinction between one and more than one is more salient for animates than for
inanimates, so that speakers are more likely to make use of available plural markers when
they refer to a plurality of animates” (Haspelmath 2013). In §2.2.5, I discuss how such
preferential patterns of language use become structurally obligatory as a consequence of
grammaticalization.
Finally, it is worth mentioning that, if a language has associative plural constructions
of the type “X + those surrounding X”, these are often used at least with kinship
terms, which occupy the top-ranking position of the Animacy Hierarchy proposed by
Haspelmath (2013) (see Daniel & Moravcsik 2013 and Moravcsik 2003 on the typology
of associative plural constructions; on the peculiar grammatical properties of kinship
terms, see Dahl & Koptjevskaja-Tamm 2001). The same constructions are also often
used with proper person names, which, however, do not figure as an independent type
on the hierarchy.13 Associative plural constructions express non-additive plurality: they
do not refer to a homogeneous set of referents but rather to the combination of a focal
referent (the one referred to by the kin term or the proper name) and a group of referents
associated with it.
2.2.4 The morphosyntax of number: indexation
As mentioned in §2.2, nominal number distinctions can be marked via affixation on
nouns, plural words or via indexation in the sense discussed in §2.1.2.1. The most common number-indexing targets in the languages of the world are demonstratives and verbs.
In general, indexation is described as less central for the definition of the number system
of a language than it is for gender. In some languages, however, syntactic manifestations
of number are crucial for encoding number distinctions since these are not obligatorily
coded on nouns. In languages in which both gender and number indexation coexist, the
two systems can either work in parallel or compete to some extent with each other. As
13
On the opportunity of including proper names in the hierarchy, see Koptjevskaja-Tamm (2002) and
Moravcsik (2003).
23
2 Background: gender, number, evaluative morphology and relevant interactions
outlined in chapter 1, interactions between gender and number in the indexation domain
are one of the research foci of this dissertation and will be discussed in detail in chapter
5.
As in the case of gender, manifestations of nominal number through indexation are
constrained by the Agreement Hierarchy:
attributive > predicate > relative pronoun > personal pronoun
Number indexation can also be either referential or lexical (see the discussion in
§2.1.2.2), and the likelihood of referential indexation increases as one moves rightward
on the hierarchy. As with gender indexation in the case of hybrid nouns, the possibility
of mismatches between indexation triggers and targets also exists in the case of number
indexation. Examples (2.7) and (2.8) illustrate cases of mismatch for different types of
triggers.
(2.7) Corporate nouns in English (Corbett 2000: 189)
The committee have decided
(2.8) Honorifics in Bulgarian (Indo-European, Slavic) (Ljuba Veselinova, personal
communication)
Vie
ste
stanal
nerven
You.pl be.2.pl become.m.sg nervous.m.sg
‘You have become nervous’
In example (2.7), the noun committee denotes a group, that is, a multiplicity of individual entities conceptualized as a unit. Plural indexation with the word committee
in example (2.7) signals that the multiplicity aspect of the semantics of the np is highlighted. Singular indexation is preferred to plural indexation by speakers of American
English, as opposed to speakers of British and New Zealand English who tend to accept both (Corbett 2000). Example (2.8) shows the co-occurrence of singular and plural
indexing targets with the same head noun, when the head noun is plural and denotes
an individual in polite address forms. Both the adjective, nerven, and the participle,
stanal, perform a predicative function and show singular indexation. Only the auxiliary
ste is inflected as plural. Comrie (1975) claims that phenomena of this type can be taken
into account by splitting the predicate position of the Agreement Hierarchy (11) into a
sub-hierarchy of the type: verb <participle <adjective <noun, where the probability of
referential indexation increases as one moves from more prototypical to less prototypical instances of predicate-like words, that is, from verbs to nouns. The two examples
illustrate that, similarly to gender indexation, the choice between different patterns of
number indexation depends on the types of indexation trigger, the types of target, and
the discourse context.
Special constraints in the distribution of patterns of number indexation are also found
for quantified nps, associative plurals and conjoined nps. In the case of conjoined nps,
languages are biased between preference toward indexation patterns with the conjunct
24
2.2 Number
that is closer to the target and preference for indexation patterns with all conjuncts.
Both possibilities can be found in one and the same language; their distribution is usually
constrained by the Agreement Hierarchy. Precedence in linear order and animacy also
play a role. Indexation triggers that precede their targets are more likely to trigger
indexation according to the referential properties of the np and even more so if animate
and/or human. For a more detailed typology of number indexation with conjoined nps
as well as for a discussion of similar matters with quantified nps and associative plurals,
see Corbett (2000: 195-216).
2.2.5 Distribution and diachrony of number systems
2.2.5.1 The distribution of number systems
Nichols (1992) hypothesises the existence of a correlation between the morphological
characteristics of languages and the types of attested number systems. Grammatical
number marking is, for example, very infrequent among the isolating languages of Southeast Asia. The WALS database shows comparable results concerning the distribution of
obligatory and non-obligatory strategies of number marking: if one looks at the world
map based on the data by Haspelmath (2013), the distribution of obligatory nominal
plural marking appears to be quite skewed. Africa and the westernmost part of Eurasia
are the areas where this type of marking is most commonly found. Similar results have
been attained by Wälchli (2012), who investigates the frequency of occurrence of nominal plural marking in a sample of 82 languages by means of parallel texts. The study
confirms that the frequency of occurrence of nominal plural marking is skewed towards
Africa and the European region of the Eurasian macro-area. It also shows that the
occurrence of nominal plurality tends to align – although not without exception – with
the predictions made by the Animacy Hierarchy. In the same study, Wälchli (2012: 255)
suggests that attrition in the nominal number domain rarely leads to complete loss.
When, as, for instance, in French, number marking on head nouns disappears due to
phonological erosion, number-indexation strategies used on demonstratives, articles or
even verbs are likely to become more prominent in discourse thus allowing for number
distinctions to be still overtly coded through syntax.
2.2.5.2 Sources of number
Grammaticalized number markers can be traced to several sources such as demonstratives, quantifiers, nouns or markers of verbal plurality. Cristofaro (2012) makes a useful
distinction between sources of nominal number markers that are etymologically connected with the encoding of plurality and sources that are instead entirely independent
of number. The latter come to be used as number markers as an accidental result of
grammaticalization. Among the first type of sources, she includes:
(1) Collectives and distributives (expressions like ‘people here and there’), which develop as plural markers in an already existing system of grammaticalized number
marking.
25
2 Background: gender, number, evaluative morphology and relevant interactions
(2) Expressions of multitudes: quantifiers such as ‘many,’ ‘a lot,’ ‘all’ or nouns denoting
groups of entities – usually humans – such as ‘people,’ ‘men’.
To the above-mentioned sources one could also add markers of nominal plurality that
originate from markers of verbal plurality. For instance, Mithun (1988) investigates the
evolution of nominal number marking in a variety of North American languages in which
the marking of number on the np (nouns and adjectives) is the result of extension from
markers of verbal number. This is not generalised to all nouns but is restricted to highly
individuated nouns only.
Among the second type of sources of nominal number marking in Cristofaro’s (2012)
classification, demonstratives and other deictic or anaphoric entities are found. For instance, Frajzyngier (1997) analyses markers of nominal (and verbal) plurality in a variety
of Chadic languages that developed as a result of grammaticalization from demonstratives. Plural marking on nouns is intimately connected to definiteness; deictic markers
that initially had scope over entire nps, and only marked definiteness, later began to
mark plurality on nouns.
Different patterns of grammaticalization of nominal number share analogous distributional properties: the overt coding of number distinctions is initially very likely to occur
with highly individuated and human or highly animate nouns only. This preference in
language use can lead, through increasing frequency, to obligatory marking. Obligatory
number marking may then stay bound to high degrees of animacy and individuation or
become extended to all types of nouns (Haspelmath 2013). Cristofaro (2012) criticises
this approach by claiming that explanations that account for synchronic distributions
of language types according to typological hierarchies (in this case the Animacy Hierarchy) cannot always be applied to diachrony. Markers of nominal plurality originate
from different and often unrelated diachronic patterns and “the hierarchy is simply a
schema which is general enough to subsume all of these processes” (Cristofaro 2012: 8).
Along these lines, I shall show in chapter 5 that, for instance, the Animacy Hierarchy
may not have the ability to account for the development of number indexation in certain
languages of the African macro-area.
2.3 Evaluative morphology
Languages have various means for expressing the semantic values of big vs. small and
good vs. bad. The markers that morphologically encode these distinctions are labelled
augmentatives (= bigger size), diminutives (= smaller size), appreciatives (= something
is good), and depreciatives (= something is bad). The same marker can express more
than one value at once – e.g., small size is often associated with appreciation/endearment
or depreciation/contempt – and it is often only on the basis of context that overlapping
meanings can be teased apart. Diminutive and augmentative markers belong to the functional and semantic domain referred to as evaluation. Morphological evaluative markers
constitute a subdomain of the domain of evaluation and are commonly categorized under the label of evaluative morphology. They can be found associated with various word
classes, e.g., nouns, verbs, adjectives, but nouns seem to be the word class most likely
26
2.3 Evaluative morphology
to undergo evaluative marking (Bauer 1997). In this thesis, I focus on the marking of
evaluative morphology on nouns only and, in particular, on those constructions that
encode diminutive and augmentative meanings.
The rest of the section is structured as follows: the semantics and pragmatics of
evaluative morphology is treated in §2.3.1; in §2.3.2, I provide an overview of the major
strategies of evaluative marking and their distribution in the languages of the world
before I move on to discuss the diachrony of diminutives and augmentatives in §2.3.3.
2.3.1 The semantics, pragmatics and functions of evaluative markers
One of the seminal crosslinguistic studies on the semantics of evaluative morphology is
Jurafsky (1996). His analysis of the semantics of diminutive markers in the languages of
the world is based on the notion of radial category, following Lakoff (1987). A simplified
version of the schema elaborated by Jurafsky as a possible representation of the universal
semantic properties of diminutives is provided in figure 2.4.
Figure 2.4: Universal semantic and pragmatics properties of diminutives according to Jurafsky
(figure adapted from Jurafsky 1996: 542)
The figure can be read both synchronically and diachronically. As explained by Jurafsky,
each node is labelled with one of the senses associated with the diminutives, whereas
the arrows stand for patterns of semantic change via inference, metaphor, generalization
and lamda-abstraction-specification.14 Jurafsky’s schema shows that the semantically
14
Jurafsky (1996) defines lambda-abstraction-specification as a mechanism of semantic change that ac-
27
2 Background: gender, number, evaluative morphology and relevant interactions
prototypical, and historically original, meaning of the construction is ‘child’ (see §2.3.3).
In addition, the figure maps aspects of the pragmatics of diminutives, that is, the way
diminutive markers are used to express the attitude of the speaker towards the referent.
Reference to the semantics of augmentatives is made in the paper, but no systematic
treatment of the category is carried out.
Another attempt at unifying semantics and pragmatics in the representation of the
function of diminutive constructions is that of Wierzbicka (1984). Wierzbicka analyses
the categories of diminutive and dispreciative in Polish and Australian English within
the framework of Natural Semantic Metalanguage.15 The different usages of diminutives
in the two languages are explained in terms of sociocultural differences between the two
speech communities.
Diminutives and augmentatives received special attention within the framework of
Morphopragmatics, which was initiated by Dressler & Barbaresi (1994) as a theory and
a method for analysing the pragmatic relevance of morphological rules. Within the
morphopragmatic approach, the pragmatics of diminutives and augmentatives is viewed
as prior and superordinate to semantics.
Grandi’s (2002) crosslinguistic study of evaluative morphology in the languages of the
Mediterranean area is one of the few existing typological investigations of evaluative
constructions. Grandi distinguishes between a descriptive and a qualitative dimension
of evaluative morphology, in which the former refers to the use of evaluatives markers in
connection to size variation, and the latter to the use of evaluative markers to express
speakers’ attitudes toward the np referents. Körtvélyessy (2012) has a similar understanding of the domain and defines evaluative morphology as the set of morphological
strategies that languages use to encode the semantic notion of “less than/more than the
standard quantity of substances, qualities, actions and circumstances, with the concept
of standard quantity being a relative one.” Both models take a unified approach to the
semantics and pragmatics of evaluative markers. In this thesis, I use the two models by
Grandi (2002) and Körtvélyessy (2012) as guidelines for the description of the semantic
properties of diminutives and augmentatives in the languages of my sample.
An interesting comparison can be made between evaluative morphemes and those
strategies for the encoding of evaluative meanings that fall outside morphology. Analytic constructions of the type “Adjective encoding size or value + noun” – and vice
versa – are generally kept outside the domain of evaluation. However, to a certain
extent, these constructions show semantic and functional characteristics that are comparable with those of evaluative morphemes. On a general basis, for instance, the basic
meanings of evaluative markers – ‘small,’ ‘big,’ ‘good,’ ‘bad,’ ’young,’ ‘old’ – correspond
counts for the rise of second-order predicates from diminutive (e.g., approximation, adverbial partitive
quantifiers, hedges).
15
Natural Semantic Metalanguage is a theory of semantics based on the assumption that linguistic meaning, both at the lexical and grammatical level, can be explained by means of semantic primitives, or
primes, and reductive paraphrasis. Semantic primitives are a set of simple, undefinable and universal
concepts; reductive paraphrasis is a method of semantic analysis that aims at reducing the complexity
of language structure to simpler concepts on the basis of semantic primes. Anna Wierzbicka and Cliff
Goddard are the main proponents of the theory.
28
2.3 Evaluative morphology
to some of the basic semantic types of prototypical adjectives in Dixon’s (1982; 2004)
classification: dimension, value, age. In languages lacking evaluative morphology, adjectives that encode the above-mentioned meanings are used in cases in which languages
with evaluative morphology would tend to use diminutives or augmentatives. Interestingly, in spite of similar functional and semantic underpinnings, evaluative morphemes
and adjectives expressing size and value are not at all related from a diachronic point of
view. As discussed in §2.3.3, there is no evidence for morphological evaluative markers
of any type being the diachronic descendents of semantically related prototypical adjectives. The two types of constructions seem to have independent and non-matching
histories. Commonalities and differences between affixal and non-affixal encoding of
evaluation surely represent a very promising area of research for the understanding of
the semantics and pragmatics of evaluation.
2.3.2 Types of marking in evaluative morphology
A valuable generalisation to make when considering types of evaluative markers and
their distribution across languages is that diminutives are generally more frequent than
augmentatives (Bauer 1997; Dahl 2006; Körtvélyessy & Stekauer 2011). Moreover, certain usages that are commonly associated with diminutives – for instance, their use as
quantifiers of uncountable nouns – are never or very rarely encountered with augmentatives.
2.3.2.1 Phonetic iconicity in evaluative morphology: universal or language-specific?
A relevant research trend in the field of evaluative morphology focusses on the existence
of iconic patterns of sounds in diminutive and augmentative marking. This tradition of
studies goes back to Jespersen (1922) and Sapir (1951), who pointed out the existence
of a relation between the high, front vowel [i] and the encoding of small size in a variety
of European languages (Jespersen) and in English (Sapir). In Universal #1926 of the
Universals Archive (Plank & Filimonova 2009), such patterns are claimed to correspond
to a universal tendency :
There is an apparently universal iconic tendency in diminutives and augmentatives: diminutives tend to contain high front vowels, whereas augmentatives
tend to contain high back vowels.
The universal is based on Payne (1997).
Further typological investigations, such as Ultan (1978), Bauer (1996), Gregová et al.
(2010), Körtvélyessy (2011) test the hypothesis of sound symbolism in evaluative morphology on different language samples. Both vowels and consonants are considered in
these studies, the assumption being that palatal (or post-alveolar) consonants – together
with high front vowels – are more typical of diminutives than augmentatives. The results
of these studies do not confirm the existence of universal patterns of phonetic iconicity
in evaluative morphology. The initial hypothesis is thus rejected in all cases. The studies
nevertheless show that phonetic iconicity does play a role in the encoding of diminutives
29
2 Background: gender, number, evaluative morphology and relevant interactions
and augmentatives but that the patterns of phonetic iconicity tend to be areal or rather
language-specific (for an overview of patterns of phonetic iconicity in the languages of
western North America, see, e.g., Nichols 1971).
2.3.2.2 Morphosyntactic encoding of evaluative markers
The morphological encoding of evaluation varies across languages. The following types of
marking are attested in the languages of the world (an example for each of the patterns
mentioned above is provided):
(a) Affixation
(2.9) Diminutive suffixes in Italian (Indo-European, Romance) (constructed
example)
(a) tavolo
table.m.sg
‘table’
(b) tavol-ino
table-dim.m.sg
‘small table’
(2.10) Diminutive prefixes in Italian (Indo-European, Romance) (constructed
example)
(a) appartmento
apartment.m.sg
‘apartment’
b mini-appartamento
dim-apartment.m.sg
‘small apartment’
(2.11) Diminutive infixes in Standard Arabic (Afro-Asiatic, Semitic) (Grandi
2002: 220)
(a) kalb
‘dog’
(b) kulayb
dog.dim
‘puppy’
(b) Reduplication
(2.12) French (Indo-European, Romance) (Grandi 2002: 256)
gué-guerre
red-war
‘small fight between people among a group, normally without physical
violence’ (translation amended by a native speaker)
30
2.3 Evaluative morphology
The distribution of the different types of evaluative markers varies according to language specific properties (e.g., how much morphology there is in a language) and is of
course conditioned by the degree of grammaticalization of individual evaluative markers
within a given language (see §2.3.3).
In languages with highly elaborated systems of evaluative morphology, several different
strategies can be found. Different markers can be found associated with a particular subset of nouns or with the encoding of particular meanings. In addition, evaluative markers
can co-occur on the same noun and express (1) extremes in size, whether very small or
very large, or (2) the combination of endearment/contempt meanings and size-related
meanings. Finally, on a general basis, evaluative markers do not trigger indexation. The
only clear exception is represented by those languages in which diminutives and augmentatives are part of the inventory of gender distinctions. This is discussed in detail
in chapter 6.
2.3.3 Distribution and diachrony of evaluative markers
2.3.3.1 The distribution of evaluative markers
Evaluative morphology is, in many respects, still an under-investigated phenomenon,
and there is no tradition of studies on the distribution of evaluative markers in the
languages of the world. A very recent and promising exception to this trend is the
work by Körtvélyessy (2012). Based on a sample of world languages stratified according
to macro-areas, Körtvélyessy (2012) examines the stability and diffusion of evaluative
morphology worldwide. Her ultimate aim is to establish whether evaluative morphology
can be counted as one of the defining features of Standard Average European.16 The
data were collected on the basis of a questionnaire designed by the author as well as
on the basis of reference grammars. An index calculation formula was then used as
an instrument for data analysis. The formula measures the complexity of evaluative
morphology within languages, which Körtvélyessy refers to as Evaluative Morphology
Saturation:
EMS =
(VWF +VSC +VWC )
.
3
For each of the sampled languages, the Word Formation Value (VW F ) stands for the
number of word-formation processes attested for the encoding of evaluation. The value
for each language is computed as follows: one point is assigned if a language has one or
two word-formation processes in connection with evaluation, two points are assigned if
the word-formation values are three or four and so on. The Semantic Category Value
(VSC ) represents the types of semantic categories (descriptive or qualitative) expressed
by evaluative markers, whereas the Word Class Value (VW C ) refers to the word classes
that undergo evaluative morphology. One point is assigned to each of the semantic categories and word classes that are found in one language. The degree of EMS is calculated
16
The notion of Standard Average European was introduced by Whorf (1941) in reference to the languages of Europe when viewed as a linguistic area. An inventory of the features shared by the
languages of Standard Average European and their relative distribution in the area is presented in
Haspelmath (2001). See also Dahl (1990) and the volume by Heine & Kuteva (2006).
31
2 Background: gender, number, evaluative morphology and relevant interactions
independently for diminutives and augmentatives; the total EMS for each language is a
sum of the two. The results of the investigation show that Standard Average European
features rather complex systems of evaluative marking. Thus, according to Körtvélyessy
(2012), evaluative morphology should be included among the defining characteristics of
the European linguistic area.
2.3.3.2 The diachrony of evaluative markers
Nouns are the most common source of diminutives and augmentative markers. A nearly
universal source of diminutives in the languages of the world is the word for ‘child’
(Jurafsky 1996), which usually originates as a sort of classificatory noun used to denote
the young age of animate entities and is gradually extended to inanimate nouns for which
it is used to mark small size in countable nouns and small quantity in uncountable nouns.
Another fairly common source of diminutives are affixes that express relational meaning
or resemblance and develop as markers of approximation and then as diminutives. This
is, for instance, the case of the Italian diminutive suffix -ino/-ina, derived from the
Latin suffix -inus/-ina, which originally meant something like ‘related to X’ (Grandi
2002). Interestingly, in Jurafsky’s terms (1996: 553 ), the ‘related to X’ meaning may
also be indicative of late stages of grammaticalization of diminutive markers originally
expressing size. These markers develop more abstract relational meanings according to
the pattern: small size > small type of > related to. This may be indicative of the
existence of some sort of bidirectionality in the grammaticalization of diminutives, a
possibility that would certainly need to be further investigated. Finally, as pointed out
at the end of §2.3.1, and as also noticed by Jurafsky (1996) and Dahl (2006), there is no
evidence for affixal diminutives deriving from modifiers meaning ‘small.’ These can be
productively used in analytic constructions of the type “Modifier + noun” (as in English
small and little) but are never attested as sources of affixal diminutives.
The origin of augmentative markers has not been investigated in detail in the literature
on evaluative morphology. The exceptions are very few, as in Grandi’s (2002) overview
of the development of augmentatives in the languages of the Mediterranean area, or in
the case of Matisoff’s (1992) study of diminutives and augmentatives, in some languages
of Southeast Asia. In the latter case, the origin of augmentative markers is said to be
the word for ‘mother.’
2.4 Gender and its interaction with number and evaluative
morphology
Gender, number and evaluative morphology have been introduced in §§2.1, 2.2 and 2.3.
In this second and last part of the chapter, I introduce the specific object of study of
this dissertation.
As already mentioned in chapter 1, this dissertation has three main research foci: (1)
interactions between gender and number; (2) interactions between gender and evaluative
morphology; and (3) interactions of gender and grammatical complexity. The following
32
2.5 The interaction between gender and number: state of the art
research questions are addressed within each of the three research foci (a more finegrained formulation of the research questions is presented in §2.8).
(1) Interactions between gender and number
– Do gender and number have the same encoding? How are gender distinctions
mapped across number values?
– Do gender and number compete through indexation patterns?
– Do gender and number have the same relevance to nouns?
(2) Interactions between gender and evaluative morphology:
– Can size be a criterion for gender assignment?
– How do diminutive and augmentative genders interact with the other gender
distinctions of a language?
(3) Interactions of gender and grammatical complexity:
– Is it possible to measure the grammatical complexity of gender systems?
– Can interactions of gender with other domains of grammar be accounted for
by such a complexity metric?
– What is the role of these interactions in the overall complexity of a gender
system?
Previous studies on the interactions between gender and number are discussed in §2.5,
whereas §2.6 focusses on interactions between gender and evaluative morphology. The
notion of grammatical complexity and previous studies on the complexity of gender
systems are tackled in §2.7.
2.5 The interaction between gender and number: state of the
art
That gender and number are engaged in a special relationship with each other is an
assumption that linguists from different theoretical backgrounds and with different research agenda share with great confidence.
By examining noun classification systems in a quite diverse sample of languages,
de la Grasserie (1898) observed that number functions as a sort of catalyst for noun
classification. According to his analysis, this applies both to classifier systems and gender systems, whose function appears to be often related to the quantification of nouns.
A similar claim is made in more recent work by Crisma et al. (2011), which investigated
three types of noun classification strategies – non-sex-based gender, sex-based gender,
and numeral classifiers – in three different genealogical groupings – Bantu, Romance and
Sinitic. The purpose of the study was to verify the existence of any common functional
ground shared by different noun classification systems. The authors conclude that noun
classification strategies in the languages investigated in the paper (particularly Swahili,
33
2 Background: gender, number, evaluative morphology and relevant interactions
Italian, and Cantonese) “play an important role in the process of singling out individuals” and quantifying them (Crisma et al. 2011: 292).
Two parameters that are often used to classify types of interaction between gender
and number are exponence (cumulative vs. noncumulative) and syncretism. These are
discussed in detail in the next two sections.
2.5.1 Exponence
In Bickel & Nichols’ (2007) model of inflectional morphology, exponence is a dimension of semantic density. This can be defined as the amount of meaning that is stored
by a morphological unit. Exponence measures semantic density at the level of morphemes, whereas synthesis measures semantic density at the word level. Technically,
exponence can be cumulative, noncumulative (also referred to as separatist) or portmanteau. Cumulative exponence is found when two or more grammatical features are
encoded by nonsegmentable markers. Similar to cumulative exponents, portmanteau
exponents also express more than one category at once. However, the difference between
the two types of exponence lies in the fact that in portmanteau exponents “each of the
categories expressed corresponds to a separate formative that also exists in the language”
(Bickel & Nichols 2007: 188), whereas this is not the case for cumulative exponents. For
instance, similar to the French example given by Bickel & Nichols, in Italian, the definite
preposition del ‘of the’ is a portmanteau morpheme that has two corresponding separate
morphemes in the language: di ‘of’ and il, the Masculine Singular Definite Article. On
the other hand, on adjectival modifiers, the suffix -a encodes feminine and singular cumulatively, but there are no gender-only and number-only corresponding suffixes in the
language. Cumulative, noncumulative and portmanteau exponents can coexist in one
and the same language and within one and the same grammatical domain. The distribution of such splits in types of exponence is not always random (Plank 1999). According
to Plank, the first relevant morphological split is the one between derivational and inflectional features of grammar: derivational features rarely display cumulation, whereas
inflectional features are “more prone to licence at least some cumulation, though never
with derivational categories” (1999: 292). In his paper, Plank investigates morphological splits between agglutinative and flexive patterns of encoding of case and number,
and focusses on splits due to exponence (cumulative vs. noncumulative) and variance
(morphological alternations within paradigms that cannot be phonologically motivated).
Gender and number often have cumulative exponence. This is asserted in two of the
most well-known typological surveys of the two grammatical features:
Number is of more importance for gender also because it is the category most
often realized together with gender. (Corbett 1991: 189)
And, similarly,
Number is often marked morphologically in this way, cumulated with case,
gender or person, depending on the word-class. (Corbett 2000: 145)
34
2.5 The interaction between gender and number: state of the art
To the best of my knowledge, however, there are no studies that systematically survey
cumulative exponence of gender and number, and their implications on our understanding of the two grammatical domains. On the other hand, gender and number-related
patterns of syncretism have received more attention in the literature. And it is to syncretism that I now turn.
2.5.2 Syncretism
Syncretism is a type of paradigmatic asymmetry whereby certain grammatical distinctions are neutralized or reduced under certain conditioning factors. Patterns of syncretism always feature at least two grammatical domains. The grammatical domain
that conditions the occurrence of syncretism is often referred to as the context of syncretism. Thus, linguists often talk about syncretism in the context of case, syncretism
in the context of person, or, as is the case in this thesis, syncretism in the context of
number.
The most recent comprehensive survey of patterns of syncretism in the languages of
the world and across different grammatical domains is Baerman et al. (2005). According
to this model, syncretism differs from neutralization in that the lack of distinctions is
partial in the former and total in the latter. For instance, in Seneca (Iroquoian), Masculine, Feminine and Neuter Gender are distinguished under singular reference; however,
only two gender distinctions are preserved under plural reference, as the Feminine and
Neuter are conflated. This is an example of syncretism (Baerman et al. 2005: 84). In
Russian (Indo-European, Slavic), Masculine, Feminine and Neuter Gender are distinguished only under singular reference: gender is not syntactically relevant under plural
reference (Baerman et al. 2005: 28-29). This is an example of neutralization. In languages with two gender distinctions under singular reference, lack of distinction under
nonsingular number values will always lead to neutralization. In this dissertation, I
do not distinguish between partial or total lack of distinction because both phenomena
have similar implications for the relationship between gender and number. I refer to both
types of asymmetries as syncretism, but I specify when a certain pattern of syncretism
leads to neutralization.
Two types of syncretism can be identified based on how the syncretic forms relate to
their non-syncretic counterparts:
(1) The syncretic marker formally differs from each of its component values.
(2) The syncretic marker is formally the same as one of its component values.
These two possible patterns of syncretism are illustrated in figure 2.5. Given two grammatical domains X and Y, and their respective values (X1 , X2 , Y1 , Y2 ), X is subject
to undergo syncretism in the context of Y when the value of domain Y is 2. This syncretism may result into two alternative patterns depending on whether the value of X in
the context of Y2 differs from (as in 2.5a), or is the same as (as in 2.5b), one of the values
of X under Y1 . In the literature on syncretism (see, among others, Baerman et al. 2005;
Carstairs 1984, 1987; Carstairs & Stemberger 1988), the pattern in figure 2.5b is also
35
2 Background: gender, number, evaluative morphology and relevant interactions
referred to as take-over, since it functions as if one of the values of X under Y1 spreads
across paradigmatic cells. In the model of syncretism designed by Baerman et al. (2005),
these phenomena are explained in terms of directional effects and are formally accounted
for by means of directional rules.
(a) Syncretism 1
(b) Syncretism 2 (or take-over )
Figure 2.5: Possible patterns of syncretism
Previous studies have built on the idea that syncretism is concerned with markedness
relationships.17 In the case of gender and number, for instance, it has been often noted
that fewer gender distinctions tend to appear with more marked number values. This
tendency was formalised as an implicational universal by Greenberg in his Universal
37:18
A language never has more gender categories in nonsingular numbers than
in the singular. (1963b: 95)
Greenberg’s generalisation is based on a sample of 36 languages and was further tested on
a broader dataset (300 languages) by Plank & Schellinger (1997). Plank & Schellinger
(1997: 59) argue that, if Universal 37 is taken as having scope over languages in their
entirety, its predictive power is stronger than if it is taken as having scope at the word
level. At the word level, exceptions tend to be more frequent, both within and across
languages. Individual sets of nouns in individual languages may exhibit peculiar markedness preferences with respect to number values. This may come as a result of diachronic
accident or as an effect of functional pressure. Examples of diachronic accident are
17
In this thesis, I follow Haspelmath’s (2006) critical reappraisal of the notion of markedness. The notion
of markedness as a property of grammar or an explanatory tool is used in the dissertation only when
reviewing studies that explicitly worked with such notions. The words marked or unmarked are
otherwise only used as synonymous with overtly-coded and zero-coded, respectively.
18
In Universal 45, Greenberg formulates a similar generalisation about the marking of gender distinctions
on singular and plural pronouns: “If there are any gender distinctions in the plural of the pronoun,
there are some gender distinctions in the singular also” (1963b: 96). Universal 45 is not discussed in
detail in this thesis.
36
2.5 The interaction between gender and number: state of the art
those cases in which gender distinctions enter the nonsingular part of a paradigm when
words distinguishing biological gender are grammaticalized as nonsingular number markers. On the other hand, functional pressure can be observed when certain gender distinctions are used to solve gender conflicts that can only arise with nonsingular numbers (Plank & Schellinger 1997: 93). Moreover, it is suggested that with nouns that
are semantically associated with nonsingularity, nonsingular number values may be less
marked than the singular, and this may also reflect on the way gender distinctions are
parsed across number values (1997: 94).
According to Baerman et al.’s (2005) model, when gender undergoes syncretism across
number values, gender distinctions are likely to be realigned on a semantic basis. For
instance, all nouns denoting animate entities might receive the same kind of plural
marking irrespective of the gender they are assigned to under singular reference. This
effect is referred to as semanticization of gender distinctions. Thus, syncretic genders
tend to be more semantically transparent than non-syncretic genders. Crosslinguistic
evidence is presented in support of this claim both from languages with large and small
gender systems (in terms of number of distinctions). I shall discuss this in greater detail
in chapter 5.
2.5.3 Cumulative exponence, syncretism and relevance hierarchies
Another classic typological generalisation concerning the relationship between gender
and number is Greenberg’s Universal 36:
If a language has the category of gender, it always has the category of number.
(1963b: 102-103)
Universal 36 establishes an implicational relationship between the two grammatical domains that, according to Greenberg, is also related to markedness. In this view, markedness relationships exist not only within but also across grammatical domains. Greenberg
suggests that markedness relationships across domains can be formalized in the form of
proximity hierarchies. According to Greenberg, proximate grammatical categories tend
to be encoded closer to the root or to the stem of the word they attach to, and are functionally more relevant than non-proximate ones. They also tend to be morphologically
more elaborated and crosslinguistically more frequent. Given a grammatical domain, the
existence of a less proximate category in one language implies that of the more proximate
one. With respect to nominal categories, Greenberg suggests that grammatical gender
(as encoded on the indexing targets) and case are less proximate than number. Thus,
the presence of gender (and/or case) in one language implies the existence of number
(Universal 36). Moreover, phenomena of syncretism are more likely to affect the implying and less proximate categories rather than the more proximate ones (Universal 37,
see §2.5.2). Greenberg’s notion of proximity hierarchies for nominal categories closely
resembles the notion of relevance hierarchy elaborated later by Bybee (1985) with respect to verbal categories. As far as I know, however, Greenberg’s notion of proximity
hierarchy has not had any follow-up in the literature.
37
2 Background: gender, number, evaluative morphology and relevant interactions
Twenty years after Greenberg’s work, Andrew Carstairs-McCarthy published two
studies on case syncretism in which he suggested that a relevance hierarchy of the type
proposed by Bybee (1985) for the verbal domain may also exist for nominal features.
Carstairs (1987) examines a set of 43 patterns of syncretism (mostly involving case and
number) from 16 different languages and shows that syncretic paradigms are almost
always also cumulative paradigms. In Carstairs & Stemberger (1988), computational
evidence is brought in support of the crosslinguistic data examined by Carstairs (1987).
Beside providing evidence for the fact that case syncretism in the context of number
is more likely when the two domains have cumulative exponence, Carstairs (1987) and
Carstairs & Stemberger (1988) also tentatively suggest that syncretism is one of the ways
in which the nominal relevance hierarchy reflects on the structure of paradigms. Number
is higher on the relevance hierarchy, and thus it is less prone to undergo syncretism. On
the other hand, case is located at a lower level in the hierarchy and is thus more likely
to undergo syncretism in the context of number.
Similar conclusions were drawn much later by Vafaeian (2013) in a study focussing
on a different grammatical phenomenon: nominal suppletion. In her typological investigation of patterns of adjectival and nominal suppletion, Vafaeian shows that number
is the nominal category that is more likely to trigger suppletion in nominal paradigms
crosslinguistically, followed by possession, vocative and accusative/ergative case. Since
gender is an inherent property of nouns whose grammatical evidence is only displayed
through indexation, it falls outside the scope of nominal suppletion (Vafaeian 2013: 123).
Taking inspiration from Veselinova (2006), where suppletion on verb paradigms is shown
to illustrate the effect of the relevance hierarchy on verbs, Vafaeian (2013) suggests that
her results on nominal suppletion might also be considered to reflect a hypothetical
relevance hierarchy for nouns. Following this suggestion, number would have a higher
position in the hierarchy than the other grammatical features considered in the study.
To summarise, by exploring different grammatical phenomena (syncretism, cumulation and suppletion), and on the basis of independent datasets, Greenberg (1963b),
Carstairs (1987), Carstairs & Stemberger (1988), and Vafaeian (2013) suggest the existence of a relevance hierarchy for nouns. These studies also agree that, within a
hypothetical nominal relevance hierarchy, number has greater importance to nouns than
other features – i.e., grammatical gender and case (Greenberg 1963b), case (Carstairs
1987; Carstairs & Stemberger 1988), possession, vocative and accusative/ergative case
(Vafaeian 2013). Apart from Greenberg (1963b), I do not know of any study that has
specifically focussed on relevance effects in the relationship between gender (as marked
on the indexing targets) and number.
2.5.4 The interaction between gender and number in the languages of
Africa
The relationship between gender and number in the African languages has been discussed
in the context of individual languages or individual genealogical groupings within the
macro-area. Reference to these studies is made throughout the dissertation. In this
section, I focus on a few studies that are particularly relevant to the formulation of my
38
2.6 The interaction between gender and evaluative morphology: state of the art
research questions (see §2.8).
In their (2008) paper on Africa as a morphosyntactic area, Creissels et al. survey
patterns of areal convergence in African languages by taking into consideration a set
of morphosyntactic features that include gender and number. The following claims are
made concerning the relationship between the two grammatical domains in the African
macro-area:
(a) Languages devoid of a gender system frequently have a single plural
marker with the morphological status of a phrasal affix, and such plural
markers tend to be used on a “pragmatic basis”, i.e., to be employed
only when plurality is both communicatively relevant and not implied
by the context, at least in the case of nouns that do not refer to persons
[...].
(b) Languages that have gender generally have a morphologically complex
plural marking, characterized by a fusion of gender and number markers, and variation in gender and number manifest themselves through
morphemes affixed to the head noun and to (some of) its modifiers, in
an agreement relationship. In these languages [...] plural markers tend
to be present in every np referring to a plurality of individuals [...].
(Creissels et al. 2008: 119)
Ultimately, what Creissels et al. (2008) claim is that there exists some sort of correlation
between presence or absence of grammatical gender and type of encoding of nominal
plurality and that this correlation is spread throughout the African macro-area. The
major limitation of these claims is that they are not formulated on the basis of the
investigation of a stratified sample of African languages. Rather, they are based on the
authors’ personal knowledge of African languages. As formalized in §2.8, testing the
validity of these generalizations on a stratified sample of African languages is one of the
goals of this dissertation.
Less broad in scope but general enough to be anticipated here is the question of
whether or not gender and number compete in the indexation domain. This question
holds particular relevance for those languages in which dedicated patterns of plural
indexation are not used with all plural nouns but with a restricted subset of plural nouns.
A case in point in the literature on African languages are, for instance, Cushitic (AfroAsiatic) languages. It is a tradition among experts of Cushitic to consider dedicated
patterns of plural indexation as a manifestation of a gender value, the so-called “plural
gender” (see Mous 2008: for a general overview). This issue is discussed in detail in
chapter 5.
2.6 The interaction between gender and evaluative
morphology: state of the art
The relationship between grammatical gender and evaluative morphology has not been
as widely investigated as the one between gender and number. A review of the state of
39
2 Background: gender, number, evaluative morphology and relevant interactions
research in this field is thus bound to be short.
In the literature on noun classification, size-related meanings are nearly unanimously
mentioned among the possible semantic values of a gender system (Allan 1977; Corbett
1991; Croft 1994; Aikhenvald 2003). However, it is also noticed that, among the most
typical semantic underpinnings of gender - i.e., sex, animacy, shape, size - size is the least
likely to occur as an independent criterion for the categorisation of nouns. Interestingly,
in his seminal work on noun classification systems, Allan (1977: 303) claims that noun
classes and gender distinctions based on size alone “only appear in African languages
[...].”19
In languages in which gender marking is functional to the encoding of size, gender
assignment is not rigid. Variation in size can be expressed by shifting a noun from one
gender to the other (see notions of free/manipulable vs. fixed/rigid gender assignment
and recategorization as discussed in §2.1.1). Gender shifts may also express qualitative
evaluative meanings, such as bad vs. good. It has been observed that in languages with
sex-based gender, noun class and gender shifts of this type are driven by stereotypical
semantic associations between each sex and each size value (Croft 1994; Aikhenvald
2003). For example, many speech communities associate ‘female’ with ‘small’ and ‘male’
with ‘big.’ Finally, it has been noticed that the extent to which gender can be used
to mark physical properties of nouns, among them size, depends on the animacy of
nouns (Aikhenvald 2003). Nouns are more likely to be classified according to physical
properties if they denote inanimate entities. As mentioned in §2.1.1, Heine (1982) reports
that gender shifts of this type are very common in African languages. He also mentions
that the possibility of assigning nouns to multiple genders is not totally unconstrained
in African languages and that, at least for some nouns, assignment preferences do exist.
The nature and scope of these assignment preferences is, however, left unspecified in his
paper.
To date, neither the nature of the relationships between gender and evaluative morphology, nor the variables that determine them, have been investigated in a systematic
fashion. One of the few exceptions is Grandi (2001), in which the relationships between
evaluative morphology and gender, on the one hand, and evaluative morphology and
number, on the other, are investigated using data from Greek, Romance and South
Slavonic languages. See also Grandi (forthcoming-b) for general typological overview of
patterns of interaction between gender and evaluative morphology (as well as evaluative
morphology and number). The use of gender shifts as a means of encoding biological
gender and size in Moroccan Berber has been recently discussed by Kossman (2014).
2.7 Grammatical complexity of gender systems
The study of language complexity is a very fascinating and controversial field of research
within linguistics. For an overview of some of the most recent approaches to the study
of language complexity, see Dahl (2004) as well as the different contributions in the
volume edited by Miestamo et al. (2008). Two main approaches to the study of language
19
In Allan’s (1977) paper, the noun class systems typical of African languages are referred to as classifiers.
40
2.7 Grammatical complexity of gender systems
complexity currently exist: the relative approach and the absolute approach. Within the
relative approach, language complexity is assessed from the perspective of its users, that
is, in terms of processing, learnability and usage (see, among others, Kusters 2003, 2008).
Within the absolute approach, which I follow in this dissertation, language complexity
is viewed as an objective property of a given system rather than as a measure of the
costs and difficulties that the users of that system experience when manipulating it (see,
among others, Dahl 2004; Miestamo 2008; Sinnemäki 2011).
Taken in this absolute sense, grammatical complexity is defined “as the number of
parts in a system or the length of its description” (Miestamo 2008: 28). Methodologically,
Miestamo also suggests, given the practical impossibility to measure the grammatical
complexity of a language as a whole, that metrics for measuring grammatical complexity
crosslinguistically be local (rather than global) in their scope. Typologists should focus
on individual functional domains and compare them across languages, trying to account
for their “minimal, average and maximal complexity” (2008: 31). One of the possible
outcomes of an enterprise of this type, he continues on the same page, is to enable
typologists to examine “typological correlations between the complexities of different
domains” and inspect whether or not “complexity in one domain is compensated by
simplicity in another.” Miestamo (2008) and Sinnemäki (2011) suggest two major criteria
for measuring complexity of individual grammatical domains:
The Principle of Fewer Distinctions. This principle is concerned with the number
of obligatory distinctions within a given functional domain: the higher the number
of grammaticalized distinctions, the more complex the domain.
The Principle of One-Meaning–One-Form. This principle is concerned with the
relationship between meaning and form within a particular functional domain.
Discontinuous morphemes or portmanteau morphemes represent a typical violation
of the One-Meaning–One-Form principle. In the case of discontinuous morphemes,
one meaning is distributed over a plurality of forms. In the case of portmanteau
morphemes, one form is associated with a plurality of meanings.
Modelling absolute complexity of individual grammmatical domains is still a novel
field in language typology. Yet, some work has already been initiated, and gender systems figure as one of the domains under investigation. Audring (2014) proposes a set of
cues to assess the grammatical complexity of the gender systems of individual languages.
The study is mostly a general programmatic discussion of possible ways of exploring the
grammatical complexity of gender and is not based on a stratified language sample.
However, different types of gender systems from different areas of the world are referred
to throughout the paper. One of the ultimate purposes of Audring’s work is to explore
whether and, if so, how the absolute complexity of gender systems relates to their learnability and processing from the perspective of language users (i.e., relative complexity).
Toward this aim, Audring (2014) individuates three dimensions for assessing the absolute
complexity of gender systems:
I The number of gender values
41
2 Background: gender, number, evaluative morphology and relevant interactions
II The number and nature of assignment rules
III The amount of formal marking (i.e., number of gender-indexing targets).
The gender system of English would rank low with respect to all three dimensions:
it has three gender values, minimal indexation (the only indexes of gender agreement
being pronouns), and the assignment rules are both semantic and broad in scope (males
are masculine, females are feminine and inanimate nouns – with a few exceptions – are
neuter).
Concerning mutual relationships between the three dimensions, Audring (2014) focusses on the way in which I and II interact with III. She argues that pervasive gender
indexation (III) is expected to be found in languages with a high number of gender values (I) and/or complex assignment rules (II). This is due to the fact that complexity in
formal marking (pervasive indexation) facilitates the processing of grammatical gender
in terms of learnability: the more manifestations of gender there are in discourse, the
easier it is to learn gender values and assignment rules.
Neither of the two principles suggested by Miestamo (2008) and Sinnemäki (2011), nor
the three dimensions of gender complexity suggested by Audring (2014), explicitly tackle
the issue of how morphosyntactic interactions between grammatical domains affect the
absolute complexities of the individual domains. Audring (2014) mentions cumulative
exponence (with number and/or other features of nominal morphology) and syncretism
as two factors that increase the absolute grammatical complexity of gender systems
at the level of formal marking (dimension III). However, within her model, it is not
clear how the occurrence of the two phenomena is to be counted in the assessment of the
overall grammatical complexity of gender systems. The issue of language complexity and
interactions between grammatical domains, as well as its specific relevance to gender,
are extensively dealt with in chapter 7.
Previous studies on the role that cumulation and syncretism play in paradigm structures have mostly discussed their implications on the learnability and processing of language. In §2.5.3, I mentioned two studies by Carstairs (1987) and Carstairs & Stemberger
(1988) with respect to the notion of nominal relevance hierarchy. The results of these
studies are also interpreted by the authors in terms of language processing. The occurrence of syncretism is said to have a simplifying effect on cumulative paradigms in
terms of learnability and usage. Similarly, Plank & Schellinger (1997) reflect on the role
of markedness in the management of paradigm sizes and of the relationships between
inflectional features of grammar:
[I]nflectional systems ought to remain within certain limits of complexity,
and the problem that may arise then is how to cut down on them when a
language indulges in inflectional elaboration to the extent that these limits
are in danger of being exceeded. Quantitatively, the obvious measure to take
is simply not to have as many distinct forms as there are paradigmatic distinctions – which is the rationale for syncretism and defectivation, especially
where the exponents of two or more categories are expressed cumulatively
rather than separately. (Plank & Schellinger 1997: 58; emphasis mine)
42
2.8 Research questions
The notion of complexity that Plank & Schellinger work with is also of the procedural
type: constraints on paradigm sizes are seen as optimising the way speakers handle
inflectional morphology.
How can cumulation and syncretism be accounted for within an absolute approach
to language complexity? When two obligatory features have cumulative exponence, the
overall number of forms in the paradigm gets reduced. At the same time, since one
marker is associated with at least two grammatical meanings, the mapping between
form and function is less straightforward than in the case of noncumulative exponence
(Principle of One-Meaning-One-Form). Ultimately, cumulation decreases linearity in the
internal structure of morphemes.
Dahl (2004) examines patterns of syncretism in the context of a broader discussion of
the development of nonlinear structures in language (the label featurization is used to
refer to such processes). According to Dahl, the occurrence of syncretism presupposes the
lack of a one-to-one correspondence between surface and underlying language structures
(“[i]n general, syncretism implies that the relationship between surface structure and
underlying representations is not one-to-one,” 2004: 188). The occurrence of syncretism
per se, however, is not necessarily equal to an increase in absolute complexity (Östen
Dahl, personal communication). This interpretation will be followed and developed
throughout the dissertation.
Semantic interactions between gender and number and gender and evaluative morphology are, to date, a very poorly investigated field in the study of gender systems
complexity. I believe that the notion of gender shift is essential for exploring and classifying these types of interactions. This will be thoroughly discussed in chapter 7.
2.8 Research questions
This dissertation aims to answer the following questions:
(2.13) Research questions concerning gender and number
Q 1: How common is cumulative exponence of gender and number in the
languages of the sample?
Q 2: What are the formal and semantic factors that trigger gender syncretism in
the context of number? Does gender syncretism in the context of number
presuppose cumulative exponence?
Q 3: What are the implications of cumulative exponence and syncretism on the
absolute complexity of gender and number systems?
Q 4: Can these types of interaction between gender and number be seen as a
reflex of a nominal relevance hierarchy?
Q 5: Can gender and number compete through indexation patterns?
Q 6: Is there any correlation between types of encoding of gender and types of
encoding of number?
43
2 Background: gender, number, evaluative morphology and relevant interactions
Q 7: What type of semantic interactions can be found between gender and
number?
(2.14) Research questions concerning gender and evaluative morphology
Q 8: How frequently does size occur as an independent gender value? How stable
and how widely distributed is this phenomenon within genenealogical units?
Q 9: Do the interactions between gender and evaluative morphology differ across
types of gender systems and/or strategies of gender assignment (e.g.,
sex-based vs. non-sex-based gender systems, or rigid vs. manipulable gender
assignment)?
(2.15) Research question concerning all three domains
Q 10: How do interactions between gender and number and gender and evaluative
morphology affect the grammatical complexity of gender systems? Is it
possible to measure the role that these interactions play in the absolute
complexity of individual gender systems?
In the next chapter, the methods of data collection and analysis that I followed in the
attempt to answer these research questions are discussed. Questions 1 to 7 are addressed
in chapter 5, whereas questions 8 and 9 are addressed in chapter 6. Finally, question 10
is approached in chapter 7.
2.9 Summary of the chapter
The aim of this chapter has been to introduce the grammatical domains investigated in
this dissertation: gender, number and evaluative morphology. Each of the domains has
been defined along four dimensions: semantics and functions, morphosyntax, typological
distributions and diachrony.
Previous studies on the interactions between gender and number, and gender and
evaluative morphology, as well as on the grammatical complexity of gender systems
have been surveyed in the second part of the chapter.
Finally, the research questions addressed in the dissertation have been outlined and
situated in the context of previous and ongoing debate on the typology of gender systems,
their interactions with other domains of grammar and their absolute complexity.
44
3 Method: language sampling, data
collection and organization
The aim of this chapter is to illustrate the sampling methodology followed in the dissertation and to relate it to the research questions outlined in §2.8. The chapter is
structured as follows. The sampling methodology designed for this thesis is described in
§3.1. In §3.2, I discuss the state of the art on the genealogical classification of African
languages, whereas major linguistic areas within Africa are surveyed in §3.3. A practical
description of the procedure of language selection is found in §3.4. Data collection and
organization are briefly discussed in §3.5. A summary of the chapter is presented in §3.6.
3.1 Sampling methodology
Based on the configuration of its sample, this dissertation may be classified as an instance
of what Haspelmath (2012) calls continent-wide typology. The label refers to typological
studies that restrict their sample space to the continent level, which, in this specific case,
is Africa. It does not refer to a specific sample procedure but is merely used to describe
a recent trend in typological studies towards continent-based language samples. One
could also refer to studies of this type as typological investigations of macro-areas, in
Dryer’s (1989a; 1992) terms. Dryer describes macro-areas as large geographic groupings
consisting of multiple genera. He defines a genus as a genealogical unit “roughly comparable to the subfamilies of Indo-European, like Germanic and Romance” (1989a: 267).
The five macro-areas identified by Dryer (1989a) are defined on the basis of continental
zones: Africa, Eurasia, Australia-New Guinea, North America and South America.
Sampling is a crucial and thoroughly debated issue in typological research (for an
overview of the specific challenges of language sampling in linguistic typology, see Bakker
2011). As in any kind of investigation based on empirical data, the construction of a
typological sample must reflect the nature of the research questions a study aims to
answer (Bakker 2011: 106). As outlined in §2.8, this dissertation explores the interactions that occur between gender and number, and gender and evaluative morphology.
I am interested in accounting for the types and nature of possible interactions between
these domains, as well as their distribution and stability throughout linguistic areas and
genealogical units. Toward this aim, I decided to build my sample with the purpose
of combining intra- and intergenealogical observations of a restricted, and somewhat
self-consistent, area of the world. The following considerations led me to elaborate a
sampling design of this type:
(1) Via intragenealogical comparison we can map synchronic distributions of linguistic
45
3 Method: language sampling, data collection and organization
phenomena throughout closely related languages. On the basis of such distributional data, we can attempt to estimate the stability of the observed phenomena.
In the best-documented cases, we may also infer aspects of the diachronic evolution of the phenomena under investigation. This procedure has been extensively
practised in typology since its early years (see the study conducted by Greenberg
1980 on word order in Semitic Ethiopian and Iranian languages) but has become
more popular within the last couple of decades. Croft (2003: 247) describes intragenealogical typology as the application of the historical-comparative method
for typological purposes, and as the basic methodological ground for diachronic
typology, which he also refers to as dynamic typology.
(2) Via intergenealogical comparison we are able to account for the distribution of a
linguistic phenomenon throughout unrelated languages, as well as to illustrate and
motivate the limits of crosslinguistic variation with respect to that phenomenon.
When practised on a restricted geographic area, this method may lead to the
discovery of patterns of areal convergence that cut across genealogical affiliation.
Research in areal typology has been very productive in the past two or three
decades. As Koptjevskaja-Tamm (2011: 577-578) puts it, one of the major contributions of areal typology to the field of linguistic typology is to show that “it can
be of limited value to search for a possible human language without simultaneously
investigating its genetically and areally determined manifestations and trying to
uncover the possible historical reasons behind this variation.”
The dataset that resulted from this combined sample procedure consists of 100 languages selected from different genealogical groupings of Africa.20 For each of the selected
genealogical units, a convenience subsample was created; as a rule of thumb, the number
of languages for each subsample was decided on the basis of the size of the individual
genealogical units.21 The availability of descriptive materials on a given language also
played a role in the process of language selection. A detailed description of the internal
structure of the language sample and the principles that guided language selection is
found in §3.4.
The choice of Africa as the sampling area was directly connected to the research
questions that the dissertation aims to answer (see §2.8). One of the most important
criteria for a language being included in my sample is the presence of grammatical gender
in the language. Given that Africa has been defined as one of the world’s hotbeds of
grammatical gender (see, among others, Corbett 2013b; Nichols 1992: 130), I decided to
confine my research to this area of the world. As shall be illustrated in detail in chapter
4, 84 languages in the sample have grammatical gender. The remaining 16 languages
do not have grammatical gender. They have been included in the sample as control
20
The language sample also contains languages such as Hebrew and Maltese, which are actually spoken
outside Africa. As Dryer (1989a: 268) points out, all Semitic languages can be seen as part of the
same large linguistic area because “their genetic relationships go in that direction.”
21
This method is likely to result in genealogical biases and is thus not suited for large-scale typological
samples. See the discussion in Veselinova (2006: fn 2).
46
3.2 African languages: genealogical classification
languages in order to investigate whether any aspect of the grammatical encoding of
number and evaluative morphology correlates with the presence or absence of gender.
Before exploring the details of the structure of the language sample created for this
thesis, I give an overview of the current state of the art on the genealogical classification
of African languages and the major linguistic areas within Africa.
3.2 African languages: genealogical classification
According to the most recent counts, about 2000 languages are estimated to be currently
spoken in Africa (Heine & Nurse 2000; Mous 2003b). These figures need to be taken
with caution for more than one reason. First of all, as noticed by Heine & Nurse (2000),
many African languages are still underdescribed and others are dying as their last native
speakers die. In addition, different figures may result depending on where the line
between languages and dialects is drawn, which is often done on the basis of sociopolitical
rather than linguistic grounds.
The genealogical classification of the African languages proposed by Greenberg in
1963 has been for several decades undisputed both among africanists and typologists.
Greenberg (1963a) divides the languages of Africa into the following four macro-families:
Niger-Kordofanian (or Congo-Kordofanian), Nilo-Saharan, Afro-Asiatic, and Khoisan.
All further subdivisions within each of the four families are listed in figure 3.1 (next
page).
In recent years, and on the basis of extensive historical-comparative investigation,
Greenberg’s classification has been strongly criticised by specialists of individual languages and genealogical units of Africa. The effects of this debate on the typological
community have apparently been relatively minor, considering that some of the canonical
typological resources, as the WALS database (Dryer & Haspelmath 2013) and Ethnologue (Lewis et al. 2013) are still quite faithful to Greenberg’s classification with respect
to the sampling of the African macro-area.
As Dimmendaal (2008: 850) puts it, among the four language families posited by
Greenberg, Afro-Asiatic is the one whose classification is the most secure. With respect
to the other three macrogroups, it is now believed that more genealogical diversity
than the one assumed by Greenberg needs to be posited. In the rest of this section,
I provide a short survey of the latest developments in the genealogical classification of
the African languages, mostly based on the useful overview by Dimmendaal (2008). I
especially consider those aspects of the ongoing debate which have been more crucial in
the creation of my own sample.
Most africanists would nowadays agree that one of the weakest and most problematic points in Greenberg’s (1963a) classification is the idea of Khoisan as a homogeneous genealogical unit (see, for example, Güldemann 2008). Greenberg believed that
the Khoisan family included all the click languages of Africa with the exception of those
classified as Bantu (e.g., Xhosa) or Cushitic (e.g., Dahalo). Today, three independent language families and two isolates are posited under the label Khoisan. The term Khoisan
is still used as a label of convenience by which no genealogical reality is intended. Table
47
3 Method: language sampling, data collection and organization
Congo-Kordofanian (= Niger-Kordofanian)
Niger-Congo
(West) Atlantic
Mande
Voltaic (= Gur)
Kwa
Benue-Congo
Adamawa-Eastern (= Adamawa-Ubangi)
Kordofanian
Nilo-Saharan
Songhai
Saharan
Maban
Fur (Isolate)
Chari-Nile
Eastern Sudanic
Central Sudanic
Berta (Isolate)
Kunama
Koman
Afro-Asiatic
Semitic
Egyptian
Berber
Cushitic
Chadic
Khoisan
Southern African Khoisan
Northern
Central
Southern
Sandawe (Isolate)
Hadza (Isolate)
Figure 3.1: Greenberg’s (1963a) genealogical classification of the African languages
48
3.2 African languages: genealogical classification
3.1 summarises the current state of the art on the genealogical relationships among the
languages of the Khoisan area. The classification presented in the table is the one followed by the Glottolog database (Nordhoff et al. 2013) and is based on some of the most
recent contributions in the field of historical-comparative reconstruction on the Khoisan
area. See appendix A for a complete list of the languages included in the sample from
the individual genealogical units listed in table 3.1.
Table 3.1: Genealogical relationships within Khoisan
Genealogical Unit Level
Hadza
Isolate
Sandawe
Isolate
Khoe-Kwadi
Family
Kxa
Family
Tuu
Family
Another important revision of Greenberg’s classification concerns the place of Omotic
within Afro-Asiatic. When Greenberg elaborated his classification of the African languages, Omotic was considered a subgroup of Cushitic and was commonly referred to as
West Cushitic. Fleming (1969) was the first to propose considering the Omotic/Western
Cushitic languages as an independent subgrouping directly descending from Afro-Asiatic.
Greenberg never adjusted his classification accordingly. Nowadays, the different subdivisions within Omotic are classified either as independent subgroupings within AfroAsiatic or as independent languages families with no relationship to Afro-Asiatic. By
the latter interpretation, shared features with other Afro-Asiatic languages or between
languages of different Omotic families are explained as the result of language contact.
The Afro-Asiatic membership of Omotic, the genealogical relationships between the different Omotic subgroupings and their possible position within the rest of the family are
the topic of a very big, and as yet unresolved, controversy within Afro-Asiatic linguistics
(for an overview of different classifications of the Omotic languages, see Amha 2012).
The genealogical relationships within Greenberg’s Niger-Kordofanian macrogroup (i.e.,
Niger-Congo or, following Glottolog, Atlantic-Congo22 ) have also been thoroughly revisited in recent years. Although I cannot go into the details of this debate here, a few
major points need to be clarified.
First of all, since the evidence supporting the affiliation of the Mande languages with
the Atlantic-Congo family has proved to be scarce, most scholars today prefer to consider these languages as an independent group outside Atlantic-Congo (Dimmendaal
2008: 842).23
22
It is worth mentioning that the two terms – Niger-Congo and Atlantic-Congo – are not synonymous.
Certain subgroupings that are traditionally counted as Niger-Congo (e.g., Mande) are not conceived
as part of the Atlantic-Congo group in the Glottolog (see further discussion in this section). The
different labelling reflects non-overlapping classifications.
23
Dimmendaal (2008) also considers the Ubangi languages to be outside Atlantic-Congo, but this is not
49
3 Method: language sampling, data collection and organization
Second, the status of the Atlantic languages within Atlantic-Congo is still vigorously
debated. The individual subgroupings within Atlantic are so different from each other
that, similarly to what has been observed for Khoisan, some scholars have gone as far
as to suggest that there is no such thing as an Atlantic genus within Atlantic-Congo.
According to these scholars, the label Atlantic should be taken as a cover term for a language area rather than for a genealogically coherent grouping (for a brief discussion, see
Dimmendaal 2008: 842). This attitude is reflected in recent work by Segerer (2010), who,
based on cognacy judgements (mainly) of lexical items, establishes a new genealogical
classification of the Atlantic languages. Three major genealogically coherent subgroupings are distinguished: Northern Atlantic, Bak, and Mel (former Southern Atlantic).
Gola and Limba, traditionally conceived of as part of the southern subbranch, are classified as isolates within Atlantic-Congo. Accordingly, the term Atlantic is retained only
for its areal significance, whereas the affiliation of these languages to the Atlantic-Congo
family is unquestioned.
Finally, no agreement has been reached on the internal stratification of the other genealogical units within Atlantic-Congo; their status as members of the family is, however,
generally not questioned (Dimmendaal 2008: 843).
The status of Nilo-Saharan as a linguistic unit within Africa is also considered to
be very controversial. As in the case of the other macrofamilies, many scholars have
harshly criticised Greenberg’s lumping approach and have instead enhanced the genealogical diversity of the languages grouped under the Nilo-Saharan label. Interestingly,
Dimmendaal (2008: 843-844), himself an expert on Nilo-Saharan, defends Greenberg’s
work on the classification of the Nilo-Saharan languages. He claims that establishing
Nilo-Saharan as a language family was Greenberg’s “most important contribution to the
classification of African languages.” Yet, based on new evidence from the most recent
descriptive and comparative research, Greenberg’s genealogical tree for Nilo-Saharan has
also been revised. In particular, two subbranches, Songhai and Koman, are no longer
considered as part of the family but as independent genealogical units (Dimmendaal
2008).
To summarise, as conceived today, the genealogical classification of the languages of
Africa assumes more diversity than Greenberg did in 1963. This is evident in the map
presented as figure 3.2, which illustrates the genealogical classification of Africa proposed by Dimmendaal (2008: 245). As shown earlier in this section, no consensus has
been reached yet on the internal stratification of some of the individual genealogical units
(e.g., Atlantic within Atlantic-Congo). In fact, more comparative and descriptive work is
needed in order to reach a better understanding of the nature of the mutual relationships
between individual languages and language groupings within the African macro-area. It
is also reasonable to think that in some areas of the continent, genealogical diversity
used to be even greater in the past and that its decrease came as a consequence of the
expansion of certain speech communities over others (e.g., the expansion of the Bantu
languages towards the Khoisan area). An overview of the ecological and sociohistoriwidely accepted; on the other hand, many scholars agree on the idea that Ubangi does not constitute
a coherent group within Atlantic-Congo (Harald Hammarström, personal communication).
50
3.2 African languages: genealogical classification
Figure 3.2: African language families and isolates according to Dimmendaal (2008). Copyright
[Dimmendaal (2008)]; this material is reproduced with the permission of John Wiley
& Sons, Inc.
51
3 Method: language sampling, data collection and organization
cal foundations of the distribution of linguistic diversity within Africa can be found in
Dimmendaal (2008: 845-855).
3.3 African languages: large-scale language contact
As observed in the previous section, historical-comparative work on African languages
has almost exclusively focussed on establishing genealogical relationships among individual languages and groups of languages within the continent. On the other hand, little
has been done on the investigation of large-scale language contact and areal convergence
within the African continent (Güldemann 2010; Heine & Nurse 2008).24 Early attempts
at investigating the distribution of linguistic features within Africa as a result of areal
convergence are Greenberg (1959, 1983) and Heine (1975, 1976). Greenberg (1959, 1983)
attempts to establish a set of features whose distribution identifies the core of Africa as a
linguistic area. These features cut across different domains of grammar and the lexicon
(phonology, morphology, lexical semantics). As pointed out by Güldemann (2010), one
of the major shortcomings of Greenberg’s investigation is the great overlapping between
the four linguistic areas identified through this method and the four linguistic macrofamilies identified in his (1963a) work on genealogical relationships among African languages
(see discussion in §3.2). Heine (1975, 1976) instead focusses on the investigation of areal
patterns in the distribution of word order types based on a large sample of African languages. He identifies four major word order types whose distribution is, at in least in
part, areally skewed (for an overiew, see Güldemann 2010).
Over the last decade, research on large-scale language contact within the African continental area has received renewed attention. A seminal publication in this sense is the
volume edited by Bernd Heine and Derek Nurse in 2008, A Linguistic Geography of
Africa. The general assumption behind the individual contributions in the volume is
that unravelling the dynamics of large-scale language contact within Africa can be crucial for understanding structural properties of African languages and their distribution
(Heine & Nurse 2008: 2). The volume is organized as a set of two types of case studies:
(1) case studies of individual grammatical phenomena and their distribution throughout
Africa (see, for instance, Christa König’s chapter on the nominative-marked languages
of eastern Africa); (2) case studies of individual linguistic areas and their characteristic
features (see, for instance, Tom Güldemann’s chapter on the Macro-Sudan belt linguistic
area). Two other chapters are more general in their scope and discuss patterns of areal
convergence between African languages at the levels of phonology (Clements & Rialland
2008) and morphosyntax (Creissels et al. 2008).
24
Studies of language contact at the micro-areal level have always been very productive in African
linguistics. African speech communities are often highly multilingual, and studies of this type have
especially focussed on describing and modelling the social dynamics of language contact among two
or more speech communities (see, among many others, the studies by Heine 1970 on African lingua
francas and the volume edited by McLaughlin 2009 on multilingualism in African urban areas).
52
3.4 Sampling procedure
An important contribution to the identification of large-scale linguistic areas within
Africa comes from accomplished and ongoing research by Tom Güldemann, who has
worked extensively on this topic during the past fifteen years. In his work, Güldemann
has focussed on the characterization of specific linguistic areas within Africa (see, for
instance, Güldemann 1998 on the Kalahari-Basin area, and Güldemann 2008 on the
Macro-Sudan belt) as well as on the overall classification of the languages of Africa in
terms of areal rather than geneological groupings. The macro-areal profile of the African
content is investigated by Güldemann (2010). Methodologically, the paper takes into
account the distribution of two types of linguistic features: typologically rare phenomena
that are remarkably frequent in the African macro-area as opposed to the rest of the
world (e.g., click phonemes), and typologically less rare features that are nevertheless
fairly typical of African languages (e.g., nasal vowels). The following macro-areas are
identified by Güldemann (2010) within Africa:
I Berber spread zone
II Chad-Ethiopia
III Macro-Sudan belt
IV Bantu spread zone
V Kalahari Basin
The geographic distribution of the individual areas is shown in the map in figure 3.3
(next page).
Core and peripheral members of each of the five areas are listed in Güldemann (2010)
together with a survey of the grammatical features that are shared by the languages of
each of the sub-areas. The Berber spread zone and the Chad-Ethiopia zone are the least
genealogically diverse: all the languages assigned to these areas are, in fact, Afro-Asiatic.
Possible correlations between the natural geography of Africa and the distribution of
the five areas throughout the continent are also discussed in Güldemann’s paper. For
instance, it is pointed out that the northernmost and southernmost areas (i.e., the Berber
spread zone and the Kalahari Basin) are less diverse in terms of genealogical composition
and number of languages, whereas the highest language diversity is concentrated around
the Tropics.
3.4 Sampling procedure
Data collection for this dissertation was conducted on the basis of established genealogical relationships between African languages. Accounting for areal groupings within the
African macro-area became especially relevant when analysing the data and interpreting
the results of the investigation.
53
3 Method: language sampling, data collection and organization
Figure 3.3: Linguistic macro-areas within Africa according to Güldemann (2010). Copyright
[Mouton de Gruyter]; this material is reproduced with the permission of Mouton de
Gruyter
54
3.4 Sampling procedure
Since this thesis is mainly a typological investigation of the languages of Africa, I
make no claims concerning the status of the genealogical classification of the languages
of the macro-area. When establishing the coding conventions for my database, I used
the language coding convention of Ethnologue (Lewis et al. 2013) because it is accessible and also widely used by other typological databases (e.g., WALS and Glottolog).
With respect to language classification in the strict sense, I follow the classification
provided by Glottolog (Nordhoff et al. 2013) because, in my opinion, it is the one that
more closely reflects the most recent developments in the genealogical classification of
African languages (see discussion in §3.2). The following levels of classification are used
in Glottolog: isolate, top-level family, subfamily. The label subfamily is used to refer
to all levels of classification below top-level family. Thus, when comparing Glottolog
with WALS, a Glottolog subfamily can correspond either to a subfamily or a genus of
WALS. For instance, in Glottolog, Benue-Congo and Bantoid are both labelled as subfamilies. Benue-Congo is counted as a subfamily within Atlantic-Congo (Niger-Congo
in WALS) and Bantoid as a subfamily of Benue-Congo. On the other hand, in WALS,
Benue-Congo is labelled as a subfamily of Atlantic-Congo and Bantoid as genus. “Practically, all WALS genera exist as subgroups in Glottolog” (Harald Hammarström, personal
communication): the genealogical stratifications followed in the two databases are thus
largely comparable.
Technically, the sampling procedure that I designed for this thesis approximates what
is commonly referred to as a variety sample (Bakker 2011; Veselinova forthcoming). Variety samples are created with the purpose of investigating a linguistic phenomenon “in
its greatest possible variability” (Veselinova forthcoming). Thus, for a language to be
included in a variety sample, it needs to display the phenomenon under investigation in
a given study. In this case, the presence of gender is the main condition for inclusion
in the sample. However, my language sample differs from a canonical variety sample in
that: (1) only one area of the world is investigated; (2) several languages from the same
genealogical unit are considered; (3) languages lacking the variable under investigation
(i.e., gender) are also included as a control group. I began by looking at those genealogical groupings of Africa whose gender systems had already been investigated in their
interaction with number and evaluative morphology. I thus started by sampling Bantu
and Cushitic languages. A vast amount of literature is available on the Bantu gender
systems and interactions with number, and evaluative morphology have often been discussed. Similarly, the interaction between gender and number in Cushitic languages has
been at the core of an intense scientific debate during recent decades. From this first
nucleus, the sample was expanded to include:
(1) Additional genealogical groupings characterised by the presence of gender systems similar to the Bantu type (e.g., North-Central Atlantic languages and SElEE25
within Kwa)
(2) Additional genealogical groupings characterised by the presence of gender systems
similar to the Cushitic type (e.g., Chadic, Semitic and Eastern Nilotic)
25
The spelling Selee is used in Glottolog (Nordhoff et al. 2013).
55
3 Method: language sampling, data collection and organization
(3) Additional genealogical groupings characterised by the absence of gender (e.g.,
Mande and Western Nilotic, or Akan and Ewe within Kwa26 ).
A subsample was created for each of the selected genealogical units. These subsamples
can be defined as convenience samples. I did not follow any mathematical procedure to
establish the exact number of languages that should be selected from each subgrouping.
Thus, the language sample created for this thesis cannot serve as a basis for statistical
analysis of the inferential type, that is, to formulate predictions over preferred typological
patterns. The statistical analysis that is applied to the data of the language sample is
rather of the descriptive type: the frequency distributions of the attested crosslinguistic
patterns are computed and presented in the results chapters. In chapter 7, correlation
coefficients between the features of my complexity metric are presented. Also in this
case, the figures are used as a means for exploring and describing the behaviour of my
complexity measure rather than to formulate crosslinguistic generalizations (see chapter
7 for details).
Accessibility of descriptive resources and/or availability of contact with experts of individual languages played a major role in guiding language selection. Bigger subgroupings
are generally represented by a higher number of languages. For instance, there are 23
Bantu languages and only 6 Berber languages in the sample. In the Glottolog database,
558 languages are classified as (Narrow) Bantu and 28 as Berber. The number of Bantu
and Berber languages included in the sample is thus proportionate to the size of the
respective genealogical units. In addition, I tried to organise each of the subsamples in
a way that could reflect the internal diversity of a given genealogical unit. For instance,
in the case of the Cushitic languages, I made sure to include at least one language from
each of the internal sub-groupings within the sub-family: Beja, South, East and Central
(see Nordhoff et al. 2013 and appendix A).
The individual genealogical units represented in the sample are shown in table 3.2.
The table provides the following pieces of information:
(1) Name of the genealogical unit
(2) Level of genealogical classification, according to Glottolog
(3) Areal grouping that a unit belongs to (if applicable)
(4) Superordinate genealogical grouping that a unit belongs to, according to Glottolog
(if applicable)
(5) Number of languages per genealogical unit.
It is worth mentioning that, in the case of the Omotic languages, no superordinate
genealogical grouping is claimed in Glottolog. However, since, as discussed in §3.2, the
genealogical relationships between the Omotic groups and their affiliation to Afro-Asiatic
26
Nominal prefixes are attested both in Ewe and Akan as remnants of a now estinct gender system. On
the loss of the noun class system in Akan, see, for instance, Osam (1993).
56
Table 3.2: Genealogical units in the sample
Total
Genealogical unit
Glottolog level of classification Areal grouping Supeordinate genealogical unit
No. of lngs.
Berber
Chadic
Cushitic
Semitic
Subfamily
Subfamily
Subfamily
Subfamily
–
–
–
–
Afro-Asiatic
Afro-Asiatic
Afro-Asiatic
Afro-Asiatic
6
8
13
7
Dizoid
South Omotic
Ta-Ne-Omotic
Top-level family
Top-level family
Top-level family
Omotic ?
Omotic ?
Omotic ?
Afro-Asiatic ?
Afro-Asiatic ?
Afro-Asiatic ?
1
1
4
(Narrow) Bantu
Defoid
Igboid
Kwa
Mel
North-Central Atlantic
Subfamily
Subfamily
Subfamily
Subfamily
Subfamily
Subfamily
–
–
–
–
Atlantic
Atlantic
Bantoid, Benue-Congo, Volta-Congo, Atlantic-Congo
Atlantic-Congo
Atlantic-Congo
Atlantic-Congo
Atlantic-Congo
Atlantic-Congo
23
1
1
3
3
7
Eastern Nilotic
Western Nilotic
Subfamily
Subfamily
–
–
Nilotic
Nilotic
3
6
Mande
Top-level family
–
–
4
Khoe-Kwadi
Kxa
Tuu
Top-level family
Top-level family
Top-level family
Khoisan
Khoisan
Khoisan
–
–
–
5
1
1
Hadza
Isolate
Khoisan
–
1
Sandawe
Isolate
Khoisan
–
1
100
3 Method: language sampling, data collection and organization
are still debated issues among specialists, I added Omotic as an areal cover term and
Afro-Asiatic as a superordinate genealogical grouping. A question mark is added next
to the two labels to signify that their use is not established.
A complete list of the languages of the sample, the lower levels of classification and
sources are provided in table A.1. An alphabetical index of the languages of the sample
is presented in table A.2.
3.5 Data collection and organization
Information about the languages to be included in the sample was gathered by attempting at combining three different types of sources:
(1) Reference grammars and other types of descriptive materials
(2) Consultation of experts of individual languages and language families
(3) Consultation of native speakers.
Unfortunately, not all three sources were available for each of the sampled languages.
For all the languages in the sample, data were gathered at least on the basis of reference grammars and descriptive materials of various types (e.g., dictionaries, articles in
academic journals, teaching resources). Experts of individual languages and language
families were consulted for all the genealogical units represented in the sample, except
Igboid, Defoid and the Isolate Sandawe. For these languages, only descriptive materials were used. A two-week research visit at the French research institute LLACAN
(Langage, Langue et Culture d’Afrique Noire) was instrumental in collecting data on the
following genealogical groupings: Bantu, Berber, Cushitic, Dizoid, Mel, North-Central
Atlantic, South Omotic, and Ta-Ne-Omotic. Only for two of the sampled languages,
Amharic (Semitic) and Ewe (Kwa), was it possible to rely on the direct judgments of
native speakers. Desalegn Hagos Asfawwesen (Stockholm University) helped me with
Amharic and Yvonne Agbetsoamedo (Stockholm University) with Ewe. A complete list
of the sources corresponding to each of the languages in the sample is found in appendix
A.
The data were collected with the support of a coding sheet that is provided in appendix
B. For each of the sampled languages, data collected through the coding sheet were
then stored in a relational database. The database (which I call Gender, Number and
Evaluative Morphology in Africa, henceforth GNEAF) is organised as follows. Basic
linguistic and paralinguistic information (e.g., genealogical classification, geographical
area, number of speakers, relevant references) is stored for each of the sampled languages.
This is linked to information on the grammatical phenomena investigated in the thesis:
gender, number, evaluative morphology, and their mutual interactions. The attributes
associated with the three grammatical domains are explained in chapter 4, where the
gender, number and evaluative morphology systems attested in the language sample are
surveyed. Attributes regarding interactions between gender and number, and gender
and evaluative morphology, were elaborated on the basis of the research questions that
58
3.6 Summary of the chapter
this thesis aims to answer (see §2.8). A set of attributes is meant to capture the presence
or absence of cumulative exponence between gender and number as well as syncretism
of gender in the context of number. Another set of attributes is meant to capture
relationships between evaluative morphology and grammatical gender. Attributes called
“Notes” are used to store language-specific patterns of interactions between gender and
number and gender and evaluative morphology that are difficult to capture in the form
of pre-set values. Finally, the complexity scores for the gender systems of the sampled
languages are also calculated via the database and stored in it.
3.6 Summary of the chapter
The sampling method followed in this thesis was designed to combine intra- and intergenealogical typology. The language sample that resulted from this sampling methodology
is biased in at least two ways: genealogically and bibliographically. The sample is genealogically biased because it is built on the basis of genealogical relationships between
subsets of languages. The sample is bibliographically biased in that, despite internal
subdivisions within genealogical units being reflected by subsamples, these subsamples
were created on the basis of available resources and not by means of mathematical methods. These two biases impact data analysis in that the language sample cannot be used
for making statistical predictions over preferred typological patterns within and outside
the African macro-area. However, making statistical predictions is not the purpose of
this investigation. Rather, this investigation aims at to account for the frequency of, the
stability of and that distribution of grammatical phenomena concerning the interaction
between gender and number, and gender and evaluative morphology. My assumption is
that these aims are better attained by looking at multiple subsets of related languages
within a self-consistent area of the world. I also assume that a sampling method of
this type may allow typologists to observe crosslinguistic patterns that a world-based,
balanced language sample could not always detect with the same accuracy (see Iemmolo
2011 and Wälchli 2005 for similar considerations on language sampling).
59
4 Gender, number and evaluative
morphology in the languages of the
sample: an overview
4.1 Introduction
In this chapter, I survey the types of gender, number and evaluative morphology systems
that are found in the 100 languages of the sample. The typology outlined in the chapter is
synchronic. Diachronic data are discussed when useful for explaining peculiar synchronic
distributions in the language sample.
Each section within the chapter is devoted to one of the three domains under investigation. §4.2 describes the types of gender systems attested in the languages of the
sample; §4.3 classifies the languages of the sample according to nominal number, and an
overview of the attested evaluative morphology systems is given in §4.4. A summary of
the chapter is found in §4.5.
4.2 Gender
As already mentioned in chapter 3, 84 languages within the sample have grammatical
gender. The remaining 16 languages do not have gender and are thus not discussed in
this section.
In my database (see appendix B), the gender systems of the individual languages of
the sample are classified according to the following criteria:
1. Sex-based and non-sex-based gender systems
2. Number of genders
3. Gender assignment
4. Number of gender-indexing targets
5. Occurrence of gender marking on nouns.
The first three criteria are based on WALS and, more specifically, on the chapters
by Corbett (2013b,c,d). Criteria numbers 3 and 4 are my own. Criterion 3, “Number
of indexing targets,” has been introduced because, together with criteria 1 and 2, it is
particularly useful for comparing how gender and number interact through indexation
(chapter 5) and for investigating the absolute complexity of gender systems (chapter 7).
61
4 Gender, number and evaluative morphology in the languages of the sample: an overview
Criterion 4, “Occurrence of gender marking on nouns,” has been introduced because
of its usefulness to the study of exponence of gender in comparison with number (see
chapter 5). Additional criteria concern interactions of gender and number and gender
and evaluative morphology; these criteria are discussed in detail in chapters 5 and 6.
On the basis of criterion 1, the languages of the sample are classified into two macrotypes:
languages with sex-based gender and languages with non-sex-based gender. This classification is based on Corbett (2013c). The following values are assigned to this feature
in WALS and in my own database:
• Sex-based gender
• Non-sex-based gender
• No gender.
Sex-based gender systems are those whose semantic core is based on biological sex.
Non-sex-based gender systems are not based on biological sex but rather on some notion
of animacy (Corbett 2013c; see also discussion in §2.1.1). Table 4.1 provides the number
of languages with sex-based and non-sex-based gender systems; percentages with respect
to the languages with gender (Rel. %) and the total number of languages in the sample
(Abs. %) are also given. Finally, the distribution of the different system types per
genealogical unit is presented in the following way: the name of the genealogical unit
is given first; this is followed by the number of languages from that unit that belong to
a given type, as compared to (the symbol “/” is used) the total number of languages
sampled from that unit. For instance, in table 4.1 the formulation Chadic (6/8) next to
the value sex-based means that six out of the eight Chadic languages of my sample have
sex-based gender. The same type of layout is used throughout the dissertation for the
tables that summarise typological distributions.
In my sample, the languages with sex-based gender outnumber the languages with
non-sex-based gender. A more detailed discussion of the sex-based and non-sex-based
gender systems attested in the languages of the sample is found in §§4.2.2 and 4.2.3.
62
4.2 Gender
Table 4.1: Sex-based and non-sex-based gender systems in the language sample
Gender system
Total
No. of
lngs.
Rel. %
Abs. %
Genealogical
groups
Sex-based
48
57%
48%
Berber (6/6)
Chadic (6/8)
Cushitic (13/13)
Dizoid (1/1)
Hadza (1/1)
Khoe-Kwadi (5/5)
Eastern
Nilotic
(3/3)
South Omotic (1/1)
Sandawe (1/1)
Semitic (7/7)
Ta-Ne-Omotic
(4/4)
Non-sex-based
36
43%
36%
Bantu (23/23)
Kxa (1/1)
Kwa (1/3)
Mel (3/3)
North-Central Atlantic (7/7)
Tuu (1/1)
No gender
16
–
16%
Chadic (2/8)
Defoid (1/1)
Igboid (1/1)
Kwa (2/3)
Mande (4/4)
Western
Nilotic
(6/6)
100
100%
100%
63
4 Gender, number and evaluative morphology in the languages of the sample: an overview
Criterion 2, “Number of genders,” classifies languages according to the number of
gender distinctions that they display. As discussed in §2.1.2.2, the number of genders
in a language is established on the basis of the identification of indexation classes (see
examples (2.1) and (2.2) on the indexation classes of Italian). The classification that
I follow in my database is based on the one used by Corbett (2013b). Languages are
divided in subgroups according to the following cut-off points:
• Two genders
• Three genders
• Four genders
• Five or more
• No gender.
The distribution of number of gender distinctions in the languages of my sample is illustrated in table 4.2. As shown in the table 4.2, 50% of the languages with gender in my
sample have two genders, whereas the remaining 50% is split between seven languages
with three genders (8.3%), one language with four genders (1.2%) and thirty-four languages with five or more genders (40.5% of the total number of gendered languages).
In addition, in the languages of my sample, systems with two gender distinctions are
all sex-based with the exception of the Bantu language Bila, whose gender system is
based on the opposition between animate and inanimate gender. Gender systems with
three distinctions are also sex-based, whereas non-sex-based gender systems pattern with
larger systems (from four to more than five distinctions). It is worth mentioning that
this tendency cannot be generalised to the whole African continent. In his study of
gender in African languages, Heine (1982) points out that the Ubangi languages Zande
and Ma – not included in my sample – have a gender system that consists of four noun
classes: Masculine, Feminine, Animal and Inanimate. Heine (1982) refers to this type
of gender system as a mixed gender system, where sex-based and non-sex-based (or
“nature-based,” as he calls them) criteria of classification co-exist. However, in Heine’s
typology of the African noun class systems, this type of gender system is claimed to be
very rare.27
Finally, in the literature on gender, as well as in reference grammars, it is rather
common to refer to gender systems with four, five or more than five distinctions as noun
class systems. This tradition is also followed by Corbett (2013b) and Corbett (2013c),
and is discussed in §2.1. In this thesis, I use the cover term gender for both larger and
smaller systems. The term noun class is used to refer to individual markers within a
large gender system.
27
Heine (1982) discusses various criteria for the classification of the gender systems of the African languages. Many of these criteria overlap to a large extent with the ones used in this thesis. See, for
instance, his features “Sex-based vs. nature-based gender systems” and “Overt vs. covert noun
classes” in comparison with my features “Sex-based vs. non-sex-based gender systems’ and “Occurrence of gender marking on nouns.”
64
4.2 Gender
Table 4.2: Number of genders in the language sample
No. of genders
Total
No. of
lngs.
Rel. %
Abs. %
Genealogical
groups
Two
42
50%
41%
Bantu (1/23)
Berber (6/6)
Chadic (6/8)
Cushitic (13/13)
Dizoid (1/1)
Eastern
Nilotic
(1/3)
Hadza (1/1)
South Omotic (1/1)
Sandawe (1/1)
Semitic (7/7)
Ta-Ne-Omotic
(4/4)
Three
7
8.3%
7%
Eastern
Nilotic
(2/3)
Khoe-Kwadi (5/5)
Four
1
1.2%
1%
Kxa (1/1)
Five or more
34
40.5%
34%
Bantu (22/23)
Kwa (1/3)
Mel (3/3)
North-Central Atlantic (7/7)
Tuu (1/1)
No gender
16
–
16%
Chadic (2/8)
Defoid (1/1)
Igboid (1/1)
Kwa (2/3)
Mande (4/4)
Western
Nilotic
(6/6)
100
100%
100%
65
4 Gender, number and evaluative morphology in the languages of the sample: an overview
The third criterion of classification, “Gender assignment,” is based on Corbett (2013d)
and attempts at classifying the criteria according to which nouns are allocated to the
genders available in one language (for a general discussion of gender assignment in the
languages of the world, see §2.1.1). The following values are assigned by Corbett to this
feature:
• Semantic assignment
• Semantic and formal assignment
• No gender.
The distribution of gender assignment types in the languages of my sample is shown
in table 4.3. The distribution of systems of gender assignment within my sample reflects
that of the sample created by Corbett (2013d) for the WALS database, even though
the two samples do not overlap much in their composition. Corbett’s sample has 49
languages from the African macro-area. Of these 49 languages, 39 have gender: the
majority of the gendered languages (34/39) has formal and semantic gender assignment
whereas, the gender assignment of the remaining 5 languages is based strictly on semantic
properties. In sum, according to both Corbett’s sample and my sample, the distribution
of systems of gender assignment in the African macro-area is highly skewed towards the
mixed type (i.e., formal + semantic).
The fourth and fifth criteria of classification that I use in my database, “Number of
gender-indexing targets” and “Occurrence of gender marking on nouns,” are concerned
with the formal marking of gender in the languages of the sample.
The values assigned to the feature “Number of gender-indexing targets” are:
• One indexing target
• Two indexing targets
• Three indexing targets
• Four or more
• No gender.
The following notions were used as general guidelines for identifying and counting
the gender-indexing targets in each language: adjectives (in the sense of dedicated word
classes for property-words), demonstratives, determiners, verbs, numerals, copulas, complementizers, adpositions. I did not examine in detail the behaviour of different subtypes
within each of these macrotypes of targets. For instance, when I counted demonstratives
as one of the possible gender indexing targets in a language, I did not consider different
types of demonstratives as different indexes. A more thorough analysis of the behaviour
of different types and subtypes of gender-indexing targets would be desirable but falls
outside the scope of this investigation. The distribution of number of gender-indexing
targets in the languages of my sample is shown in table 4.4.
66
4.2 Gender
Table 4.3: Systems of gender assignment in the languages of the sample
Assignment system
No. of
lngs.
Rel. %
Abs. %
Genealogical
groups
Semantic
6
7.1%
6%
Bantu (1/23)
Chadic (1/8)
Cushitic (1/13)
Dizoid (1/1)
Eastern
Nilotic
(1/1)
South Omotic (1/1)
Semantic and formal
76
90.5%
76%
Bantu (22/23)
Berber (6/6)
Chadic (5/8)
Cushitic (12/13)
Eastern
Nilotic
(2/3)
Hadza (1/1)
Khoe-Kwadi (4/5)
Kxa (1/1)
Mel (3/3)
North-Central Atlantic (7/7)
Sandawe (1/1)
Semitic (7/7)
Ta-Ne-Omotic
(4/4)
Tuu (1/1)
No gender
16
–
16%
Chadic (2/8)
Defoid (1/1)
Igboid (1/1)
Kwa (2/3)
Mande (4/4)
Western
Nilotic
(6/6)
No
information
available
2
2.4%
2%
Chadic (1/8)
Khoe-Kwadi (1/5)
Total
100
100%
100%
67
4 Gender, number and evaluative morphology in the languages of the sample: an overview
Table 4.4: Number of gender-indexing targets in the language sample
No. of targets
Total
68
No. of
lngs.
Rel. %
Abs. %
Genealogical
groups
One
5
6%
5%
Chadic (2/8)
Khoe-Kwadi (1/5)
Kxa (1/1)
North-Central
Atlantic (1/7)
Two
16
19%
16%
Bantu (2/23)
Chadic (1/8)
Cushitic (4/13)
Khoe-Kwadi (4/5)
Mel (1/3)
South Omotic (1/1)
Ta-Ne-Omotic (3/4)
Three
28
33.3%
28%
Bantu (3/23)
Berber (5/6)
Chadic (1/8)
Cushitic (8/13)
Dizoid (1/1)
Eastern Nilotic (1/3)
Hadza (1/1)
North-Central
Atlantic (1/7)
Semitic (5/7)
Sandawe (1/1)
Ta-Ne-Omotic (1/4)
Four or more
33
39.3%
32%
Bantu (17/23)
Berber (1/6)
Chadic (2/8)
Cushitic (1/13)
Eastern Nilotic (2/3)
Kwa (1/3)
Mel (1/3)
North-Central
Atlantic (5/7)
Semitic (2/7)
Tuu (1/1)
No data
2
2.4%
2%
Mel (1/3)
Bantu (1/23)
No gender
16
–
16%
Chadic (2/8)
Defoid (1/1)
Igboid (1/1)
Kwa (2/3)
Mande (4/4)
Western Nilotic (6/6)
100
100%
100%
4.2 Gender
Table 4.4 shows that even though there is considerable variation within and across genealogical groupings in terms of the number of gender-indexing targets, the distribution
is skewed toward either “three” or “four or more indexing targets.” The languages of
the sample thus tend to have rather pervasive gender indexation.
Finally, it is worth mentioning that only four languages within the sample have a
pronominal gender system (for a definition of pronominal gender system, see §2.1.2.2):
Ju|’hoan (Kxa), Kwadi (Khoe-Kwadi), Mwaghavul (Chadic) and Pero (Chadic). This
distribution reflects the crosslinguistic generalisation whereby pronominal gender systems are a very rare phenomenon worldwide (see Corbett 2013b and chapter 2 for discussion). Very little is known about the gender system of Kwadi, a language spoken in
Angola that recently became extinct, and the Nigerian language Pero (see Güldemann
2004 for a description of the Kwadi gender system in the Khoe-Kwadi context, and
Frajzyngier 1989 for a description of the gender system of Pero). On the other hand,
the gender systems of Ju|’hoan and Mwaghavul are well-described and will be discussed
in greater detail later in this thesis.
The fifth and last criterion for classifying the gender systems of the languages sample
is “Occurrence of gender marking on nouns.” The following values are assigned to this
feature:
• Gender marking on nouns: Yes
• Gender marking on nouns: No
• No gender.
The distribution of overt gender marking in the languages of my sample is shown in
table 4.5 (next page).
Table 4.5 shows that the large majority of the languages with gender in the sample
have overt coding of gender on nouns. As mentioned at the beginning of the section,
this feature is especially relevant for analysing types of exponence of gender on nouns in
comparison with number. This is discussed in detail in §5.2.1.4.
4.2.1 Criteria of classification of gender systems: summary
In this section, I presented the criteria that I use to classify the gender systems of the
languages of the sample in my database and throughout the dissertation. Five criteria
have been introduced:
1. Sex-based and non-sex-based gender systems
2. Number of genders
3. Gender assignment
4. Number of gender-indexing targets
5. Occurrence of gender marking on nouns.
69
4 Gender, number and evaluative morphology in the languages of the sample: an overview
Table 4.5: Overt coding of gender on nouns in the languages of the sample
Gender on
nouns
Total
70
No. of
lngs.
Rel. %
Abs. %
Genealogical groups
Yes
69
82%
69%
Bantu (23/23)
Berber (6/6)
Chadic (2/8)
Cushitic (8/13)
Dizoid (1/1)
Eastern Nilotic (3/3)
Hadza (1/1)
Khoe-Kwadi (5/5)
Kwa (1/3)
Mel (3/3)
North-Central
Atlantic
(6/7)
Semitic (6/7)
Ta-Ne-Omotic (3/4)
Tuu (1/1)
No
15
18%
15%
Chadic (4/8)
Cushitic (5/13)
Kxa (1/1)
North-Central
Atlantic
(1/7)
Sandawe (1/1)
Semitic (1/7)
South Omotic (1/1)
Ta-Ne-Omotic (1/4)
No gender
16
–
16%
Chadic (2/8)
Defoid (1/1)
Igboid (1/1)
Kwa (2/3)
Mande (4/4)
Western Nilotic (6/6)
100
100%
100%
4.2 Gender
The distribution of the values of each of the five criteria in the languages of the sample
has been shown.
Even though the five criteria are equally important, for the sake of simplicity I use
criterion 1, “Sex-based and non-sex-based gender systems,” as a tool to refer to the
gender systems of the languages of the sample throughout the dissertation. Reference
to the other criteria is made whenever relevant to the topics under discussion.
The distribution of the sex-based and non-sex-based gender systems within the languages sample is illustrated in the map in figure 4.1 (next page). As shown in figure
4.1, non-sex-based gender systems are localised in a sub-area of Africa that extends
southward from the Sub-Saharan region. Sex-based gender systems are more scattered
around the continent, but their concentration is higher all throughout northern and central Africa (eastern and central Africa more than western Africa). Isolated sex-based
gender systems are also found in southern Africa among the languages of the Khoe-Kwadi
family.
I now turn to an overview of the types of sex-based and non-sex-based gender systems
attested among the languages of the sample.
4.2.2 Sex-based gender
Of the 48 languages with sex-based gender, 41 have two gender distinctions (i.e., masculine vs. feminine), whereas only 7 languages have three (i.e., masculine vs. feminine
vs. neuter/common). Gender systems of the latter type are characteristic of the KhoeKwadi28 and Eastern Nilotic languages. In the Khoe-Kwadi languages, nouns can be
assigned to the Feminine, the Masculine or the Common Gender. No noun is assigned
to the Common Gender by default: in fact, the third gender is only used when the
gender of a noun is left undefined, or in case of conjoined masculine and feminine nouns.
The former usage is shown in example (4.1) from Nama, where the noun for ‘animal’ is
marked as common and plural since the np refers to the animal world as consisting both
of male and female referents.
(4.1) Nama (Khoe-Kwadi) (adapted Hagman 1977: 153)
xaḿ-i
ke
’a
|úrú-ǹ
há-ń
t k’o-’ao
lion-3.m.sg decl cop animal-3.c.pl all-3.c.pl of rule-man
‘The lion is the king of all beasts’
Gender distinctions in Khoe-Kwadi languages are expressed by means of cumulative
suffixes that also encode number distinctions; gender distinctions are usually overtly
coded on nouns (with the exception of Kwadi) or on the last member of the np. Gender
indexation is usually also found on pronouns and verbs. In Kwadi, pronouns are the
only targets of gender indexation.
28
The existence of a third gender – the Common Gender – in Kwadi has been called into question by
Güldemann (2004), who claims that the morphological evidence for a third gender in the language
is, at best, very weak.
71
4 Gender, number and evaluative morphology in the languages of the sample: an overview
Figure 4.1: Sex-based and non-sex-based gender systems in the languages of the sample
72
4.2 Gender
In two Eastern Nilotic languages that are included in the sample, namely, Turkana
and Karamojong, nouns can be assigned to the Feminine, Masculine or Neuter Gender.
Both in Turkana and Karamojong, the Neuter Gender is, overall, less productive than
the Masculine and the Feminine. Masai (Eastern Nilotic) has been classified as having
two genders. A third gender is also found in the language, but it is only used to encode
location. Payne (1998) refers to the Locative Gender of Masai as a very marginal noun
class whose manifestations are only associated with one noun wwéjı̀ ‘place.’ The Locative
Gender in Masai qualifies as an inquorate gender in Corbett’s (1991: 170) terms, that
is, as a gender that is “postulated on the basis of an insufficient number of nouns”
that are to be counted as lexical exceptions. In all Eastern Nilotic languages, gender is
encoded by means of cumulative prefixes that also encode number distinctions. Gender
indexation is generally internal to the np.
Sex-based gender systems with two genders are characteristic of all the Afro-Asiatic
branches included in my sample – Berber, Chadic, Cushitic, and Semitic – as well as
the Omotic groups (Dizoid, South Omotic and Ta-Ne-Omotic) and the two isolates
Hadza and Sandawe. Examples from a language with this type of system, Kambaata
(Cushitic), are given in examples (4.2) and (4.3). Example (4.2) illustrates semantically
motivated gender assignment, whereas (4.3) illustrates gender indexation on verbs with
two inanimate nouns whose gender assignment is not semantically motivated.
(4.2) Kambaata (Cushitic) (Treis 2008: 127)
(a) hizóo-ha
sibling-m
‘brother’
(b) hizóo-ta
sibling-f
‘sister’
(4.3) Kambaata (Cushitic) (Treis 2008: 128)
xórb-u
barcum-íichch aaz-íin
afuu’ll-itée’u
ball-f.nom chair-m.abl
interior-m.icp sit-3f.pfv
‘The ball is [lit. ‘is sitting’] under the chair’
In Berber, the Feminine Gender is overtly coded by means of a circumfix. The Masculine Gender is zero-marked. Gender distinctions are generally overtly coded on nouns,
pronouns, adjectives and verbs. As for the other Afro-Asiatic branches within the sample, as well as the Omotic groups, those languages in which gender distinctions are
overtly coded on nouns usually have suffixal gender markers. Gender marking on nouns
is suffixal also in the two isolates Hadza and Sandawe. The number and types of genderindexing targets vary from language to language and are thus difficult to summarise
here. Reference to language-specific patterns of gender indexation is made at different
places throughout the thesis (see also table 4.4).
73
4 Gender, number and evaluative morphology in the languages of the sample: an overview
The origin of the sex-based gender of some of the languages considered in this thesis
has been object of extensive investigation. In all the documented cases, the grammaticalization of sex-based gender distinctions is connected with deictic (demonstratives or
definiteness markers) or anaphoric (personal pronouns) entities. I now turn to a short
survey of some of these grammaticalization paths.
In the case of the Eastern Nilotic languages, sex-based gender is an innovation that
opposes them to the closely related languages of the Western Nilotic branch, where there
is no gender. The gender system of Eastern Nilotic languages originated from prenominal modifiers derived from the noun for ‘member/person,’ in the case of the Masculine
Gender, and from the word for ‘girl/daughter,’ in the case of the Feminine Gender.
The prenominal modifiers later grammaticalized as demonstratives and merged with the
demonstrative paradigm inherited from Proto-Nilotic. The new sets of demonstratives
initially encoded only deixis but later became gender-indexing targets. A detailed analysis of the origin of gender in Eastern Nilotic languages is found in Heine & Vossen
(1983). Similarly, the overt coding of gender distinctions on nouns in Eastern Cushitic
languages originated from the grammaticalization of gender-sensitive demonstratives as
gender markers (Treis 2008).
The sex-based gender of the Berber languages is the result of reanalysis of markers
that were at first used to single out entities within a part-whole type of relationship
and only later began to mark biological gender (Mettouchi 2000). Finally, the system
of gender marking in the Khoe-Kwadi languages originated from anaphoric pronouns
encoding third person reference (on the grammaticalization of gender/number markers
in Kxoe, see Kilian-Hatz & Heine 2010).
Typically, in languages with sex-based gender, gender assignment is more semantically
transparent with human and animate nouns. This is in line with the generalisations on
the typology of gender systems discussed in chapter 2. In chapters 6 and 7, I shall
show that a component of semantic predictability in the gender system of many of these
languages is related to the encoding of size variation and/or variation in the countability
properties of nouns.
4.2.3 Non-sex-based gender
As mentioned before, the non-sex-based gender systems within my sample are in nearly
all cases large systems with four, five or more than five gender distinctions.
Ju|’hoan, a Kxa language spoken in Botswana and Namibia, is the only language of my
sample with four noun classes. Gender distinctions in Ju|’hoan are never overtly coded on
nouns but only appear on anaphoric pronouns, demonstratives and possessee pronouns.
These vary according to gender, but no pattern of affixation is identifiable since none of
the pronouns appears to be morphologically complex (Heine & König 2011: 154-155). No
gender indexation is found on verbs. Five genders are found in !Xóô, a Tuu language also
spoken in Botswana. Noun class markers in !Xóô are etymologically related to person
pronouns and are overtly coded on nouns, adjectives and verbs (object indexation) by
means of suffixes (Güldemann 2000).
If the two systems mentioned above, and the Bantu language Bila (with only two
74
4.2 Gender
genders) are excluded, the most robust type of non-sex-based gender system within my
sample has more than five distinctions. It is found in 33 out of 36 languages with non-sexbased gender (see table 4.1). These languages, with the exception of the Kwa language
SElEE, all belong to the North-Central Atlantic, Bantu and Mel genealogical groups. The
noun class systems of the languages of these groups are very similar to each other; these
similarities are genealogically motivated since the three groups constitute independent
branches of the Atlantic-Congo family. The noun class system of the Bantu languages is,
however, considered more conservative with respect to the Proto-Atlantic-Congo system
(Katamba 2003; Nurse & Philippson 2003), and the affiliation of a language or a set of
languages with the Atlantic-Congo family is often evaluated on the basis of the presence
(or absence) of a Bantu-like noun classification system. In this section, I discuss those
characteristics of the noun class systems of the Bantu, North-Central Atlantic, and Mel
languages that are shared by all three groups. Differences between Bantu, on the one
hand, and North-Central Atlantic and Mel, on the other, are discussed separately in
§§4.2.3.1 and 4.2.3.2. For the sake of simplicity, I use the label Atlantic as a cover term
to refer to North-Central Atlantic and Mel. As explained in §3.2, the term does not have
any genealogical reality because the two groups – North-Central Atlantic and Mel – are
classified as independent branches of Atlantic-Congo.
The prototypical noun class-marking strategies in the Atlantic and Bantu languages
consist of the combination of affixes marked on the noun stem and a number of indexing
targets, the most frequent of which are verbs, adjectives, pronouns (of different kinds)
and numerals. Noun class markers are entirely prefixal in Bantu, whereas in Atlantic,
they can be either prefixal or suffixal (see §4.2.3.2). The individual noun classes in Bantu
and Atlantic are traditionally described as being either singular or plural. According to
the Bantu and Atlantic tradition, pairs of singular and plural class markers form genders
(see also §2.1). Nouns are assigned by default to a gender/noun class pair or, in the case
of nouns that do not participate in number distinctions, to a noun class. In addition,
nouns can be assigned to noun classes other than their default class. This process is
often referred to in the literature as noun class shift or gender shift (see §2.1.1). The
formation of diminutives, augmentatives, and collectives, as well as personifications of
animal nouns, and – in some cases – the encoding of location are all realised by means of
processes of this type. It is common among specialists of Bantu and Atlantic languages
to define the change from singular to plural classes also as an instance of noun class
shift. Thus, one could say that the notion of noun class shift can be operationalised
both in a broad and narrow sense. An example of noun class shift in the broad sense is
given in (4.4): the example illustrates singular-plural alternations in the Bantu language
Kirundi.
(4.4) Kirundi (Bantu) (adapted from Mel’ čuck & Bakiza 1997: 286)
(a) umun-tu mu-bi
cl1-man cl1-amazing
‘an amazing man’
75
4 Gender, number and evaluative morphology in the languages of the sample: an overview
(b) aba-ntu ba-bi
cl2-man cl2-amazing
‘amazing men’
A crucial issue when dealing with Bantu and Atlantic gender systems is their semantic
basis. Experts of Bantu and Atlantic languages generally argue in favour of one of the
following two hypotheses, both based on synchronic analysis:
1. Gender systems in Bantu and Atlantic are devoid of semantic content. Attempts at
reconstructing the underlying semantics of the individual noun classes are neither
satisfactory nor convincing (see, for example, Van de Velde 2006: 192-195).
2. Semantically driven gender assignment is still recoverable in contemporary Bantu
and Atlantic languages despite that the semantics of some individual noun classes
and genders has undergone massive bleaching through time (see, among others Breedveld 1995; Contini-Morava 1997; Denny & Creider 1976; Givón 1971;
Katamba 2003; Sagna 2012). A good strategy for depicting semantic productivity
in the Bantu and Atlantic noun class systems is to analyse the word-formation
processes that are based on class shifts (Desmuth 2000; Sagna 2012).
In this thesis, I consider the second hypothesis the most enlightening one. In fact,
even though the semantic motivation for gender assignment is in many cases no longer
synchronically transparent, mechanisms of word-formation based on noun class shift are,
to a certain extent, grounded in semantics. Based on these premises, two questions are
worth asking: (1) does the semantics of the Bantu and Atlantic individual noun class
markers condition their grammatical behaviour? and (2) do the different classes have
the same status within the system? The two questions are addressed in §§4.2.3.1 and
4.2.3.2, as well as in chapter 6, where the peculiar properties of the evaluative noun
classes in Bantu and Atlantic languages are discussed.
4.2.3.1 Noun Classes in Bantu
The gender systems of the Bantu languages are very well studied. Among the most
recent contributions, a comprehensive comparative account of the gender systems of a
sample of 300 Bantu languages is found in Maho (1999). In this section, I shall survey
aspects of the Bantu gender systems that are not shared with North-Central Atlantic
and Mel.
Table 4.6 provides a complete overview of the noun class morphology of Kirundi:
noun class prefixes on nouns (ncp), gender indexes on adjectives (adj), pronouns (pro),
subject (sbj) and object (obj) markers on verbs are noted. Notice that the Bantu noun
class markers are usually numbered in such a way that singular prefixes are referred to
by odd digits and plural prefixes by even digits. In table 4.6, the singular classes are
typed in boldface.
It is believed that the Proto-Bantu system was larger than any system attested in
contemporary Bantu languages. This explains the notation of the noun class marker
76
4.2 Gender
Table 4.6: Noun classes and indexation patterns in Kirundi (adapted from Meeussen 1959;
Mel’ čuck & Bakiza 1997)
Class
ncp
adj
pro
sbj
obj
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15/17
16
mubamumiri-, ømakibiny-, øny-, ørukatubukuha
mubamumirimakibinynyrukatubukuha
ubauiriakibiizirukatubukuha-
abauirimakibi
izirukatubukuha-
mubauirimakibiizirukatubukuha-
numbered as 15/17 in the table: in Kirundi, the two Proto-Bantu classes numbered as
15 and 17 merged into one class marker, ku-. Similar to Kirundi, none of the attested
Bantu languages has preserved the system reconstructed for Proto-Bantu (Katamba
2003). Variation is found in the number of genders that has been inherited or lost in the
individual languages of the group.29 It is generally assumed among Bantu specialists that
languages with the most canonical type of system have (a) approximately six individual
class markers paired according to singular and plural alternations, and (b) a certain
number of classes (more or less up to six) that are not paired according to number
distinctions.
In many Bantu languages, gender prefixes can be preceded by a prefix that is traditionally referred to as augment, pre-prefix or initial vowel. The function of the augment
varies from language to language and multiple functions can be found in one and the
same language. In general, the function of augments is related to definiteness, specificity
and/or focus. For an overview, see Katamba (2003) and Maho (1999).
Two types of noun class prefixes can be distinguished in Bantu languages: replacive
and additive. This distinction is particularly relevant for the analysis of the interactions
between gender and evaluative morphology but not for the relationship between gender
and number. The difference between replacive and additive classes is mainly concerned
with the way in which gender is marked on nouns and how this affects indexation pat29
A few languages, like Komo (Bantu, D.30), do not have any gender distinctions at all (Katamba
2003: 109).
77
4 Gender, number and evaluative morphology in the languages of the sample: an overview
terns. In the case of the replacive classes, noun class shifts determine the replacement
of the original class marker on the noun as well as on the indexing targets; in the case of
the additive classes, the prefix of the class to which the noun is shifted is added to the
original prefix and, typically, triggers indexation. As will be shown in detail in chapter 6,
diminutive and augmentative class markers can be of the additive type. Example (4.5)
illustrates an additive class marker in Herero, Class 7, which is used to derive augmentatives (Herero is not part of my language sample). Only the additive class marker is
reflected on the indexing target (in this case the possessive for ‘my’).
(4.5) Herero (Bantu) (Crisma et al. 2011: 259)
otji-ru-vy
tj-nde
cl7-cl11-knife cl7-my
‘my big knife’
The difference between replacive and additive classes already answers the two questions
raised at the end of the previous section: (1) Does the semantics of the (Bantu) individual
noun classes condition their grammatical behaviour? and (2) Do the different classes
have the same status within the system? The status of the individual noun classes within
the system is not equal: differences in their grammatical behaviours may be semantically
conditioned.
4.2.3.2 Noun classes in Atlantic
In the Atlantic languages, the gender markers can be either prefixal, suffixal or both.
Suffixes and prefixes are etymologically related to each other, but the suffixes represent
an innovation in the languages of the area; the origin of suffixal noun class markers in
Atlantic is briefly described at the end of this section. Table 4.7 illustrates the individual
noun class prefixes of the North-Central Atlantic language Bandial, as marked both
on nouns (ncp) and indexing targets (in order: determiners, demonstratives, personal
pronouns, subject markers, relative pronouns, and adjectives).
The singular classes are typed in boldface, as in the case of Kirundi (table 4.6). Notice
that the last three classes, which express location, do not distinguish number. In line
with the tradition followed for the Bantu languages, the singular classes of Bandial are
referred to by odd digits with the only exception being Class 12, which is also singular
but is labelled with an even digit (Sagna 2012). Contrary to what is observed for the
Bantu languages, however, there is no conventionalised numbering of the individual
noun classes of the Atlantic languages. The numbering of the markers often varies from
language to language. Nonetheless, at least for gender markers that are are used to
mark human nouns, that is, classes 1 and 2, there is a substantial overlapping in the
numbering adopted for the different languages.
The number of the individual noun classes of the Atlantic languages can vary a great
deal: as reported by Wilson (1989: 96), for example, Kobiana (North-Central Atlantic)
has nearly 40 classes, whereas Nalu (an isolate within Atlantic-Congo and the Atlantic
area) has only three. The morphophonology of the markers also varies from language to
78
4.2 Gender
Table 4.7: Noun classes and indexation patterns in Bandial (adapted from Sagna 2012)
Class
ncp
def.det
dem
pro
sbj
rel
adj
1
2
abugguuesue-, ysu-, si-, sbu-, bi-,
b-, bau-,
wwawu
fu, fi-, f-,
fagu, ga-, ggamu-, mi-,
ma-, mju-, jañatdn-
øahu
bugagu
gagu
wawu
yayu
sasu
yayu
sasu
babu
umubugu-bugubugubugu-buguyusub-
øbugbugbugbugbugysb-
aguguguguguesubu-
ø
gggggysb-
aguguguguguesubu-
uw-
ø
w-
u-
w-
u-
fafu
uf-
f-
fu-
f- fu-
gagu
gagu
mamu
ugugum-
ggm-
gugumu-
ggm-
gugumu-
jaju
ñañu
tatu
dadu
–
ujuñutud–
jñtdn-
juñutudu–
jñtd–
juñu–
–
–
3
4
5
6
7
8
9
10
11
12
13
14
15
language. However, languages from the same genealogical unit have, as expected, rather
similar noun class inventories. In spite of this diversity in the shape and number of the
noun class exponents, virtually every language of the Atlantic area has an o-like marker
for singular humans, a ba-like marker for plural humans, and a m/ma-like marker for
masses and liquids (Wilson 1989: 98).
An interesting characteristic of gender marking in the Atlantic languages is consonant
alternation, whereby the quality of the (initial) consonants of the noun class markers
phonetically conditions the quality of the (initial) consonants of the noun stems. Consonant alternation is not found everywhere within the area: it is a prominent feature
of North-Central Atlantic languages – such as Maasina Fulfulde, Serer, Nuclear Wolof
(although less in the latter language than in the first two) – and of other subgroups, such
as the Tenda languages. Only a few traces of consonant alternation are found in the
languages of the Mel group. In Maasina Fulfulde, where noun class marking is suffixal,
consonant alternation on the initial consonant of the noun stem is the only trace of the
now lost noun class prefixes.
As mentioned in the beginning of the section, in the languages of the Atlantic area,
79
4 Gender, number and evaluative morphology in the languages of the sample: an overview
gender marking can be either prefixal, suffixal or both. Suffixal class markers are found,
for example, in Kisi (Mel) and in the languages of the Fula subgroup (North-Central
Atlantic). The synchronic distribution of the two types of noun class affixes has been
explained as the result of a diachronic process of gender renewal, whereby, as mentioned
before, the suffixes represent the innovation and the prefixes the original type of marking.
A short digression into diachrony may be useful for understanding the distribution of
this phenomenon in the languages of the area.
Gender renewal in Kisi, and the Mel languages in general, has been investigated by
Childs (1983) using the method of intragenealogical typology (see chapter 3). Childs
compares gender marking in Kisi with gender marking in the other languages of the
Mel branch (Southern Atlantic, according to his classification) and postulates an areal
continuum, ranging from strongly prefixal gender systems – as in Timne – to solely
suffixal gender systems – as in Kisi. He also provides evidence for remnants of noun
class prefixes in Kisi. In Childs’ analysis, the different types of languages attested in
this area represent synchronic snapshots of what the diachronic stages of the process
of affix renewal may have looked like in the languages that underwent it. According
to Childs, suffixal class marking originates from the repetition of noun class indexation
after a relative clause or a fronted np. The expansion of suffix use is accompanied by
the phonetic erosion of the older prefixes.
To my knowledge, the origin of the suffixal gender markers in North-Central Atlantic,
and Fula languages in particular, has not been investigated in a systematic fashion.
A possible explanation for the origin of gender renewal in Fula might be seen in the
behaviour of definite and indefinite nouns in the closely related language Serer (NorthCentral Atlantic) (Guillaume Segerer, personal communication). In Serer, noun class
marking is strongly prefixal. However, in addition to their respective class prefixes,
definite nouns are accompanied by a clitic that expresses definiteness, proximity and
class at the same time.
(4.6) Serer (North-Central Atlantic) (McLaughlin 1992)
(a) o-tew
cl1-woman
‘woman’
(b) o-tew=oxe
cl1-woman=cl.def
‘the woman’
If the interpretation of the Serer data is correct, the overt coding of definiteness on nouns
could be seen as a plausible trigger of the grammaticalization of noun class suffixes in
Fula (Guillaume Segerer, personal communication). This hypothesis calls for further
investigation.
80
4.3 Number
4.3 Number
In my database (see appendix B), number systems are classified according to the following criteria:
1. Type of nominal number systems in terms of obligatoriness
2. Number values
3. Number of number-indexing targets.
I designed the three criteria based on those aspects of the typology of nominal number
systems that are more relevant to the purposes of this investigation. Criterion 1 is used
to control if the relationship between gender and number is subject to vary depending
on whether or not nominal number is obligatorily marked. Criterion 2 accounts for
the types and number of oppositions within a number system and is used to control if
these affect in any way the interactions between gender and number. Finally, criterion
3 is crucial for investigating how gender and number interact through indexation. It is
worth mentioning that two other features are used to explore the typology of nominal
number systems in WALS: “Occurrence of nominal plurality” (Haspelmath 2013) and
“Coding of nominal plurality” (Dryer 2013). The first feature accounts for the extent
to which plurality is overtly coded on nouns in a given language; the second feature
classifies languages according to the morphosyntactic encoding of plurality on nouns.
Since the two features are not strictly relevant to the study of the interactions of gender
and number, I do not use them as a parameter of classification for the languages of my
sample.
On the basis of criterion 1, “Type of nominal number system in terms of obligatoriness,” languages are classified into two macrotypes:
(1) Languages with obligatory number: languages in which speakers need to choose
the number value of a noun whenever they use it.
(2) Languages with general number: languages in which (some or all) nouns can be
outside the system of number distinctions.
This classification is based on Corbett’s (2000) discussion of the major systems of number
marking in the languages of the world. It is worth mentioning here that Corbett’s
classification is grounded on the number morphology of nouns rather than on the patterns
of number indexation triggered by nouns (see chapter 2 for a general overview). Table
4.8 illustrates the distribution of the two systems in the languages of the sample; the
percentage of each type of system with respect to the total number of languages in the
sample is also given in the table. For a discussion of general number and obligatory
number, see §§2.2.2.1 and 2.2.2.2.
81
4 Gender, number and evaluative morphology in the languages of the sample: an overview
Table 4.8: Nominal number systems in the languages of the sample
Type
No. of lngs.
%
Obligatory number
71
71%
Bantu (23/23)
Berber (6/6)
Cushitic (2/13)
Dizoid (1/1)
Eastern Nilotic (3/3)
Hadza (1/1)
Khoe-Kwadi (5/5)
Kwa (3/3)
Mel (3/3)
North-Central
Atlantic
(7/7)
Sandawe (1/1)
Semitic (6/7)
South Omotic (1/1)
Ta-Ne-Omotic (2/4)
Tuu (1/1)
Western Nilotic (6/6)
General number
29
29%
Cushitic (11/13)
Chadic (8/8)
Defoid (1/1)
Igboid (1/1)
Kxa (1/1)
Mande (4/4)
Semitic (1/7)
Ta-Ne-Omotic (2/4)
100
100%
Total
Genealogical groups
As shown in table 4.8, the distribution of types of nominal number systems in terms
of obligatoriness is genealogically skewed in the sense that languages from the same
genealogical units have roughly the same type of number system. Outliers are Awngi
and Beja with respect to Cushitic; Bench and Koorete with respect to Ta-Ne-Omotic;
and Amharic with respect to Semitic. Awngi and Beja are the only Cushitic languages of
the sample with obligatory number. Similarly, Bench and Koorete are the only languages
with general number within my Ta-Ne-Omotic sample, and Amharic is the only Semitic
language of the sample where general number is productive (see §5.2.1 for nouns with
general meaning in Maltese).
Criterion 2, “Number values,” identifies the types of number distinctions that are
found in a language. The values associated with these features are based on the discussion
of the Number Hierarchy outlined in §2.2.2.2, where the distribution of number values
across the languages of the world has been considered.
• Singular vs. plural
• Singular vs. plural vs. dual
82
4.3 Number
• Singular vs. plural vs. dual vs. trial
• Singular vs. plural vs. dual vs. paucal.
The distribution of number values in the languages of the sample is shown in table
4.9.
Table 4.9: Number values in languages of the sample
Number values
singular vs. plural
Total
Number
of lngs.
% Genealogical groups
89
89% Bantu (23/23)
Berber (6/6)
Cushitic (12/13)
Chadic (7/8)
Defoid (1/1)
Dizoid (1/1)
Estern Nilotic (3/3)
Hadza (1/1)
Igboid (1/1)
Kwa (3/3)
Kxa (1/1)
Mande (4/4)
Mel (3/3)
North-Central Atlantic
(7/7)
Semitic (4/7)
Sandawe (1/1)
South Omotic (1/1)
Ta-Ne-Omotic (3/4)
Tuu (1/1)
Western Nilotic (6/6)
singular vs. plural vs. dual
9
singular vs. plural vs. trial
0
singular vs. plural vs. paucal
2
100
9% Chadic (1/8)
Khoe-Kwadi (5/5)
Semitic (3/7)
– –
2% Cushitic (1/13)
Ta-Ne-Omotic (1/4)
100%
As shown in table 4.9, 89% of the sampled languages have a “singular vs. plural” type
of contrast. The distribution of the less common systems shows some kind of genealogical
and areal skewing. The “singular vs. plural vs. dual” type of opposition is found in all
the Khoe-Kwadi languages of the sample as well as in three very closely related Semitic
languages – Standard Arabic, Moroccan Arabic and Maltese. The “singular vs. plural
vs. paucal” type of opposition is extremely rare within my language sample and only
83
4 Gender, number and evaluative morphology in the languages of the sample: an overview
attested in one Cushitic and one Ta-Ne-Omotic language, Baiso and Koorete. The two
languages are not genealogically related (especially if Ta-Ne-Omotic is not viewed as
part of Afro-Asiatic; see discussion in §3.2), but they have a long-standing history of
contact since they are spoken in the same area of Ethiopia (see §5.5.2 for a more detailed
overview).
The third and last criterion of classification of number systems in the languages of the
sample is “Number of number-indexing targets.” The targets of number indexation in
the languages of the sample have been counted following the same method adopted for
gender indexation (see §4.2 for a detailed description). The values associated with this
feature are:
• One indexing target
• Two indexing targets
• Three indexing targets
• Four or more indexing targets
• No number indexation.
The distribution of number of indexing targets in the languages of the sample is presented in table 4.10. For a general discussion of indexation and nominal number, see
§2.2.4. As shown in table 4.10, 60% of the languages of the sample has rather pervasive
nominal number indexation (between three and four indexing targets). The relationship between presence of gender and pervasiveness of number indexation is discussed in
greater detail in section 5.6. For the sake of comparability, the distribution of different combinations of number of gender- and number of number-indexing targets in the
languages of the sample is presented in appendix C.
As in the case of gender, the three criteria for the classification of nominal number
systems are equally important. However, for the sake of clarity, I use the opposition
between languages with general number and languages with obligatory number as a
way to refer to the number systems of the languages of the sample. The geographical
distribution of the individual types of number systems within the African macro-area is
also illustrated in the map in figure 4.2.
Strategies of number marking in languages with general number are discussed in detail
in §4.3.1. In §4.3.2, I provide an overview of a very interesting type of obligatory number
system that is found only among the Nilotic languages of the sample and is commonly
labelled as tripartite number system. Those properties of number marking that correlate
with the marking of gender distinctions are accounted for in chapter 5.
84
4.3 Number
Table 4.10: Number of number-indexing targets in the languages of the sample
No. of targets
Total
Number
of lngs.
%
Genealogical groups
One
11
11% Chadic (1/8)
Cushitic (1/13)
Igboid (1/1)
Khoe-Kwadi (1/5)
Kwa (1/3)
Kxa (1/1)
Mande (4/4)
North-Central Atlantic (1/7)
Two
21
21% Bantu (2/23)
Chadic (4/8)
Cushitic (3/13)
Defoid (1/1)
Dizoid (1/1)
Khoe-Kwadi (4/5)
Kwa (1/3)
Mel (1/3)
South Omotic (1/1)
Ta-Ne-Omotic (3/4)
Three
26
26% Bantu (3/23)
Berber (5/6)
Cushitic (8/13)
Chadic (1/8)
Eastern Nilotic (1/3)
North-Central Atlantic (1/7)
Sandawe (1/1)
Semitic (5/7)
Ta-Ne-Omotic (1/4)
Four or more
34
34% Bantu (17/23)
Berber (1/6)
Chadic (2/8)
Cushitic (1/13)
Eastern Nilotic (2/3)
Hadza (1/1)
Kwa (1/3)
Mel (1/3)
North-Central Atlantic (5/7)
Semitic (2/7)
Tuu (1/1)
No indexation
6
6% Western Nilotic (6/6)
No data
2
2% Bantu (1/23)
Mel (1/3)
100
100%
85
4 Gender, number and evaluative morphology in the languages of the sample: an overview
Figure 4.2: General and obligatory number in the languages of the sample
86
4.3 Number
4.3.1 General number
In chapter 2, I pointed out that in a language with general number and a singular/plural
opposition, ideally, each noun should have three distinguished forms: (1) the general
form, interpretable both as singular and plural; (2) a form interpretable as singular and
(3) a form interpretable as plural. However, as pointed out by Corbett (2000), in most
languages, this type of tripartite system is often found only with a restricted number of
nouns. In other words, the form used for the general meaning with some nouns is also
used for the singular or the plural with other nouns. Corbett further points out that the
most frequent general number system is the one in which nouns have a form that can
be interpreted either as general or as singular, and another form that is interpreted only
as plural. This system is labelled by Corbett as “general/singular vs. plural system.”
The other logically possible system, whereby the plural and the general are encoded in
the same way, and there exists a dedicated marker only for the singular (“general/plural
vs. singular,” following Corbett’s terminology), is never attested on its own. Within
one language, this latter type of system tends to be restricted to a subset of nouns that
are usually semantically inherently associated with plurality (Corbett 2000: 17). The
languages of my sample reflect this crosslinguistic pattern.
Among the 29 languages with general number in my sample, only the Cushitic language
Baiso has a unique general number form for nearly all nouns. This is shown in example
4.7. Notice that, in addition to Singular and Plural, Baiso also has a Paucal Number.
(4.7) Number distinctions in Baiso (Cushitic) (adapted from Corbett 2000: 11)
(a) lúban
lion.general
‘lion(s)’
(b) lubán-titi
lion-sg
‘a lion’
(c) luban-jaa
lion-pauc
‘a few lions’
(d) luban-jool
lion-pl
‘lions’
In Baiso, there is also a small subset of nouns that are number-unmarked but semantically associated with plurality. These nouns are extensively discussed in §5.5.2.
In the remaining 28 languages with general number, the same form that is used to
express general meaning is largely used to encode singular reference as well and is usually
number-unmarked. These languages thus display a system of the “general/singular vs.
plural” type. Lexical splits, whereby the general form is the same as the singular of
certain nouns and the plural of other nouns, are especially common among the Cushitic
87
4 Gender, number and evaluative morphology in the languages of the sample: an overview
languages of the sample. In Cushitic, number-unmarked forms can be interpreted as
either general, inherently singular or inherently plural.30
Cushitic languages have a rich inventory of morphemes that are used to specify the
number value of nouns with general number as well as to form the plural of inherently
singular nouns or the singular of inherently plural nouns. These morphemes are often described in reference grammars as word-formation strategies. Number marking in Cushitic
is thus seen mainly as a derivational phenomenon. Examples of such word-formation
processes are shown in tables 4.11 and 4.12.
Table 4.11: Singular marking in Kambaata, Borana-Arsi-Guji Oromo and Tsamai
Language
Basic form
Singular marking
Source
Kambata
Borana-Arsi-Guji
Oromo
Tsamai
meentú ‘women’
k’áallitS ‘priest(s)’
meenticcúta ‘woman’
k’áallu’ ‘the priest’
Treis 2008
Andrzejewski 1960
Pukaè-e ‘eggs’
Pukaèitte ‘one egg’
Savà 2007
Table 4.12: Plural marking in Kambaata, Borana-Arsi-Guji Oromo and Tsamai
Language
Basic form
Plural marking
Kambata
Borana-Arsi-Guji
Oromo
Tsamai
meseleta ‘girl’
ábba ‘father’
masaláakkaáta ‘girls’
Treis 2008
ábbooti, ‘fathers, pater- Andrzejewski 1960
nal uncles’
gurlaââe ‘cats’
Savà 2007
gurlo ‘cat(s)’
Source
Multiple strategies of nominal plural formation are also found among the Chadic languages. For instance, 40 different plural markers, which can be in turn reduced to a dozen
distinct morphological patterns of plural formation, are found in Hausa (Newman 1990,
2000). For an overview of nominal (and verbal) plurality in Chadic, see Newman (1990).
Only one plural suffix is found in Bench (Ta-Ne-Omotic), in which the number-unmarked
form can be interpreted as either general or singular (Rapold 2006). In Koorete (TaNe-Omotic), number-marked forms are also either general or singular. Overtly marked
number values are the Plural and the Paucal (Teketal 2004).
Some of the Western African languages included in the sample, such as Yoruba and
Igbo, as well as the four Mande languages Bambara, Dyula, Mann and Susu have quite
similar number systems. In these languages, plurality is overtly coded by means of
suffixes or clitics that are diachronically related to the third person plural pronouns.
Number-unmarked nouns are interpreted as general or singular. In Igbo, general number
is restricted to inanimate nouns. With animate nouns, number-unmarked forms can
only be interpreted as singular (Carrell 1970; Welmers 1973). These languages have, in
general, rather poor morphologies.
30
Number-unmarked nouns can also be mass or abstract nouns. These noun types can interact with
number morphology in special ways. For instance, in Kambaata, mass nouns are rarely marked for
number, and abstract nouns are never marked for number (Treis 2014).
88
4.3 Number
In Ju|’hoan (Kxa), nominal plurality can be encoded by means of suppletion, clitical
and/or suffixal markers, and associative plurals. None of these strategies is obligatory,
and the overt coding of plurality is very frequently omitted in actual language use, especially with inanimate nouns (Heine & König 2011: 166). Interestingly, among the southeastern dialects of Ju|’hoan, the marking of nominal number is becoming increasingly
more systematic and is nearly mandatory for nouns denoting human entities; nonhuman
nouns are still generally not marked for number (Heine & König 2011: 372).
In §4.3, I mentioned that in Amharic, number-unmarked nouns can be interpreted
either as general or singular and that this is quite exceptional among the Semitic languages of the sample. I also mentioned that in Maltese, there exist nouns that may be
described as displaying general number. In descriptive grammars of Maltese, and in line
with the tradition of Arabic grammars, these nouns are usually called collectives. They
are unmarked for number and mostly denote fruits, vegetables and small animals (e.g.,
insects). Depending on their meaning, they display different countability properties and
can be interpreted either as a sort of generic nouns denoting a whole class of entities,
as mass nouns or as plurals. Gil (1996) and Corbett (2000) suggest the possibility of
analysing collective nouns as lying outside the singular/plural opposition, similarly to
general number nouns in Cushitic languages. The status of the collective nouns in Maltese and their relationship with the same category in the Arabic dialects is not discussed
in detail in this dissertation.
4.3.2 Tripartite number systems
Tripartite number system is the label conventionally used in reference grammars to refer
to the strategies of number marking attested in Nilotic or, more broadly, in Nilo-Saharan
languages. Nilotic languages have obligatory number; the overt coding of number distinctions in these languages follows three patterns:
1. singular marking of inherently (i.e., number-unmarked) plural nouns
2. plural marking of inherently (i.e., number-unmarked) singular nouns
3. replacement marking: both singular and plural reference are morphologically overtly
coded.
Tripartite number systems are morphologically very elaborated systems in the sense that
a variety of markers are available for the individual number values, and the distribution
of such markers is both semantically and formally motivated. Examples of tripartite
number systems in Western and Eastern Nilotic languages, the two branches of Nilotic
represented in my sample, are provided below. The first set of examples is taken from
Mabaan, a language of the Western Nilotic branch.
89
4 Gender, number and evaluative morphology in the languages of the sample: an overview
(4.8) Singular formation in Mabaan (Western Nilotic) (Storch 2005: 123)
(a) ñÓÒk
‘lice’
(b) ñÓÓk-cn
”
lice-sg
‘louse’
(4.9) Plural formation in Mabaan (Western Nilotic) (Storch 2005: 123)
(a) jwÓm
‘monkey’
(b) jwÓm-g2
monkey-pl
‘monkeys’
(4.10) Replacement pattern in Mabaan (Western Nilotic) (Storch 2005: 123)
(a) Pú2t-à
worm-sg
‘worm’
(b) Pú2t-n
”
worm-pl
‘worms’
Eastern Nilotic languages have grammatical gender (see §4.2.2) and a tripartite number system. Interestingly, the prefixes that mark gender also encode number distinctions.
Hence, in these languages, number values are encoded twice on noun stems. In spite of
this redundancy, the two strategies are equally active and productive in language use
(Dimmendaal 2000: 233). This type of system is illustrated with a set of examples from
Turkana.
(4.11) Singular formation in Turkana (Eastern Nilotic) (Dimmendaal 1983: 224)
(a) Ni-sikin
m.pl-breast
‘breasts’
(b) E-sikin-a
m.sg-breast-sg
‘one breast’
(4.12) Plural formation in Turkana (Eastern Nilotic) (Dimmendaal 1983: 224)
(a) a-Nasep
f.sg-placenta
‘placenta’
90
4.4 Evaluative morphology
(b) Na-NsÈp-a
f.pl-placenta-pl
‘placentas’
(4.13) Replacement pattern in Turkana (Eastern Nilotic) (Dimmendaal 1983: 224)
(a) e-kk-ut
m.sg-chicken-sg
‘chicken’
(b) Ni-kuku-i
m.pl-chiken-pl
‘chickens’
A more thorough discussion of the number systems of Eastern and Western Nilotic
languages is found in chapter 5.
4.4 Evaluative morphology
In my database, the criteria that I use to classify types of evaluative morphology systems
are:
1. Availability of morphological evaluatives and types of evaluative distinctions
2. Presence or absence of interaction with the encoding of gender
3. Type of marking.
Criterion 1 is the most relevant here, whereas types of interactions with gender (criterion 2) and types of evaluative markers (criterion 3) are more extensively discussed in
chapter 6. The following values are associated with criterion 1:
• Diminutive and augmentatives
• Only diminutives
• Only augmentatives
• None
• No data.
The figures relative to criterion 1 are presented in table 4.13. Morphological evaluative
markers are found in the majority of the languages of the sample (74%). In most cases
(45%), languages have relatively elaborated systems of evaluative morphology in which
both diminutive and augmentative values are grammaticalized. If only one type of
evaluative meaning is grammaticalized, this will always be the diminutive. Thus, the
data from the languages of the sample confirm the typological generalisation according
91
4 Gender, number and evaluative morphology in the languages of the sample: an overview
Table 4.13: Evaluative morphology systems in the languages of the sample
Evaluative distinctions
92
% Genealogical groups
Diminutives and augmentatives
45
45% Bantu (14/23)
Berber (6/6)
Chadic (2/8)
Cushitic (3/13)
Eastern Nilotic (2/3)
Hadza (1/1)
Khoe-Kwadi (3/5)
Kwa (1/3)
Mande (1/4)
North-Central
Atlantic
(5/7)
Semitic (3/7)
Ta-Ne-Omotic (2/4)
Western Nilotic (2/6)
Only diminutives
29
29% Bantu (7/23)
Chadic (1/8)
Cushitic (1/13)
Dizoid (1/1)
Eastern Nilotic (1/3)
Khoe-Kwadi (1/5)
Kwa (2/3)
Kxa (1/1)
Mande (2/4)
North-Central
Atlantic
(2/7)
Sandawe (1/1)
Semitic (3/7)
South Omotic (1/1)
Ta-Ne-Omotic (2/4)
Tuu (1/1)
Western Nilotic (2/6)
Only augmentatives
0
None
2
No data
Total
No. of
lngs.
100
24
100%
0 –
2% Defoid (1/1)
Igboid (1/1)
24% Bantu (2/23)
Chadic (5/8)
Cushitic (9/13)
Khoe-Kwadi (1/1)
Mande (1/1)
Mel (3/3)
Semitic (1/7)
Western Nilotic (2/6)
4.5 Summary of the chapter
to which diminutives are the most frequent type of evaluative morphemes (see, among
others, Dahl 2006).
The geographical distribution of the types of evaluative distinctions across the sample
languages is shown in the map in figure 4.3 (next page). The map in figure 4.3 does not
reveal any significant genealogical and areal skewing. On the other hand, as I shall show
in chapter 6, the distribution of the types of morphological marking (criterion 3) is, in
some cases, areally and genealogically skewed.
Criterion 2, “Presence or absence of interaction with the encoding of gender,” accounts
for interactions between the morphological encoding of diminutives and augmentatives,
on the one hand, and the encoding of gender distinctions, on the other. Evaluative
markers appear to be connected with the marking of gender in 51 languages out of 100.
Two types of interactions are found in the languages of the sample:
(1) Languages have dedicated diminutive and augmentative genders
(2) Languages do not have dedicated evaluative genders, but gender shifts (see 4.2.3 for
a definition of gender shift) are equally used to encode diminutive and augmentative
meanings.
Interactions between gender and evaluative morphology, their relation with types of
evaluative marking and their distribution throughout the languages of the sample are
thoroughly analysed in chapter 6.
Finally, it should be pointed out that the data available on evaluative morphology in
the languages of my sample are somewhat less accurate than in the case of gender and
number systems. No relevant information on evaluative morphology was found in the
sources for 24 languages out of 100. Filling this gap would contribute to a more complete
survey of the attested systems, their distribution and their properties.
4.5 Summary of the chapter
In this chapter, I provided an overview of the gender, number and evaluative morphology systems that are attested in the languages of my sample. The features used in
my database for the classification of gender, number and evaluative morphology in the
languages of the sample have been introduced, and their values described. The distribution of the languages of the sample with respect to each of the features has also been
discussed.
93
4 Gender, number and evaluative morphology in the languages of the sample: an overview
Figure 4.3: Types of evaluative morphology systems in the languages of the sample
94
5 Gender and number
5.1 Introduction
The present chapter illustrates morphosyntactic and semantic interactions between gender and number in the 100 languages of my sample. The research questions addressed
in the chapter are listed in (5.1).
(5.1) Gender and number: research questions (same as (2.13))
Q 1: How common is cumulative exponence of gender and number in the
languages of the sample?
Q 2: What are the formal and semantic factors that trigger gender syncretism in
the context of number? Does gender syncretism in the context of number
presuppose cumulative exponence?
Q 3: What are the implications of cumulative exponence and syncretism on the
absolute complexity of gender and number systems?
Q 4: Can these types of interaction between gender and number be seen as a
reflex of a nominal relevance hierarchy?
Q 5: Can gender and number compete through indexation?
Q 6: Is there any correlation between types of encoding of gender and types of
encoding of number?
Q 7: What types of semantic interactions can be found between gender and
number?
The first four research questions, which are concerned with the issues of exponence
and syncretism, are addressed in §§5.2, 5.3, and 5.4. As discussed in detail in §2.5.1,
cumulative exponence is found when the values of at least two grammatical domains – in
our case, gender and number – are encoded by nonsegmentable markers. The label syncretism is used to describe a phenomenon whereby the values of a particular domain (e.g.,
in the case of gender, masculine and feminine) share a common morphosyntactic realisation in a certain conditioning environment (e.g., that of plural reference) (Baerman et al.
2005; Corbett 1991). One could think of cumulation and syncretism as having scope on
two different dimensions. Cumulation acts on the syntagmatic level, as it has to do
with the way in which gender and number are encoded on words in context. Syncretism
functions at the paradigmatic level, as it relates to the number of gender distinctions
that are available across different number values, and vice versa. An illustration of the
two dimensions is provided in figure 5.1, where I represent cumulation and syncretism
95
5 Gender and number
(a) Cumulation
(b) Syncretism
Figure 5.1: The syntagmatic and paradigatic level of the interaction between gender and number
in a hypothetical language with a sex-based gender (Masculine vs. Feminine) and two
obligatory number values (Singular vs. Plural). It is worth mentioning that the schemes
proposed in figure 5.1 should not be interpreted as having any diachronic reality. Particularly in the case of cumulation, I do not assume that cumulative encoding of gender
and number necessarily presupposes coalescence of two previously distinguished markers. Moreover, even though independent in principle, cumulation and syncretism merge
with each other frequently. As observed in §2.5.3, previous studies by Carstairs (1984),
Carstairs (1987) and Carstairs & Stemberger (1988) have shown that there exists a correlation between cumulative exponence of case and number, on the one hand, and the
occurrence of case syncretism in the context of number, on the other. Similarly, genderand number-related syncretisms could be expected to be more frequent in languages
where the two grammatical domains have cumulative encodings. In this chapter, I test
this hypothesis on the languages of the sample.
The question of competing gender and number indexation patterns (research question
5 in (5.1)) is relevant for those languages with sex-based gender where there are either no
dedicated plural indexation patterns or the use of dedicated plural indexation patterns
is semantically constrained. An example of the first condition would be a language in
which plural nouns always trigger the same indexation pattern, for instance, as feminine
singular nouns. An example of the second condition would be a language in which
a dedicated plural indexation pattern is used, for instance, only when the head noun
is a plural noun denoting a human entity, or with personal pronouns with a human
antecedent. This issue is discussed in §5.5.
Correlations between the presence of gender and type of number marking (question 6
in (5.1)) are investigated in order to test the claim by Creissels et al. (2008), according to
which African languages without gender would tend to display less-elaborated strategies
96
5.2 Cumulation
for the marking of nominal plurality (see also §2.5.4). In §5.6, I try to verify the validity of
this claim on the languages of my sample. Finally, semantic interactions between gender
and number (question 7 in (5.1)) are discussed in §5.7. A summary of the patterns
investigated in the chapter is found in §5.8.
5.2 Cumulation
Ideally, one could think of cumulation as a yes/no parameter, whereby languages can
be classified as either cumulative or noncumulative with respect to the encoding of two
or more grammatical features. This notion of cumulation goes hand in hand with the
idea that the morphology of languages can be classified holistically either as agglutinative (and thus noncumulative) or as flexive (and thus cumulative) (for a discussion
of the notion of holistic morphological typology see, among others, Haspelmath 2009;
Plank 1999). Typological investigations of various grammatical domains have shown
that languages rarely, if ever, function so neatly. With respect to exponence, cumulative
and non-cumulative strategies may co-exist in one and the same language and within
one and the same grammatical domain (see discussion in §2.5.1).
In the case of grammatical features that trigger indexation, such as gender and number,
splits may occur between types of exponence on the indexation triggers and on the
indexing targets. Since gender is bound to indexation, cumulation between gender and
number can truly be assessed only by looking at types of exponence on the indexes.
However, as discussed in §5.2.1.4, in those languages in which gender is overtly coded
on nouns, and nouns inflect for number, patterns of exponence on nouns are also worth
investigating.
One relevant question when investigating types of exponence of grammatical features is
how to treat zero-coded values. The problem is discussed by Plank (1999) (see also Bybee
1994 for a study of the meaning of zeroes within tense and aspect systems; McGregor
2003 for a caveat against the proliferation of zeroes in linguistic analysis; and Dahl 2004
for an account of zeroes and grammatical complexity). According to Plank, the way
zeroes are interpreted – i.e, “in combination or opposition with something” (1999: 303)
– is more relevant than actual morphological segmentability when accounting for presence
or absence of cumulation. Ultimately, what is crucial to understand is how zero-coded
forms interact with the rest of the paradigm and what types of oppositions there are
in the system. In this thesis, cumulation values for each of the sampled languages are
controlled for zeroes in the following way. As a general rule, when zero-coded values
are in paradigmatic opposition to noncumulative, overtly-coded values, the zeroes are
counted as noncumulative exponents. For instance, if language X has zero-marking of
the Masculine Gender and the Feminine is overtly coded by means of a marker that does
not cumulate with number, the zero-coded Masculine is also counted as noncumulative
(examples are discussed in detail in §5.2.1.3).
The following values are associated with the feature “Cumulation” in my database:
1. Cumulative with all indexing targets
97
5 Gender and number
2. Cumulative with some indexing targets
3. Noncumulative
4. No gender
5. No number indexation.
Consequently, languages are classified according to total cumulation, partial cumulation
and absence of cumulation, the ultimate aim being to identify and attempt to motivate
possible distributional preferences in the occurrence of morphological splits related to
the exponence of gender and number. The results are presented and discussed in the
next section.
5.2.1 Cumulation between gender and number: results
Table 5.1 provides the figures relative to patterns of cumulation between gender and
number on indexing targets. Cumulation values, percentages over the number of languages with gender (Rel. %) and over the total number of languages in the sample
(Abs. %) are listed together with the distribution of each value throughout genealogical
groupings.
As shown in table 5.1, the distribution of the sampled languages with respect to
cumulation between gender and number is highly skewed: 81% of the languages with
gender have cumulative encoding of gender and number on all indexes. Only two languages within the sample completely lack cumulation between gender and number on
the indexes. Finally, none of the languages with gender in the sample completely lacks
number indexation (Western Nilotic languages lack number indexation, but they also
lack gender).
As mentioned before, the languages of the sample differ a great deal with respect to
the number and the types of possible indexes of gender. Such differences are, however,
not relevant to the data discussed in this section (see chapter 7 for the relationship
between number of indexes and grammatical complexity of gender systems).
98
5.2 Cumulation
Table 5.1: Cumulation between gender and number on the indexing targets
Cumulation value
Cumulative with all indexes
No. of
lngs.
Rel.%
Abs.% Genealogical
groups
68
81%
68% Bantu (23/23)
Chadic (5/8)
Cushitic (6/13)
Dizoid (1/1)
Eastern
Nilotic
(3/3)
Hadza (1/1)
Khoe-Kwadi (5/5)
Kwa (1/3)
Kxa (1/1)
Mel (3/3)
North-Central Atlantic (6/7)
Semitic (7/7)
South Omotic (1/1)
Ta-Ne-Omotic
(3/4)
Tuu (1/1)
Sandawe (1/1)
Cumulative with some
indexes
14
16.6%
14% Berber (6/6)
Cushitic (6/13)
Chadic (1/8)
Ta-Ne-Omotic
(1/4)
Noncumulative
No gender
No number indexation
Total
2
2.4%
2% Cushitic (1/13)
North-Central Atlantic (1/7)
16
–
16% Chadic (2/8)
Defoid (1/1)
Igboid (1/1)
Kwa (2/3)
Mande (4/4)
Western
Nilotic
(6/6)
0
–
100
100%
– –
100%
99
5 Gender and number
5.2.1.1 Cumulative encoding of gender and number with all the indexes
Cumulation between gender and number on the indexes is found both in languages with
obligatory number marking (5.2) and in languages in which number marking on nouns
is not obligatory (5.3).
(5.2) Gender and number indexation in Bandial (North-Central Atlantic) (Sagna
2008: 190)
(a) imbi eno fi-ttix
fafu
fu-mug-i-mux
perm if
cl7-war cl7:def cl7.3sg-kill-2sg.obj-dup
‘lit: If it happens that the war kills you’ (If you die during the war)
(a) imbi eno gu-ttix
gagu
gu-mug-ul-mux
perm if
cl8-war cl8:def cl8.3pl-kill-2pl.obj-dup
‘lit: If it happens that the wars kill you’ (If you (pl die during the wars)
(5.3) Gender and number indexation on demonstratives in Gidar (Chadic) (adapted
from Frajzyngier 2008: 319)
(a) â´@f `@n-ká
man this-m.sg
‘this man’
(b) â´@k
k´@-n-k´@
woman f.sg-this-f.sg
‘this woman’
(c) âı́
ı̀n-kı́
men this-pl
‘these men’
As in many other languages of the sample, in Gidar, gender distinctions are neutralized
under plural reference (see §5.3 for more details on syncretism of gender in the context
of number).31
The noun class system of Ju|’Hoan (Kxa) can be considered as a peculiar instance of
cumulative exponence of gender and number, and deserves a short note. Ju|’Hoan has
four noun class markers:
• Class 1: ha
• Class 2: sı̀
• Class 3: yı̀
• Class 4: ka
31
The proximate demonstrative marker in Gidar is -n; `@n- and ı̀n- are two different allomorphs of -n.
100
5.2 Cumulation
With animate nouns, the gender markers also serve the purpose of encoding number
distinctions. Inanimate nouns are outside the number system; with such nouns, the
gender markers do not bear any relationship with the encoding of singular and plural
reference. Table 5.2 illustrates the semantics of the gender system of Ju|’Hoan and, in
the case of animate nouns, its relationship with the encoding of number distinctions (see
Güldemann & Vossen 2000; Güldemann 2000; Heine & König 2011 for a more detailed
account of gender in Ju|’Hoan, as well as §5.3.2 for gender syncretism in Ju|’Hoan).
Table 5.2: The semantics of the noun class system of Ju|’hoan (adapted from Heine & König
2013: 156)
Gender
Number
Salient meanings
cl1/cl2
cl1/cl3
Singular/Plural
Singular/Plural
cl1
General number
cl3
cl4
General number
General number
Human nouns
Many animals, inanimate
nouns
Many plants and plant
products,
inanimate
nouns
Mostly inanimate nouns
Inanimate nouns, body
part nouns,
abstract
nouns
5.2.1.2 Cumulative encoding of gender and number with some indexes
As shown in table 5.1, in 16.6% of the languages with gender in the sample, cumulation
between gender and number is not found with all the possible targets of gender and
number indexation but only with some of them. This pattern of partial cumulation is
mainly attested among languages of the Afro-Asiatic family. The number of cumulative
and noncumulative indexes can vary a great deal across individual languages.
An interesting system is found in the Berber languages. Gender and number are
noncumulative on adjectival modifiers where they are encoded in the same way as on
nouns. The Masculine Gender is zero-marked, whereas the Feminine is overtly coded
by means of the t morpheme. This, depending on the indexing target, can occur as a
prefix, infix or suffix. The marking of gender and number on different types of pronouns
(personal, possessive, demonstrative) and on the verbs tend to also be noncumulative.
However, instances of cumulation are also found. Consider, for instance, the paradigm
of the Independent Personal Pronouns in Kabyle, as represented in table 5.3.
101
5 Gender and number
Table 5.3: Independent Personal Pronouns in Kabyle (Mettouchi 2000)
Person
1sg
1pl
2sg
2pl
3sg
3pl
Masculine Feminine
nekk
nekw ni
nekw nti
kecc
kemm
kunwi
kunnemti
neťťa
neťťat
nutni
nutenti
Cumulation value
–
Noncumulative
Cumulative
Partially Cumulative
Noncumulative
Noncumulative
Idiosyncrasies in inflectional patterns, as the one illustrated in table 5.3, are the result of language-internal developments whose description falls outside the scope of this
investigation.
In Cushitic languages, gender and number may not be overtly coded on the same
indexes. In 5 of the 13 Cushitic languages of my sample, gender and number are encoded
cumulatively at least on the Independent Third Person Pronouns. A more thorough
discussion of gender and number indexation in Cushitic is found in §5.5.
In Koorete (Ta-Ne-Omotic), the Demonstrative Modifiers inflect neither for gender
nor for number. On the other hand, the Demonstrative Pronouns inflect both for gender
and for number, but gender and number values are encoded by separate exponents. This
is illustrated in example (5.4).
(5.4) Demonstrative Pronouns in Koorete (Ta-Ne-Omotic) (adapted from table in
Teketal 2004: 168)
(a) há-Pis-a
this-m-sg.indf.abs
‘this’
(b) haá-Pis-iâe
this-m-pl.indf.abs
‘these’
5.2.1.3 Noncumulative encoding of gender and number on the indexing targets
Only two languages have been found in the sample in which the encoding of gender and
number on the indexes is never cumulative. These are the Cushitic language Beja and
the North-Central Atlantic language Wamey. The two languages have different types
of gender systems, with Beja having a sex-based gender and Wamey a non-sex-based
gender.
In Beja, similarly to the Berber languages, the encodings of the Masculine Gender and
the Feminine Gender on the indexes differ from each other only for the presence/absence
of the feminine marker t. This is added to the individual singular and plural indexes
without there being any interaction with the encoding of singular and plural reference
(Roper 1928). The paradigm of the Definite Article in Beja is presented in table 5.4 as
an illustration of this pattern of encoding.
102
5.2 Cumulation
Table 5.4: The Definite Article in Beja (Roper 1928: 9)
Case
Masculine Feminine
sg
pl
sg
pl
nom and voc
ū
ā
tū
tā
acc and other cases
o
E
to
tE
In Wamey, gender and number indexes are also non-cumulative. Figure 5.2 provides
the inventory of the singular and plural noun class markers in Wamey, and their corresponding singular-plural pairings (these markers are used both on the targets and the
nouns themselves). The noun classes of Wamey have CV syllable structure, whereby C
is always a labial consonant (v or w ). The vocalic component of the CV template is, in
nearly all cases, the vowel of the corresponding singular class. The ordinal numbers that
occur after the noun class prefixes in figure 5.2 indicate the type of consonant alternation
that the prefix triggers on the noun stem (see §4.2.3.2 for a definition of consonant alternation in the Atlantic languages). It is worth mentioning that certain class pairs (e.g.,
5/6 and 7/8) are identical with each other in terms of their phonological appearance
but trigger different patterns of consonant alternation on the nouns and on the indexing
targets. They are thus classified by Santos (1996) as different genders.
Concerning the morphology of the plural class markers in Wamey, it has been suggested (Santos 1996; Ferry & Pozdniakov 2001; Pozdniakov 2010, 2013) that the labial
component of the plural noun classes – v- (and w-) – is, in fact, a plural marker that
precedes the actual noun class marker. This generalised plural morpheme is built on the
model of the plural noun class for human nouns, [email protected] At some point in the history of the
language, the labial consonant must have been interpreted as a plural morpheme and
was then extended to the entire paradigm, independently of the animacy of the noun.
Konstantin Pozdniakov (personal communication) suggests that the phenomenon might
have originated from the use of markers of Classes 1 and 2 as generalised anaphoric
pronouns. In figure 5.2, the labial component of the plural noun classes is presented in
boldface.
In sum, as a result of innovation in the encoding of plural reference, in Wamey plurality
is always overtly coded by means of a labial consonant, whereas zero-coding of number
is interpreted as expressing singular reference. Gender and number are thus no longer
coded cumulatively.
103
5 Gender and number
Figure 5.2: The noun class system of Wamey (North-Central Atlantic) (Santos 1996: 145)
104
5.2 Cumulation
5.2.1.4 Cumulation between gender and number on nouns
When gender is overtly coded on nouns (see data and discussion in §4.2) and nouns inflect
for number, interesting facts can be observed with respect to the exponence of gender and
number values. Even though this phenomenon is not central to the current investigation
of gender systems, I tried to account for patterns of cumulation between gender and
number on nouns in my database by adding an additional feature: “Cumulation between
gender and number on nouns.” The following values are assigned to this feature:
1. Cumulative
2. Cumulative sg vs. noncumulative pl
3. Noncumulative with specific gender (G) and number (N) markers
4. Noncumulative
5. No gender marking on nouns
6. No gender.
Table 5.5 provides the figures relative to the distribution of patterns of cumulation between gender and number on nouns across the languages of the sample. As shown in the
table, 15 out of the 84 languages with gender in the sample do not have overt marking of
gender on nouns and are thus excluded from the count of patterns of cumulation between
gender and number on nouns. In slightly more than half of the languages with gender
in the sample, the encoding of gender and number on nouns is cumulative. Cumulative
encoding of gender and number on nouns is found throughout the individual genealogical
groupings represented in the sample. Immediately after come the 17 languages with noncumulative marking, which constitute the 20.2% of the languages with gender. These 17
languages belong to the following genealogical units: Berber, Chadic, Cushitic, Dizoid,
North-Central Atlantic and Ta-Ne-Omotic. Eight of these seventeen languages mark
number obligatorily on nouns, whereas nine have optional number marking on nouns.
The remaining 9.5% of the gendered languages of the sample is characterised by some
sort of split system whereby gender and number values on nouns can have both cumulative and noncumulative exponence. In my database, these languages are classified as
belonging to one of the two groups in between “cumulative” and “noncumulative” that
are listed in table 5.5. These two groups are discussed in detail in the two next following
sections.
105
5 Gender and number
Table 5.5: Cumulation between gender and number on nouns
Cumulation
value
No. of
lngs.
Rel.%
Abs.% Genealogical groups
44
52.4%
44% Bantu (22/23)
Cushitic (2/13)
Hadza (1/1)
Khoe-Kwadi (5/5)
Kwa (1/3)
Mel (3/3)
North-Central Atlantic
(5/7)
Semitic (2/7)
Ta-Ne-Omotic (2/4)
Tuu (1/1)
Cumulative sg vs.
noncumulative pl
1
1.2%
1% Bantu (1/23)
Noncumulative
with specific G
and/or N markers
7
8.3%
7% Eastern Nilotic (3/3)
Cumulative
Semitic (4/7)
Noncumulative
17
20.2%
17% Berber (6/6)
Chadic (2/8)
Cushitic (6/13)
Dizoid (1/1)
North-Central Atlantic
(1/7)
Ta-Ne-Omotic (1/4)
No gender marking
on nouns
15
17.9%
15% Chadic (4/8)
Cushitic (5/13)
Kxa (1/1)
North-Central Atlantic
(1/7)
Sandawe (1/1)
Semitic (1/1)
South Omotic (1/1)
Ta-Ne-Omotic (1/6)
No gender
Total
106
16
100
100%
16% Chadic (2/8)
Defoid (1/1)
Igboid (1/1)
Kwa (2/3)
Mande (4/4)
Western Nilotic (6/6)
100%
5.2 Cumulation
5.2.1.4.1 Cumulative singular vs. noncumulative plural This is attested only in one
language of the sample: Kinshasa Lingala (Bantu). The noun class system of Kinshasa
Lingala is a reduced version of the system that is found in the standard variety of the
language, Makanza Lingala. Examples (5.5) and (5.6) illustrate the similarities and
differences among the two varieties.
(5.5) Makanza Lingala (Bantu) (Bokamba 1977: 184)
(a) mo-to
cl1-person
‘person’
(b) ba-to
cl2-person
‘people’
(c) li-loba
cl5-word
‘word’
(d) ma-loba
cl6-word
‘words’
(e) n-dako
cl9-house
‘house’
(f) n-dako
cl10-house
‘houses’32
(5.6) Kinshasa Lingala (Bantu) (Bokamba 1977: 184)
(a) mo-to
cl1-person
‘person’
(b) ba-to
cl2-person
‘people’
(c) li-loba
cl4-word
‘word’
(d) ba-ma-loba
pl-cl6-word
‘words’
(e) n-dako
cl9.house
‘house’
(f) ba-n-dako
pl-cl10-house
‘houses’33
The striking difference between Makanza and Kinshasa Lingala is that, along with
nouns that do not belong to Gender 1/2, Kinshasa Lingala seems to mark plurality
twice: (1) by regular class shift and (2) by adding the prefix ba-, that is, the plural class
prefix for nouns of Gender 1/2. The plural marker initially used only with human nouns
has been generalised as a plural morpheme for all nouns. This has given rise to a situation
in which number marking on nouns is somewhat redundant, and gender and number on
nouns are, at least in part, noncumulative. The generalised use of Class 2 as a plural
marker on nouns is also found in other Bantu languages, where it is usually semantically
32
In Lingala, as in other Bantu languages, Classes 9 and 10 are not distinguishable on nouns. They do,
however, trigger different indexation patterns on the indexing targets.
33
See footnote 29.
107
5 Gender and number
constrained in the sense that it is restricted to animate nouns only (Maho 1999: 134136). One could speculate that, initially, the use of class 2 as a generalised nominal plural
marker in Kinshasa Lingala might have been also semantically constrained. Interestingly,
even though the patterns of gender and number indexation of Kinshasa Lingala are quite
impoverished, gender and number are still encoded cumulatively on the indexing targets
(Bokamba 1977 as well as §7.7).
5.2.1.4.2 Noncumulative with specific gender and/or number markers This pattern
is found in the Eastern Nilotic and some Semitic languages of the sample. It only
concerns nouns and does not have any parallels on the indexing targets. The Eastern
Nilotic languages Turkana, Masai and Karamojong mark number twice. Every noun is
marked by prefixes that express gender and number cumulatively. In addition, number
is marked by singular and plural suffixes, which do not cumulate with gender. This is
illustrated in (5.7) with an example from Turkana.
(5.7) Gender and number marking in Turkana (Eastern Nilotic), same example as
(4.13) (Dimmendaal 1983: 224)
(a) e-kk-ut
m.sg-chicken-sg
‘chicken’
(b) Ni-kuku-i
m.pl-chiken-pl
‘chickens’
Such a redundancy in number marking is considered to be the result of historical
developments that are internal to the Eastern Nilotic genealogical group. The singular
and plural suffixes of the Eastern Nilotic languages are shared with the other languages
of the Nilotic genus and are inherited from the protolanguage. On the other hand,
the gender/number cumulative prefixes are a later innovation within Nilotic, and are
only attested in the Eastern Nilotic branch. A more thorough discussion of the number
system of the Nilotic languages is found in §5.6.
In Standard Arabic, Moroccan Arabic, Maltese and Tigre, nominal plurality on nouns
can be encoded either by suffixes or by patterns of vowel alternation on the nouns stem
(also known as Broken Plurals). Broken Plurals in these languages never convey information about gender. Example (5.8) illustrates one type of Broken Plural in Standard
Arabic.
108
5.3 Syncretism
(5.8) Broken Plurals in Standard Arabic (Semitic) (Ryding 2005: 146)
(a) Taalib
student\[sg]
‘students’
(b) Tullaab
student\pl
‘students’
5.2.1.5 Summary
In this section, I discussed the occurrence of cumulative exponence of gender and number
in the languages of the sample (thus addressing question 1 in (5.1)). The data show that
the large majority of the languages of the sample have cumulative encoding of gender
and number on all indexing targets. Noncumulative, or partially cumulative, encodings
of gender and number tend to be found in languages where nominal number marking
is not obligatory, but they are also attested in languages with obligatory number (e.g.,
Berber). As far as exponence of gender and number on nouns is concerned, with the
exception of the Berber languages and the North-Central Atlantic language Wamey,
noncumulative encodings tend to be found in languages where number marking on nouns
is not obligatory but gender is. More hybrid patterns are also attested and are usually
the result of language-internal developments.
5.3 Syncretism
Syncretism is a type of paradigmatic asymmetry that features formal identity of cells
within an inflectional paradigm (see Baerman et al. 2005 as well as §§2.5.2 and 5.1).
Gender distinctions are particularly prone to undergo syncretism across number values.
For instance, the three gender values of language X can be conflated under dual reference
but still be distinguished under singular and plural. In addition, the occurrence of gender
syncretism may also be restricted by indexing targets. For instance, the demonstratives
in language Y may be insensitive to the distinction between the feminine and neuter
gender, which is otherwise overtly coded on other indexing targets, such as adjectives
and verbs. In this thesis, I mainly investigate instances of syncretism of the first type.
In §2.5.2, I surveyed previous crosslinguistic studies on the occurrence of gender syncretism across number and observed that most of these studies capitalise on the idea
that gender and number-related syncretism reflects markedness relationships within and
across the two grammatical domains. In particular, I showed that gender distinctions
tend to be syncretic in the context of number. In general, gender distinctions are reduced or neutralized under nonsingular number values. I now turn to an overview of the
results.
109
5 Gender and number
5.3.1 Syncretism of gender in the context of number: results
Overall, 67 languages among the 84 gendered languages in the sample have syncretism.
In addition, the data suggest that there is a strong correlation between patterns of
syncretism between gender and number, on the one hand, and cumulative exponence of
gender and number, on the other. This is shown in table 5.6.
The results suggest that in the languages of my sample, the possibility of having syncretism without cumulation is nearly ruled out. The only two languages where syncretism
is found in the absence of cumulation are Beja and Wamey (see §5.2.1.3). Otherwise, in
my sample, all the languages with syncretism also have either total or partial cumulative encodings of gender and number. As discussed later in §5.4, this is in line with the
results of Carstairs (1984), Carstairs (1987), and Carstairs & Stemberger (1988) with
respect to case syncretism in the context of number.
Two types of gender syncretism occur under plural reference in the languages of my
sample:
(1) The syncretic plural marker is different from each of the correspondent singular
markers (Figure 5.3a, examples follow throughout the section)
(2) The syncretic plural marker is the same as one of the correspondent singular markers (Figure 5.3b, examples follow throughout the section).
(a) Syncretism 1
(b) Syncretism 2
Figure 5.3: Types of gender syncretism across number
Patterns of syncretism in which the syncretic marker is formally identical to one of its
component values – as in (2) – are generally less frequent than patterns of syncretism
in which there is no such formal identity – as in (1). The same crosslinguistic tendency
is reflected by the languages of my sample. Syncretisms of the type described in (1) are
the most frequent.
110
5.3 Syncretism
Table 5.6: Gender syncretism and its relationship with cumulation in the language sample
Syncr.
Cumulation
value
X
Cum.: all indexes
X
Cum.:
dexes
some in-
No. lngs.
Rel.% Abs.% Genealogical
groups
57
68%
8
9.5%
57% Bantu (22/23)
Chadic (5/8)
Cushitic (6/13)
Dizoid (1/1)
Eastern
Nilotic
(3/3)
Kwa (1/3)
Mel (3/3)
North-Central Atlantic (6/7)
Sandawe (1/1)
Semitic (5/7)
South Omotic (1/1)
Ta-Ne-Omotic
(3/4)
8% Chadic (1/8)
Cushitic (6/13)
Ta-Ne-Omotic
(1/4)
X
Noncum.
Cum.: all indexes
Cum.:
dexes
No
gender
some in-
No gender
2
2.4%
2% Cushitic (1/3)
North-Central Atlantic (1/7)
11
13.1%
6
7%
6% Berber (6/6)
16
–
16% Chadic (2/8)
11% Bantu (1/23)
Hadza (1/1)
Khoe-Kwadi (5/5)
Kxa (1/1)
Semitic (2/7)
Tuu (1/1)
Defoid (1/1)
Igboid (1/1)
Kwa (2/3)
Mande (4/4)
Western
Nilotic
(6/6)
Total
100
100%
111
5 Gender and number
For instance, in Turkana (Eastern Nilotic), nouns are assigned either to the Masculine,
the Feminine or the Neuter Gender. The Masculine and Neuter genders are syncretic
under plural reference. The plural marker for masculine and neuter nouns differs from
both the corresponding masculine singular and neuter singular markers. This is shown
in table 5.7 in which the paradigm of the restrictive markers is provided.34
Table 5.7: Gender syncretism in Turkana restrictive markers (adapted from Dimmendaal
1983: 217)
Singular Plural
f.
m.
n.
aei-
NaNiNi-
On the other hand, gender syncretisms of the type described above in (2), and represented in figure 5.3b, are quite rare. Only 8 of the 67 languages with syncretism in my
sample have syncretism of type (2). They are, in order of genealogical affiliation: Baiso
(Cushitic), Borana-Arsi-Guji Oromo (Cushitic), Dasaanach (Cushitic), Iraqw (Cushitic),
Kambaata (Cushitic), Bench (Ta-Ne-Omotic) and Standard Arabic (Semitic). Moreover,
it seems that syncretism of type (2) never exists on its own in a language. In fact, in all
the cases where this pattern is attested within my sample, either syncretism of type (1)
or no syncretism at all are also found. For instance, in Standard Arabic, plural nouns
trigger different types of gender and number indexation depending on the animacy of
the np referent. Two patterns are attested:
(1) Plural nouns denoting human beings do not undergo gender syncretism under
plural reference (5.9).
(5.9) Gender and number indexation with plural human nouns in Standard
Arabic (Semitic) (Ryding 2005: 126)
c askariyy-uuna
(a) qaadat-un
pl\leader-nom military-m.pl
‘military leaders’
(b) al-nisaaO -u
l-mutaqaddim-aatu fii l-sinni-i
def-women-nom def-avanced-f.pl in def-age-gen
‘women of advanced age’
34
The notion of restrictiveness in Turkana is “allied with, but not exactly equivalent to, definiteness”
(Baerman et al. 2005: 84, fn 25).
112
5.3 Syncretism
(2) Nonhuman plural nouns trigger the same indexation pattern as feminine singular
(5.10)
(5.10) Gender and number indexation with non-human plural nouns in Standard
Arabic (Semitic) (Ryding 2005: 126)
al-funuun-u
l-muc aaSir-atu
pl\def-art-nom def-contemporary-f.sg
‘contemporary arts’
Sometimes, when construed as abstract nouns (Ryding 2005: 126), human collectives also trigger feminine indexation:
(5.11) Gender and number indexation with human collectives in Standard Arabic
(Semitic) (Ryding 2005: 126)
al-suluTaat-u
l-ruumaaniyy-atu
pl\def-autorithy-nom def-Roman-f.sg
‘the Roman authorities’
The gender and number system of Standard Arabic is thus structured in such a way that
gender distinctions under plural reference are maintained only when the referent of the
np is human. Instead of being distributed arbitrarily to one gender or the other – as it
happens under singular reference – all nonhuman plural nouns behave in the same way
with respect to the indexation patterns that they trigger. Thus in Standard Arabic, the
occurrence of syncretism of type (2) is constrained by the semantics of the head noun.
The languages with syncretism of type (2), and the way their distribution competes
with that of languages with type (1) syncretism, or no syncretism at all, are surveyed
in §5.5 in the context of a discussion on restrictions on the use of dedicated patterns of
plural indexation.
5.3.2 Semanticization of gender distinctions under syncretism
One of the typological generalisations on gender syncretism formulated by Baerman et al.
(2005) is that in languages with large gender systems, not only are gender distinctions
reduced under plural reference, but they also tend to be consolidated and realigned on
the basis of more transparent semantic criteria than the corresponding singular forms.
Baerman et al. refer to this phenomenon as semanticization of gender distinctions under
plural reference (2005: 86). Their generalisation has been tested on the 34 languages
with large gender systems in my sample. Only four of these languages confirm this
tendency: the Maasina Fulfulde (Nort-Central Atlantic, North), Nuclear Wolof (NorthCentral Atlantic, North), Kisi (Mel) and Gola (Mel). Two of these languages, Maasina
Fulfulde and Kisi, are also part of the sample created by Baerman et al. (2005). In
all four languages, the increase in the semantic predictability of gender distinctions in
the plural is animacy-driven. The patterns of gender syncretism across number in the
Maasina Fulfulde and Kisi were previously discussed by Baerman et al. (2005: 87-89). In
the rest of this section, I mostly concentrate on the systems of Gola and Nuclear Wolof.
113
5 Gender and number
The gender system of Gola is outlined in table 5.8. Besides the class pairs illustrated
in the table, Gola also has one locative class that does not distinguish number.
Table 5.8: Gender and number distinctions in Gola (adapted from Fachner 1990: 10)
Singular Plural Semantics of the plural classes
wowONkeeke-
aNmaNmaNmaN
le-
Animate
Inanimate
Sgv. vs. Clv.35
As shown in table 5.8, the marking of nominal plurality in Gola is bound to the opposition
between animate and inanimate nouns. Gender distinctions are reshuffled under plural
reference on the basis of this clear-cut semantic opposition.
A similar pattern of gender syncretism across number is found in Nuclear Wolof.36
Nuclear Wolof has eight singular classes and two plural classes, which are illustrated in
table 5.9.
Table 5.9: Gender and number distinctions in Nuclear Wolof (adapted from McLaughlin
1992: 200)
Singular Plural
kbgjlmsw-
ñyyyyyyy-
Class k- and ñ- are very restricted in use. Interestingly, the semantically opposite nouns
nit ‘person’ and këf ‘thing’ are the only nouns assigned by default to class k- when
singular. Class ñ- is the regular plural for nit and, depending on the speakers, is also
used as the plural marker for the following highly animate nouns: jigéen ‘women,’ góor
‘men’ and gaa ‘people.’ All other nouns in Nuclear Wolof, including këf ‘thing,’ form
their plural in Class y- (McLaughlin 1992: 200-201). The massive reduction of the
number of class distinctions under plural reference is accompanied, at least for some
speakers, by some sort of semantic realignment whereby the most generic human nouns
– ‘man,’ ‘woman’ and ‘person’ – receive the same type of plural marking.
35
This is an extremely rare gender in Gola. Class le- expresses collective meanings, and Class ke- is
sometimes used to derive singulative-like nouns from collective nouns, e.g., le-kuo ‘a collection of
eggs,’ ke-kuo ‘a single egg’ (Fachner 1990: 10).
36
The variety of Wolof referred to in this thesis is the one described by McLaughlin (1992, 1997). In
Glottolog (Nordhoff et al. 2013), this is referred to as Nuclear Wolof.
114
5.3 Syncretism
In Kisi, the plural classes are also semantically more clear-cut than the singular. What
is typologically more exceptional in the gender system of Kisi is that the singular noun
classes are more syncretic than the plural. There are three distinct noun classes under
singular reference and five under plural. In Baerman et al.’s (2005) typological study
of syncretism, Kisi stands out as a fairly unique instance of gender-related syncretism.
This applies to my language sample, too (for a more detailed discussion of the gender
system of Kisi, see Baerman et al. 2005: 88-89).
In the majority of cases, gender syncretism under plural reference in languages with
large gender systems does not seem to work on the basis of semantically transparent
criteria. For instance, the North-Central Atlantic language Noon has six genders with six
singular classes and two plural classes. Synchronically, the distribution of the two plural
classes over the singular is not semantically motivated. Animate nouns receive extra
marking on the determiners under both singular and plural reference. Such markers
are, however, extraneous to the gender system (Soukka 2000: 65-66). Maho (2003)
carried out a survey of syncretic plural classes in the Bantu languages that led to similar
results. In his paper, he refers to plural classes that pair with more than one singular
class as polyplural classes. The use of the syncretic plural classes in Maho’s sample is
not semantically constrained nor is their distribution specifically restricted to individual
areas within the Bantu-speaking world. Thus, syncretism of certain gender distinctions
under plural reference can be placed far earlier in the history of the Bantu languages,
and its semantic motivation, if any, is impossible to trace (Maho 2003: pp.165-169).
One last interesting case among the languages of the sample with non-sex-based gender
is Ju|’Hoan (Kxa). Based on Dickens (1992), Baerman et al. (2005) describe the noun
classes of the language as a very rare type of system, where gender distinctions under
plural reference outnumber gender distinctions under singular reference, and the latter
are semantically more transparent than the former. In fact, as I mentioned in §5.2.1, in
Ju|’Hoan, only with animate nouns do gender markers encode number as well. Inanimate
nouns have general number, and, for them, gender assignment is not relevant to the
encoding of number distinctions. Ultimately, one could suggest that the system of gender
indexation of Ju|’Hoan falls outside the scope of syncretism.
Baerman et al. (2005: 84-86) suggest that the semanticization of gender distinctions
under non-singular reference can also be found in languages with smaller gender systems. In a tripartite sex-based gender system of the “masculine vs. feminine vs. neuter
type,” the masculine and feminine genders are generally more likely to undergo syncretism because, on the basis of their respective semantic cores, they constitute a sort of
natural class with shared semantic features such as human or animate (Baerman et al.
2005: 84). However, the opposite is also found in the languages of the world. For instance, among the languages of my sample, as well as in the sample investigated by
Baerman et al., Turkana has a tripartite gender system of the type described above, but
the Masculine Gender is syncretic with the Neuter rather than with the Feminine (see
Table 5.7).
Beside gender syncretism across number, an interesting case of gender syncretism restricted by target is found among the languages with non-sex-based gender in my sample.
The phenomenon is traditionally known as animate concord (Wald 1975; Bokamba 1977;
115
5 Gender and number
Maho 1999) and occurs when, at least for some targets, gender indexation is determined
on the basis of the animacy of nouns rather than on the basis of their morphological
properties. As pointed out in §5.3.1, gender syncretism is restricted by targets when
two nouns that in principle belong to two different genders trigger the same gender indexation pattern only with some targets. Animate concord is a type of semantically
motivated syncretism restricted by targets. It thus triggers semanticization of gender
distinctions in Baerman et al.’s (2005) terms. Animate concord is found in 11 of the
35 languages with non-sex-based gender in my sample37 . These languages differ in the
number and types of indexing targets that show animate concord. In urban varieties
of Swahili, for example, animate concord is highly pervasive, since it is used with all
targets but possessive modifiers (Wald 1975). This is shown in example (5.12): gender
indexation on the verb is based on the referential properties of the np (Class 1 indexation is selected because the np referent is animate), whereas gender indexation on the
possessive modifier is based on the lexical gender of the noun.
(5.12) Swahili (Bantu) (Wald 1975)
Rafiki
y-angu
a-me-fika
cl9.friend cl9-of.me cl1-prf-arrive
‘My friend has arrived’
In Nyanja, animate concord is quite marginal since it is only attested with the Personal
Pronouns, and in pragmatically marked contexts (Corbett & Mtenje 1987: 13-14). All 11
languages with animate concord in the sample are either North-Central Atlantic, Bantu
or Mel, with the exception of !Xóô, which is a Tuu language. There is a tendency for
languages with animate concord to cluster areally (Wald 1975; Childs 1983; Ström 2011).
Interestingly, in my sample, animate concord is found among the Languages of Wider
Communication38 (e.g., Lingala, Nuclear Wolof, Swahili, etc.) or among languages that
have intense contact with Languages of Wider Communication (e.g., Ndengereko, with
respect to Swahili). According to Maho (1999: 123), it is reasonable to think that the
presence of animate concord might be underrepresented in reference grammars due to
the fact that it is often perceived as a substandard phenomenon.
5.3.3 Summary
In this section, I discussed the occurrence of gender syncretism in the context of number
in the languages of the sample. The data from the language sample suggest that the odds
of gender being syncretic across number are higher when gender and number values are
encoded cumulatively. In addition, they also suggest that patterns of gender syncretism
37
These languages are: Gola (Mel), Kinshasa Lingala (Bantu), Kisi (Mel), Maasina Fulfulde (NorthCentral Atlantic), Ndengereko (Bantu), Noon (North-Central Atlantic), Nyanja (Bantu), Swati
(Bantu), Swahili (Bantu), Timne (Mel), !Xóô (Tuu).
38
Languages of Wider Communication are languages that are used by individuals with different mother
tongues to communicate with each other (Bokamba 2009; Heine 1970; Mufwene 2003).
116
5.4 Cumulation and syncretism: summary and discussion
across number are usually difficult to motivate from a semantic point of view. There are
a few exceptions whereby gender syncretism across number is driven by the animacy of
the nouns (§5.3.2).
5.4 Cumulation and syncretism: summary and discussion
Cumulative exponence of gender and number turned out to be very frequent in the
languages of my sample and so did syncretism. In addition, I found that syncretism
of gender in the context of number largely entails cumulative exponence of the values
of the two domains. These results confirm what has already been found by Carstairs
(1984), Carstairs (1987), and Carstairs & Stemberger (1988) with respect to patterns of
case syncretism in the context of number. Syncretism of the values of one grammatical
domain (e.g., gender or case) in the context of another domain (e.g., number ) is more
likely to occur when the two domains have cumulative exponence.
In the rest of this section, I discuss cumulative exponence and syncretism in relationship with language complexity (question 3 in (5.1)) and consider the significance of these
findings for the notion of nominal relevance hierarchy (question 4 in (5.1)).
5.4.1 Cumulation, syncretism, and grammatical complexity
As mentioned in §2.7, in this thesis I interpret language complexity in an absolute sense,
that is, as an objective property of a given system (Dahl 2004; Miestamo 2008; Sinnemäki
2011). Absolute complexity can be measured on the basis of the number of parts that
a system is made up of or the length of its description. One of the assumptions of this
approach to grammatical complexity is that the least complex grammatical domain is one
in which each form is associated with only one meaning (this is referred to as Principle
of One-Meaning–One-Form in Miestamo 2008). Within this assumption, cumulation
increases the absolute complexity of grammatical domains. A cumulative marker is
associated with at least two grammatical meanings, and this determines asymmetries in
the relationship between meaning and form.
With respect to syncretism, my results show that syncretic paradigms also tend to
be cumulative paradigms that in turn feature an increase in absolute complexity. The
presence of syncretism thus presupposes absolute complexity but does not necessarily
add further complexity to the structure of inflectional paradigms. These notions are
operationalised in chapter 7, where a metric for gender complexity is proposed (see also
Dahl: 188, and the discussion in §2.7).
5.4.2 Syncretism and relevance hierarchy for nominal features
A few studies (Carstairs 1987; Carstairs & Stemberger 1988; Greenberg 1963b; Vafaeian
2013), chronologically quite distant from each other and conducted on the basis of independent datasets, have suggested that there exist asymmetries in the behaviour, and
mutual interactions, of different features of nominal morphology (see §2.5.3 for an introduction). Such asymmetries have been interpreted as pointing at the existence of a
117
5 Gender and number
relevance hierarchy for nominal features similar to the one proposed by Bybee (1985)
for the verbal domain. The above-mentioned studies focus on two phenomena:
(1) Syncretism, based on the assumption that the most relevant features are the
least likely to undergo syncretism (Carstairs 1987; Carstairs & Stemberger 1988;
Greenberg 1963b).
(2) Suppletion, based on the assumption that the higher a feature is on a relevance
hierarchy, the more likely it is to trigger suppletion (Vafaeian 2013, based on the
results of Veselinova 2006 on patterns of verbal suppletion).
The results of these studies converge in pointing to number as the highest ranking feature of a possible nominal hierarchy for nouns. Greenberg (1963b), Carstairs (1987) and
Carstairs & Stemberger (1988) show that number – as opposed to grammatical gender
(Greenberg 1963b) and case (Greenberg 1963b; Carstairs 1987; Carstairs & Stemberger
1988) – is the grammatical feature which is less likely to undergo syncretism. Similarly,
Vafaeian (2013) shows that number is the grammatical feature that more frequently
triggers suppletion in nominal paradigms (followed by possession, accusative/ergative
case, and vocative).
The investigation of syncretism conducted in this thesis continues the line of research
described in (1). In particular, the results discussed in this section provide further
support for the argument of Greenberg (1963b) with respect to gender and number.
Number is not likely to undergo syncretism in the context of gender, whereas the opposite
is very frequent. Thus the relationship between the two features is strongly asymmetrical.
If these asymmetries are interpreted as pointing to a relevance hierarchy, then the results
of this investigation suggest that nominal number has higher relevance to nouns than
gender.
5.5 Split plural indexation systems
As observed throughout chapter 2, manifestations of grammatical gender are bound to
indexation, whereas nominal number may (but needs not to) be manifested through
indexation. In §§5.2 and 5.3, I showed that if a language has both gender and number
indexation, the two are likely to interact via exponence, syncretism or both. In this
section, I consider those rare cases in which languages with sex-based gender exhibit a
split in the indexation patterns triggered by plural nouns (question 5 in (5.1)). I suggest
referring to these systems as split plural indexation systems. In general, split plural
indexation systems function in such a way that:
(1) Some plural nouns trigger the same indexation pattern as either feminine singular
or masculine singular nouns, or the same indexation pattern as their correspondent
singular.
(2) Some plural nouns trigger an indexation pattern that differs both from feminine
and from masculine. Given that this indexation pattern is exclusively used to
signal plurality, I refer to it as Dedicated Plural Indexation (henceforth, DPI).
118
5.5 Split plural indexation systems
Eight languages of my sample have split plural indexation systems. They all have
bipartite sex-based gender and belong to different subgroupings of the Afro-Asiatic language family or to Ta-Ne-Omotic. They are (in order of genealogical affiliation): Miya
(Chadic), Baiso (Cushitic), Borana-Arsi-Guji Oromo (Cushitic), Dasaanach (Cushitic),
Iraqw (Cushitic), Kambaata (Cushitic), Bench (Ta-Ne-Omotic) and Standard Arabic
(Semitic). In these languages, the distribution of DPI depends on the type of indexation
trigger or the type of indexing target. Restrictions on the use of DPI based on the type
of target are only found in the Cushitic languages Kambaata and Dasaanach. These
languages are discussed in §5.5.2.3. Restrictions on the use of DPI based on indexation
trigger types are found in the remaining six languages. They are constrained either by
the animacy or by the lexical plurality of nouns. Restrictions based on animacy are
discussed in §5.5.1; restrictions based on the lexical plurality of nouns are discussed in
§5.5.2. In the following sections, I will show that these phenomena, as rare as they are,
can be very relevant for understanding how gender and number indexation systems come
to interact and to be mutually constrained.
5.5.1 Animacy-based uses of Dedicated Plural Indexation
In §5.3, I surveyed patterns of gender syncretism in the context of number and analysed
the distribution of plural indexation in Standard Arabic. In Standard Arabic, DPI is
only used with plural human nouns, that is, with nouns denoting entities that are located
at the top of the Animacy Hierarchy. When the plural noun is nonhuman, the feminine
singular indexation pattern is used. Consider the following examples, already discussed
in §5.3.1.
(5.13) Indexation with human plural nouns in Standard Arabic (Semitic) (Ryding
2005: 126) (same example as (5.9))
c askariyy-uuna
(a) qaadat-un
leader\pl-nom military-m.pl
‘military leaders’
(b) al-nisaaO -u
l-mutaqaddim-aatu fii l-sinni-i
def-women-nom def-avanced-f.pl in def-age-gen
‘women of advanced age’
(5.14) Indexation with nonhuman plural nouns in Standard Arabic (Semitic) (Ryding
2005: 126) (same example as (5.10))
al-funuun-u
l-muc aaSir-atu
def-art\pl-nom def-contemporary-f.sg
‘contemporary arts’
In the nonstandard dialects of Arabic, the distribution of DPI is fuzzier than in the
standard. Belnap & Shabaneh (1994), Belnap (1999) and Corbett (2000) investigate
variation in the indexation patterns selected by plural nps in Cairene Arabic. They
119
5 Gender and number
show that, even though in Cairene Arabic DPI is more likely to be used when the np
refers to a plurality of humans, animals or individuated entities, or when the target
is distant from its trigger, the indexation patterns associated with plural nps are still
relatively unpredictable. Furthermore, based on a corpus of Old Arabic texts from the
6th and 7th century, the studies by Belnap & Shabaneh (1994) and Belnap (1999) show
that in Old Arabic the use of feminine indexation with non-human plural nps was also
less predictable (and less frequent) than in Standard Arabic. The more categorical
use of feminine singular indexation with non-human plural nps is thus described by
the authors as an innovation of Standard Arabic, not equally shared by the nonstandard
varieties (i.e., New Arabic). In Moroccan Arabic, the only nonstandard dialect of Arabic
included in my sample, DPI with plural nps is described as the most frequent option
(Harrell 1965: 158). Feminine indexation is rarely used and is described by Harrell as
an idiosyncratic pattern. In those varieties of Moroccan Arabic where feminine singular
indexation with plural nps is more widespread, it is always in free variation with DPI
(Harrell 1965: 158).
Animacy-based restrictions on the use of DPI in Miya (Chadic) have been previously
investigated by Schuh (1989), Corbett (1998), and Corbett (2000). In Miya, overt plural
marking on nouns is obligatory only with animate nouns and, in such cases, it always
triggers plural indexation. This is shown in (5.15).
(5.15) Plural indexation with animate nouns in Miya (Chadic) (Corbett 2000: 72-73,
quoted from Schuh (1989))
nýkin dzá[email protected]
this.pl man.pl
‘these men’
With inanimate nouns, number marking on nouns is optional and plural indexation
never occurs. Even when plurality is overtly coded, inanimate nouns trigger the same
indexation as their correspondent singular (i.e., either masculine or feminine). This is
shown in (5.16).
(5.16) Indexation with plural-marked inanimate nouns in Miya (Chadic) (Corbett
2000: 72-73, quoted from Schuh (1989))
ná[email protected]
vı́yaúwawáw
this.sg.m fireplace.pl (vı́yaúwawáw ‘fireplace’ is masculine)
‘these fireplaces’
Thus in Miya, the overt coding of plurality on inanimate nouns is optional, and its
syntactic manifestation via plural indexation is ungrammatical. Interestingly, the two
inanimate nouns for ‘thing,’ [email protected] and ham, and the inanimate interrogative pronoun maa,
‘what?’ always trigger plural indexation in Miya (Schuh 1989: 179). They are inanimate
but control plural indexation because they are lexically associated with plurality.
Lexical plurality is the factor that controls the distribution of DPI with plural nouns
in some languages of the Cushitic group. This is discussed in detail in the next section.
120
5.5 Split plural indexation systems
5.5.2 Dedicated Plural Indexation in Cushitic
5.5.2.1 Introduction
In this section, I examine the uses of DPI in the Cushitic languages of the sample.
Before examining in detail the indexation patterns associated with plural nps in Cushitic
languages, the main characteristics of their gender and number systems are outlined.
There are 13 Cushitic languages in my sample. They all have grammatical gender
and distinguish between masculine and feminine gender. Gender assignment is largely
arbitrary except for nouns denoting humans and animate entities: in general, nouns
denoting male entities are assigned to the masculine gender and nouns denoting female
entities to the feminine gender. As for nominal number, all Cushitic languages distinguish between singular and plural. In addition, the number system of Cushitic languages
can be classified according to two main parameters:
(1) Obligatoriness
(2) Indexation patterns associated with plural nouns.
In many Cushitic languages, nominal number marking is not obligatory (see also §4.3.1).
Nouns with general number, inherently singular nouns or nouns that are overtly marked
as singular trigger indexation according to their gender, that is, either masculine or
feminine. On the other hand, plural nouns can trigger different indexation patterns
in different languages. In some languages, DPI is used with all plural nouns, whereas
other languages have split plural indexation systems, and DPI is used with only a subset
of plural nouns. In order to investigate the uses of DPI in the Cushitic languages of
my sample, I considered only those indexing targets in which plural indexation is in
opposition with masculine and feminine indexation. In all the languages of the Cushitic
sample, this opposition is found at the very least on the verbs and/or the third person
plural pronouns.39
Finally, in Cushitic languages, as in many other languages of the world, nouns can
be plural either morphologically or lexically. Morphological plurals are nouns for which
plurality is overtly coded by means of plural morphemes. Lexical plurals are nouns that
are not marked for number but are associated with plurality by virtue of their lexical semantics (Acquaviva 2008). The following types of nouns are particularly likely to exhibit
lexically plurality: abstract nouns, nouns denoting activities involving multiple actions
or participants, collectives, body parts nouns, nouns denoting diseases, expressions of
location and place names, masses, nouns denoting internally complex objects, and expressions of time and time intervals. Lexical plurals are both animate and inanimate
and thus feature discontinuous areas of the Animacy Hierarchy.
The number of lexical plurals in the Cushitic languages of my sample varies. For
instance, according to my sources, there are at least 70 lexical plurals in Iraqw, 19 in
Borana-Arsi-Guji Oromo, 12 in Rendille and 11 in Baiso. These are listed in appendix D.
39
In many Cushitic languages, adjectives are reduplicated when they modify plural nouns (Mous 2008).
This pattern is not taken into account in my study because its distribution does not compete with
that of masculine and feminine indexation.
121
5 Gender and number
The noun lists provided in the appendix are organised according to semantic subfields.
The lists have been elaborated on the basis of previous literature on lexical plurals in
other language families of the world (see Kibrik 1992 on lexical plurals in Daghestanian
languages and Koptjevskaja-Tamm & Wälchli 2001 on lexical plurals in the languages
of the Circum-Baltic area).
In some of the Cushitic languages of my sample, lexical plurals and morphological plurals can trigger different indexation patterns. This is discussed in detail in the following
subsections.
5.5.2.2 The distribution of DPI in the Cushitic sample
The marking of nominal plurality is obligatory only in two languages of my Cushitic
sample: Awngi and Beja. I refer to these languages as Group 1 languages. These
languages, which will not be examined in detail here, have generalised uses of DPI:
all plural head nouns – together with the third person plural pronouns – trigger an
indexation pattern that differs from both masculine and feminine indexation.
In the remaining 11 languages of my Cushitic sample, the marking of nominal number
is not obligatory. Two of these eleven languages, Kambaata and Dasanaach, have rather
idiosyncratic uses of DPI and are thus discussed separately in §5.5.2.3. The remaining
nine languages differ in the distribution of the DPI in more systematic ways, which are
discussed in the rest of this section.
In Dahalo, Dirasha, Qimant, Somali and Tsamai, DPI is used with third person plural
pronouns, morphological plurals and lexical plurals. I refer to these languages as Group
2 languages. Plural indexation with lexical plurals and morphological plurals in Tsamai
is illustrated by (5.17).
(5.17) Indexation with lexical plurals and derived plurals in Tsamai (Cushitic)
(a) Lexical plurals (Savà 2005: 51)
gore
žiP-e
people eat-3pl.unm
‘The people ate’
(b) Morphological plurals (Savà 2007: 10)
èezg-aââe li-ä-e
star-pl
go.out-3pl.unm
‘The stars went out’
Languages of this type represent the dominant pattern in my Cushitic sample.
Baiso, Iraqw and Rendille have split plural indexation systems. They differ from each
other with respect to the cut-off points of these splits. In Baiso, Iraqw and Rendille, morphological plural nouns trigger the same indexation pattern as either masculine (Baiso)
or feminine nouns (Iraqw, and Rendille). I refer to these languages as Group 3 languages.
In Iraqw, DPI is reserved to the Third Person Plural Pronouns, the lexical plurals and
some morphologically plural nouns (Mous 1993). In Rendille, DPI has a rather broad
122
5.5 Split plural indexation systems
range of uses: it is found with Third Person Plural Pronouns, lexical plurals and most of
the nouns that are morphologically marked as plural. The only exception is the plural
suffix -Ce/-nye, which triggers feminine singular indexation (Oomen 1981: 60). A more
elaborated system is found in Baiso.
In Baiso, as in all other Cushitic languages with non-obligatory number, general number nouns and morphological singular nouns trigger indexation according to their gender.
The morphological plural nouns trigger the same indexation pattern as masculine nouns
(Corbett 2000; Corbett & Hayward 1987). This is illustrated in examples (5.18) and
(5.19), where the indexation patterns associated with a masculine and a feminine noun
under general, singular and plural reference are illustrated.
(5.18) Number indexation with masculine nouns in Baiso (Cushitic) (adapted from
Corbett 2000: 181)
(a) lúban
hudure
lion.m.general slept.m
‘The lion(s) slept’
(b) lubán-titi hudure
lion.m-sg slept.m
‘The lion slept’
(c) luban-jool hudure
lion-pl
slept.m
‘The lions slept’
(5.19) Number indexation with feminine nouns in Baiso (Cushitic) (adapted from
Corbett 2000: 182)
(a) kimbı́r
hudurte
bird.f.general slept.f
‘The bird(s) slept’
(b) kimbı́r-titi hudurte
bird.f-sg slept.f
‘The bird slept’
(c) kimbir-jool hudure
bird-pl
slept.m
‘The birds slept’
As shown in the two examples, whenever a noun is marked by the plural suffix -jool, it
triggers the same indexation pattern as masculine singular nouns.
Baiso is the only Cushitic language of my sample to have Paucal as a number value.40
The paucal suffix -jaa is used to specify reference to a small group of entities. As the
40
A paucal marker is also found in the Ta-Ne-Omotic language Koorete, which is spoken in the same
area where Baiso is spoken. It has been suggested that the morphological encoding of paucality in
Koorete – but not the marker as such – is the result of borrowing from Baiso (Teketal 2004).
123
5 Gender and number
Singular and the Plural, it is not an obligatory number value. It has been suggested
(Corbett & Hayward 1987: 17-18) that, historically, the paucal suffix in Baiso resulted
from the reanalysis of an originally plural suffix. Interestingly, the paucal suffix in Baiso
always triggers the same indexation pattern as the Third Person Plural Pronouns, that
is, DPI. This is illustrated in (5.20), where the same nouns as in (5.18) and (5.19) are
used.
(5.20) DPI with paucal nouns in Baiso (Cushitic) (adapted from Corbett
2000: 181-182)
(a) luban-jaa hudureene
lion-pauc slept.pl
‘A few lions slept’
(b) kimbir-jaa hudureene
bird-pauc slept.pl
‘A few birds slept’
Corbett (2000: 216-217) refers to the indexation patterns associated with paucal-marked
nouns in Baiso as an instance of semantic indexation (or “semantic agreement” in Corbett’s own terms): smaller groups are more individuated than larger and thus more
likely to trigger semantically justified indexation (that is, DPI).
DPI in Baiso is also used with a small subset of seven morphological plural nouns.
These are listed in table 5.10.
Table 5.10: Morphologically plural nouns taking plural indexation in Baiso (adapted from
Corbett & Hayward 1987: 13)
Number-unmarked form Gender Plural-marked form Meaning
baal
suul
fer
gilib
nébe
aayo
ilkoo
m
m
m
m
f
f
–
baalallo
suulallo
fererroo
gilboo
nebebboo
aayoos
ilkool
‘feather, leaf’
‘nail, claw’
‘finger, toe’
‘knee’
‘ear’
‘mother’
‘tooth/teeth’ 41
Corbett & Hayward observe that the nouns in table 5.10 denote entities that are likely
to occur in pairs or in small groups: “one talks of two knees and two ears more frequently
than of larger numbers, and similarly nails, fingers and mothers are found in smaller
rather than larger groups” (1987: 13). Accordingly, I propose to classify the indexation
patterns associated with these nouns also as an instance of semantic indexation or, to
use the terminology introduced by Dahl (2000a), referential indexation.
Finally, in Baiso DPI is used with the lexical plural nouns listed in table 5.11.
41
The noun for ‘tooth/teeth’ is the only one that triggers plural indexation both when number-unmarked
and when marked as plural. See also table 5.11 and Corbett & Hayward (1987: 13).
124
5.5 Split plural indexation systems
Table 5.11: Lexical plurals in Baiso (Corbett & Hayward 1987: 9)
Semantic groupings
Nouns
Animate collectives
saé ‘cattle’
Body parts
Objects coming in pairs
ilkoo ‘tooth, teeth’; kalaljaa ‘kidneys’; luk
ffk
ff aa ‘foot, feet, leg(s)’;
ilffoo ‘eye(s)’; ogorroo ‘hair’; moo ‘hips, lumber region’
keferoo ‘sandals’
Mass nouns
eenoo ‘milk’; soo ‘meat’; udú ‘faeces’
As mentioned before, similar to Baiso, Iraqw and Rendille, Borana-Arsi-Guji Oromo
also has split plural indexation. In this language, however, DPI is only found with Third
Person Plural Pronouns and the lexical plurals. Morphological plurals always trigger
the same indexation as feminine singular nouns. I refer to this language as Group 4.
Examples (5.21) and (5.22) illustrate indexation with morphological and lexical plurals
in Borana-Arsi-Guji Oromo.
(5.21) Indexation with morphological plurals in Borana-Arsi-Guji Oromo (Cushitic)
(adapted from Stroomer 1995: 51)
shift-aa-ni
taani k’awwee d’aan-te
bandit-pl-sbj those guns
shoot-3f.pst
‘Those bandits shot guns’
(5.22) Indexation with lexical plurals in Borana-Arsi-Guji Oromo (Cushitic) (adapted
from Stroomer 1995: 51)
bisaani ibidda d’aamf-uu
didani
water fire
estomguish-nom refuse-3pl.prs
‘The water refuses to extinguish the fire’
It is worth mentioning that, even though they have the same endings, the two nouns –
shiftaani and bisaani – are analysed in different ways in my source. Shiftaani is described
as a morphological plural, marked by the plural suffix -aa and the subject marker -ni,
whereas bisaani is described as a lexical plural noun, morphologically nonsegmentable.
To summarise, the Cushitic languages of the sample differ in their use of DPI. Most
languages have generalised uses of DPI. In other languages, the use of DPI is constrained
by the type of linguistic entity that triggers indexation. Figure 5.4 represents the type
of plural indexation in each of the languages analysed in this section. The figure is based
on the group subdivisions outlined throughout the section:
125
5 Gender and number
Group 1: Awngi, Beja
Group 2: Dahalo, Dirasha, Qimant, Somali and Tsamai
Group 3: Baiso, Iraqw, Rendille
Group 4: Borana-Arsi-Guji Oromo
Languages of groups 1 and 2, which differ from each other in the obligatoriness of
number marking but share the same indexation patterns for plural nouns, are represented
together in the figure. As shown by the figure, in the Cushitic languages with split plural
indexation, the distribution of DPI is constrained by the lexical plurality of nouns.
Figure 5.4: Triggers of DPI in the Cushitic sample
The geographic distribution of the four groups represented in figure 5.4 is shown in
the map in figure 5.5.
It is noteworthy that Beja, one of the two languages with obligatory number, is spoken at the very edge of the Cushitic-speaking area, as represented in my sample. The
peculiar status of Beja within the Cushitic language family is briefly discussed by Mous
(2012: 345), who describes the language as “quite deviant compared to the rest of the
Cushitic languages.” Awngi, on the other hand, together with the other languages of the
Agaw subgroup, is still understudied and little is found in the literature with respect to
its status within the language family.
126
5.5 Split plural indexation systems
Figure 5.5: DPI in the Cushitic sample
127
5 Gender and number
5.5.2.3 Two peculiar languages within the Cushitic sample: Kambaata and
Daasanach
The use of DPI in Kambaata and Daasanach is very marginal.
In Kambaata, the (Independent) and (Dependent) Personal Pronouns as well as the
Anaphoric Demonstratives are the only indexing targets that inflect for number. Otherwise, plural indexation triggers in Kambaata (lexical plurals, morphological plural nouns
and the Third Person Plural Pronouns themselves) always trigger the same indexation
patterns as feminine singular nouns on all targets. Verbal indexation with plural nouns
is shown in (5.23).
(5.23) Feminine indexation with plural nouns in Kambaata (Cushitic) (Treis 2014: 131)
alı́
wud-ı́
yabúr-r-assa
aaqq-án-t-iyan
upper.m.gen side-m.gen lip-pl-f.nom-3.pl.poss take-pass-3.f.pco-ds
hor-i-ssá
inq-áak-ant
hı́nn y-itoo’u
all-m.gen-3.pl.poss teeth-pl-f.nom<n> smirk say-3.f.pfv
‘the upper lips were grabbed (lit. ‘taken’) and the teeth of all of them smirked’
The example also shows two occurrences of the Dependent Third Person Plural Pronoun
-ssa. In both occurrences – yabúrrassa and horissá – the suffix is interpreted as a
possessive pronoun. In Kambaata, when the Dependent Personal Pronouns are attached
to nouns, they always function as possessives (Treis 2008: p.348).
Personal Pronouns (Dependent and Independent) as well as the Anaphoric Demonstratives are the only targets of plural indexation in Kambaata. When the antecedent is a
lexical plural or a morphological plural, the plural form of these pronouns is selected. The
Independent Personal Pronouns are only used when the antecedent is human whereas
the Anaphoric Demonstratives are used with all other types of antecedent (Treis 2014).
The use of DPI in Dasaanach is even more marginal than it is in Kambaata. There
are no grammaticalized Third Person Pronouns, and plural nouns in Dasaanach always
trigger the same indexation as masculine singular nouns (Tosco 2001). The generic nouns
mı́/mú ‘man,’ máa ‘man, person,’ minni ‘woman’ and gáal ‘people’ are used as sort of
anaphoric words with human antecedents (Tosco 2001: 212). Interestingly, it is only with
these anaphoric words that an incipient pattern of plural indexation is found, since gáal
‘people’ will be used when the human antecedent is plural. It is not clear what type of
anaphoric words are used with nonhuman antecedents. Tosco (2001: 227) only mentions
that the generic noun Pée ‘thing’ can be used when an np lacks an overt nominal head.
5.5.3 Discussion
In this section, I considered the issue of whether or not gender and number compete
through indexation (see question 5 in (5.1)) and focussed on a small subset of the languages of the sample with split plural indexation. In these languages, the use of DPI
is constrained by the lexical semantic properties of the indexation triggers and is in
128
5.5 Split plural indexation systems
complementary distribution with masculine and feminine indexation (these languages
all have bipartite sex-based gender).
In §5.5.1, I considered the case of Standard Arabic and Miya, in which the use of DPI
is constrained by the animacy of the np referent. These types of semantic restrictions
on the use of DPI go hand in hand with the well-documented role that the Animacy
Hierarchy plays on the overt coding of nominal plurality on nouns across the languages
of the world (see Corbett 2000; Haspelmath 2013; Smith-Stark 1974, as well as §2.2).
In §5.5.2, I investigated the uses of DPI in Cushitic. In Cushitic languages with
split plural indexation systems, the Animacy Hierarchy does not seem to play a major
role in predicting the types of indexation triggers that are more likely to select DPI.
In these languages (with the exception of Kambaata and Dasaanach), at the very least
the third person plural pronouns trigger DPI. In addition to these pronouns, the most
common triggers of DPI are the lexical plurals, that is, nouns that are morphologically
unmarked for number but inherently associated with plurality in virtue of their lexical
semantics. In languages with less restricted or non-restricted uses of DPI, the next
available indexation triggers after the lexical plurals are nouns that are morphologically
marked as nonsingular. If read from the bottom to the top, Figure 5.4 suggests that the
number of triggers of DPI grows somewhat incrementally across the Cushitic languages
of the sample. I have found no languages in which morphological plurals trigger DPI,
but lexical plurals do not do so. Similarly, I found no language in which morphological
singular nouns trigger DPI.
The Cushitic data analysed in this section suggest that the Animacy Hierarchy may not
be the only tool to explain the distribution of optional and obligatory number marking
on the indexing targets (see also Cristofaro 2012 on the limitations of the Animacy
Hierarchy in understanding the grammaticalization of nominal number). More difficult
is it to establish the diachronic significance of these data. The situation attested in some
of the Cushitic languages of the sample could be interpreted as the result of the expansion
of the triggers of DPI from the personal pronouns to the lexical plurals, and from there
to the morphological plurals. On the other hand, the opposite path, i.e., a reduction
of the domain of use of DPI, is also a plausible diachronic scenario. A more thorough
historical-comparative investigation of the Cushitic languages would be needed in order
to understand the diachronic development of this type of system. This lies outside the
scope of the present investigation.
Finally, an interesting interplay between gender and number indexation is found in
Bench (Ta-Ne-Omotic), where the distribution of DPI closely resembles the situation
observed in some of the Cushitic languages of the sample. In Bench, plural-marked
nouns normally trigger masculine indexation. DPI can be found only with “indicative
final, polar question, imperative, and relative verbs” as well as with demonstratives
(see Rapold 2006: 540), but it is always optional. Rapold (2006: 540-541) describes
DPI in Bench as an instance of semantic indexation that competes with morphological
indexation (masculine and feminine). The distribution of DPI in Bench is described as
both semantically and pragmatically constrained, but not much is said in my source
about the exact nature of these constraints.
129
5 Gender and number
5.6 Presence of gender and type of number marking
In this section, I discuss if and how the presence of gender correlates with any aspect
of the nominal number marking system that is found in a language (see last research
question in (5.1)). As explained at the beginning of the chapter, this research question
was elaborated in order to verify whether two generalisations on morphosyntactic properties of African languages formulated by Creissels et al. (2008) can be applied to the
languages of my sample. The two generalisations are as follows:
(a) Languages devoid of a gender system frequently have a single plural
marker with the morphological status of a phrasal affix, and such plural
markers tend to be used on a “pragmatic basis,” i.e., to be employed
only when plurality is both communicatively relevant and not implied
by the context, at least in the case of nouns that do not refer to persons
[...].
(b) Languages that have gender generally have a morphologically complex
plural marking, characterized by a fusion of gender and number markers, and variation in gender and number manifest themselves through
morphemes affixed to the head noun and to (some of) its modifiers,
in an agreement relationship. In these languages [...] plural markers
tend to be present in every np referring to a plurality of individuals [...]
(Creissels et al. 2008: 119).
The claims made in (b) describe a very common pattern in the languages of the sample
that have obligatory number. In fact, the properties enumerated in (b) – cumulation
(fusion) of gender and number markers, and marking of gender and number on nouns
and through indexation (agreement) – have been at the bulk of this chapter.
The language type described in (a) is also attested in the languages of my sample
but has not been considered in detail so far. Of the 16 genderless languages of my
sample, 10 have strategies of nominal number marking that very closely reflect the
generalisation expressed in (a). These languages are: Yoruba (Defoid), Igbo (Igboid),
Bambara (Mande), Susu (Mande), Mann (Mande), Mandinka (Mande), Hdi (Chadic),
Mina (Chadic), Akan (Kwa) and Ewe (Kwa). Let us consider Yoruba as an illustration
of this type of system. The description of the number system of Yoruba is based on
Ajı́bóyè (2010) and Bamgbose (1966).
In Yoruba, there is no grammatical gender, the marking of nominal plurality is optional with all nouns. Number-unmarked nouns can be interpreted either as singular
or plural. When speakers want to specify that the referent of a np is plural, several
strategies are available: the plural word àwo.n (formally identical to the Third Person
Plural Pronoun), the plural morpheme wò.n (which can only be used to pluralize demonstratives), quantifiers such as púpò., ‘many,’ numerals, and, finally, when the noun occurs
together with an adjectival modifier, adjectival reduplication. Thus, from a morphosyntactic point of view, the only indexes of the number values of the np referent are the
Third Person Pronouns and the Demonstratives. When plurality is overtly coded on
130
5.6 Presence of gender and type of number marking
the demonstratives, the whole np is obligatorily interpreted as having a plural referent.
Plural marking with the plural word àwo.n is illustrated in (5.24).
(5.24) Plural marking in Yoruba (Defoid): the plural word àwo.n (Ajı́bóyè 2010: 143)
Mo kı́
àwo.n o.kúnrin tı́
ó
wà
nı́bè.
1sg greet pl
man
that rpbe there
‘I greeted the men that were there’
One property that seems to be shared by each of the ten languages mentioned above
is the lack of pervasive systems of number indexation. Table 5.12 illustrates how many
number-indexing targets are found in these languages.
Table 5.12: Number-indexing targets in Creissels et al.’s (2008) type (a) languages
Number of
number indexes
Number of Genealogical groups
languages
1
6
Kwa (1/3); Igboid (1/1) Mande
(4/4)
2
4
Kwa (1/3); Chadic (2/8); Defoid
(1/1)
The figures in table 5.12 show that all ten languages classifiable as type (a) in the
classification proposed by Creissels et al. (2008) have very minimal number indexation
with only one or a maximum of two number indexes. In addition, if a language has one
number index only, this is the third person pronouns.
The remaining six genderless languages in the sample are the Western Nilotic languages Acoli, Belanda Bor, Dinka, Luwo, Mabaan and Nuer. They all lack grammatical
gender but are characterised by a rather elaborated system of number marking on nouns
that also serves a classificatory function. This type of number system is known in the
literature as the tripartite number system 42 because three types of nouns exist in these
languages (see also the discussion in §4.3.2):
(1) Nouns where only singular reference is overtly coded
(2) Nouns where only plural reference is overtly coded
(3) Nouns where both singular and plural reference are overtly coded (also known as
replacement pattern).
42
Tripartite number systems with classificatory functions are also found in the three Eastern Nilotic
languages of the sample – Turkana, Masai and Karamojong. In these languages, the tripartite
number systems co-exist with cumulative encoding of gender and number. This was discussed in
§5.2.1.4.2.
131
5 Gender and number
Since all nouns, both those that are overtly coded for number and those that are not,
are associated with a number value, the tripartite number system of Western Nilotic
is an obligatory number system. The three patterns described above are exemplified
in examples (5.25), (5.26) and (5.27). These examples have already been discussed in
§4.3.2, where a general description of tripartite number systems was outlined.
(5.25) Singular marking in Mabaan (Western Nilotic), same example as (4.8) (Storch
2005: 123)
(a) ñÓÒk
‘lice’
(b) ñÓÓk-cn
”
lice-sg
‘louse’
(5.26) Plural marking in Mabaan (Western Nilotic), same example as (4.9) (Storch
2005: 123)
(a) jwÓm
‘monkey’
(b) jwÓm-g2
monkey-pl
‘monkeys’
(5.27) Replacement pattern in Mabaan (Western Nilotic), same example as (4.10)
(Storch 2005: 123)
(a) Pú2tworm-sg
‘worm’
(b) Pú2t-n
”
worm-pl
‘worms’
Dimmendaal defines the tripartite number systems spread throughout Nilotic and,
more broadly, Nilo-Saharan languages as “a classificatory technique” (2000: 214), whereby
nouns with analogous morphophonological or semantic properties result in similar number morphology. The same interpretation is proposed by Storch (2005, 2007) in her
investigation of the structure, properties and diachrony of nps in Western Nilotic. She
defines the morphological strategies that Western Nilotic languages use to categorise
nouns in semantically homogeneous groups as nominal classifiers following Aikhenvald’s
(2003) sense of the term. Beside signalling number distinctions, these markers express
the following meanings: biological gender, animacy, tactile perception, shape and size.
Even though the tripartite number system is preserved in all Western Nilotic languages,
the classifier-like morphemes through which it functions are not equally spread among
the languages of the group. Storch (2005, 2007) suggests that the systems attested in
all contemporary Western Nilotic languages are a reduced version of what the original
protosystem must have looked like. This reduction is explained by Storch as the result
of a general propensity towards the simplification of the number marking system, and
towards the generalisation of a few singular and plural markers to all nouns, independently of their lexical semantics. These generalised number markers are called by Storch
132
5.6 Presence of gender and type of number marking
imperialistic number markers (2005; 2007). The inventory of the singular and plural
markers in three of the Western Nilotic languages included in the sample – Mabaan,
Dinka and Luwo – is given in appendix E.
Interestingly, similarly to the ten languages examined above, in the Western Nilotic
languages of the sample number indexation is also minimal or null (Storch 2005).
To summarise, the languages of my sample generally reflect the generalisations proposed by Creissels et al. (2008) with respect to gender and number in African languages.
However, what my data suggest is that the grammatical phenomenon that tends to be
affected the most by the co-presence of gender and number is indexation. Languages
in which gender and number are both attested tend to have very pervasive indexation
systems. On the other hand, if only number is present, indexation patterns tend to be
very poor, and number distinctions are mostly marked at the phrasal level (i.e., only one
marker per np is found). This pattern is illustrated in figure 5.6 (see also appendix C
for a more detailed overview).
Figure 5.6: Gender- and number-indexing targets in the languages of the sample
Two relevant patterns can be detected in figure 5.6: (1) the majority of the languages of the sample have pervasive indexation systems both for gender and number;
(2) the majority of the languages of the sample have the same number of gender- and
number-indexing targets. However, with respect to (2), it is worth mentioning that
same number of gender- and number-indexing targets does not always mean same type
of indexing targets or same type of marking strategy. For instance, in Cushitic languages
number distinctions on adjectives tend to be signalled by means of reduplication patterns, whereas gender distinctions tend to be signalled only by means of affixation. A
systematic of overview of such differences in the distribution of type of indexing targets
and types of marking strategies per domain constitutes a promising area of research;
however, this falls outside the scope of this investigation.
133
5 Gender and number
To conclude, in light of the data examined in this section and in the rest of the chapter,
I suggest reformulating the generalisations by Creissels et al. (2008) in the following way:
(a1 ) At least in African languages, the development of rich and pervasive nominal indexation systems seems always to involve both gender and number. These systems
exhibit a strong tendency for cumulative exponence and related interactions (i.e.,
syncretism).
(b1 ) When gender marking is absent, number marking tends to remain optional and
follow the principle “one marker per np.”
5.7 Gender shifts and semantic interactions between gender
and number
In §2.1.1, I discussed the phenomenon whereby in certain languages, nouns can be assigned to more than one gender depending on the way the np referent is construed, and
I briefly examined previous studies on the topic. I proposed to use the label manipulable gender assignment to refer to the possibility of assigning nouns to multiple genders
and the label rigid gender assignment to refer to those languages in which nouns are
invariably assigned to one gender. As also observed in §2.1.1, the process whereby nouns
can be assigned to genders others than their default is referred to in this dissertation as
gender shift.
In this section, I discuss a set of data from my language sample that suggests that,
in some languages, gender assignment can be manipulated in order to indicate variation
in the countability properties of nouns. The distribution of this phenomenon in the
languages of the sample is shown in table 5.13.
Table 5.13 shows that gender shifts for the encoding of variation in the countability
properties of nouns are attested in 41.6% of the gendered languages in the sample. The
distribution of the phenomenon is, however, highly skewed from a genealogical point
of view. Gender shifts for the encoding of variation in the countability properties of
nouns are attested only among the following genealogical groups: Bantu, Berber, Eastern
Nilotic, North-Central Atlantic and Semitic. The majority of the languages displaying
this phenomenon are Bantu, followed by Berber, North-Central Atlantic, Semitic and
Eastern Nilotic, where the phenomenon is attested in only one language, Turkana.
The rest of this section is organised as follows. In §5.7.1, I discuss data from the Bantu
and North-Central Atlantic languages because the two groups have a rather similar
gender system and are also genealogically related. In §5.7.2, I consider the occurrence
of this phenomenon in Berber, Semitic and in the Eastern Nilotic language Turkana.
Similarly to Bantu and North-Central Atlantic languages, these languages also share the
same type of gender system (i.e., sex-based gender).
134
5.7 Gender shifts and semantic interactions between gender and number
Table 5.13: Distribution of the use of gender shifts for the encoding of variation in the countability
properties of nouns
Use
No. of
lngs.
Rel.%
Abs.% Genealogical groups
Present
35
41.6%
35% Bantu (21/23)
Berber (6/6)
Eastern Nilotic (1/3)
North-Central Atlantic
(4/7)
Semitic (3/7)
Absent
45
53.6%
45% Bantu (2/23)
Cushitic (13/13)
Chadic (6/8)
Dizoid (1/1)
Eastern Nilotic (2/3)
Hadza (1/1)
Khoe-Kwadi (5/5)
Kwa (1/3)
Kxa (1/1)
Mel (1/3)
North-Central Atlantic
(2/7)
Sandawe (1/1)
Semitic (4/7)
South Omotic (1/1)
Ta-Ne-Omotic (4/4)
No data
4
4.8%
4% North-Central Atlantic
(1/7)
Mel (2/3)
Tuu (1/1)
16
–
100
100%
No gender
Total
16% Chadic (2/8)
Defoid (1/1)
Igboid (1/1)
Kwa (2/3)
Mande (4/4)
Western Nilotic (6/6)
100%
135
5 Gender and number
5.7.1 Gender shifts, number/countability in Bantu and North-Central
Atlantic
The two obligatory number values in most North-Central Atlantic and Bantu languages
are singular and plural. The individual noun classes of these languages are associated
either with singular or with plural reference. As observed in 4.2.3, in the most regular
cases, the singular classes have systematic correspondents among the plural classes.
Examples (5.28) and (5.29) illustrate gender and number marking in Swahili (Bantu)
and Bandial (North-Central Atlantic).
(5.28) Swahili (Bantu)(Thompson & Schleicher 2001: 39)
(a) m-tu
cl1-person
‘person’
(b) wa-tu
cl2-person
‘people’
(5.29) Bandial (North-Central Atlantic) (Sagna 2008: 275)
(a) a-rokka
cl1-worker
‘worker’
(b) u-rokka
cl2-worker
‘workers’
Nouns with low countability properties in Bantu and North-Central Atlantic languages
are always assigned to a noun class and often do not undergo number distinctions.
Moreover, since noun class values are always cumulated with number values, uncountable
nouns are either lexically singular or lexically plural. In Swahili, nouns denoting abstract
notions are assigned to Class 11 and lack a plural counterpart. An example of a Class
11 abstract noun is given in (5.30).
(5.30) Swahili (Bantu)(Contini-Morava 2000: 6)
u-singizi
cl11-sleep
‘sleep, sleepiness’
In Bandial, nouns denoting liquids and masses are typically assigned to the plural class
mu-, Class 10, and do not have a singular counterpart. An example of a Class 10 mass
noun is given in (5.31).
136
5.7 Gender shifts and semantic interactions between gender and number
(5.31) Bandial (North-Central Atlantic) (Sagna 2008: 272)
mu-u
cl10-salted.water.from.the.river
‘salted water from the river’
If the gender systems of the Bantu and North-Central Atlantic languages are interpreted
as rigidly grounded in the notion of gender as singular-plural pairing, lexical singular
and lexical plural nouns should be represented as displaying defective paradigms. As
shown later in this section, other irregularities can be observed in these systems.
Tables 5.14 and 5.15 illustrate the regular singular-plural pairings of Swahili and Bandial. Nonpairing noun classes – as the locative classes43 – are excluded from the table.
Table 5.14: Noun class and singular-plural pairs of Swahili (adapted from Contini-Morava 2000)
Class Singular Class Plural
1
3
5
7
9
11
11
mmjikinuu-
2
4
6
8
10
6
10
wamimavinman-
Table 5.15: Noun class and singular-plural pairs of Bandial (Sagna 2012: 136)
Class Singular Class Plural
1
3
5
7
9
11
12
aebufugajuñu-
2
4
6
8
10
6
6
bugsuugumuuu-
In addition to the usages discussed so far, in the Bantu and North-Central Atlantic languages of my sample, noun class marking can be manipulated to encode non-obligatory
43
Many Bantu and North-Central Atlantic languages have noun classes that are used to encode location
(locative classes are also found in the languages of the Mel group). The locative noun classes have
rather peculiar properties as compared to others. For instance, they have no basic members if not
the word for ‘place’ and tend to trigger noncanonical patterns of subject marking, also known in the
literature as locative inversions (for an overview of this matter, see Maho 1999). In a sense, they
can be seen as rather nonprototypical members of the gender systems of the North-Central Atlantic,
Mel and Bantu languages (see, for example, Marten 2010: 252, on locative classes in Bantu). The
locative classes are generally insensitive to number distinctions.
137
5 Gender and number
number values such as collectives, distributive plurals or singulatives. In Bandial, for
instance, several class markers can be used to form collectives from nouns that otherwise have regular singular-plural noun class pairings. One of these markers is ba-, a
singular class that is defined as a subclass of Class 5 and is described as semantically
associated with assemblages (Sagna 2008, 2011). When nouns with other regular class
pairs are assigned to Class 5, the resulting meaning is what Sagna (2011) defines as
diminutive collective, that is, a noun that denotes “a group of small entities as a unit”
(Sagna 2011: 243). This is shown in example (5.32).
(5.32) Bandial (North-Central Atlantic) (Sagna 2011: 243)
(a) e-vval
cl3-stone
‘stone’
(b) si-vval
cl4-stone
‘stones’
(c) ba-vval
cl5-stone
‘pile of stones’
Class 4 is the regular plural of the noun for ‘stone’ (5.32b). When assigned to Class 5
(5.32c), the noun denotes a collection of stones.
Similarly, in Swahili the noun class prefix ma-, Class 6, has several usages: (1) it is
the regular plural correspondent of Class 5; (2) it is the default noun class for lexical
plurals denoting masses and liquids; and (3) it is used to form the collective of nouns
that form their regular plural in other classes. The latter use is illustrated in example
(5.33).
(5.33) Swahili (Bantu) (Shadeberg 2003: 84)
(a) simba
cl9.lion
‘lion’
(b) ma-simba
cl6-lion
‘a pride of lions’
In Swahili, nouns assigned to Class 9 when singular are usually assigned to Class 10
when plural. When assigned to Class 6, as in (5.33b), they are interpreted as collectives.
The phenomena illustrated in examples (5.32) and (5.33) show that in some Bantu
and North-Central Atlantic languages class assignment can be manipulated in order to
modify the countability properties of nouns and encode non-obligatory number values.
These phenomena thus challenge the idea that the gender system of these languages can
138
5.7 Gender shifts and semantic interactions between gender and number
be understood solely on the basis of the singular-plural pairings. In the North-Central
Atlantic and Bantu languages where these phenomena are attested, the possibility of manipulating the countability properties of nouns is a fundamental property of the gender
system (see also Acquaviva 2008: 41-42).
5.7.2 Gender shifts, number/countability in Berber, Semitic and Eastern
Nilotic
The feminine marker of the Berber languages can be used to encode variation in the
countability properties of nouns. In particular, uncountable nouns are marked by the
feminine marker (and trigger feminine indexation) when a singulative interpretation is
intended for the noun. This is illustrated in example (5.34), from the Berber language
Nefusi.
(5.34) Nefusi (Berber) (Beguinot 1942)
(a) ettefâh
" (collective)
‘apples’
(b) t-attefâh-t
"
f-apples-f
‘one apple’
The circumfixing of the feminine marker to the collective noun ettefâh results in the
"
formation of a noun that is semantically singular and grammatically triggers
feminine
indexation. This noun is countable, and, as such, it can undergo regular plural marking,
as illustrated in example (5.35).
(5.35) Nefusi (Berber) (Beguinot 1942)
t-attefâh-în
"
f-apples-f.pl
‘apples’ (plural)
The singulative function of the feminine marker t is attested in all the Berber varieties of my language sample. As shall be illustrated in §6.4, in the Berber languages
of my sample, gender assignment can also be manipulated to encode diminutive and
augmentative meanings. Inherently masculine nouns are shifted to the Feminine Gender
when a diminutive interpretation is intended for the noun (see example (6.22)). Similarly, inherently feminine nouns are shifted to the Masculine Gender when an augmentative interpretation is intended for the noun (see example (6.23)). Specialists of Berber
languages point out that the interpretation of the feminine marker as either feminine
proper, singulative or diminutive depends on the interplay between the lexical semantics
of the noun, the discourse context and the speaker’s attitude towards the np referent
(see, among others, Penchoen 1973b and Mettouchi 2000 with respect to Tachawit and
139
5 Gender and number
Kabyle, respectively). Mettouchi (2000) also suggests that the diminutive and singulative function of the t marker in the Berber languages was diachronically prior to the
feminine proper function. As pointed out already in §4.2.2, Mettouchi’s idea is that the
original function of the marker was to identify an entity with respect to a reference point,
the relationship between the two (the entity and the reference point) being inherently
hierarchical and roughly of the part-whole type. According to Mettouchi, only at a later
stage, and starting with animate entities, was this polarity reanalysed as a “masculine
vs. feminine” type of opposition. The diminutive and augmentative uses of the gender
markers in the Berber languages are discussed in detail in §6.4.1 (on the mutually exclusive relationship between the different meanings of gender shifts – biological gender,
size and countability –, see also Kossman 2014).
Similarly to the Berber languages, in Classical and Standard Arabic, Moroccan Arabic
and Maltese, the feminine marker functions as a singulative when used with uncountable
nouns, both masses and collectives. Example (5.36) illustrates how in Classical Arabic,
the interpretation of the suffix -AT varies based on the countability properties of the
base noun.
(5.36) The suffix -AT and the countability properties of nouns in Classical Arabic
(Semitic) (adapted from Hämeen-Anttila 2000: 602)
(a) kalb- ‘dog (male)’
kalb- + AT ‘bitch, female dog’
(b) h.amām- ‘pigeons’
h.amām- + AT ‘one pigeon’
In (5.36a), the base noun is animate and countable. It is interpreted as denoting both
female and male dogs, and syntactically behaves as masculine, in the sense that it triggers
masculine indexation. When this base noun is marked by -AT, the resulting meaning
can only be feminine and the noun triggers feminine indexation. On the other hand, in
(5.36b), the base noun is uncountable and syntactically behaves as a masculine noun.
When this base noun is marked by -AT, the result is a singulative, countable noun that
triggers feminine indexation and can be regularly pluralized. Interestingly, masculine and
feminine gender with the noun in (5.36b) can be encoded only periphrastically, that is, by
using the modifiers for ‘male’ and ‘female’ together with the noun for ‘pigeons/pigeon.’
This is shown in example (5.37).
(5.37) Biological gender with collectives and singulatives in Classical Arabic (Semitic)
(adapted from Hämeen-Anttila 2000: 602)
(a) dhakar- al-h.amām- ‘the male of pigeons’
(b) al-h.amā+at adh-dhakar- ‘one male pigeon’
In this thesis, I follow the tradition of most descriptive grammars of Arabic dialects
in which the collective, uncountable form is considered the basic form as opposed to
the singulative-marked form. Elie Wardini (personal communication) suggests that at
least in the spoken varieties of Arabic, the singulative-marked forms may actually be
considered the basic forms. Following this suggestion, the noun h.amāmat ‘pigeon’ should
140
5.7 Gender shifts and semantic interactions between gender and number
be interpreted as an inherently feminine countable noun analogous to the nonsingulative
noun qittat ‘cat.’ The uncountable noun h.amām- ‘pigeons’ should rather be interpreted
˙˙
as the result
of some sort of subtractive morphological rule, whereby presence vs. absence
of overt gender marking is a means for manipulating the countability properties of nouns.
Wardini’s hypothesis could be verified on the basis of behavioural data from native
speakers of different Arabic dialects. This research line lies, however, beyond the primary
scope of the present investigation.
What is more important to underline here is that, as in the Berber languages, the
above-mentioned Semitic languages productively use manipulation of gender assignment
as a means for encoding variation in the countability properties of nouns. Uncountable
nouns – both masses and collectives – tend to behave as masculine nouns syntactically.
When marked by the feminine marker, these nouns behave syntactically as feminine and
are semantically interpreted as countable and singular. While the semantic denotation of
the noun stays unchanged, the construal of its countability properties changes radically
when the indexation class changes.
Similarly to Mettouchi’s (2000) proposal on the diachrony of gender in the Berber
languages, it has been proposed that the gender system of the Arabic dialects was also
originally non-sex-based. If this is the case, it is reasonable to hypothesise that the
feminine marker originated from “a more complex system of classes within which the
category of number has to be included as well [...]” (Moscati 1964: 86).
The feminine marker is also used to encode diminution in Moroccan Arabic and Maltese, similarly to the Berber languages, but not in Standard Arabic (for a more detailed
description of evaluative morphology in the Semitic languages and its relationship with
gender, see chapter 6).
In Turkana (Eastern Nilotic), feminine and masculine nouns can be shifted to the
Neuter Gender to encode variation in the countability properties of nouns or in the
size/age of the np referent. The interpretation of shifts to the Neuter Gender largely
depends on the countability properties of the base noun. If the base noun is countable,
the meaning resulting from the gender shift is diminutive; if the base noun is uncountable,
shifts to the Neuter Gender are used to encode singulative meanings. Examples are
discussed in §6.4.1. See also §7.7 for a discussion of the grammatical complexity of
gender in Turkana.
5.7.3 Gender shifts and semantic interactions between gender and number:
summary
In this section, I examined the use of gender shifts as a strategy for encoding variation in
the countability properties of nouns in languages with manipulable gender assignment.
This was counted as an instance of semantic interaction between gender and nominal
number. The phenomenon is attested in 34 of the 84 gendered languages is the sample
but its distribution is highly genealogically skewed. The role that this type of interaction
plays in the absolute complexity of gender systems is discussed in detail in chapter 7.
141
5 Gender and number
5.8 Summary of the chapter
The purpose of this chapter has been to provide a systematic overview of patterns of
interaction between gender and number in the languages of the sample along the following dimensions: exponence, syncretism, competing indexation patterns, correlations
in type of marking, and semantic interactions through gender assignment. The research
questions addressed in the chapter are listed in (5.1).
Both cumulative exponence of gender and number were found to be very frequent in
the languages of the sample. In addition, the occurrence of gender syncretism in the
context of number was found to correlate with the presence of cumulative paradigms for
the encoding of gender and number.
The presence of cumulative exponence of gender and number has been interpreted as
an increase in the absolute complexity of gender and number systems. Furthermore, the
results of the investigation, in line with what has been claimed in previous literature,
have been interpreted as suggesting that, in a hypothetical relevance hierarchy for nouns,
number would need to be ranked higher than gender.
The data from the language sample seem to disprove the claim by Baerman et al.
(2005) according to which syncretism leads to the semanticization of gender distinctions
across number values (particularly nonsingular number values). In the majority of the
cases attested in the sample, syncretism of gender in the context of number does not
yield any semantically transparent reorganisation of gender values. In the few cases in
which semanticization does occur, it is always animacy-driven.
The question of how gender and number interact through patterns of indexation was
tackled by considering those (rare) cases in which a dedicated pattern of plural indexation (DPI) exists, but its uses are restricted by the semantic properties of the indexation
triggers. Two types of semantic restrictions were found: (1) restrictions based on the
animacy of the indexation triggers, whereby only human or animate nouns trigger DPI,
and (2) restrictions based on the inherent plurality of the indexation triggers, whereby
lexical plurals are one of the few available triggers of DPI besides the third person plural
pronouns. These results suggest that the Animacy Hierarchy might not be the only
valuable predictor of the use of number indexation in a language. They also suggest
that there might be an opportunity for revising existing models pertaining to the grammaticalization of nominal number. One could perhaps think of a split whereby animacy
is the best predictor for the use of obligatory number marking on nouns (as suggested
by Corbett 2000, Haspelmath 2013, and Smith-Stark 1974), whereas both animacy and
lexical plurality play a role in constraining the distribution of number indexation. Further research is needed in order to verify the crosslinguistic validity of this hypothesis.
A commonly acknowledged major functional divide between gender and number is that
gender is a lexical feature, whereas number is referential. The Cushitic data discussed
in this chapter suggest that competing indexation patterns between gender and number
tend to be associated with nouns for which this functional divide is not as clear-cut, that
is, with the lexical plurals. The crosslinguistic relevance of these phenomena should also
be further investigated.
In §5.6, I investigated possible correlations between presence/absence of gender and
142
5.8 Summary of the chapter
type of number marking by using the generalizations of Creissels et al. (2008) as a reference. I showed that, as a rule, the languages of my sample reflect these generalisations.
In addition, I showed that, a relevant implication of the co-presence of gender and number is the existence of very pervasive indexation systems. In languages without gender,
number indexation is null or, at best, very poor. In addition, the languages of the sample
tend to exhibit the same number of gender and number indexing targets.
Finally, in §5.7, I discussed a particular type of semantic interaction between gender
and number, that is, the use of gender shifts to encode variation in the countability
properties of nouns. I showed that in languages where nouns can be assigned to multiple
genders depending on the construal of the np referent, gender shifts can be used to mark
uncountable nouns as countable and singular.
143
6 Gender and evaluative morphology
6.1 Introduction
In this chapter, I discuss the interactions between gender and the morphological encoding of evaluation in the languages of my sample. Within the grammatical domain of
evaluation, only diminutive and augmentative markers are considered. The following
research questions are addressed in the chapter:
(6.1) Research questions concerning gender and evaluative morphology (same as (2.14))
Q 1: How frequently does size occur as an independent gender value? How stable
and how widely distributed is this phenomenon within genenealogical units?
Q 2: Do the interactions between gender and evaluative morphology differ across
types of gender systems and/or strategies of gender assignment (e.g.,
sex-based vs. non-sex-based gender systems, or rigid vs. manipulable gender
assignment)?
Two kinds of interaction are found in my data: (1) interactions between grammatical
gender (both sex-based and non-sex-based) and evaluative morphology, and (2) interactions between evaluative morphology and the encoding of biological gender (male vs.
female) in languages with non-sex-based gender or without grammatical gender. A major part of the present chapter is devoted to the analysis and discussion of the first type
of interaction. This is discussed in §§6.2, 6.3 and 6.4. The second type of interaction,
which was observed in a smaller number of languages, is shortly discussed in §6.5. Absence of interaction between the two domains is discussed briefly in §6.6. A summary of
the chapter is provided in §6.7.
6.2 Gender and evaluative morphology: overview of results
Two major kinds of interactions between grammatical gender and evaluative morphology
are found in the languages of the sample:
(a) If a language has non-sex-based gender, which generally implies a high number
of gender distinctions (see chapter 4), diminutives and augmentatives are part
of the inventory of gender distinctions. Nouns are generally not assigned to the
diminutive and augmentative genders by default but only when a diminutive or
augmentative interpretation is intended. I refer to these languages as Type 1.
(b) If a language has sex-based gender, which generally implies two or three gender
distinctions (see chapter 4), a change in the default gender of a noun (e.g., from
145
6 Gender and evaluative morphology
masculine to feminine, from feminine to masculine or from masculine/feminine
to neuter) may express diminutive or augmentative meanings. I refer to these
languages as Type 2.
The distribution of each type of interaction within the languages of my sample is
summarised in table 6.1.
Table 6.1: Interactions between grammatical gender and evaluative morphology
Type
Total
146
No. of
lngs.
Rel. %
Abs. % Genealogical groups
Type 1
23
27.4%
23% Bantu (15/23)
Kwa (1/3)
North-Central Atlantic
(7/7)
Type 2
28
33.3%
28% Berber (6/6)
Chadic (2/8)
Cushitic (3/13)
Dizoid (1/1)
Eastern Nilotic (3/3)
Hadza (1/1)
Khoe-Kwadi (3/5)
Semitic (3/7)
Sandawe (1/1)
South Omotic (1/1)
Ta-Ne-Omotic (4/4)
No interaction
13
15.5%
13% Bantu (6/23)
Chadic (1/8)
Cushitic (1/13)
Kxa (1/1)
Semitic (3/7)
Tuu (1/1)
No data
20
23.8%
20% Bantu (2/23)
Chadic (3/8)
Cushitic (9/13)
Khoe-Kwadi(2/5)
Mel (3/3)
Semitic (1/7)
No gender
16
100
16% Chadic (2/8)
Defoid (1/1)
Igboid (1/1)
Kwa (2/3)
Mande (4/4)
Western Nilotic (6/6)
100%
100%
6.3 Type 1: Diminutive and augmentative genders
Table 6.1 shows that languages of Type 2 are slightly more frequent than languages
of Type 1. The two types of interactions are discussed in two separate sections: Type 1
is discussed in §6.3, whereas Type 2 is the focus of §6.4.
As mentioned in §2.1.1, Corbett (2013a) proposes to refer to Type 2 phenomena as
instances of recategorization. In addition, as observed in 4.2.3, both Type 1 and Type
2 phenomena are often described in the literature as gender shifts. Since the notion of
recategorization and related terminology has not yet been established (Greville Corbett,
personal communication), the notion of gender shift – or, depending on the system, noun
class shift – is used in this dissertation to refer to both Type 1 and Type 2 phenomena.
To date, very little has been done to explore the crosslinguistic distribution of size-related
(and value-related) gender shifts (see discussion in §§2.1.1 and 2.6). To my knowledge,
this dissertation provides the first extensive account of the crosslinguistic distribution of
this phenomenon over a large sample of languages from one of the world’s macro-areas.
6.3 Type 1: Diminutive and augmentative genders
As outlined in §6.2, in the languages of Type 1 (23 out of 100; see table 6.1), diminutives
and augmentatives are part of the inventory of gender distinctions. In my analysis, I
count as diminutive and augmentative markers internal to a gender system either the
singular-plural class pairs or, if no number distinctions can be found, the individual
classes that are used to express diminutive and augmentative meanings. Within my
sample, this type is attested among the North-Central Atlantic and Bantu languages
as well as, to a certain extent, in the Kwa language SElEE (table 6.1). Typically, the
formation of diminutives and augmentatives is done by means of gender shift. Speakers
shift nouns from their default gender to the diminutive or the augmentative genders. As
a results, nouns trigger different indexation patterns (the diminutive and augmentative
indexation patterns) and are themselves marked in a different way. Examples (6.2)
and (6.3) illustrate the mechanisms of noun class shift for diminutive and augmentative
formation in a Bantu and a North-Central Atlantic language.
(6.2) Tonga (Bantu) (Carter 2002: 21)
(a) mu-sankwa
cl1-boy
‘boy’
b tu-sankwa
cl12-boy
‘small boy’
(6.3) Wamey (North-Central Atlantic) (Santos 1996: 160)
(a) ı̀-ñí
cl5-elephant
‘elephant’
147
6 Gender and evaluative morphology
(b) [email protected]̃í
cl18-elephant
‘big elephant’
It is relatively common for a Type 1 language to have several diminutive and augmentative noun class markers. This is discussed in §6.3.1 and illustrated with examples
from North-Central Atlantic and Bantu languages. In addition, the diminutive and augmentative genders may trigger a special type of marking on nouns: this is discussed in
§6.3.2 with examples from Bantu and North-Central Atlantic. §6.3.3 provides an insight
into diachrony and discusses the renewal of the evaluative genders in the southeastern
Bantu languages. Finally, in §6.3.4, I discuss evaluative genders in the Kwa language
SElEE, where the use of diminutive gender markers is obligatorily combined with the use
nominal diminutive suffixes.
6.3.1 Distribution of diminutive and augmentative genders and general
characteristics
Within my sample, the languages of Type 1 differ in the number of diminutive and augmentative genders. For instance, some languages may have only one diminutive and one
augmentative gender, whereas others may have more than one diminutive gender and
more than one augmentative gender. When accounting for the distribution of diminutive and augmentative genders in my database, I only differentiate between “One” and
“More than one” diminutive gender and “One” and “More than one” augmentative gender. However, within my sample, languages with more than one diminutive or more
than one augmentative always have maximum two genders per evaluative type. The
only exception is Venda (Bantu), which has three diminutive genders (but only two
augmentative genders) (Poulos 1990). The distribution of diminutive and augmentative
genders in the languages of Type 1 is summarised in table 6.2.
Table 6.2 shows that none of the attested distributional patterns is strikingly dominant. However, the most frequent pattern is the one with one diminutive gender and
no augmentative gender and is attested in 7 out of 23 languages. Interestingly, in 6 of
the 23 Type 1 languages – Bandial (North-Central Atlantic), Bemba (Bantu), Bidyogo
(North-Central Atlantic), Lega (Bantu), Shona (Bantu) and Wamey (North-Central Atlantic) –, the system of augmentative marking is richer than the diminutive (more class
distinctions are available). Out of these 23 languages, 5 have one diminutive and one
augmentative gender. The combination of more than one diminutive and more than one
augmentative gender is found only in two languages of the sample (the Bantu languages
Lega and Venda). Finally, a system with more than one diminutive gender and no augmentative genders is only attested in two languages (the Bantu language Chiga and the
Kwa language SElEE). Overall, the augmentative genders within my sample tend to be
less differentiated than the diminutive genders from a morphological point of view. If
one of the evaluative genders in a language lacks plural marking, this is the augmentative
– as in Kirundi (Bantu) or Serer (North-Central Atlantic). Moreover, as shown in table
6.2, languages may have diminutive gender(s) but lack augmentative gender(s) – as in
148
6.3 Type 1: Diminutive and augmentative genders
Table 6.2: Distribution of diminutive and augmentative genders in Type 1 languages (23/100)
Dim.
Aug.
More than one
More than one
2
8.7% Bantu (2/23)
More than one
One
1
4.4% North-Central
Atlantic (1/7)
One
More than one
6
26% Bantu (3/23)
North-Central
Atlantic (3/7)
One
One
5
21.7% Bantu (4/23)
North-Central
Atlantic (1/7)
More than one
None
2
8.7% Bantu(1/23)
Kwa (1/3)
One
None
7
30.5% Bantu (5/3)
North-Central
Atlantic (2/7)
Total
No. of
lngs.
23
Abs. % Genealogical
groups
100%
Nuclear Wolof (North-Central Atlantic) or Noon (North-Central Atlantic). The opposite
is not attested in the languages of the sample: all the languages with augmentative genders have at least one diminutive gender (see also §4.4 for more general tendencies in the
distribution of diminutive and augmentative markers in the languages of the sample).
If a language has more than one evaluative gender per evaluative meaning (that is,
more than one diminutive and/or augmentative gender), the different available markers
may differ from each other in terms of distributional preferences. For instance, they
may specialise in the encoding of different size nuances, i.e., ‘small’ vs. ‘tiny’ or ‘big’
vs. ‘huge.’ This is the case in the Bantu language Lega, where the diminutive marker
ka-, Class 12, encodes small size, whereas the Diminutive Class si»-, Class 19, encodes
extremely small, or tiny, size. The uses of the two prefixes are illustrated in (6.4).
(6.4) Lega (Bantu) (Botne 2003: 430)
(a) mu-ntu
cl1-person
‘person’
(b) ka-ntu
cl12-person
‘small person’
149
6 Gender and evaluative morphology
(c) si»-ntu
cl19-person
‘tiny person’
Another possibility is that some of the individual diminutive and augmentative classes
are used with restricted types of nouns as in Wamey (North-Central Atlantic), where
the augmentative marker ga-, Class 20, is used with nouns originally assigned to Classes
1 and 3 only (example (6.5)), whereas the Augmentative Class [email protected], Class 18, is used to
form the augmentative of nouns from any other class (example (6.3)).
(6.5) Wamey (North-Central Atlantic) (Santos 1996: 160)
(a) à-sæ̀n
cl1-man
‘man’
(b) ga–sæ̀n
cl20-man
‘big man’
Such patterns of differential marking are usually neutralized under plural reference.
Thus, in Lega, the Diminutive Plural Class correspondent to the Diminutive Singular
Classes 12 and 19 is Class 13, which corresponds to the marker tu-; likewise, in Wamey,
the Augmentative Plural Class correspondent of the Augmentative Singular Classes is
va- (Class 19).
Synchronically, two types of evaluative nouns classes can be distinguished:
(1) Noun classes that are extremely polysemous and productive in language use, and
are also used to derive diminutives and augmentatives when nouns assigned by
default to other classes are shifted to them. I call these Extended Evaluative
Classes (henceforth EECs)
(2) Noun classes that are only used to derive diminutives and augmentatives. Generally, there are no (or very few) nouns that are assigned by default to any such
classes; these essentially function as word-formation strategies. I call these OnlyEvaluative Classes (henceforth OECs).
Languages can have both EECs and OECs or only one of them. Bidyogo (NorthCentral Atlantic) and Shona (Bantu) provide a very good illustration of the coexistence
of the two types of evaluative classes.
In Bidyogo, Class E- and KO-44 are a good example of EECs. Class E- is one of the
most frequently used and one of the most bleached from a semantic point of view (it is
44
In his grammar of Bidjogo, Segerer (2002) uses capital letters to refer to the phonological realisation
of a class marker; this orthographic convention is also used in the glosses as a means for referring to
the individual noun classes and the indexation patterns associated with them. The noun classes are
not numbered.
150
6.3 Type 1: Diminutive and augmentative genders
also the noun class to which loanwords are assigned by default). Nouns assigned to class
E- are most commonly shifted to class KO- to encode number distinctions. However,
the interpretation of the two class markers in terms of number distinctions is subject
to vary. With some nouns, E- works as the singular and KO- as the plural, whereas
with other nouns the two markers work in exactly the opposite way. E- and KO- also
function as evaluative markers: the polarity between the two classes is maintained, but,
again, the semantic values associated with each class vary according to the type of noun
that they classify and depending on the context. Thus, E- is both a diminutive and an
augmentative and so is KO-. Consider the following examples:
(6.6) Bidjogo (North-Central Atlantic) (Segerer 2002: 103)
(a) kO-kOñ
KO-palm.leaf
‘palm leaf’
(b) E-kOñ
E-palm.leaf
‘small palm leaf’
(c) E-man
E-rice
‘rice’
(c) kO-man
KO-rice
‘grain of rice’
In opposition to Class E- and KO-, Class BA- is a perfect example of OEC. It is only
used to derive augmentatives, and there are no nouns that are inherently assigned to it.
(6.7) Bidjogo (North-Central Atlantic) (Segerer 2002: 125)
(a) jO-kO
JO-house
‘house’
(b) ba-kO
BA-house
‘big house’
In Shona, singular augmentatives can be encoded either by shifting a noun to Class
5 or to Class 21. Class 5, (6.8), is one of the most productive and least semantically
transparent noun classes in the language. It is thus another good example of an EEC.
In Fortune’s (1955) analysis, the underlying form of Class 5 is ru-, but the only phonetic
realisation of the prefix is the voicing of the initial consonant of a noun stem – if unvoiced
– as in (6.8). On the other hand, Class 21, (6.9), is used only to form augmentatives and
is thus an instance of OEC. Interestingly, Classes 21 and 5 trigger the same indexation
(Fortune 1955: p.105).
151
6 Gender and evaluative morphology
(6.8) Shona (Bantu) (Fortune 1955: 77)
(a) mu-kwasha
cl1-son-in-law
‘son-in-law’
(b) gwasha
cl5.son-in-law
‘big son-in-law’
(6.9) Shona (Bantu) (Fortune 1955: 104)
(a) rw-eNga
cl11-frying.pan
‘frying pan’
(b) z-eNga
cl21-frying.pan
‘large frying pan’
Nouns assigned by default to Class 5 are shifted to Class 21 to form their augmentatives. The opposition between Class 5 and Class 21 in the encoding of augmentatives is
neutralized under plural reference: plural augmentatives are always in Class 6.
Interestingly, with respect to diminutive formation, Shona only has OECs. Diminutives are encoded by shifting nouns from their default genders to Class 12, ka-, for
diminutive singular, and Class 13, tu-, for diminutive plural. These classes are exclusively used for diminutive formation and have no default members (Fortune 1955: 9495).
The evaluative classes of Swahili (Bantu) are closer to the EEC type. Shifts to Classes
7, ki-, and 8, vi-, are used for the formation of singular diminutives and the plural diminutives. Classes 5, ji-, and 6, ma-, are used for the formation of singular augmentatives
and plural augmentatives. The two class pairs encode evaluative meanings only as a
result of gender shift. Otherwise they are two of the most productive genders in Swahili.
Based on the description by Contini-Morava (2000), a wide range of nouns, covering a
wide range of semantic fields, is assigned by default to these genders. In the case of the
5/6 class pair, there is apparently no connection between the semantics of the nouns
assigned by default to this class pair and the augmentative meanings that this encodes
via gender shift. On the other hand, Contini-Morava describes the semantics of the
nouns assigned by default to the 7/8 pair as being generally associated to “smallness of
size” (Contini-Morava 2000: 10).
6.3.2 Multiple or additive class marking with diminutives and augmentatives
An interesting morphological property characterises the evaluative classes of some Bantu
languages: nouns that are shifted to the evaluative classes do not necessarily lose their
original class markers. On the contrary, they retain them and combine them with the
evaluative class markers, which ultimately control indexation. The phenomenon has
152
6.3 Type 1: Diminutive and augmentative genders
already been introduced in §4.2.3.1, where I discussed the distinction between additive
and replacive types of class markers in Bantu. In this section, I examine a couple of
examples from Bemba (Bantu) and Kikuyu (Bantu), which illustrate this phenomenon.
Two patterns are attested in these languages:
1. Nouns from Classes 1 and 2 (mostly humans), and 3 and 4 (generally plants and
some animals) maintain their original class prefix when shifted to the diminutive
class. This is shown in examples (6.10a) and (6.10b).
2. The original prefix of the noun is not retained. This applies to the majority of the
nouns undergoing diminution and is shown in examples (6.10c) and (6.10d).
(6.10) Kikuyu (Bantu) (Stump 1993: 8-9)
(a) mû-raata
cl1-friend
‘friend’
(b) ka-mû-raata
cl12-cl1-friend
‘small friend’
(c) i-rima
cl5-hole
‘hole’
(d) ka-rima
cl12-hole
‘small hole’
Multiple class marking is found associated with the marking of evaluative morphology
elsewhere in Bantu. According to Kavari & Marten (2009), for instance, multiple class
marking in Otjiherero is associated with the marking of diminutive, augmentative and
locative meanings, as well as with some kinds of plural formation.
A systematic survey of multiple class marking in Bantu is still missing, and the occurrence of the phenomenon in individual languages is usually explained in morphophonological terms. At a higher level of abstraction, the phenomenon of multiple class marking
is an interesting indicator of the fact “that the lexical use of noun classes is at least
in some instances morphologically distinguished from the derivational uses of classes”
(Crisma et al. 2011).
I did not find any example of multiple class marking in the North-Central Atlantic
and Mel languages of the sample. An arguable case is found in Bandial (North-Central
Atlantic). In Bandial tales, animal names are personified by means of the prefix ja-. The
prefix is analysed by Tendeng (2000) as a combination of the diminutive class marker
ju- – Class 11 – and the class marker a-, which is characteristic of nouns denoting
humans. Further on, the author states that the two class prefixes merge as the result
of a phonological rule of vowel deletion, whereby /u/ is dropped before the a- prefix.
153
6 Gender and evaluative morphology
Semantically, the diminutive class marker added to the marker for humans – Class
1 – functions as a sort of attenuator: animals are not just construed as human but
as “slightly human.” If Tendeng’s (2000) hypothesis is accepted, then personification
constructions in Bandial can be seen as an instance of multiple class marking. The
explanation offered by Tendeng has received mixed reactions in the literature. For
instance, Sagna (2008: 227) suggests that the cancellation of /u/ before a- may not be a
synchronically productive phonological rule of the language. Example (6.11) illustrates
the pattern of personification of animal names in Bandial.
(6.11) Bandial (North-Central Atlantic) (Sagna 2008: 228)
já-munduno a-jeg-or
me
bú-sol
maa...
cl11b-Hyena cl1.3sg-turn-recp subord cl5a-back like
this
‘Every time Hyena turns back like this’
In addition, Tendeng’s argument seems to be further weakened by the fact that in
Bandial, diminutivized human nouns do not retain their default class prefix, as illustrated
by example (6.12).
(6.12) Bandial (North-Central Atlantic) (Sagna 2008: 255)
(a) a-ññil
cl1-child
‘child’
(b) ju-ññil
cl11-child
‘baby’
Following Sagna’s (2008) convention, the alleged “diminutive + human class” combination is treated as an allomorph of the Diminutive Class and is referred to as Class 11b.
Personified nouns trigger Class 1 indexation, which could be explained as an instance of
referential indexation.
6.3.3 The renewal of evaluative morphology in the southeastern Bantu
languages
In §6.3, I showed that two types of evaluative classes can be found in the languages of
Type 1: the EECs and the OCSs. The former are noun classes with a very productive
range of uses and very loose semantics, which are also used to encode diminutives and
augmentatives. Conversely, the OECs are noun classes that are only used to express
evaluative meanings. In this section, I focus on the diachronic stability of the OECs in
the Bantu languages. In fact, there is evidence to believe that, at least in some Bantu
languages, the OECs are diachronically less stable than the EECs. When the system of
noun class marking of a Bantu language erodes, these classes – often together with the
locative classes – disappear. In such cases, either only the EECs survive – as in Swahili
154
6.3 Type 1: Diminutive and augmentative genders
– or new evaluative markers develop. These new evaluative markers may be completely
independent of grammatical gender or may coexist (and sometimes co-occur) with the
EECs. In the rest of this section, the diachronic development and the usages of such
markers are discussed.
The marking of diminutives and augmentatives underwent massive change in Bafia,
Eton, Northern Sotho, Shona, Swati, Tswana, Venda and Zulu. The innovations follow
two paths: on the one hand, Northern Sotho, Shona, Swati, Tswana, Venda and Zulu
– all spoken in the southeastern region of the Bantu-speaking area – have developed
diminutive and augmentative suffixes; on the other hand, Bafia and Eton have developed strategies for the encoding of diminutives and augmentatives that differ both from
class marking and from suffixation. I begin with an overview of the origin and uses of
diminutive and augmentative suffixes in the southeastern Bantu languages and move on
to Bafia and Eton thereafter.
Previous studies (Creissels 1999; Güldemann 1999) have shown that the diminutive
suffix in southeastern Bantu originates from the Proto-Bantu word for ‘child’ *jana
(Creissels 1999: 34) or *yana (Güldemann 1999). Table 6.3 provides an overview of the
suffixes attested across the six languages and notable facts concerning their most typical
uses. Information about the co-occurrence of diminutive suffixes and diminutive noun
classes is also given.
Table 6.3: The diminutive suffix in some southeastern Bantu languages
Language
Dim. suffix
Notable facts
Dim. noun classes
Northern Sotho -ana, -nyana
Lobedu dialect: noun class shift
+ suffix
Yes
Shona
-ana
Very incipient strategy, often cooccurring with the prefix
Yes
Swati
-ana, -anyana
-ana= small size ; -anyana= tiny
size
No
Tswana
-ana,-nyana
With animate nouns: -ana=
young age; -nyana= small size
No
Venda
-ana; -nyana
Noun class shift + suffix
Yes
Zulu
-ana
–
No
As shown in the table, languages vary in their use of the diminutive markers and in
the extent to which the diminutive noun class prefixes are maintained. However, some
general patterns can be observed throughout the data. Four languages (Northern Sotho,
Swati, Tswana and Venda) have two diminutive suffixes, the regular form and a form
that is described by grammars as “reduplicated.” While the reduplicated form is used
for expressing tiny size in Swati, in Tswana, both forms alternate in order to express
age (-ana) or size (-nyana) with animate nouns. This is exemplified in (6.13). With
inanimate nouns, the two suffixes are interchangeable.
155
6 Gender and evaluative morphology
(6.13) The diminutive suffix in Tswana (Bantu) (Cole 1955: 109)
(a) nku
‘sheep’
(b) nku-ana
sheep-dim
‘young sheep’
(c) nku-nyana
sheep-dim
‘small sheep’
In the Lobedu dialect of Northern Sotho, Shona and Venda, the Diminutive Classes
are still very productive and the two strategies (prefixal and suffixal) often co-occur. In
Venda, different sizes are expressed by the prefix on its own (either Class 7 or 20) and
by the combination of prefix and suffix.
(6.14) Diminutive marking in Venda (Bantu) (Poulos 1990: 88)
(a) tshi-kali
cl7-clay.pot
‘small clay pot’
(b) tshi-kal-ana
cl7-clay.pot.dim
‘very small clay pot’
(c) ku-kali
cl20-clay.pot
‘small clay pot’
(d) ku-kali-ana
cl20-clay.pot-dim
‘very small clay pot’
One of the few cases in which the Venda diminutive suffix occurs on its own is given in
(6.14).
(6.15) Venda (Bantu) (Poulos 1990: 86)
(a) kholomo
‘cow’
(b) nam-ana
cow-dim
‘calf’
156
6.3 Type 1: Diminutive and augmentative genders
In (6.15), the diminutive suffix is attached to a different noun stem: such kinds of
stem alternations are cross-linguistically very common for differentiating age in domestic
animals (Ljuba Veselinova, personal communication). Moreover, the word for ‘calf’ in
Venda may also be interpreted as a case of lexicalization, whereby speakers of Venda do
not actually analyse it as a diminutive noun any longer (Mark Van de Velde, personal
communication).
In addition to the diminutive suffix, an augmentative suffix is also found among the
southeastern Bantu languages. However, its use as an augmentative marker is restricted
to Swati, Zulu, and, more rarely, Tswana. In Shona, Northern Sotho, and Venda the
same suffix is attested but is only used to derive nouns denoting females. The augmentative affix originated from the proto-Bantu noun *kádi ‘woman’. Examples of the
augmentative suffix in Swati and Zulu follow.
(6.16) Swati (Bantu) (Taljard et al. 1991: 141)
(a) li-tje
cl5-stone
‘stone’
(b) li-tje-kati
cl5-stone-aug
‘rock’
(6.17) Zulu (Bantu) (Poulos & Bosh 1997: 9)
(a) umu-ntu
cl1-person
‘person”
(b) umu-ntu-kazi
cl1-person-aug
‘huge person’
Bantu nominal morphosyntax is prominently prefixal. Thus, the development of nominal suffixes such as the diminutive, the feminine/augmentative, and the locative suffixes
among the southeastern Bantu languages poses a relevant problem in the general understanding of the typology of Bantu nominal morphology. In Güldemann (1999), the
process is explained as a result of areal contact with predominantly suffixal languages
spoken in the same geographical area. According to his analysis, the best candidate
contact languages are Khoe and non-Khoe languages spoken in southern Africa (such as
Nama or Ju|’Hoan). In these languages, diminutive, augmentative and locative meanings are encoded by means of suffixes or postposed relational nouns. The southeastern
Bantu languages discussed in this section acquired suffixal strategies to mark evaluative morphology but not the markers as such. The phenomenon came as an effect of
intense cultural, social and economic contact (Güldemann 1999). The sociocultural context which fostered the grammaticalization of the diminutive and augmentative/feminine
suffixes is also discussed in §6.5.
157
6 Gender and evaluative morphology
I now turn to a short overview of the evaluative markers of the Bantu languages Bafia
and Eton. As mentioned at the beginning of this section, the evaluative markers of
these two languages are neither prefixal nor suffixal. In Bafia, diminutives are encoded
periphrastically: the noun to be diminutivized is preceded by the word mán, ‘child’,
Gender 1/2 (plural: áOń). This behaves as a fully lexical noun and triggers Gender 1/2
indexation. This is illustrated in example (6.18).
(6.18) Bafia (Bantu) (Guarisma 2003: 318)
m-á
zaÜ
á
ń-púp
cl1-child cl9chicken cl1 cl1-white
‘small white chicken’
In (6.18) the noun that undergoes diminution, zaÜ, is by default assigned to Gender
9/1045 but the nominal modifier for ‘white’ indexes the gender of the diminutive marker
(which, as mentioned before, is a noun belonging to Gender 1/2). Augmentatives are
also periphrastic in Bafia: the noun for ‘thing’ (cóm / áyóm, Gender 7/8) is used in
combination with the noun to be augmented and, as in the case of the diminutive marker,
triggers indexation.
In Eton, diminutives and augmentatives are encoded by means of proclitic words. The
lexical source of the diminutive proclitic, mO= (plural bO), is the noun for ‘child’(mONO/
bONO), from which the diminutive marker differs in virtue of some phonological erosion
and the floating high tone. The augmentative proclitic, mòd (plural bǒd), is formally
identical to the word for ‘person’, and differs from it only for its final floating high tone
(Van de Velde 2008: 208).
From a morphotypological point of view, the innovations in Bafia and Eton follow the
expected pattern of Bantu nominal morphology: the nouns and the clitics that encode
evaluative meanings in Bafia and Eton, respectively, precede the nouns that undergo
diminution or augmentation, as in the case of the prefixal diminutives and augmentatives.
Bafia and Eton are both spoken in Cameroon and belong to the same Bantu sub-area.
It might be that their innovations in the marking of diminutives and augmentatives are
geographically restricted similarly to those encountered in the southeastern languages.
6.3.4 Diminutive marking in SElEE
In this section, I focus on the diminutive constructions of SElEE, a Kwa language of
the Ghana-Togo-Mountain subgroup that is spoken in the Volta region of Ghana. In
contrast with the other two Kwa languages in the sample, Akan and Ewe, which do not
have gender, SElEE has a non-sex-based gender system similar to the Bantu and NorthCentral Atlantic type. Besides, SElEE has a very interesting inventory of diminutive
constructions, whereby diminutive markers internal to the gender system are always used
in combination with diminutive suffixes marked on nouns. These suffixes are extraneous
45
In Bafia, when nouns assigned to Gender 9/10 begin with a voiced consonant, the noun class prefixes
are not overtly coded on nouns (Guarisma 2003).
158
6.3 Type 1: Diminutive and augmentative genders
to the gender system and have historical origins similar to the diminutive suffixes of
the southeastern Bantu languages (see 6.3.3). A detailed description of the gender
system of SElEE is found in Agbetsoamedo (2014), whereas Agbetsoamedo & Di Garbo
(forthcoming) account for SElEE evaluative morphology.
According to Agbetsoamedo & Di Garbo (forthcoming), SElEE does not have morphological augmentatives and the attested diminutive constructions conform to two main
constructional patterns:
(1) Gender shift + suffixation. Nouns are shifted either to Gender 5/8 or to gender
3/7, and are marked by one of the following suffixes: -bi, -mii, E, -nyi.
(2) Suffixation. -bi is the only suffix that can occur without gender shift. This happens
when the diminutive of inanimate nouns is encoded.
It is worth mentioning that, among the diminutive suffixes mentioned above, -bi is
the only one for which a clear grammaticalization path has been established. The suffix
originates from the noun obi ‘child,’ which is still a productive lexical item in the language
(for more details, see Agbetsoamedo & Di Garbo forthcoming).
In the rest of this section, I focus on the diminutive constructions of the type described
in (1), as they are the only ones where interaction between the marking of gender and
evaluation occurs. Examples (6.19), (6.20) and (6.21) illustrate some instances of the
pattern described in (1). Note that for each gender, only the singular forms are given.
(6.19) The Diminutive in SElEE (Kwa): shift to Gender 5/8 + suffix -bi
(Agbetsoamedo & Di Garbo forthcoming)
(a) O-kla
cl1-cat
‘cat’
(b) lE-kla-bi
cl5-cat-dim
‘kitten’
(6.20) The Diminutive in SElEE (Kwa): shift to Gender 3/7 + suffix -mi
(Agbetsoamedo & Di Garbo forthcoming)
(a) kansiE46
cl1.bird
‘bird’
(b) ka-kansiE-mii
cl3-bird-dim
‘small bird’
46
In SElEE, some nouns within Gender 1/2 may lack overt coding of gender. These are mostly borrowings.
See Agbetsoamedo (2014) for details.
159
6 Gender and evaluative morphology
(6.21) The Diminutive in SElEE (Kwa): shift to gender 3/7 + suffix -nyi
(Agbetsoamedo & Di Garbo forthcoming)
(a) o-ti
cl1.person
‘person’
(b) ka-ti-nyi
cl3-person-dim
‘tiny person’ (derogatory)
Agbetsoamedo & Di Garbo (forthcoming) show that not all the possible combinations
between gender shift and suffixation are grammatical or equally productive in SElEE. For
instance, suffix -bi tends to combine with shift to Gender 5/8 whereas the suffixes -mi,
E and -nyi always combine with shift to Gender 3/7. Such distributional preferences
are also semantically relevant. With nouns referring to animal nouns, the diminutive
construction “shift to Gender 5/8 + -bi ” encodes young age or offspring (6.19), whereas
the construction “shift to Gender 3/7 + -mii (or -E)” expresses small size (6.20). Finally,
the construction “shift to Gender 3/7 + -nyi ” is attested only with two nouns, ‘person’
– as in (6.21) – and ‘hand,’ and is always derogatory.
Such a division of labour between the different possible diminutive constructions involving gender shift is similar to the situation described in §6.3.3 for the southeastern
Bantu languages. The major difference between SElEE and the southeastern Bantu languages is that in SElEE, none of the noun class markers involved in diminutive marking can express diminution independently of suffixation. Agbetsoamedo & Di Garbo
(forthcoming) suggest that in SElEE – similarly to the Bantu languages – “gender shifts
might have been the original strategy for the encoding of diminutives” and that, later
on, this started being combined (at the beginning, presumably only optionally) with a
set of diminutive suffixes. Interestingly, the most recent and semantically transparent
diminutive suffix in SElEE, -bi, is the only one that, at least in some cases, can occur
without gender shift.
6.3.5 Type 1: summary
In this section, I discussed languages for which diminutives and augmentatives are part of
the inventory of gender distinctions (Type 1 in the classification presented in §6.2). The
distribution of Type 1 languages within my sample is highly genealogically skewed: they
are only attested among Bantu, North-Central Atlantic and Kwa, that is, in languages
with large and non-sex-based gender systems. Two types of evaluative noun classes can
be found in Type 1 languages: the EECs, which are not only used to express diminutive
and augmentative meanings, and the OECs, which are only used to express evaluative
meanings. Languages differ as for whether or not they have both or only one type of
evaluative class. In §6.3.2, I discussed instances of multiple class marking that are related
to the use of the diminutive and augmentative classes in the languages of Type 1. The
diachronic stability of Type 1 languages was addressed in §6.3.3, where I discussed the
160
6.4 Type 2: Diminutives and augmentatives in languages with sex-based gender
renewal of evaluative classes in southeastern Bantu languages. Similar phenomena were
discussed in §6.3.4 in connection with the Kwa language SElEE. In SElEE, the diminutive
noun classes cannot be used on their own but always co-occur with the diminutive
suffixes.
6.4 Type 2: Diminutives and augmentatives in languages with
sex-based gender
Type 2 languages (28 out of 100; see table 6.1) have sex-based gender. Depending
on the number of distinctions that they exhibit in their gender systems, Type 2 languages use shifts from masculine to feminine, from feminine to masculine, or from masculine/feminine to neuter as a strategy for expressing evaluative meanings. This is
summarised in table 6.4.
Table 6.4 shows that Type 2 languages with two genders most commonly use gender shifts from masculine to feminine to encode diminutive meanings but lack a similar
mechanism for the encoding of augmentative meanings. This pattern is found in 12
languages. Slightly less common are systems in which the possibility of gender shift
is bidirectional. In these systems, when nouns are shifted from masculine to feminine,
the resulting meaning is diminutive; on the other hand, when nouns are shifted from
feminine to masculine, the resulting meaning is augmentative. This pattern is attested
in 10 of the 28 Type 2 languages in my sample. Finally, in the isolate Hadza, gender
shifts from the Masculine to the Feminine Gender are used for encoding augmentative
meanings (Edenmyr 2004: 17). In two languages with three gender distinctions, gender
shifts from the masculine and/or feminine to the neuter gender are used as a strategy for encoding diminutive meanings. These languages are Karamojong and Turkana,
both belonging to the Eastern Nilotic group. An interesting pattern is attested in the
Khoe-Kwadi languages of the sample. Even though these languages have three genders
– Masculine, Feminine and Common – the Common Gender is never involved in gender
shifts encoding size variation. In two languages, Kxoe and Naro, shifts from the Masculine to the Feminine encode diminutive meanings, whereas shifts from the Feminine to
the Masculine encode augmentative meanings. In Nama, only shifts from the Masculine
to the Feminine Gender are documented in my sources and they are used to express
diminution. Examples from Nama are given in §6.4.1.
The general characteristics of Type 2 languages are discussed in §6.4.1. As opposed
to Type 1 languages, which are only attested among the North-Central Atlantic, Bantu
and Kwa languages of the sample, Type 2 languages have a wider distribution across
genealogical groupings. This is examined in detail in §6.4.2.
161
6 Gender and evaluative morphology
Table 6.4: Number of gender distinctions and type of attested gender shifts in Type 2 languages
(28/100)
No. of
genders
Type of shift
2
M → F = Dim.
F → M = Aug.
No. of
lngs.
10
Abs. % Genealogical
groups
35.8% Berber (6/6)
Eastern Nilotic
(1/3)
Semitic (1/7)
Ta-Ne-Omotic
(2/4)
2
M → F = Dim.
12
43% Cushitic (3/13)
Chadic (2/8)
Dizoid (1/1)
Sandawe (1/1)
Semitic (2/7)
South Omotic
(1/1)
Ta-Ne-Omotic
(2/4)
2
M → F = Aug.
1
3.5% Hadza (1/1)
3
M → F = Dim.
2
7.1% Khoe-Kwadi
(2/5)
F → M = Aug.
Total
162
3
F → M =Dim
1
3.5% Khoe-Kwadi
(1/5)
3
M/F → N =
Dim.
2
7.1% Eastern Nilotic
(2/3)
28
100%
6.4 Type 2: Diminutives and augmentatives in languages with sex-based gender
6.4.1 Examples of Type 2 languages
Afro-Asiatic languages may be defined as hotbeds for Type 2 phenomena (see §6.4.2 for
more details). To give an example, in all the the Berber languages of my sample, shifts
from the Masculine to the Feminine Gender and vice versa are used to encode variation
in size with inanimate nouns. The evaluative value conveyed by gender shifts depends
on the default gender of a noun and the construal of the np referent with respect to
the parameter of size. Inanimate masculine nouns are shifted to the Feminine Gender
when a diminutive construal is intended. On the other hand, inanimate feminine nouns
are shifted to the Masculine Gender when an augmentative construal is intended. If
the default gender of a noun meaning, for example, ‘basket’ is Feminine, the Masculine
Gender may be used to refer to a basket that is bigger than expected. Gender shifts
are mostly used with objects that appear in different sizes, for instance, a small and
a big pot, a small and a big jewel, etc. (Amina Mettouchi, personal communication).
Examples (6.22) and (6.23) show how size-related gender shifts in Tachawit.
(6.22) Tachawit (Berber) (Penchoen 1973a: 12)
(a) aq-nmuš
[m-]sg-pot
‘pot’
(b) taq.nmuš-t
f-sg-pot-f.sg
‘small pot’
(6.23) Tachawit (Berber) (Penchoen 1973a: 12)
(a) taG-nžak-t
f-sg-spoon-f
‘spoon’
(b) aG-nž
[m-]sg-spoon
‘big spoon, ladle’
In example (6.22), a masculine noun is shifted to the Feminine Gender to encode reduced
size, whereas in example (6.23), an inherently feminine noun is shifted to the Masculine Gender to express increased size. What happens then if speakers want to express
notions like ‘big pot’ or ‘small spoon’ ? In other words, how do speakers of Tachawit
(and other Berber languages) construe size values for which the gender shift strategy
cannot be used? In such cases, speakers would use adjectives for ‘big’ and ‘small’ in
combination with the noun. The notion of ‘pot’ and ‘small pot,’ and ‘spoon’ and ‘big
spoon,’ as expressed by the nouns in examples (6.22) and (6.23), are, to a certain extent,
lexicalized insofar as they are built into gender assignment. Thus, speakers can combine
the adjective for ‘big’ and ‘small’ with any of these nouns. For instance, the adjective
for ‘big’ could co-occur with the noun taqnmušt ‘small pot’ (6.22b), when speakers want
163
6 Gender and evaluative morphology
to make a distinction between two pots of different sizes and point at the big small pot
(Amina Mettouchi, personal communication). The same principle holds for the use of
the adjective for ‘small’ with the noun aGnž ‘big spoon, laddle’ (6.23).
In the Berber languages, nouns denoting humans and higher animals can also undergo
gender shifts. However, in such cases, the gender shift is interpreted as encoding sex
rather than size. As pointed out by Grandi (forthcoming-a) with nouns that “designate
living beings without sexual dimorphism,” such as reptiles, fish, and bugs, gender shifts
can be interpreted both as sex or size-related, depending on the context. An example
is given in (6.25), where the feminine form of the noun for ‘lizard’ can be interpreted as
referring either to a female lizard or to a small one.
(6.24) Tachelhit (Berber) (as cited in Grandi forthcoming-a)
t-a-herdan-t
f/dim.f-sg-lizard-f/dim.f
‘small lizard’ or ‘female lizard’
Lexical restrictions to this pattern are also found. For instance, again in Tachelhit,
when the noun for ‘bird’ is treated as feminine, this can only be interpreted as encoding
diminution. This is illustrated in (6.25).
(6.25) Tachelhit (Berber) (as cited in Grandi forthcoming-a)
F-a-fruq-F
dim.f-sg-bird-dim.f
‘small bird’
In Masai (Eastern Nilotic), speakers can use gender shifts with human nouns to convey
size differences (and derogation) and not only biological gender. This is shown in (6.26).
(6.26) Masai (Eastern Nilotic) (Payne 1998: 166)
(a) Enk-anáshÈ
f.sg-sister
‘sister’
(b) Onk-anáshÈ
m.sg-sister
‘very large sister’ (pejorative)
(c) O-aláshÈ
m.sg-brother
‘brother’
(d) Enk-aláshÈ
f.sg-brother
‘weak brother’
164
6.4 Type 2: Diminutives and augmentatives in languages with sex-based gender
The pattern illustrated in (6.26) is crosslinguistically very rare: the use of gender
shifts to encode diminutives and augmentatives is usually restricted to inanimate nouns,
as in the case of the Berber languages (see also data in Aikhenvald 2003). In my sample,
Masai is the only language where this pattern is attested with animate nouns. The
gender system of Masai is very peculiar also in other respects. Masai has two productive genders: Masculine and Feminine.47 In addition, as shown by Payne (1998) and
Shirtz & Payne (2012), nouns in Masai can be classified according to their compatibility
with the Masculine Gender and the Feminine Gender:
[s]ome nouns stems seem equally compatible with both. Other seem to have a
‘default’ gender and reserve the other for pejorative or unusual size readings.
Only a small subset of the lexicon is compatible with only one gender. Finally,
there is a small set of noun stems which is completely incompatible with the
category of gender. (Shirtz & Payne 2012)
Based on this distinction, for the majority of nouns in Masai, gender assignment is
not lexically specified but is the result of the interplay between the denotation of a
noun and the speaker’s construal of the np referent. According to Payne (1998) and
Shirtz & Payne (2012), the Masai gender system challenges existing models of gender
assignment whereby nouns are allotted to a gender on the basis of their meaning, their
inflectional class or a combination of both. I shall come back to the relevance of their
argument in the final discussion chapter (see chapter 8).
Finally, in nearly all Type 2 languages with two genders, size-related gender shifts are
driven by the association between feminine gender and small size, and masculine gender
and big size. The opposite (feminine is large/big and masculine is small) is attested
in one language, the isolate Hadza. Notice that Masculine is the unmarked gender in
Hadza.
(6.27) Hadza (isolate) (Edenmyr 2004: 16)
(a) Pato
small.axe.m
‘small axe’
(b) Pato-ko
axe-f
‘large axe’
In the languages with three gender distinctions (masculine, feminine, neuter/common),
gender shifts that are relevant for the encoding of evaluative meanings can be restricted
to the masculine and the feminine genders only (as in the Khoe-Kwadi languages) or can
extend to the third gender. This is the case of the Eastern Nilotic languages Turkana
and Karamojong. The third gender in Turkana and Karamojong – labelled as Neuter
47
There is also a third gender, the Locative, which is, however, very rare and can be regarded as an
instance of inquorate gender (see §4.2.2)
165
6 Gender and evaluative morphology
Gender in grammars – is the lexical gender of a very limited number of nouns as opposed to the more productive Masculine Gender and Feminine Gender. In Turkana and
Karamojong, inherently neuter nouns are nouns that denote the offspring of animate
entities or individual members/instances of groups. In Turkana, animate and inanimate
feminine and masculine nouns can be shifted to the Neuter Gender to encode young age
(6.28 a and b), small size (6.28 c and d), or small quantity (6.28 e and f), depending on
the semantics of the noun and its countability properties.
(6.28) Turkana (Eastern Nilotic) (Dimmendaal 1983: 218)
(a) a-kale
f.sg-goat
‘she-goat that has not yet produced young’
(b) I-kle
n.sg-goat
‘kid-goat’
(c) e-dya
m.sg-boy
‘boy’
(d) i-dya
n.sg-boy
‘small boy’
(e) Na-kOt
f.pl-blood
‘blood’
(f) Ni-kOt
n.pl-blood
‘A little blood’
In Karamojong, shifts to the Neuter Gender are only used with animate nouns and
indicate offspring (Novelli 1985).
As mentioned above (see also table 6.4), in the Khoe-Kwadi languages, size-related
gender shifts do not involve the third gender, traditionally labelled as the Common
Gender. Gender shifts from Masculine to Feminine and from Feminine to Masculine are
attested in Kxoe and Naro. In Nama, the only examples of gender shifts mentioned
in the grammar involve shifts from Feminine to Masculine, which are used to express
diminutive meanings. These are presented in (6.29).
(6.29) Nama (Khoe-Kwadi) (Hagman 1977: 23)
(a) ’om-s
house-f.sg
‘house’
166
6.4 Type 2: Diminutives and augmentatives in languages with sex-based gender
(b) ’om-i
house-m.sg
‘big house, apartment or office building’
(c) xãa-s
penis-f.sg
‘penis’
(d) xãa-p
penis-m.sg
‘small penis’
In Nama, size-related gender shifts are restricted to inanimate nouns and “express that
there is something unusual about the referent of the noun” (Hagman 1977: 23). Hagman
states that the nature of the semantic shift is hard to pin down since it depends on the
semantics of the noun that undergoes gender shift and on the general discourse context
in which the noun is used. He continues by saying that if the typical appearance of a
np referent is not large/big, gender shifts encode increased, large/big size; on the other
hand, if it is usual for a noun referent to be large/big, gender shifts mark unexpected
small size. Usually, if the largeness or smallness of size are undesirable for the object in
question, the encoding of dimensional variation also entails derogation. Nama has also
specialised diminutive and augmentative suffixes (see discussion in §6.4.2).
6.4.2 The distribution of Type 2 languages
As mentioned in §6.4, Type 2 languages are found virtually among all the genealogical
groupings with sex-based gender: Berber, Chadic, Cushitic, Dizoid, Khoe-Kwadi, Eastern Nilotic, South Omotic, Semitic, Ta-Ne-Omotic and the isolates Hadza and Sandawe
(table 6.1). The distribution of size-related gender shifts within the individual genealogical groups varies: they are attested in all the sampled Berber varieties, in all the sampled
Omotic groups, and in all the sampled Khoe-Kwadi languages (with the exception of
{Ani and Kwadi). The existence of size-related gender shifts appears to be less pervasive in Chadic, Cushitic, and Semitic. The phenomenon was explicitly mentioned by my
sources only in the case of two Chadic languages out of eight (Gidar and Hausa), and
three Cushitic languages out of thirteen (Awngi, Beja and Daasanach). I cannot exclude
that similar phenomena might also exist in other Cushitic and Chadic languages of the
sample for which I could not retrieve any information in the sources that I consulted.
Noteworthy are the findings from Semitic. The only languages that can be clearly
classified as Type 2 are Amharic, Tigre and Maltese. In his survey of diminutive constructions in Hebrew, Bolozky (1994: 50) mentions the existence of a set of lexicalized
diminutives that differ from their correspondent nondiminutivized forms only in gender (the lexicalized diminutives are all feminine). However, he continues, gender shifts
are no longer perceived as a productive strategy of diminutive formation by speakers
of Israeli Hebrew (1994: 51). No relationship between gender shift and evaluation is
167
6 Gender and evaluative morphology
found in Standard Arabic48 and Moroccan Arabic, whereas no information was at all
recoverable on evaluative constructions in the Aramaic language Lishan Didan. Due to
intense contact with Romance languages, the evaluative morphology system of Maltese
displays an interesting interplay between indigenous (Semitic) and borrowed (Romance)
evaluative markers. In this section, I mainly focus on those aspects of Maltese evaluative
morphology that concern its interaction with gender marking (see Grandi 2002: 221-231,
for a complete inventory of the evaluative constructions of Maltese). As is typical of
other Semitic languages, diminutives in Maltese are formed by means of discontinuous
infixal morphemes, that is, by a set of vowel templates that are inserted in the root in
different ways, mostly depending on the number of consonants that a given root has. In
some cases, vowel infixation is accompanied by a shift in gender, notably from Masculine
to Feminine. This is exemplified in (6.30).
(6.30) Discontinuous infixal diminutive + gender shift in Maltese (Semitic) (adapted
from Grandi 2002: 225)
(a) xat
beach.m
‘beach’
(b) xtajta
beach:dim.f
‘small beach’
In other cases, the diminutive of a noun can be encoded only by means of gender shift,
that is, without infixation of the diminutive vowel template (Grandi 2002: 226). The
relation between feminine and diminutive reference seems to have spread to some nouns
of Romance origin. This is shown in (6.31).
(6.31) Gender shift as a strategy for diminutive formation with nouns of Romance
origin in Maltese (Semitic) (adapted from Grandi 2002: 230)
(a) forno
oven.m
‘oven’
(b) forn-a
oven-dim.f
‘small oven’
In (6.31), the diminutive of the borrowed noun forno ‘oven’ is formed by changing the
gender of the base. More commonly, nouns of Romance origin are diminutivized by
means of the diminutive suffix -in, also borrowed from the Romance contact languages.
This is shown in example (6.32).
48
In his typological survey of evaluative constructions in the languages of the Mediterranean area, Grandi
(2002: 218-248) notices that evaluative morphology is generally not very productive in Standard Arabic and Hebrew. Diminutives and augmentatives are more frequently used in nonstandard varieties
of Arabic.
168
6.4 Type 2: Diminutives and augmentatives in languages with sex-based gender
(6.32) Romance diminutives in Maltese (Semitic) (adapted from Grandi 2002: 229)
(a) biskott
biscuit.m
‘biscuit’
(b) biskutt-in
biscuit.m-dim
‘small biscuit’
There are no morphological augmentatives of Semitic origin in Maltese. The augmentative suffix -un has been borrowed from the Romance contact languages and is productively used with Semitic nouns as well (Grandi 2002: 229-231). Gender shift does not
play any role in the formation of augmentatives in Maltese.
To summarise, given that size-related gender shifts are attested throughout all the
Afro-Asiatic genealogical subgroupings included in the sample, there is evidence to claim
that they are an inherited feature of nominal morphology across the language family,
rather than an independent innovation of the individual subgroupings.
Gender shifts may not be the only strategy available in a Type 2 language to encode
evaluation. In a handful of languages of my sample, evaluative markers independent of
gender coexist with size-related gender shifts. This is the case of Male (Ta-Ne-Omotic),
Nama (Khoe-Kwadi) and, as described earlier in this section, Maltese. In Male, gender
marking with inanimate nouns always implies size. In addition, the diminutive suffix
-ómma is found (Amha 2001: 70-71). In Nama, as previously mentioned, there exist
a diminutive and an augmentative suffix, -ró and -kra. Interestingly, the diminutive
suffix works independently of gender shifts, whereas the augmentative is often used in
combination with gender shifts to refer to extra-large size (Hagman 1977: 27).
Not much can be said about the diachrony and stability of the phenomenon of sizerelated gender shifts. What is known is that they tend to occur in languages in which
gender assignment is not rigid. In such languages nouns can be assigned to different
genders on the basis of the context in which they occur. The possibility of manipulating
gender assignment is, however, conditioned by certain semantic properties of nouns. On
a general basis, nouns that denote human beings have a tighter relation with their default
gender.
6.4.3 Type 2: summary
In this section, I discussed languages with sex-based gender in which gender assignment
is not rigid and gender shifts are used to encode diminutive and augmentative meanings
(Type 2 in the classification proposed in §6.2). In §6.4.1, I surveyed the general characteristics of languages of Type 2 and discussed examples from various languages within
Type 2. The distribution of Type 2 languages within my sample has been discussed
in §6.4.2, where I showed that this type of interaction between gender and evaluative
morphology has a wide diffusion across different genealogical groupings of Africa.
169
6 Gender and evaluative morphology
6.5 Evaluative morphology and biological gender in languages
without (sex-based) gender
In this section, I describe a pattern that I found attested in only 8 of the 100 languages
of my sample. This pattern is concerned with the interaction between the linguistic
encoding of biological gender (i.e., strategies to characterise animate entities as females
or males) and evaluative morphology. Although it is not entirely relevant for the general
typological picture of the relationship between grammatical gender and evaluative morphology in the languages of Africa, this pattern deserves some discussion. The pattern
was found in languages with non-sex-based gender and in languages without grammatical
gender, that is, in languages where the encoding of biological gender is outside the gender
system, or where there is no gender system at all. Five of these nine languages belong
to the Bantu group and are the languages discussed in §6.3.3 with respect to evaluative
morphology renewal; they have non-sex-based gender: Northern Sotho, Swati, Tswana,
Venda and Zulu. The remaining three languages are Luwo, Acoli and Akan. Luwo and
Acoli belong to the Western Nilotic genealogical unit. Akan is an Atlantic-Congo language of the Kwa subgroup. None of the three languages mentioned last in this list has
grammatical gender.
6.5.1 Bantu
Bantu languages do not encode biological gender within the noun class system. The most
common strategies for conveying information about biological gender are suppletion or
constructions of the types female cattle for ‘cow’ (Güldemann 1999: 59). Nonetheless, in
some of the southeastern Bantu languages, morphological strategies for the overt coding
of female and male reference exist and are somewhat related to the domain of evaluation
in two ways:
1. The diminutive suffix derived from *jana ‘child’ can be used as a marker of female reference in very specific contexts, mostly with animal names or with colour
adjectives modifying animal names.
2. The suffix resulting from the grammaticalization of *kádi ‘woman’ becomes productive for the derivation of feminine nouns. As shown in 6.3.3, in fewer languages
within the southeastern Bantu area, this suffix is also used as augmentative.
Selected examples from the above-mentioned six languages will now be discussed to
illustrate how these usages developed diachronically.
In Sotho, the suffix -hadi is used to encode female reference when added to nouns
denoting animals but functions as an augmentative when used with nouns denoting
human beings (Creissels 1999). Table 6.5 illustrates how the meaning of the suffix varies
depending on the meaning of the base noun.
Poulos & Louwrens (1994) present a different set of facts with respect to the Northern
Sotho variety. Both -ana (the Diminutive) and -gadi (the Augmentative-Feminine) are
said to be used as feminine suffixes. However, differently from what reported by Creissels
170
6.5 Evaluative morphology and biological gender in languages without (sex-based) gender
Table 6.5: The meanings and function of -hadi in Sotho (based on Creissels 1999: 32)
Base noun
pere,‘horse’
monna, ‘man’
+
+
Suffix
Suffix meaning
Resulting form
-hadi
-hadi
f
aug
pere-hadi, ‘she-horse’
monna-hadi, ‘big man’
(1999), according to Poulos & Louwrens (1994: 68-69), the suffix -gadi is no longer used
as an augmentative marker in Northern Sotho, and its use as a feminine marker is also
very rare. The diminutive suffix, -ana, on the other hand, can express feminine meaning
but only when used with animal nouns or adjectives denoting colour. The same pattern
is found in Tswana, geographically very close to Sotho.
Grammatical descriptions of Tswana (see, for instance, Cole 1955: 141-144) report
that, when colour adjectives modify nouns denoting domestic animals, the diminutive
suffix is always used to mark female reference on the adjective. When the adjective is not
marked by the diminutive suffix, the combination “Noun+ adjective” is interpreted as
having male reference. Creissels (1999: p.34) provides an interesting example in which
such a construction is attested being used to denote a human entity rather than an
animal noun. In Creissels’ example the colour adjective for ‘yellow’ occurs with the
word for ‘woman,’ mosadi, and is marked by the diminutive suffix -ana.
(6.33) Tswana (Bantu) (Creissels 1999: 34)
(a) monna
yomo-setlha
cl1.man cl1-yellow
‘a man with clear skin’
(b) mosadi
yomo-setlh-ana
cl1.woman cl1-yellow-f
‘a woman with clear skin’
The use of -ana with adjectives modifying human nouns is not mentioned in grammatical
descriptions of Tswana and is described as fairly atypical by Tore Janson (personal
communication). What is relevant about the example in (6.33b) is that two different
indexation patterns appear on the adjective. The adjective indexes the gender of the
noun by means of Class 1 prefix. In addition, the occurrence of the suffix -ana on the
adjective can be read as some sort of referential indexation pattern. The suffix signals
that the np referent is female. This usage can be interpreted as a step forward in the
grammaticalization of the diminutive/feminine suffix. The phenomenon is also attested
in Zulu, both with the diminutive and the feminine/augmentative suffix (Güldemann
1999: 74).
The suffixes -gadi and adi in Tswana have a meaning “associated with the feminine,
the marriage institution or the idea of the opposite sex” (Cole 1955: p.110). The two
suffixes are rarely used to mark biological gender with animal nouns (e.g., tau ‘lion,’
171
6 Gender and evaluative morphology
taugadi ’she-lion’). The suffix is more productive in the neighbouring (and closely related) languages Southern Sotho and Zulu. Similar patterns are also attested in Swati
and Venda.
To summarise, the overt coding of female reference in the southeastern Bantu languages is the result of two recent developments, which do not seem to have reached the
same stage of grammaticalization in all the languages where they appear:
(1) diminutive > feminine
(2) feminine > augmentative
The rise of markers for the encoding of female reference follows the development of the
new diminutive morpheme in the case of suffixes derived from *jana, ‘child’ (1). On the
other hand, in the case of the suffixes derived from *kádi, ‘woman’, the most plausible
pattern is: from marker of female reference to augmentative (2). The two patterns reveal
a substantial interaction between evaluative morphology and the marking of biological
gender.
As shown in the present section, the overt coding of female reference is often limited
to a narrow range of occurrences, such as gender specification with animal nouns or,
as in Tswana, colour adjectives used to differentiate different kinds of animals, mostly
cows (Tore Janson, personal communication). Thus, the use of these suffixes as markers
of biological gender is limited to very salient domains of experience for societies whose
economic subsistence is very often based on hunting, livestock farming and selling. As
already mentioned in §6.3.3, the development of diminutive and augmentative/feminine
suffixes in the southeastern Bantu languages is explained by Güldemann (1999) as the result of contact with Khoe and non-Khoe languages spoken in southern Africa. Livestock
breeding and hunting play a crucial role in the economy of many (Bantu and non-Bantu)
societies of southern Africa:
this conceptual orientation towards animals suggests that such closely associated features as sex and size are prone to be expressed linguistically
(Güldemann 1999: 71).
Regular and transparent encodings of such notions is thus quite likely to be borrowed
cross-culturally across languages.
6.5.2 Western Nilotic
In Western Nilotic languages, there exists an inventory of morphological strategies for
the categorisation of nouns in semantically homogeneous groups (see discussion in §5.6).
In this section, I examine a few cases of interactions between the word-formation
strategies that are used in these languages for the encoding of biological gender, on the
one hand, and size, on the other. The languages considered are Acoli and Luwo.
In Luwo, the diminutive marker is the prefix ñÌ-, which literally means ‘daughter of.’
Interestingly, the prefix is also found in ethnonyms where it expresses female reference.
The two possible uses of the prefixes are illustrated in examples (6.34) and (6.35).
172
6.5 Evaluative morphology and biological gender in languages without (sex-based) gender
(6.34) Diminutive formation in Luwo (Western Nilotic) (Storch 2005: 295)
ñÌ-bòò
dim-banana
‘small quantity of banana’
(6.35) Feminine ethnonyms in Luwo (Western Nilotic) (Storch 2005: 295)
ñÌ-jáN
f-dinka
‘Dinka woman’
When used as diminutive, the prefix tends to express only small quantities as in (6.34).
Diminution in size is constructed by combining the use of the diminutive prefix with the
adjective mát”ÍÍn
” ‘small,’ as in example (6.36).
(6.36) Luwo (Western Nilotic) (Storch 2005: 295)
ñÌ-t”ÍÍn
”
mát”ÍÍn
”
dim-child small
‘small child’
Similarly to Luwo, in Acoli the diminutive prefix, ñ-, is etymologically related to
the noun for ‘child, daughter of.’ Augmentatives are constructed by juxtaposing mı́in,
‘mother/female of,’ to the noun to be augmented (Storch 2005: 358).
6.5.3 Akan
The Kwa language Akan does not have grammatical gender. The diminutive marker
of Akan is the suffix -ba/-wa – depending on the dialect – which, as described by
Appah & Amfo (2011), originates from the noun Oba, ‘child’ (in the sense of offspring).
Appah & Amfo (2011) discuss the connection between the use of the diminutive suffix
and the encoding of female reference. In Akan, many common and proper names denoting women are marked by the suffix -ba/-wa. Three nouns of this type are listed in
(6.37).
(6.37) Diminutive marking and female reference in Akan (Kwa) (Appah & Amfo 2011)
• abaayewa ‘young woman’
• aborOwa ‘European woman’
• Egyrba female name (the correspondent male name is Egyr )
173
6 Gender and evaluative morphology
With the exception of the first noun in the list, the semantics of these nouns is not
related to diminution in any evident way. The suffix -ba/-wa only encodes female reference. Interestingly, in the Asante and Akuapem dialects of Akan, female names can also
be formed by using the suffixes -waa and -bea, which are cognates derived from the noun
Obaa/Obea ‘female’ (Appah & Amfo 2011: 91). Appah & Amfo (2011: 91) propose that
the diminutive and the feminine suffixes derive from two different but morphophonologically very similar nouns – Oba ‘child’ and Obaa/Obea ‘female,’ respectively – and are used
interchangeably by speakers of Akan to derive female names from the correspondent
male nouns. No such connection between diminutive and female reference is found in
the two other Kwa languages of the sample, Ewe and SElEE.
6.5.4 Summary
In this section, I surveyed interactions between the encoding of biological gender and
evaluative markers in eight languages with non-sex-based gender or no gender. In these
languages, the encoding of female reference shows associations with both diminutive
markers originated from the word for ‘child’ and augmentative markers originated from
the words ‘woman’ or ‘daughter.’ These interactions are very rare in my language sample,
but correspond to a rather frequent pattern worldwide (Bauer 2002; Jurafsky 1996;
Matisoff 1992).
6.6 Absence of interaction between gender and evaluative
morphology
As shown in table 6.1, the marking of diminutives and augmentatives shows no interaction with the gender domain in 13 of the 100 languages of the sample. The evaluative
constructions of these languages are not surveyed in detail here. However, a few examples are discussed as an illustration of the types of evaluative markers that are attested
in such languages (see also §6.3.3 for a discussion of evaluative morphology in the Bantu
languages Bafia and Eton).
In {Ani (Khoe-Kwadi), Ju|’hoan (Kxa) and !Xóô (Tuu), only diminutives are found.
These are suffixal and all etymologically related to the word for ‘child.’
In Kambaata (Cushitic), the encoding of diminutive and augmentative meanings does
not interact with gender marking but rather with number. In certain contexts, the
singular and plural suffixes can be used as diminutives and augmentatives/pejoratives,
respectively.
(6.38) Singular marking as diminutive in Kambaata (Cushitic) (Treis 2011: 6)
mat-íta
ann-uhúu
ám-atii
yoo-ba’í
one-f.acc father-m.nom.crd mother-f.nom.crd cop.3-neg.rel
wotar-ch-úta aass-íi
iitt-an-tóo’u
foal-sg-f.acc give-m.dat decide-pass-3fpfv
174
6.7 Summary of the chapter
‘It was decided to give [them] a [tiny] donkey foal which had no father and
mother’
(6.39) Plural marking as augmentative/pejorative in Kambaata (Cushitic) (Treis
2008: 144)
hor-i-ssá
inq-áakk-ant
hínn y-itóo’u
all-m.gen-3.pl.poss tooth-pl-f.nom<n> smirk say-3.f.pfv
‘[As they said, their upper lips were lifted up and] the teeth of all of them
smirked’
The use of the plural suffix in (6.39) can be interpreted as either signalling the big size
of the donkey’s teeth or “as indicating that the teeth are an object of ridicule” (Treis
2008: 144).
6.7 Summary of the chapter
The purpose of this chapter has been to survey patterns of interaction between gender
and evaluative morphology in the languages of the sample. In particular, the chapter
has aimed at examining (1) the frequency of occurrence of size as an independent gender
value and (2) how different types of gender systems (sex-based vs. non-sex-based) and
strategies of gender assignment (rigid vs. manipulable gender assignment) condition
the attested patterns of interactions between gender and evaluative morphology. The
research questions addressed in the chapter are listed in (6.1).
In general, the data presented in this chapter show that grammatical gender and the
morphological encoding of evaluation interact in relevant ways in the languages of Africa.
This is in line with what has been pointed out by previous investigations such as Allan
(1977) (see also §2.6).
The occurrence of size as an independent gender value was found in 23 of the 84
languages with gender in my sample. The distribution of the phenomenon is restricted
to the following genealogical groups: North-Central Atlantic, Bantu and Kwa. These
languages are characterised by non-sex-based gender systems and a large number of gender distinctions (see §6.3). The most frequent type of interaction between gender and
evaluative morphology, which was found in 28 of the 84 languages with gender, does not
entail the presence of dedicated diminutive and augmentative genders. In languages with
sex-based gender systems, gender shifts from masculine/feminine to neuter or from masculine to feminine (and vice versa) can be used to encode diminutive and augmentative
meanings (see §6.4).
The data discussed in the chapter thus show that the nature of the interactions between
gender and evaluative morphology varies according to the type of gender system that a
language has. The relevant variables are: type of gender system (sex-based vs. non-sexbased gender), number of distinctions within the system and degree of manipulability
of gender assignment. The data also confirm that size can be a productive criterion for
175
6 Gender and evaluative morphology
the classification of nouns but is never centrally prominent, neither synchronically nor
diachronically, in the two types of gender systems that are found among the languages
of the sample.
In languages with large gender systems, evaluative meanings are (1) part of the semantic potential of very polysemous classes, and/or (2) associated with noun classes
that are only used to form diminutives and augmentatives. The latter type of evaluative
classes may exhibit peculiar morphosyntactic properties. For instance, in some Bantu
languages, they can be added to the default class marker of a noun (see §6.3.2). From
a diachronic point of view, the evaluative classes of the type described in (2) tend to be
less stable (see §6.3.3).
As shown in §6.4, in languages with sex-based gender, size-related gender shifts are
almost always restricted to inanimate entities: the polarity between masculine and feminine genders is variously exploited to encode polar notions within the domain of size
(‘big’ vs. ‘small’) . The only exception is Masai, where shifts from the Masculine to the
Feminine Gender also involve animate nouns. In Turkana and Karamojong, which have
a tripartite gender system, the Neuter Gender is used to encode young age of animate
entities and small size of inanimate.
Animacy-driven restrictions on the encoding of evaluative meanings by means of gender shifts are only attested in languages with sex-based gender, where sex and size-related
gender shifts are realised by the same markers. In the case of animate nouns, speakers
seems to conceive sexual dimorphism as the most salient semantic property and any
size-related gender shift is excluded. With inanimate nouns, the relevance of sexual
dimorphism is instead ruled out. One could say that sex and size-related gender shifts
are competing types of semantically driven gender assignment, insofar as they are both
determined by the semantic properties of nouns and/or the cultural representation of
the np referent.
It is noteworthy that diminutive and augmentative markers that are not related to
sex distinctions show rather strong preferences for highly animate referents. For instance, previous studies on the diachronic development of diminutives (see, among others, Grandi 2002, 2011;Jurafsky 1996) show that the first function of diminutive constructions that originates from the word for ‘child’ is to encode young age with nouns
denoting humans or higher animals. Only at the second stage of grammaticalization
do these constructions start marking small size with inanimate entities. Similarly, a
very common source for the grammaticalization of augmentative markers are the animate nouns ‘mother’ and ‘woman’ (Matisoff 1992), and a diachronic connection between
augmentative markers and markers of animacy has also been noted (Grandi 2002, 2011).
176
7 Gender and grammatical complexity
7.1 Introduction
The aim of this chapter is to propose a complexity metric for gender that can account
for the interactions between gender and number and gender and evaluative morphology
investigated in this thesis. The research question addressed in the chapter is:
(7.1) Research question concerning grammatical complexity (same as (2.15))
Q 9: How do interactions between gender and number and gender and evaluative
morphology affect the grammatical complexity of gender systems? Is it
possible to measure the role that these interactions play in the absolute
complexity of individual gender systems?
The chapter is structured as follows. In §7.2, I provide some background on the approach to grammatical complexity adopted in the dissertation and discuss the dimensions
of gender complexity suggested by Audring (2014). In §7.3, I discuss how the interactions between gender and number and gender and evaluative morphology considered in
this dissertation can be added to Audring’s criteria. In §7.4, I propose a method for
maximally local complexity measures. The complexity metric used in this thesis is presented in §7.5, whereas §7.6 describes the method followed to calculate the complexity
of the gender systems attested in the language sample. Results and discussion of the
results are found in §7.7, whereas a summary of the chapter is provided in §7.8.
7.2 Background
The notion of grammatical complexity has been introduced in §2.7 and discussed in
§5.4.1 in connection with my results on patterns of exponence and syncretism of gender
and number. As pointed out on those occasions, I interpret grammatical complexity in
its absolute sense, that is, as an objective property of a grammatical domain. Absolute
complexity can be measured either based on the number of parts that make up a system, or based on the length of the description of that system. In addition, measures of
absolute complexity can only be local, in the sense that they should address individual
grammatical domains rather than languages in their entirety. As also discussed in §2.7,
Miestamo (2008) and Sinnemäki (2011) suggest two major principles as general guidelines for elaborating complexity metrics for any grammatical domain: the Principle of
Fewer Distinctions and the Principle of One-Meaning–One-Form. The two principles are
repeated below in a simplified form:
177
7 Gender and grammatical complexity
The Principle of Fewer Distinctions. The higher the number of grammaticalized
distinctions, the more complex the domain.
The Principle of One-Meaning–One-Form. The least complex domain is the one
where there is a one-to-one mapping between meaning and form.
In §7.3, I argue that the two principles do not suffice to account for the whole range of
phenomena that can be relevant to the study of language complexity. By taking gender
as a case in point, I suggest that a third principle can be added. In addition, in §7.4, I
suggest how the three principles can be further specified in order to allow for maximally
local complexity measures.
In her pioneering paper on gender as a complex feature, Audring (2014) proposes three
criteria to assess the grammatical complexity of gender systems throughout the world’s
languages:
I The number of gender values
II The number, nature and scope of assignment rules
III The amount of formal marking (i.e., how much gender indexation there is in a
language).
The three dimensions are discussed by Audring by using a set of examples from different types of gender systems attested in different areas of the world (see §2.7 for a
detailed description of the methods and aims of Audring’s work on gender as a complex
feature).
Audring’s dimensions of gender complexity provide us with a baseline to calculate
the complexity of gender systems along the three most characteristic properties of this
grammatical domain: classification, assignment and indexation. Interactions of gender
and other domains of grammar are not in the scope of Audring’s investigation. In §7.3,
I suggest a way to include the patterns of interactions investigated in this dissertation
into a metric for gender complexity.
7.3 How to integrate interactions of gender into a metric for
gender complexity
In this section, I suggest that the absolute complexity of a grammatical domain is affected
by the number of interactions that it has with other domains in the following way: the
more interactive a grammatical domain, the higher its absolute complexity. In addition,
I suggest that this aspect of the absolute complexity of grammatical domains cannot be
handled by the Principle of Fewer Distinctions nor by the Principle of One-Meaning–
One-Form and rather calls for the introduction of a third principle.
The interactivity of a grammatical domain cannot be accounted for by the Principle
of Fewer Distinctions because it is not strictly concerned with the inventory of grammaticalized distinctions within an individual domain. Similarly, it cannot be accounted
178
7.3 How to integrate interactions of gender into a metric for gender complexity
for by the Principle of One-Meaning–One-Form because it is not strictly concerned with
the mapping between meaning and form. The interactivity of a grammatical domain
rather has to do with how isolated or independent a domain is from other domains. Its
relevance to language complexity can be assessed as follows:
The least complex grammatical domain is the one whose description is independent of semantic and functional properties of other domains. The more
interactive the domain, the less independent. The less independent the domain, the higher its absolute complexity.
I propose to refer to this principle as the Principle of Independence.
A short overview of the types of morphosyntactic and semantic interactions of gender
and number and gender and evaluative morphology explored so far, as well as their role
in the complexity of gender systems, is in place.
Concerning morphosyntactic interactions, in chapter 5, I examined morphosyntactic
interactions of gender and number by looking at exponence and syncretism. Even there,
I argued that cumulative exponence features higher grammatical complexity because it
makes the mapping between meaning and form less straightforward. Similarly, in chapter
6, I discussed morphosyntactic interactions of gender and evaluative morphology in the
languages where diminutives and augmentatives are part of the inventory of gender
distinctions. In principle, the presence of diminutive and augmentative genders can also
be seen as increasing the absolute complexity of gender systems. This is discussed in
detail in §7.5.
Concerning semantic interactions, the following phenomena were counted as instances
of semantic interaction between gender and number (§5.7), and gender and evaluative
morphology (chapter 6) throughout the thesis:
(1) Manipulations of gender assignment to express variation in the countability properties/number values of a noun (see §5.7).
(2) Manipulations of gender assignment to express variation in the size of the referent
or the speaker’s attitude towards the referent (see chapter 6).
The role that the two types of interactions play in the absolute complexity of gender
systems has not been discussed so far. As argued throughout this dissertation already,
gender assignment can be rigid (i.e., nonmanipulable) or manipulable. Manipulable
gender assignment is found in those languages in which nouns can be shifted from one
gender to another in order to encode a different construal of the np referent according
to patterns of the type described in (1) and (2). The possibility of manipulating gender
assignment may be seen as piling on top of the default gender assignment rules that
are productive in a language. In languages with manipulable gender assignment, gender
markers have default and add-on meanings. These add-on meanings are dependent on
semantic associations between gender and other grammatical domains, notably countability and evaluation. Thus, based on the Principle of Independence introduced above,
their presence features an increase in the absolute complexity of gender. Gender assignment is not only given in the lexicon for each and every noun, but it is also subject to
change depending on associations with other functional domains.
179
7 Gender and grammatical complexity
An alternative analysis would be to interpret gender systems with manipulable assignment as being semantically more predictable and thus less complex than gender systems
with rigid assignment. Nouns are assigned to different genders depending on the way the
np referent is construed (e.g., big vs. small or countable vs. uncountable), that is, based
on pragmatic grounds. An analysis of this type, however, would not account for the
fact that default gender assignment in languages with manipulable gender assignment
is often semantically arbitrary. For instance, as shown in chapter 6, in the Berber languages, inanimate nouns can be both masculine and feminine (see e.g., Tachawit where
aqnmuš ‘pot’ is masculine and taGnžakt ‘spoon’ is feminine). Only in case of gender
shift is gender assignment manipulated according to semantically transparent principles
of the type:
(1) Masculine Inanimate → Feminine = Diminutive
(2) Feminine Inanimate → Masculine = Augmentative
To summarise, manipulable gender assignment is an add-on device, which acts on the
default mechanisms of gender assignment according to semantically predictable patterns
that are dependent on the interactions of gender with other domains. The presence of
manipulable assignment increases the absolute complexity of gender systems.
7.4 How to elaborate maximally local complexity measures
As observed in §§2.7 and 7.2, Miestamo (2008) and Sinnemäki (2011) suggest that language complexity can only be measured on a local scale, by focussing on individual areas
of grammar. This locality constraint lies on two fundamental problems related to the
assessment of language complexity: representativity and comparability (see discussion in
Miestamo 2008). On the one hand, it is practically impossible to elaborate complexity
measures that can be fully representative of the grammar of a language (representativity). On the other hand, it is extremely difficult to make sure that the individual
grammatical phenomena within a complexity metric can be compared with each other
in such a way that their contribution to the global complexity of a language can be
assessed (comparability).
Following a suggestion by Dahl (2011: 156), in this section I propose that, in order
to guarantee maximal locality, the formulation of the three principles of absolute complexity – the Principle of Fewer Distinctions, the Principle of One-Meaning–One-Form
and the Principle of Independence – should be based on ceteris paribus comparisons. In
other words, the maximally local way to measure absolute complexity is by means of
comparisons of the type: “Everything else being equal, X is more complex than Y” or
“Everything else being equal, the addition of Z increases the complexity of X” (Östen
Dahl, personal communication). Only by introducing the “Everything else being equal”
condition can we be sure that the individual features of a complexity metric are truly
local in their scope.
180
7.5 A metric for grammatical gender complexity
When reinterpreted in terms of ceteris paribus comparisons, the Principle of Fewer
Distinctions, the Principle of One-Meaning–One-Form and the Principle of Independence
can be reformulated as described in (7.2).
(7.2) The principles of absolute complexity reformulated in terms of ceteris paribus
comparisons
Principle of Fewer Distinctions
Everything else being equal, a grammatical domain with n distinctions is
less complex than one with n+1 distinctions.
Principle of One-Meaning–One-Form
Everything else being equal, a grammatical entity with n forms is less
complex than one with n+1 forms.
Everything else being equal, a grammatical entity with n meanings is less
complex than one with n+1 meanings.
Principle of Independence
Everything else being equal, a grammatical domain that is independent of
semantic and functional properties of other domains, is less complex than a
grammatical domain that is dependent on n or n + 1 semantic and
functional properties of other grammatical domains.
In the rest of the chapter, the three principles are operationalised in order to elaborate
a complexity metric for gender systems. This is spelled out in §7.5.
7.5 A metric for grammatical gender complexity
As mentioned in §7.2, Audring (2014) formulates three dimensions along which to assess the absolute complexity of gender: classification (number of distinctions/classes),
assignment (number, nature and scope of assignment rules) and indexation (number
of indexing targets). Audring’s gender complexity dimensions are all included in my
metric. Based on what was discussed in §7.3, in order to account for morphosyntactic
and semantic interactions between gender and number and gender and evaluative morphology, two dimensions were added to my metric: morphosyntactic interactions and
semantic interactions (see §7.2).
The following features are proposed for each of the two dimensions:
• Morphosyntactic interactions
– Cumulative exponence of gender and number
• Semantic interactions
– Manipulation of gender assignment triggered by number/countability
181
7 Gender and grammatical complexity
– Manipulation of gender assignment triggered by size.
My suggestion is that the presence of each of these phenomena increases the absolute
complexity of a gender system.
One additional adjustment of Audring’s (2014) complexity criteria concerns assignment rules. According to Audring, gender assignment can be evaluated by looking at
(1) types (or nature) of assignment rules and (2) scope of assignment rules: (1) covers
whether gender assignment is semantic or is both semantic and formal; (2) measures
how general assignment rules are, that is, the number of nouns whose gender assignment
each rule is able to predict. Since it is rather difficult to evaluate the scope of gender
assignment rules based mostly on descriptive sources, I decided to restrict Audring’s
dimension II (assignment) to only the the nature of assignment rules.
In sum, the complexity metric for grammatical gender that I use in this dissertation
is an expanded, and slightly revised, version of the model developed by Audring (2014).
Table 7.1 provides a list of the features that are included in my metric and compares
them with those used by Audring (2014). The rightmost column of the table describes
how each feature is expected to affect the complexity of grammatical gender. The feature
values and their numerical interpretation are discussed in §7.6. Since the metric aims
at capturing how minimally and maximally complex gender systems can be, absence of
gender is not featured by any of the features in the metric. It follows that the metric
can only be applied to languages with gender.
It is worth mentioning that the existence of dedicated diminutive and augmentative
genders may be considered a type of morphosyntactic interaction between gender and
evaluative morphology. This aspect of the interaction between gender and evaluative
morphology is not counted as an independent feature of the metric because it is somewhat
redundant. As shown in chapter 6, languages with diminutive and augmentative genders
are languages with large gender systems where gender assignment can be manipulated to
encode size variation. Both properties are already accounted for by two other features of
the metric: (1) number of gender values (GV) and (2) manipulation of gender assignment
triggered by size (M2).
182
7.5 A metric for grammatical gender complexity
Table 7.1: Cues for assessing grammatical complexity of gender systems
Feature My features
ID
Audring’s (2014) Description
features
gv
Number of gender Number of gender Everything else being equal, a gender system
values
values
with two distinctions is less complex than a
gender system with three or more distinctions (Principle of Fewer Distinctions).
ar
Nature of assign- Number and nature Everything else being equal, a gender sysment rules
of assignment rules tem with one type of assignment rules – e.g.,
only semantic or only formal – is less complex than a gender system with two types of
assignment rules – both semantic and formal.
(Principle of Fewer Distinctions).49
ind
Number of indexing Amount of formal Everything else being equal, a gender system
targets
marking
with one indexing target is less complex than
a gender system with two or more indexing
targets (Principle of Fewer Distinctions).
cum
Cumulative expo- –
nence of gender
and number
Everything else being equal, a marker that
only signals gender is less complex than
a marker which signals gender + number
(Principle of One-Meaning–One-Form and
Principle of Independence).
m1
Manipulation
of –
gender assignment
triggered by number/countability
Everything else being equal, a gender system where gender assignment is only lexically given is less complex than a gender system where gender assignment is given in the
lexicon + can be manipulated depending on
the countability properties of the noun (Principle of Independence) or the np.
m2
Manipulation
of –
gender assignment
triggered by size
Everything else being equal, a gender system where gender assignment is only lexically given is less complex than a gender systems where gender assignment is given in the
lexicon + can be manipulated depending on
the size of the np referent (Principle of Independence).
49
As mentioned before, gender systems with only semantic assignment rules are quite common crosslinguistically, whereas gender systems with only formal assignment rules are almost never encountered.
183
7 Gender and grammatical complexity
7.6 Method
Having defined the features (or cues) for measuring the absolute complexity of grammatical gender (see table 7.1), the next step is to establish the values associated with
each feature and to convert them into numbers. This allows us to compute the absolute
complexity of the gender system of each of the languages with gender in the sample. To
achieve this purpose, I follow Parkvall (2008) who designed a method for computing the
language complexity of creoles and noncreole languages on the basis of a set of features
taken from the WALS database (Dryer & Haspelmath 2013). The values of each feature
are assigned a number between 0 and 1. Features with three values are converted into
the numerical format 0, 0.5, 1. Similarly, features with five values are converted by
Parkvall into the format 0, 0.25, 0.5, 0.75, 1. For all the features taken into account in
Parkvall’s paper, 0 stands for minimally complex and 1 for maximally complex. The total
complexity score for each language is divided by the number of features included for that
language. This is done in order to allow languages for which less information is available
on a given feature to get average scores comparable to those of the best documented
languages. The same procedure is followed in this thesis. In addition, values have been
set for each of the features in the metric. The feature values and correspondent numbers
are illustrated in table 7.2.
The composition of the metric is thus such that the least complex possible gender
system is the one that scores zero with respect to all the features of the metric and exhibit
the following properties: two gender values, semantic gender assignment, one indexing
target, no cumulation with number, no manipulation of gender assignment triggered
by number/countability and no manipulation of gender assignment triggered by size.
On the other hand, the most complex possible gender system is the one that scores 1
with respect to all the parameters considered in the metric and exhibit the following
properties: five or more genders, semantic and formal assignment, four or more indexing
targets, cumulation with number, and manipulation of gender assignment triggered by
both number/countability and size.
It is worth mentioning that the choice of the feature values assigned to IND relied
on mere (and potentially problematic) convenience choices. First, the metric accounts
for the number of indexing targets but does not allow us to measure whether different
indexing targets affect the complexity of gender in different ways. By using the feature
IND, a rough count of how pervasive gender indexation is in a language is obtained
but it is not possible to verify whether, for instance, “one indexing target” means “only
personal pronouns” or “only adjectives”, and whether this difference has relevant consequences for the overall complexity of gender. Second, indexes are identified on the
basis of distinguishable functions. Two functionally different indexes (e.g., determiners and demonstrative pronouns) can have the same formal realisation in one language.
However, the metric cannot account for the implications of these patterns of identity of
forms on the complexity of individual gender systems. These are important questions
for the understanding of how gender systems work, but, unfortunately, they cannot be
explicitly addressed by the metric used in this dissertation.
184
Table 7.2: Gender complexity metric
Feature
Number of gender values (gv)
Nature of assignment rules (ar)
Feature value
Score
Two genders
Three
Four
Five or more
0
1/3
2/3
1
Semantic assignment
0
Semantic and formal assignment 1
One
Two
Three
Four or more
0
1/3
2/3
1
Noncumulative
Partially cumulative
Cumulative
0
1/2
1
Manipulation of gender assignment triggered by number/countability (m1)
Absent
Present
0
1
Manipulation of gender assignment triggered by size (m2)
Absent
Present
0
1
Number of indexing targets (ind)
Cumulative exponence of gender and number (cum)
7.6 Method
185
7 Gender and grammatical complexity
Finally, as mentioned in §7.4, a general issue to be addressed is what Miestamo (2008)
refers to as the problem of comparability: to what extent are the individual features
of a complexity metric comparable with each other? Do the individual features in a
metric contribute in the same way to the final complexity score? The two questions are
addressed in detail in §7.7.4. Based on the individual complexity scores of the languages
of my sample, I compare the features of the metric by looking at the way they correlate
with each other and by discussing whether any of these features can be considered as
the best predictor of the final score.
Before presenting the results of my calculations, it is worth mentioning that, in case
of missing features, the index values resulting from the calculations should be taken
with caution. In fact, even though average scores (rather than total scores) are used as
index values, the index values of languages with missing features cannot be regarded as
entirely comparable to the index values of languages for which all features are equally
documented.
7.7 Results and discussion
The grammatical complexity of the gender systems attested in the 84 gendered languages
of the sample has been calculated following the method described in §7.6. The results
are presented in table 7.3. The table is divided in two macro-columns and languages
are arranged from the most to the least complex. The leftmost columns of each macrocolumn provide the rank: languages with the same average complexity score share the
same rank. Next to the rank come the language names and their ISO code; the average
complexity score assigned to the gender system of each language is given in the rightmost
columns of the two macro-columns. I refer to the score as the Gender Complexity Score
(henceforth GCS). The complexity scores for each of the feature values in the metric, as
well the GCSs, are given in appendix F.
The following observations can be made on the basis of the results presented in table
7.3:
1. Languages from the same genealogical units, or spoken within the same areas, tend
to have similar or even identical GCSs. In many cases, areal pressure seems to be
a relevant factor in explaining the distribution of the outliers.
2. If the languages with the highest GCS (= 1) are excluded, languages may display
the same index value but arrive to it on different paths. In other words, identical
GCSs do not stand for same type of gender system. In fact, identical GCS may
result from different combinations of values for each feature in the metric.
3. Implicational relations seem to exist between some of the features in the metric.
4. Some features in the metric correlate more with each other and seem to have a
stronger impact on the GCS than others.
These four points will be discussed in §§7.7.1, 7.7.2, 7.7.3 and 7.7.4, respectively.
186
7.7 Results and discussion
Table 7.3: GCSc of the languages of the sample
Rank Language
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
2
3
3
3
3
3
3
3
3
3
4
4
4
4
5
6
7
7
7
7
7
8
8
8
Bandial
Bemba
Bidyogo
Chiga
Kagulu
Kikuyu
Lega
Maasina Fulfulde
Mongo-Nkundu
Makaa
Ndengereko
Nyanja
Shona
Serer
Swahili
Timne
Tonga
Venda
!Xóô
Tunen
Eton
Maltese
Northern Sotho
SElEE
Swati
Tswana
Turkana
Wamey
Zulu
Bafia
Dibole
Kisi
Nuclear Wolof
Tachawit
Karamojong
Kabyle
Nafusi
Tamasheq, Kidal
Tamazight, Central
Zenaga
Amharic
Gola
Noon
Isocode GCS Rank Language
Isocode GCS
bqj
bem
bjg
cgg
kki
kik
lea
ffm
lol
mcp
ndg
nya
sna
srr
swa
tem
toi
ven
nmn
baz
eto
mlt
nso
snw
ssw
tsn
tuv
cou
zul
ksf
bvx
kss
wol
shy
kdj
kab
jbn
taq
tzm
zen
amh
gol
snf
bcq
hts
hau
ary
naq
nhr
sad
arb
tig
mkf
awn
lin
mdy
wal
gax
kqy
trg
ahg
hnh
bej
gid
mas
dsh
gdl
xuu
lln
rel
mdx
heb
tsb
irk
som
dim
bsw
ktz
ktb
dal
kwz
bip
pip
sur
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
0.94
0.83
0.83
0.83
0.83
0.83
0.83
0.83
0.83
0.83
0.78
0.78
0.78
0.78
0.75
0.72
0.69
0.69
0.69
0.69
0.69
0.67
0.67
0.67
9
9
9
9
9
9
9
9
9
10
11
11
11
11
12
12
12
12
12
13
13
13
14
14
14
14
14
15
15
15
16
16
17
18
18
18
19
20
21
22
23
Bench
Hadza
Hausa
Moroccan Arabic
Nama
Naro
Sandawe
Standard Arabic
Tigre
Miya
Awngi
Lingala, Kinshasa
Male
Wolaytta
Borana-Arsi-Guji Oromo
Koorete
Lishan Didan
Qimant
{Ani
Beja
Gidar
Masai
Daasanach
Dirasha
Kxoe
Lele
Rendille
Dizin
Hebrew
Tsamai
Iraqw
Somali
Dime
Baiso
Ju|’hoan
Kambaata
Dahalo
Kwadi
Bila
Pero
Mwaghavul
0.61
0.61
0.61
0.61
0.61
0.61
0.61
0.61
0.61
0.6
0.56
0.56
0.56
0.56
0.53
0.53
0.53
0.53
0.53
0.5
0.5
0.5
0.47
0.47
0.47
0.47
0.47
0.44
0.44
0.44
0.43
0.43
0.39
0.36
0.36
0.36
0.28
0.25
0.22
0.12
0.08
187
7 Gender and grammatical complexity
Table 7.3 shows that the highest GCS is 1 and the lowest 0.08. None of the languages
of my sample thus gets the lowest possible score, 0 (see §7.6).
The results given in table 7.3 are also displayed in the graph in figure 7.1.
0
5
10
15
20
Gender Complexity Scores
0.0
0.2
0.4
0.6
0.8
1.0
0.6
0.8
1.0
we
0.0
0.2
0.4
Figure 7.1: Distribution of the GCSs
The X-axis of the histogram displays the range of attested GCSs, whereas the Y-axis
shows the distribution of the number of languages per GCS score. The box plot below
the histogram provides the distribution of the GCSs per quartiles, and thus the boldface
line represents the median. The figure shows that half of the languages of my sample
have a GCS that ranges roughly from 0.5 to 0.8. In my data sample, high GCSs are
substantially more frequent than low GCSs. The geographical distribution of the GCSs
is represented in the map provided in figure 7.2.
188
7.7 Results and discussion
Figure 7.2: Geographical distribution of the GCSs
189
7 Gender and grammatical complexity
Before continuing with further discussion, an illustration of the procedure followed to
calculate the GCSs of two languages of my sample is given. For the sake of clarity, I
consider one language for which all features are documented, Turkana (Eastern Nilotic,
rank 3 in table 7.3), and one for which two features are missing, Timne (Mel, rank 1 in
table 7.3).
In order to classify the gender system of Turkana with respect to the features of my
complexity metric, I followed the description provided by Dimmendaal (1983). Turkana
has three gender values: Masculine, Feminine and Neuter. It thus gets 1/3 with respect
to the feature GV. Gender assignment is both semantic and formal, and, as such, the
value of AR is 1. According to Dimmendaal’s description, there are three indexing
targets for gender in Turkana: determiners, adjectives and pronouns (not the Personal
Pronouns). Thus the language gets 2/3 with respect to the feature IND. In Turkana,
gender distinctions are encoded cumulatively with number (CUM = 1). Finally, in
Turkana gender shifts can be used to encode variation both in the countability properties
of nouns (M1 = 1) or in the size of the np referent (M2 = 1). In Turkana, when an
uncountable masculine or feminine noun is shifted to the Neuter Gender, the resulting
meaning is singulative. On the other hand, when countable masculine or feminine nouns
are shifted to the Neuter Gender, the resulting meaning is diminutive (see §6.4 for more
details). To summarise, for Turkana, the values assigned to each of the metric features
are:
GV = 1/3; AR = 1; IND = 2/3; CUM = 1; SYNC = 1; M1 = 1; M2 =1
, the GCS of 0.83 is obtained.
Applying the formula illustrated in §7.6, 1/3+1+2/3+1+1+1
6
The gender system of Timne is classified based on its description by Wilson (1961).
Timne has more than five genders and thus gets 1 with respect to the feature GV. Gender assignment is both semantic and formal. Therefore, Timne gets a 1 with respect
to the feature AR. According to Wilson’s description, Timne has the following indexing targets for gender: adjectives, pronouns (of various type), verbs and the Indefinite
Stabilizer, which is used with indefinite nouns in order to encode non-verbal predication
(Wilson 1961: 11); the language thus gets a 1 with respect to IND. Gender and number
are encoded cumulatively on the indexing targets (CUM = 1). The source does not provide any kind of information about gender shifts, which are, however, rather common
phenomena in languages with similar gender systems. The features M1 and M2 cannot
be documented for Timne. To summarise, for Timne, the values assigned to each of the
metric features are:
GV = 1; AR = 1; IND = 1; CUM = 1; M1 = –; M2 = –
Since two features are missing, the sum of the feature values is in this case divided by
4: 1+1+1+1
. The GCS of Timne is thus 1.
4
7.7.1 Genealogical and areal biases in the distribution of GCSs
In general, the results presented in table 7.3 suggest that closely related languages tend
to have the same or very similar gender complexity scores. To give an example, all the
190
7.7 Results and discussion
Berber languages in the sample have a gender complexity score of 0.69, with the only
exception being Tachawit (shy) whose score is 0.75 (Tachawit scores 1 with respect to
IND, whereas all the other Berber languages of the samples score 2/3; see appendix
F). This tendency towards intragenealogical homogeneity in the absolute complexity of
grammatical gender is perhaps unsurprising, given that grammatical gender tends to
be a rather conservative feature of language families (see chapter 2). Outliers can be
explained as the effect of more language-specific patterns, such as contact with other
speech communities or language-internal historical developments. Let us consider a
couple of examples.
Out of 84 languages, 19 scored 1, with all these being either Bantu, North-Central
Atlantic or Mel. As observed throughout this thesis, typically, the gender systems of
the Bantu and Atlantic type (i.e., North-Central Atlantic + Mel) exhibit features of
high grammatical complexity: high number of gender distinctions, pervasive gender
indexation, manipulability of gender assignment to express variation in the countability
properties of nouns and/or in the size of the np referents. Those Atlantic and Bantu
languages which rank lower than 1 in table 7.3 have gender systems in which one or
more of the above-mentioned features has/have been either weakened or lost. Many
such cases of reduction and loss have already been presented in the previous chapters.
For instance, as discussed in detail in §6.3.3, in 8 of the 23 Bantu languages of my sample
– Bafia, Eton, Northern Sotho, Shona, Swati, Tswana, Venda, Zulu – diminutive and
augmentative suffixes have grammaticalized from nouns. Of these eight languages, only
Venda combines the use of the diminutive and augmentative suffixes with the uses of the
diminutive and augmentative genders. In the remaining seven languages, the evaluative
genders have been lost. As a result, the absolute complexity of the gender systems of
these languages has decreased.
Two outliers with respect to the Bantu and Atlantic type of gender system are the
Bantu languages Kinshasa Lingala and Bila. Kinshasa Lingala, the variety of Lingala
spoken in the capital city of the Democratic Republic of Congo, has a gender complexity
score of 0.56, that is, roughly 50% less than a Bantu language with a more prototypical
gender system. Kinshasa Lingala preserves a variety of gender markers on nouns, but
the only opposition that is marked on the indexing targets is animate vs. inanimate. In
addition, the number of gender-indexing targets is massively reduced: gender distinctions are only marked on verbs and pronouns, and adjectives almost never index gender
with nouns (Bokamba 1977: 70). Compared to Makanza Lingala, the standard variety, Kinshasa Lingala has undergone massive grammatical semplification. As Bokamba
(1977: 75) puts it, “[t]he background linguistic heterogeneity of KL [Kinshasa Lingala]
speakers certainly constitutes” one of the reasons behind this process of simplification
and reduction.
Bila has a GCS of 0.22, the lowest score in comparison with the other languages of
the Bantu group. Bila is spoken in the northeastern part of the Democratic Republic of
Congo, which is also the northernmost corner of the Bantu-speaking area. The gender
system of Bila is very simple with respect to nearly all the dimensions of the complexity
metric used in this thesis (table 7.1). There are only two genders, the Animate Gender
and the Inanimate Gender. Gender assignment is semantic, and there is no possibil-
191
7 Gender and grammatical complexity
ity of manipulating gender assignment. In addition, contrary to other Bantu languages
where verbs are one of the most common gender-indexing targets, gender indexation in
Bila is exclusively np-internal (following Lojenga 2003: 462, the indexing targets of Bila
are: adjectives, numerals and demonstratives). Such a simplified gender system is an
innovation shared with other northern Bantu languages and explained by bantuists as
a result of intense language contact with neighbouring non-Bantu languages (Lojenga
2003). The northern part of the Bantu-speaking area is often described as a true borderland between linguistically very diverse communities that have extensive contact with
each other. In this area, Bantu speakers are surrounded by speakers of Nilo-Saharan
and Ubangi (Niger-Congo) languages (Lojenga 2003: 451-452). Due to intense mutual
contact, both the Bantu and non-Bantu languages spoken in this area are characterised
by massive lexical borrowing as well as by grammatical innovations that are not shared
with the respective cognate languages outside the area. The reduced gender system of
Bila and other neighbouring Bantu languages is one of such innovative features.
The Semitic languages are an interesting case of closely related languages with nonhomogeneous complexity scores. The highest ranking gender systems within the Semitic
sample are found in Maltese (0.83) and Amharic (0.67). Moroccan Arabic, Standard
Arabic and Tigre have the same complexity score, 0.61. The lowest ranking gender
system is found in Hebrew (0.44), whereas Lishan Didan scored 0.53. Interestingly, the
highest GCS, 0.83, is scored by Maltese, the Semitic language that is most prominent in
its long-standing history of language contact with English and some Romance languages
(Italian and Sicilian).
Two additional examples of outliers are Dahalo, with respect to the other Cushitic
languages, and Kwadi, with respect to the Khoe-Kwadi group. Dahalo has a GCS of 0.28,
and its gender system has been described by Tosco (1991: 20) as dying out as a result of
contact with the neighbouring Bantu languages. Too little is known about Kwadi, a now
extinct language of Angola. Güldemann (2004) describes its gender system as sex-based
and pronominal, but nothing is said about the mechanisms of gender assignment nor
about the use of gender shifts to encode diminutive and augmentative meanings (which
is well documented in all the other Khoe-Kwadi languages of the sample).
Finally, the two lowest ranking languages in the complexity rank given in table 7.3 are
the Chadic languages Mwaghavul (GCS = 0.08) and Pero (GCS = 0.12), both of which
are spoken in Nigeria.
Mwaghavul scores 0 with respect to all the features of the complexity metric except
for CUM, for which the score is 0.5. There are two genders in Mwaghavul (Masculine
and Feminine), gender assignment is semantic and gender indexation is only pronominal. Finally, there seems to be no possibility of manipulating gender assignment in the
language. With respect to the cumulation parameter, Mwaghavul shows at least some
patterns of interaction with number on the indexing targets. The Third Person Human
Anaphoric Subject and Object Pronouns encode gender and number cumulatively. On
the other hand, the Third Person Non-human Pronoun, n¯@, encodes neither gender nor
number distinctions (Frajzyngier & Johnston 2005). A similar type of system is found in
Pero even though, from the description provided by Frajzyngier (1989), it is not entirely
clear what type of assignment rules the language has and whether or not gender assign-
192
7.7 Results and discussion
ment can be manipulated in any of the ways considered in this thesis. The remaining
four Chadic languages with gender in the sample have higher GCSs (between 0.62 – Lele
– and 0.5 – Gidar).
To summarise, genealogically related languages tend to have similar GCSs. On the
other hand, multilingualism, language contact and second language learning seem to
be very likely triggers of changes in the gender system of a language as opposed to its
closest relatives (see, among others, Trudgill 1999; McWhorter 2001), and are also rather
ordinary sociolinguistic settings in the African macro-area. A systematic account of the
effects of language contact on the absolute complexity of gender in the languages of the
sample lies, however, outside the scope of this dissertation.
7.7.2 Same complexity score does not mean same type of gender system
As mentioned at the beginning of the section, it is important to emphasise that, with the
exception of the languages with the highest complexity score (= 1), identical complexity scores do not mean that languages have the same type of gender system. Different
gender systems may arrive to the same complexity score via different paths. These differences become clear when considering how the individual languages score with respect
to the individual features of the complexity metric (see appendix F). For instance, let
us consider the case of Noon (North-Central Atlantic, rank = 8) and Amharic (Semitic,
rank = 8). The two languages have the same GCS, 0.67, but have very different types
of gender systems. Noon has a non-sex-based gender with more than five gender distinctions, whereas Amharic has a sex-based gender system with two gender distinctions,
Masculine vs. Feminine. Thus, Noon scores 1 with respect to GV, whereas Amharic
scores 0. The situation is reversed with respect to IND: gender indexation is more pervasive in Amharic (1) than in Noon (0). The two languages share the same scores with
respect to the remaining features of the metric.
7.7.3 Implicational relationships between the features in the metric
On the basis of the results presented in table 7.3 an interesting relationship can be
observed between the features GV and AR, and AR and IND.
Strictly semantic systems of gender assignment are only found in 6 of the 84 gendered
languages within the sample: Bila (Bantu), Dahalo (Cushitic), Dime (South Omotic),
Dizin (Dizoid), Masai (Eastern Nilotic), Mwaghavul (Chadic) (see also section 4.2). All
these languages have two gender distinctions, and all but Bila have sex-based gender.
Within my language sample then, strict semantic gender assignment is only found in
languages with two or a maximum of three gender values. Moreover, there seems to be
a preference for strictly semantic gender assignment in African languages to be based on
cognitively basic oppositions such as human vs. non-human, male vs. female, animate
vs. inanimate. It would be interesting to investigate what type of preferences exist,
if they exist, in areas of the world where strictly semantic gender assignment is more
common.
Finally, it is worth mentioning that the six languages of my sample with strictly
193
7 Gender and grammatical complexity
semantic gender assignment all score less than 1 with respect to IND: thus in none of these
languages is gender indexation maximally pervasive. These results are in line with what
was suggested by Audring (2009) with respect to the correlation between pervasiveness
of indexation and type of assignment rules. Audring analyses the assignment rules of a
number of pronominal gender systems from different areas of the world, and considers
aspects of the diachrony of gender in English and Dutch. She shows that pronominal
gender systems – where manifestations of gender throughout the discourse are rather
poor – display a strong preference towards strictly semantic assignment rules. Within
my language sample, only Mwaghavul (Chadic) has pronominal gender and semantic
assignment. The remaining five languages with strict semantic assignment score either
1/3 or 2/3 with respect to IND. These results suggest that when strict semantic gender
assignment is found in non-pronominal gender systems, gender indexation is still not
maximally pervasive. In other words, semantic assignment seems to generally tolerate
lower amount of formal marking.
7.7.4 Some features may be stronger predictors of gender complexity than
others
In several occasions throughout this chapter, I mentioned that a major issue when investigating grammatical complexity is how to quantify the contribution that the individual
features of a metric bring to the overall complexity score (what Miestamo 2008 refers
to as the problem of comparability). Given that it is extremely difficult to measure the
relative weight of the individual features of a complexity metric, overall complexity results can be extracted based on a language’s average behaviour with respect to a set of
features. These results describe tendencies in the absolute complexity of a grammatical
domain with respect to an artificially designed set of cues. Therefore, they cannot be
interpreted as uncontroversial measurements (for a similar discussion in their study of
complexity in nominal plural allomorphy, see also Dammel & Kürschner 2008).
I would like to suggest here that one way of indirectly investigating the behaviour of
the features of a complexity metric is to correlate the individual features with each other.
In order to do so with my own complexity metric, I calculated the Squared Spaerman
rank correlation coefficients between the individual features of the metric. The results
are represented in the graph in figure 7.3.
Figure 7.3 is organised as follows. The individual features of the metric are displayed
both horizontally and vertically. In this way, correlations coefficients between pairs
of features can be read both row-wise and column-wise. Correlation coefficients are
visualised according to a colour scale whereby white stands for no correlation and grey
for high correlation. The grey diagonal area that cuts across the two halves of the figure
represents correlation coefficients between pairs of the same features (that is, CUM with
CUM, M2 with M2, etc.). These grey boxes correspond to a correlation coefficient that
equals to 1 since each feature obviously has the highest correlation with its own copy.
These results are thus not relevant to our analysis. With respect to correlations between
pairs of different features, the figure shows that the highest correlation coefficients are
found between IND and M1 (= 0.372), GV and M1 (= 0.324), and GV and IND (=
194
7.7 Results and discussion
M1
0.001
0.005
0.068
0.324
IND
0.025
0.024
0.036
0.232
GV
0.133
0.001
0.074
0.372
0.372
R2
0.232
0.324
Feature
0.3
0.2
0.1
AR
0.001
M2
0
CUM
CUM
0.02
0.074
0.036
0.068
0.02
0.001
0.024
0.005
0
0.001
0.133
0.025
0.001
M2
AR
GV
IND
M1
Feature
Figure 7.3: Correlation coefficients between the features of the metric
195
7 Gender and grammatical complexity
0.232).
The high correlation coefficients between IND and M1, on the one hand, and GV and
M1, on the other, can be interpreted as follows. In the languages of my sample, the
possibility of manipulating gender assignment to encode variation in the countability
properties of nouns goes hand in hand with the presence of very pervasive gender indexation or, to a slightly lower degree, high number of gender values. As shown in section
5.7, M1 is not widely distributed across the gender systems of the language sample. It is
only found in Bantu, North-Central Atlantic, Berber, a subset of the Semitic languages,
and in the Eastern Nilotic language Turkana. In a way then, both the distribution of
M1 and its correlation coefficients with IND and GV suggest that M1 is a very special
property of gender systems, which can only be found in systems with a high amount of
formal marking (IND) and/or a high number of gender distinctions (GV). On the contrary, the results show that M2, that is, manipulation of gender assignment to express
diminutive and augmentative meanings, has extremely low correlation coefficients with
both IND and GV as well as with all the other features of the metrics.
As mentioned above, the correlation coefficient between GV and IND is 0.232. The
two features correlate with each other but not as strongly as one would expect based
on Audring’s (2014) argument, whereby pervasive indexation is likely to be found in
languages with a high number of gender values (see §7.2).
Moreover, figure 7.3 shows that AR (Assignment Rules) has extremely low correlation
coefficients with all the features of the metric. These results might depend on the
fact that only 6 of the 84 languages with gender in my sample have semantic gender
assignment. In other words, nearly all the languages of the sample behave similarly with
respect to this parameter. It would be interesting to investigate the behaviour of this
feature in areas of the world where semantic gender assignment is more frequent and
compare it with my results from Africa. Finally, equally low correlations are found with
the feature CUM.
One question that is worth asking is whether the correlation coefficients presented in
figure 7.3 can tell us anything about which of these features is the best predictor of the
GCS of each language. Since the GCS is the averaged sum of the values that a language
takes for each feature in the metric, the features that show the highest correlations with
each other (M1, IND and GV) can be expected to be those which also have a stronger
impact on the final score. This can be verified by examining the associations between
the independent variables (the features in the metric) and the dependent variable (the
GCS) in a purely descriptive way, that is, by stratifying our dependent variable, the
GCSs, according to the potential predictors, the individual features in the metric (Harrell
2001: 125). This is shown in figure 7.4.
196
7.7 Results and discussion
GCSs stratified by features and feature values
N
GV
0
0.333333333
0.666666667
1
43
6
1
34
AR
6
76
2
No
Yes
Missing
IND
0
0.333333333
0.666666667
1
Missing
5
16
28
33
2
CUM
0
0.5
1
2
15
67
M1
No
Yes
Missing
45
35
4
M2
20
48
16
No
Yes
Missing
Overall
84
0.2
0.4
0.6
0.8
Average
N=84
Figure 7.4: GCSs (Average) stratified according to feature values
Figure 7.4 is organised as follows. The GCSs are displayed on the X-axis. The left
Y-axis represents the values assigned to each feature in the metric; the right Y-axis
shows the number of languages in the sample where each of the feature values is found.
The black dots represent the mean of the GCSs that languages displaying a certain
feature value have. For instance, it shows that languages that score 1/3 (0.3333333333)
with respect to GV have a GCS which, on average, ranges between 0.6 and 0.8. The
black dots thus allow us to see which of the features and feature values can trigger the
highest GCSs in the languages of the sample. As hypothesised based on the correlation
coefficients shown in figure 7.3, in the languages of my sample, the highest scores in GV,
IND and M1, trigger higher GCSs. With respect to GV, the figure shows that the impact
of the different feature values on the GCSs grows from 0 to 1/3, drastically drops at 2/3
and grows again at 1. This is likely to be an effect of the fact that only one language
within my sample has four gender distinctions, Ju}’Hoan (Kxa). As discussed in §7.7,
Ju}’Hoan has a GCS of 0.36, which is one of the lowest scores in my language sample.
197
7 Gender and grammatical complexity
To summarise, even though the quantitative analysis applied to the complexity data
does not provide a solution to the problem of comparability, it provides good tools for
describing the behaviour of the complexity metric with respect to the dataset considered
in this dissertation. Provided that my metric is a good measure for gender complexity,
the results suggest that M1, IND and GV are the features which correlate more strongly
with each other and those which seem to have the stronger impact on the final complexity
scores of the languages of my sample.
7.8 Summary of the chapter
The purpose of this chapter has been to elaborate a metric for the grammatical complexity of gender systems that could also account for the interactions that gender has
with the domains of nominal number and evaluative morphology.
The notion of grammatical complexity that I have worked with in this chapter is of
the absolute type (see §7.2). I argued that the absolute complexity of a grammatical
domain can be assessed by using three principles as major guidelines: the Principle of
Fewer Distinctions, the Principle of One-Meaning–One-Form and the Principle of Independence. While the first two principles are based on Miestamo (2008) and Sinnemäki
(2011), the Principle of Independence is my own and is used as a means of assessing
interactions between grammatical domains and their contribution to the complexity of
individual domains. Following Dahl (2011), I also argued that, in order to be maximally
local, complexity metrics ought to be based on ceteris paribus comparisons.
The bulk of my complexity metric is an expanded version of the criteria for gender
complexity suggested by Audring (2014). In order to translate the values of each features
into numbers, I followed the method designed by Parkvall (2008). The feature values,
their numerical interpretation and the methodology followed in order to compute the
grammatical gender complexity of the languages of the sample have been presented in
§7.6.
In §7.7.1, I showed that the genealogical biases within the sample are reflected by
the results on gender complexity: languages belonging to the same genealogical unit
tend to have similar complexity scores. The distribution of the outliers can be usually
explained as the effect of language contact. In §7.7.2, I showed that languages with the
same complexity score do not necessarily have the same type of gender systems. Possible
implicational relationships between some of the features in the metric were discussed in
§7.7.3. Finally, in §7.7.4, I showed that some of the features in the metric correlate more
with each other than others and can be seen as better predictors of the final complexity
score. As shown in figure 7.1, when they have gender, the languages of my sample tend
to have rather complex gender systems, whereas gender systems with lower levels of
complexity are very rare. If this sample is considered to be representative of the African
macro-area, these results suggest that gender is a complex feature of African languages.
198
8 Summary of the main results, prospects
for future research and concluding
remarks
In this chapter, I summarise the results of my investigation and assess their contribution
to the understanding of gender, viewed in its interactions with number and evaluative
morphology. The methodology followed in the study, and its relevance to linguistic
typology are also discussed. Shortcomings and limitations of the work are pointed out
together with suggestions for future research on the topic.
8.1 Summary and assessment of results
Three research foci were established in chapter 2: (1) to investigate the interactions
between gender and number; (2) to investigate the interaction between gender and evaluative morphology; and (3) to measure the impact of these interactions on the absolute
complexity of gender systems. These three research foci were tackled individually in
three independent chapters: chapters 5, 6 and 7, respectively. The research questions
addressed in each of the three chapters are here repeated for convenience. The results
and findings of each chapter are summarised and assessed thereafter.
(8.1) Research questions concerning gender and number (chapter 5)
Q 1: How common is cumulative exponence of gender and number in the
languages of the sample?
Q 2: What are the formal and semantic factors that trigger gender syncretism in
the context of number? Does gender syncretism in the context of number
presuppose cumulative exponence?
Q 3: What are the implications of cumulative exponence and syncretism on the
absolute complexity of gender and number systems?
Q 4: Can these types of interaction between gender and number be seen as a
reflex of a nominal relevance hierarchy?
Q 5: Can gender and number compete through indexation?
Q 6: Is there any correlation between types of encoding of gender and types of
encoding of number?
Q 7: What types of semantic interactions can be found between gender and
number?
199
8 Summary of the main results, prospects for future research and concluding remarks
(8.2) Research questions concerning gender and evaluative morphology (chapter 6)
Q 8: How frequently does size occur as an independent gender value? How stable
and how widely distributed is this phenomenon within genenealogical units?
Q 9: Do the interactions between gender and evaluative morphology differ across
types of gender systems and/or strategies of gender assignment (e.g.,
sex-based vs. non-sex-based gender systems, or manipulable vs. rigid gender
assignment)?
(8.3) Research question concerning gender complexity (chapter 7)
Q 10: How do interactions between gender and number and gender and evaluative
morphology affect the grammatical complexity of gender systems? Is it
possible to measure the effect of these interactions on the absolute
complexity of individual gender systems?
8.1.1 Gender and number
In chapter 5, morphosyntactic and semantic interactions between gender and number
were considered. The chapter was divided in four parts.
In the first part of the chapter (§§5.2, 5.3 and 5.4), I examined patterns of exponence of
gender and number as well as syncretism of gender in the context of number. Exponence
and syncretism are often regarded as two crucial dimensions along which gender and
number interact in language. However, to my knowledge, this dissertation offers the first
systematic survey of the patterns of exponence and syncretism of gender and number
that are attested in a large sample of languages. Cumulative exponence of gender and
number on the indexing targets turned out to be extremely frequent in the languages
of the sample as did syncretism. Since not all languages with cumulative exponence
have syncretism, but nearly all languages with syncretism have cumulative exponence,
I suggested that syncretism of gender in the context of number presupposes cumulative
encoding of the values of the two grammatical domains. I suggested that cumulation
should be viewed as a factor that increases the absolute complexity of gender systems
insofar as it introduces noise in the mapping between meaning and form. Finally, I argued
that my results on cumulation and syncretism may be read as possible indicators of the
existence of a relevance hierarchy for nominal features. In agreement with Greenberg
(1963b), Carstairs (1987) and Carstairs & Stemberger (1988), and Vafaeian (2013) –
each of whom worked on different grammatical phenomena and different datasets –
these results suggest that if a nominal relevance hierarchy existed, number would be the
highest ranking feature.
In the second part of the chapter (§5.5), I focussed on the existence of competing
patterns of gender and number indexation in a rather small subset of the language
sample. These languages all belong to the Afro-Asiatic phylum and display a bipartite
sex-based gender system. In addition, in these languages, there is a split in the indexation
patterns associated with plural nouns. Some plural nouns trigger the same indexation as
either masculine singular or feminine singular nouns, whereas other plural nouns trigger
200
8.1 Summary and assessment of results
an indexation pattern that differs both from the masculine and the feminine singular.
Since the latter indexation pattern is only used to index plurality, I proposed to refer to it
as Dedicated Plural Indexation (DPI). Furthermore, I suggested to label languages of this
type as languages with split plural indexation systems. I showed that the use of DPI may
be constrained either by the animacy (Standard Arabic, Miya) or by the lexical plurality
of the head nouns (Cushitic). A case study of the Cushitic languages of the sample
was carried out in order to examine how different languages within the group behave
with respect to the use of DPI. In the literature on Cushitic, languages with a rather
limited use of DPI are traditionally described as displaying a third indexation class,
and thus a third gender, besides masculine and feminine: the plural (for an overview,
see Mous 2008). This approach has been thoroughly criticised in more typologically
oriented literature (see, for instance, Corbett & Hayward 1987; Corbett 2000). In these
studies, however, the use of DPI has often been labelled as a mere exception without any
explanation of its idiosyncrasies being attempted (a recent exception is Corbett 2012,
who, for the first time, attempts to explain gender and number indexation in Baiso
by shifting the attention to the role of lexical plurality). In my own study, I showed
that an alternative overarching explanation for this long-debated issue becomes possible
when looking at the distribution of DPI across different languages of the family and
examining the factors that play a role in constraining its occurrence. In the Cushitic
languages with split plural indexation, DPI is only legitimate with those indexation
triggers that are inherently associated with plurality. This does not necessarily mean
that DPI is the syntactic manifestation of a gender. It simply shows that two domains,
one grammatically bound to indexation (gender) and the other potentially encodable
through indexation (number) may compete with each other on (at least) some indexing
targets. I showed that the division of labour between the two is done at different cutoff points in different Cushitic languages. In most languages nominal plurality (be it
morphological or lexical) is always indexed syntactically, in other languages only lexical
plurality is indexed.
In the third part of the chapter (§5.6), I showed that in the languages of the sample,
the development of pervasive indexation systems always involves the presence of both
gender and number and that the languages of the sample tend to exhibit the same
number of gender- and number-indexing targets.
In the fourth and last part of the chapter (§5.7), I showed that in languages with
manipulable gender assignment, that is, in languages where nouns can be assigned to
multiple genders, gender shifts can be used to encode variation in the countability property of nouns. This phenomenon has a very skewed distribution in the languages of the
sample.
8.1.2 Gender and evaluative morphology
In chapter 6, morphosyntactic and semantic interactions between gender and the morphological encoding of evaluation were considered. Through this chapter, this dissertation provides the first extensive typological account of how the encoding of size can be
incorporated in the gender system of a language.
201
8 Summary of the main results, prospects for future research and concluding remarks
Interactions between gender and evaluative morphology were explored taking different
types of gender systems as the independent variable. I found that, in the African macroarea, languages with large gender systems tend to have dedicated evaluative genders
that are used to encode diminutive and augmentative meanings. On the other hand, in
languages with smaller and sex-based gender systems, there are no dedicated evaluative
genders but the masculine, feminine or neuter genders, which may be used to encode
diminutive and augmentative meanings when a noun is shifted from its default gender
to (one of) the other(s). The two types are widespread throughout my language sample
and across genealogical groupings. In some cases, it was possible to assess their stability
as well as aspects of their diachronic development.
Ultimately, what the two types of systems have in common is the possibility to manipulate gender assignment in order to modify the construal of the np referent with respect
to the parameter of size.
8.1.3 Interactions of gender and grammatical complexity
The aim of chapter 7 was to measure the impact that the interactions between gender
and number and gender and evaluative morphology have on the absolute complexity of
a gender system. The suggestion made in the chapter is that the more interactive a
grammatical domain, the higher its absolute complexity.
Elaborating metrics to assess the absolute complexity of individual grammatical domains is a relatively novel field within language typology. In the case of gender, this has
been discussed in pioneering work by Audring (2014). The paper proposes three dimensions along which gender complexity can be computed but does not provide a method to
convert the variables associated to these dimensions into measurable values. In chapter
7, Audring’s model of gender complexity was expanded with two new aims:
(1) Accounting for the morphosyntactic and semantic interactions between gender and
number and gender and evaluative morphology explored throughout the thesis.
(2) Converting the values associated to each dimension of Audring’s model into numbers.
The second aim was achieved by using the method designed by Parkvall (2008). The
results of the calculations suggest that: (1) complexity scores are genealogically biased;
(2) languages with the same complexity score may arrive to it through different paths;
(3) some of the features in the metric interact with each other implicationally (e.g., strict
semantic gender assignment is only found in languages with small gender systems); and
(4) some of the features in the metric correlate with each other and seem to have a
higher impact on the final complexity scores of the languages of the sample.
8.2 Assessment of the sampling methodology
The sampling methodology followed in the dissertation was introduced in chapter 3,
where I discussed general matters concerning sampling in linguistic typology, my own
202
8.2 Assessment of the sampling methodology
sampling design and the procedure of data collection and organisation.
Methodologically, this dissertation was designed as an investigation of one continental
area, Africa, and as an attempt at practising crosslinguistic comparison both within and
across genealogical units. The choice of combining intra- and intergenealogical typology
in the study of one of the world’s macro-areas proved to be beneficial to answer the
research questions posited by this investigation.
For instance, having subsamples of individual genealogical units within Africa was
particularly crucial for reaching a better understanding of the following phenomena:
(1) The indexation patterns associated with plural nouns in Cushitic languages. Had I
not had a relatively extensive sample of the languages of the group, I could not have
explored the distribution of DPI throughout different members of the family. This
ultimately led me to propose an interpretation of the phenomenon that, hopefully,
sheds new light on our understanding of these languages and, more generally, on
the typology of indexation systems.
(2) The distribution of the evaluative genders in the Bantu languages. Had I not had
a relatively extensive sample of the languages of the group, I could not have shown
how common and stable these genders are throughout the family and what types
of innovations are attested in different subdivisions of the group.
On the other hand, intergenealogical comparison was essential for achieving an understanding of the distribution of the detected patterns across genealogical units. For
instance, I showed that cumulation of gender and number and syncretism of gender
in the context of number are very common throughout the languages of the sample;
the distribution of these phenomena cuts across genealogical groupings and typological
differences among the gender and number systems of the individual languages of the
sample. In addition, I showed that there is a correlation between type of gender system
and expected type of interaction with evaluative morphology and that this correlation
holds true across the different genealogical units represented in the language sample.
Large-scale genealogical biases in the distribution of individual phenomena were also
found. For instance, split plural indexation systems were found in all the genealogical
units of the Afro-Asiatic phylum included in the sample but Berber (that is, in Chadic,
Cushitic and Semitic).
Finally, in some cases, sub-areal phenomena could be detected. For instance, innovations in the gender systems of the Bantu languages were found to be distributed
according to sub-areas. The renewal of evaluative morphology – from evaluative genders
to evaluative suffixes – is concentrated in the southeastern Bantu languages (chapter 6),
whereas substantial reduction of the noun class system is mostly found in the northern
part of the Bantu-speaking area.
An intra- and intergenealogical survey of one macro-area, as the one conducted in this
thesis, cannot lead to the formulation of large-scale typological generalisations. On the
other hand, a continent-based sampling methodology provides an understanding of the
geographical distribution and diachronic stability of grammatical phenomena within restricted areas of the world that cannot be easily reached by means of large-scale language
203
8 Summary of the main results, prospects for future research and concluding remarks
comparison. The generalisations that result from continent-based typologies can only
hold at the areal level but constitute a set of assumptions that may be in turn tested on
other macro-areas of the world.
8.3 Prospects for future research
In this section, I discuss four domains of research for which more work is needed: nominal
relevance hierarchy, split plural indexation systems, gender assignment and its manipulation, and absolute complexity of gender systems. For each of the four domains, the
limitations of my own results are discussed before I provide some suggestions for further
research on the topic.
8.3.1 Nominal relevance hierarchy
On more than one occasion throughout this thesis, it was pointed out that number is
the feature of nominal morphology most likely to serve as the context of syncretism, but
which is less subject to undergo syncretism itself. In other words, certain distinctions
which are grammatically salient for nouns tend to be reduced or lost under the pressure of
number values. For instance, when nouns are marked as nonsingular, other grammatical
distinctions (such as case or gender) can be reduced or neutralized. This suggests that
nominal number has more relevance to nouns than other nominal features. However, as
Vafaeian (2013: 122) puts it, “[a] relevance hierarchy for nouns is yet to be proposed and
must be based on semantic arguments and confirmed by typological data in the same
way as was done for verbs.”
In my opinion, further research on the nominal relevance hierarchy should be carried
out on the basis of a clear-cut definition of the domains under investigation. In particular, when collecting and analysing the typological data, it may be useful to distinguish
between two major domains:
(1) Patterns of encoding of nominal features on the noun stem.
(2) Patterns of encoding of nominal features via indexation.
A systematic scrutiny of these two domains may provide different insights into the way
in which the semantic relevance of nominal categories is reflected in morphosyntax. For
instance, as shown in this thesis, since manifestations of gender are bound to indexation,
gender syncretism in the context of number can only be investigated by looking at how
the gender and number of a noun are encoded on the indexing targets. On the other hand,
as shown by Vafaeian (2013), the analysis of suppletive noun stems provides insights on
the hierarchical relationships that exist between other sets of nominal categories, such
as case, possession and, again, number. Methodologically, distinguishing between the
domains of investigation outlined in (1) and (2) could be a useful way of classifying the
typological data that may account for the existence of the nominal relevance hierarchy.
204
8.3 Prospects for future research
8.3.2 Split plural indexation systems and Dedicated Plural Indexation
The notions of split plural indexation and Dedicated Plural Indexation have proved to
be very useful in order to account for the indexation patterns associated with plural
nouns in a small subset of Afro-Asiatic languages within my sample. In addition, I was
able to show that within Cushitic languages, split plural indexation systems, where the
use of DPI is restricted to some plural triggers only, coexist with systems in which the
use of DPI is generalised to all plural nouns.
What my case study of Cushitic was not able to provide is a diachronic interpretation
of the data. What is the diachronic relationship between systems with generalised use
of DPI, systems with split plural indexation and systems where there is almost no use
of DPI? Are systems with generalised use of DPI an innovation or a retention with
respect to the two other types of systems? In order to answer these questions, a more
extensive investigation of the Cushitic languages would be needed. The sample should
be expanded, and, possibly, data from descriptive sources should be combined with
data from spoken corpora. This would allow us to verify whether the parameters that,
according to my findings, condition the use of DPI in the split plural indexation systems
of my sample also account for Cushitic languages outside the sample. Once a more
precise classification of the types of indexation systems synchronically attested within
Cushitic is reached, it may be easier to analyse their distribution from a historical point
of view and attempt at accounting for their diachronic development.
At a more general level, the validity of the two notions – split plural indexation and
DPI – should be tested outside the African macro-area. In particular, it might be useful
to verify whether the following two generalisations hold outside Africa:
(1) That split plural indexation systems are only attested in languages with sex-based
gender.
(2) That the distribution of DPI within split plural indexation systems is only constrained by the animacy and the lexical plurality of the indexation triggers.
8.3.3 Gender assignment and its manipulation
In §5.7 and chapter 6, I introduced the notion of manipulable gender assignment, and
I used it to refer to those languages where the gender of a noun can change according
to the way the np is construed. In such languages, there usually are default assignment
rules, whereby nouns have lexically specified genders, and add-on assignment rules which
allow speakers to modify the construal of the np referent by changing its gender. This
is mainly done in order to express variation in the size of the noun referent or in the
speakers’ attitude towards it. In §5.7, I showed that there are languages in which similar
mechanisms of gender shift seem to be used to encode variation in the countability
properties of nouns. This phenomenon is attested both in small and large, sex-based
and non-sex-based gender systems but is generally not very frequent across the different
genealogical groupings within the sample. Its distribution is rather concentrated within
a few genealogical groupings.
205
8 Summary of the main results, prospects for future research and concluding remarks
If both my analyses are correct, the notion of manipulable gender assignment seem to
challenge our current understanding of gender systems and, in particular, the assumption
that gender is always rigidly specified in the lexicon as an inherent property of a noun.
Further research in this field should be both empirical and theoretical in its aims. Empirically, the crosslinguistic distribution of manipulable gender assignment outside the
African macro-area should be explored in a systematic fashion. The following research
questions could be addressed:
(1) How common is manipulable gender assignment in the gender systems of the world
languages?
(2) What meanings are expressed by gender shifts crosslinguistically?
(3) In languages with manipulable gender assignment, is it always the case that nouns
have a default, lexical gender?
Theoretically, in my view, at least two questions are worth being asked:
(1) How do manipulable gender assignment systems challenge our model of gender
assignment?
(2) How can manipulable gender assignment be distinguished from mere word-formation
processes where derived nouns differ in gender from their base?
The theoretical question outlined in (1) has been rarely addressed in the literature
and, mostly, in connection with the gender system of individual languages. For instance,
Payne (1998) and, later, Shirtz & Payne (2012) point out that in the Eastern Nilotic
language Masai most nouns do not have a default gender but are potentially compatible
with any gender (see also the discussion in §6.4.1). Accordingly, they propose that
gender assignment in Masai is based on the construal of the np referent rather than
on the semantics of the noun or its lexical properties. The majority of the languages
with manipulable gender assignment in my sample display gender systems that are less
atypical than the Masai system insofar as the majority of nouns do have default, albeit
manipulable, gender. Yet, the results of the present investigation show that manipulable
gender assignment is a frequent enough phenomenon to start considering a revision of our
current understanding of gender assignment rules, their implications for the typology of
gender systems, and, on a larger scale, their interaction with word-formation processes.
8.3.4 Absolute complexity of gender systems
Research on the grammatical complexity of gender systems, and how to compute it, is
a relatively novel field within typology, and much still needs to be done.
As pointed out in §7.6, one of the shortcomings of my complexity metric is not to be
able to provide a qualitative account of how different types of indexing targets can affect
the absolute complexity of gender systems. In other words, the metric computes how
many gender indexing targets there are in a language but does not account for whether or
not it makes a difference that these targets might be, for instance, adjectives, pronouns,
206
8.4 Concluding remarks
verbs or determiners or any possible combination between them. In addition, the metric
does not account for systems with animate concord and split plural indexation systems
where, as shown throughout the dissertation, certain referential and lexical properties of
the indexation triggers interfere with gender (and number) indexation patterns.
Finally, the metric presented in chapter 7 aims to measure the impact that interactions
of gender with other nominal features have on the absolute complexity of gender systems.
The general idea behind this attempt is that the more interactive a gender system, the
higher its absolute complexity. The domains considered in this thesis are only two,
number and evaluative morphology, but, as is known, grammatical gender tends to
be deeply intertwined with other nominal features such as case and definiteness. An
overview of these patterns of interactions is needed in order to incorporate them to the
metric and assess their role on the complexity of gender systems.
A more general question that would be worth asking in future research pertains to the
correlation between grammatical complexity of gender and maturity in Dahl’s (2004)
sense. However, since very little is known about young, nonmature gender systems,
comparing the complexity of gender systems at different stages of maturation is a difficult
task to undertake. As observed many times throughout the dissertation, gender is a
special feature of grammar. It is stubbornly stable but does not tend to emerge often in
the recent history of languages.
Finally, the results of the complexity calculations presented in chapter 7 suggest that
typological distributions connected with gender systems in the African macro-area are
somewhat bimodal. Languages either have complex gender systems – and this is the case
for the majority of the languages in the sample – or they lack gender altogether. On
the other hand, low levels of complexity are very rarely associated with gender. If these
facts are interpreted in terms of stability, one can speculate that not only are noncomplex gender systems typologically infrequent but that they also represent diachronically
unstable stages in the history of languages. In order to verify this hypothesis, the absolute complexity of the gender systems attested in other macro-areas of the world should
be assessed following methods comparable to those adopted in this thesis.
8.4 Concluding remarks
This dissertation has focussed on an under-investigated aspect of the nature of gender
systems: their interactions with other domains of grammar – number and evaluative
morphology – and the relevance of these interactions to the absolute complexity of
gender. This approach has turned out to be useful for analysing the gender systems
of a large sample of African languages, challenge some pre-existing assumptions in the
literature on gender, and spark further research along these lines.
Morphosyntactic and semantic interactions between individual features of grammar
have been thoroughly investigated in the verbal domain where, as a result, grammatical
features such as tense, aspect, mood and evidentiality are now seen as a congruent whole,
both at a descriptive and theoretical level.
This investigation is certainly far from providing the ultimate methodology and theo-
207
8 Summary of the main results, prospects for future research and concluding remarks
retical justifications for an enterprise of this type to be applied to the nominal domain.
The challenges are different because the factors at stake substantially differ. The results
of this dissertation may however be seen as illustrating some of the advantages in looking at nominal features in their morphosyntactic, semantic and functional interaction,
rather than as isolated domains of grammar.
208
A The language sample
A.1 Genealogical units and internal composition of the
subsamples
Table A.1 lists the languages of the sample based on the genealogical units they are
selected from (see also table 3.2). The language ISO codes are ordered alphabetically
and are followed by the correspondent language names. Both are taken from Glottolog
(Nordhoff et al. 2013) as of August 2014. Areal groupings, lower and higher levels of
genealogical affiliation are indicated if relevant. In particular, lower subdivisions within
a genealogical unit are spelled out only if they correspond to a genus in WALS. This
is done in order to allow comparability with the WALS classification, which is genusbased. When the genealogical unit selected in my sample already corresponds to a
genus in WALS (e.g., Berber or Semitic), no lower subdivision is given. Similarly, if the
genealogical unit selected in my sample corresponds to a level below genus in WALS
(e.g., Bantu, Eastern Nilotic or Western Nilotic), no lower subdivisions are indicated in
the table. Finally, information about the sources that I consulted for each of the sampled
languages is given in the table.
The following abbreviations are used in the table to refer to different kinds of language
groupings:
AG = Areal Grouping
LGU = Lower Genealogical Unit
HGU = Higher Genealogical Unit(s)
An alphabetical index of the sampled languages is given in table A.2.
209
Table A.1: The language sample
Genealogical Unit
Bantu
LGU
ISO
code
Language
baz
Tunen
bem
Bemba
bip
Bila
bvx
Dibole
cgg
Chiga
eto
Eton
kik
Kikuyu
kki
Kagulu
ksf
Bafia
lea
Lega
lin
Lingala
lol
Mongo
mcp
Makaa
AG
HGU
Source
Bantoid,
BenueCongo, Volta-Congo,
Atlantic-Congo
Bantoid,
BenueCongo, Volta-Congo,
Atlantic-Congo
Bantoid,
BenueCongo, Volta-Congo,
Atlantic-Congo
Bantoid,
BenueCongo, Volta-Congo,
Atlantic-Congo
Bantoid,
BenueCongo, Volta-Congo,
Atlantic-Congo
Bantoid,
BenueCongo, Volta-Congo,
Atlantic-Congo
Bantoid,
BenueCongo, Volta-Congo,
Atlantic-Congo
Bantoid,
BenueCongo, Volta-Congo,
Atlantic-Congo
Bantoid,
BenueCongo, Volta-Congo,
Atlantic-Congo
Bantoid,
BenueCongo, Volta-Congo,
Atlantic-Congo
Bantoid,
BenueCongo, Volta-Congo,
Atlantic-Congo
Bantoid,
BenueCongo, Volta-Congo,
Atlantic-Congo
Bantoid,
BenueCongo, Volta-Congo,
Atlantic-Congo
Mous (2003a)
van Sanbeek (1955)
Lojenga (2003)
Leitch (2003)
Taylor (1985)
Van de Velde (2006, 2008)
Barlow (1951); Stump (1993)
Petzell (2008)
Guarisma (2003)
Botne (2003)
Guthrie (1966)
Hulstaert (1965)
Heath (2003)
Table A.1: (continued)
Genealogical Unit
Bantu
Berber
LGU
ISO
code
Language
nya
Nyanja
nso
Northern Sotho
run
Rundi
sna
Shona
swa
Swahili
ssw
Swati
toi
Tonga
tsn
Tswana
ven
Venda
zul
Zulu
jbn
kab
shy
taq
tzm
Nafusi
Kabyle
Tachawit
Tamasheq (Kidal)
Tamazight,
Central
Atlas
Zenaga
zen
AG
HGU
Source
Bantoid,
BenueCongo, Volta-Congo,
Atlantic-Congo
Bantoid,
BenueCongo, Volta-Congo,
Atlantic-Congo
Bantoid,
BenueCongo, Volta-Congo,
Atlantic-Congo
Bantoid,
BenueCongo, Volta-Congo,
Atlantic-Congo
Bantoid,
BenueCongo, Volta-Congo,
Atlantic-Congo
Bantoid,
BenueCongo, Volta-Congo,
Atlantic-Congo
Bantoid,
BenueCongo, Volta-Congo,
Atlantic-Congo
Bantoid,
BenueCongo, Volta-Congo,
Atlantic-Congo
Bantoid,
BenueCongo, Volta-Congo,
Atlantic-Congo
Bantoid,
BenueCongo, Volta-Congo,
Atlantic-Congo
Watkins (1937)
Afro-Asiatic
Afro-Asiatic
Afro-Asiatic
Afro-Asiatic
Afro-Asiatic
Beguinot (1942)
Mettouchi (2000)
Penchoen (1973a)
Heath (2005)
Penchoen (1973b)
Afro-Asiatic
Nicolas (1953)
Poulos & Louwrens (1994)
Meeussen
(1959);
Mel’ čuck & Bakiza (1997)
Fortune (1955)
Myachina (1981)
Taljard et al. (1991)
Carter (2002)
Cole (1955)
Poulos (1990)
Poulos & Bosh (1997)
Table A.1: (continued)
Genealogical Unit
LGU
ISO
code
Language
Chadic
Biu-Mandara
gid
hna
xed
lln
hau
mkf
sur
pip
HGU
Source
Gidar
Mina
Hdi
Lele
Hausa
Miya
Mwaghavul
Pero
Afro-Asiatic
Afro-Asiatic
Afro-Asiatic
Afro-Asiatic
Afro-Asiatic
Afro-Asiatic
Afro-Asiatic
Afro-Asiatic
Frajzyngier (2008)
Frajzyngier & Johnston (2005)
Frajzyngier & Shay (2002)
Frajzyngier (2001)
Newman (2000)
Schuh (1989)
Frajzyngier (1993)
Frajzyngier (1989)
bej
ahg
awn
bsw
Beja
Qimant
Awngi
Baiso
Afro-Asiatic
Afro-Asiatic
Afro-Asiatic
Afro-Asiatic
dal
dsh
gax
Afro-Asiatic
Afro-Asiatic
Afro-Asiatic
gdl
ktb
rel
som
tsb
irk
Dahalo
Daasanach
Borana-Arsi-Guji
Oromo
Dirasha
Kambaata
Rendille
Somali
Tsamai
Iraqw
Afro-Asiatic
Afro-Asiatic
Afro-Asiatic
Afro-Asiatic
Afro-Asiatic
Afro-Asiatic
Roper (1928)
Appleyard (1975)
Hetzron (1978)
Corbett
(2000);
Corbett & Hayward (1987)
Tosco (1991)
Tosco (2001)
Andrzejewski (1960); Stroomer
(1995)
Sasse (1974)
Treis (2008, 2011, 2014)
Oomen (1981)
Saeed (1987); Serzisko (1992)
Savà (2005, 2007)
Mous (2003a)
Defoid
yor
Yoruba
Atlantic-Congo
Bamgbose
(2010)
Dizoid
mdx
Dizin
Afro-Asiatic ?
Beachy (2005)
Eastern Nilotic
kdj
mas
tuv
Karamojong
Masai
Turkana
Nilotic
Nilotic
Nilotic
Novelli (1985)
Payne (1998)
Dimmendaal (1983)
hts
Hadza
Igboid
ibo
Igbo
Kxa
ktz
Ju|’hoan
East Chadic
West Chadic
Cushitic
Beja
Central Cushitic
East Cushitic
South Cushitic
Hadza
Hadza
AG
Omotic ?
Khoisan
Edenmyr (2004);
communication
Atlantic-Congo
Khoisan
(1966);
Ajı́bóyè
personal
Green & Igwe (1963)
Güldemann
(2000);
Heine & König (2013)
Table A.1: (continued)
Genealogical Unit
ISO
code
Language
AG
Khoe-Kwadi
hnh
kwz
naq
nhr
xuu
{Ani
Kwadi
Nama
Naro
Kxoe
Khoisan
Khoisan
Khoisan
Khoisan
Khoisan
Kwa
aka
ewe
Akan
Ewe
Atlantic-Congo
Atlantic-Congo
snw
SElEE
Atlantic-Congo
bam
dyu
sus
mev
Bambara
Dyula
Susu
Mann
gol
kss
tem
Gola
Kisi
Timne
Atlantic
Atlantic
Atlantic
Atlantic-Congo
Atlantic-Congo
Atlantic-Congo
Fachner (1990)
Childs (1983, 1995)
Wilson (1961)
bjg
bqj
cou
Bidyogo
Bandial
Wamey
Atlantic
Atlantic
Atlantic
Niger-Congo
Niger-Congo
Atlantic-Congo
ffm
snf
srr
wol
Maasina Fulfulde
Noon
Serer
Nuclear Wolof
Atlantic
Atlantic
Atlantic
Atlantic
Atlantic-Congo
Atlantic-Congo
Atlantic-Congo
Atlantic-Congo
Segerer (2002)
Sagna (2008, 2011, 2012)
Santos (1996);
Pozdniakov
(2010); personal communication
Breedveld (1995)
Soukka (2000)
McLaughlin (1992)
McLaughlin (1997)
sad
Sandawe
Khoisan
amh
Amharic
Mande
LGU
Western Mande
Eastern Mande
Mel
North-Central Atlantic
Central Atlantic
North Atlantic
Sandawe
Semitic
Sandawe
HGU
Source
Heine (1999)
Güldemann (2004)
Hagman (1977)
Vossen (1986)
Kilian-Hatz & Heine (2010)
Appah & Amfo (2011)
Agbedor & Agbetsoamedo
(forthcoming);
Westermann
(1930); Yvonne Agbetoamedo,
personal communication
Agbetsoamedo & Di Garbo
(forthcoming); Yvonne Agbetoamedo, personal communication
Travélé (1955)
Delafosse (1901)
Lacan (1942)
Maria Khachaturyan, personal
communication
Eaton (2010)
Afro-Asiatic
Leslau (1995); Desalegn Hagos
Asfawwesen, personal communication
Table A.1: (continued)
Genealogical Unit
HGU
Source
Standard Arabic
Afro-Asiatic
ary
heb
Moroccan Arabic
Hebrew
Afro-Asiatic
Afro-Asiatic
mlt
Maltese
Afro-Asiatic
tig
trg
Tigre
Lishan Didan
Afro-Asiatic
Afro-Asiatic
Belnap & Shabaneh
(1994);
Belnap (1999); Ryding (2005)
Harrell (1965)
Coffin & Bolozky
(2005);
Glinert (1994); Grandi (2002)
Borg & Azzopardi-Alexander
(1997); Grandi (2002)
Leslau (1945, 1948)
Garbell (1965); Khan (2008)
South Omotic
dim
Dime
Omotic ?
Afro-Asiatic ?
Seyoum (2008)
Ta-Ne-Omotic
bcq
kqy
mdy
wal
Bench
Koorete
Male
Wolaytta
Omotic
Omotic
Omotic
Omotic
Afro-Asiatic
Afro-Asiatic
Afro-Asiatic
Afro-Asiatic
Rapold (2006)
Teketal (2004)
Amha (2001)
Lamberti & Sottile (1997)
Tuu
nmn
!Xoo
Khoisan
Western Nilotic
ach
bxb
dik
lwo
mfz
nus
Acoli
Belanda Bor
Dinka (Southwestern)
Luwo
Mabaan
Nuer
Semitic
LGU
ISO
code
Language
arb
AG
?
?
?
?
?
?
?
?
Güldemann
(1994)
Nilotic
Nilotic
Nilotic
Nilotic
Nilotic
Nilo
Storch
Storch
Storch
Storch
Storch
Storch
(2000);
(2005, 2007)
(2005)
(2005, 2007)
(2005, 2007)
(2005)
(2005, 2007)
Traill
A.2 The language sample: alphabetical index
In table A.2, the languages of the sample are listed alphabetically. The language names
are followed by the ISO codes, and the names of the genealogical units that each language
is assigned to (see also table A.1).
Table A.2: The language sample: alphabetical index
Language
ISO code
Genealogical unit
Acoli
Akan
Amharic
Awngi
Bandial
Bafia
Baiso
Bambara
Beja
Belanda Bor
Bemba
Bench
Bila
Borana-Arsi-Guji Oromo
Bidyogo
Chiga
Daasanach
Dahalo
Dibole
Dime
Dinka (Southwestern)
Dirasha
Dizin
Dyula
Eton
Ewe
Gidar
Gola
Hadza
Hausa
Hdi
Hebrew
Igbo
Iraqw
Ju|’hoan
Kabyle
Kagulu
Kambaata
Karamojong
Kikuyu
ach
aka
amh
awn
bqj
ksf
bsw
bam
bej
bxb
bem
bcq
bip
gax
bjg
cgg
dsh
dal
bvx
dim
dik
gdl
mdx
dyu
eto
ewe
gid
gol
hts
hau
xed
heb
ibo
irk
ktz
kab
kki
ktb
kdj
kik
Western Nilotic
Kwa
Semitic
Cushitic
North-Central Atlantic
Bantu
Cushitic
Mande
Cushitic
Western Nilotic
Bantu
Ta-Ne-Omotic
Bantu
Cushitic
North-Central Atlantic
Bantu
Cushitic
Cushitic
Bantu
South Omotic
Western Nilotic
Cushitic
Dizoid
Mande
Bantu
Kwa
Chadic
Mel
Isolate
Chadic
Chadic
Semitic
Igboid
Cushitic
Kxa
Berber
Bantu
Cushitic
Eastern Nilotic
Bantu
215
Table A.2: (continued)
216
Language
ISO code
Genealogical unit
Kisi
Koorete
Kwadi
Kxoe
Lega
Lingala (Kinshasa)
Lele
Lishan Didan
Luwo
Masai
Maasina Fulfulde
Mabaan
Makaa
Male
Maltese
Mann
Mina
Miya
Mongo-Nkundu
Moroccan Arabic
Mwaghavul
Nafusi
Nama
Naro
Ndengereko
Noon
Northern Sotho
Nuer
Nyanja
Pero
Qimant
Rendille
Sandawe
SElEE
Serer
Shona
Somali
Standard Arabic
Susu
Swati
Swahili
Tachawit
Tamasheq (Kidal)
Tamazight (Central Atlas)
Tigre
Timne
kss
kqy
kwz
xuu
lea
lin
lln
trg
lwo
mas
ffm
mfz
mcp
mdy
mlt
mev
hna
mkf
lol
ary
sur
jbn
naq
nhr
ndg
snf
nso
nus
nya
pip
ahg
rel
sad
snw
srr
sna
som
arb
sus
ssw
swa
shy
taq
tzm
tig
tem
Mel
Ta-Ne-Omotic
Khoe-Kwadi
Khoe-Kwadi
Bantu
Bantu
Chadic
Semitic
Western Nilotic
Eastern Nilotic
North-Central Atlantic
Western Nilotic
Bantu
Ta-Ne-Omotic
Semitic
Mande
Chadic
Chadic
Bantu
Semitic
Chadic
Berber
Khoe-Kwadi
Khoe-Kwadi
Bantu
North-Central Atlantic
Bantu
Western Nilotic
Bantu
Chadic
Cushitic
Cushitic
Isolate
Kwa
North-Central Atlantic
Bantu
Cushitic
Semitic
Mande
Bantu
Bantu
Berber
Berber
Berber
Semitic
Mel
Table A.2: (continued)
Language
ISO code
Genealogical unit
Tonga
Tsamai
Tswana
Tunen
Turkana
Venda
Wamey
Wolaytta
Wolof (Nuclear)
Yoruba
Zenaga
Zulu
{Ani
!Xóô
toi
tsb
tsn
baz
tuv
ven
cou
wal
wol
yor
zen
zul
hnh
nmn
Bantu
Cushitic
Bantu
Bantu
Eastern Nilotic
Bantu
North-Central Atlantic
Ta-Ne-Omotic
North-Central Atlantic
Defoid
Berber
Bantu
Khoe-Kwadi
Tuu
217
B Database coding sheet
Information about the individual languages of the sample has been stored in the database
based on the following coding sheet.
B.1 Language data
• Language name
• ISO code
• Number of speakers
• African micro-area:
– Eastern Africa
– Central Africa
– Northern Africa
– Southern Africa
– Western Africa
• Genealogical group
• Source
B.2 Gender systems
• Language ID = ISO code
• Type of gender system
– Sex-based
– Non-sex-based
– No gender
– No data
• Number of genders
– Two genders
– Three genders
– Four genders
– Five or more
– No gender
– No data
219
• Gender assignment
– Semantic assignment
– Semantic and formal assignment
– No gender
– No data
• Number of gender-indexing targets
– One indexing target
– Two indexing targets
– Three indexing targets
– Four or more
– No gender
– No data
• Occurrence of gender marking on nouns
– Gender marking on nouns: Yes
– Gender marking on nouns: No
– No gender
– No data
• Cumulation with number on agreement targets
– Cumulative with all indexing targets
– Cumulative with some indexing targets
– Noncumulative
– No gender
– No number indexation
– No data
• Cumulation with number on nouns
– Cumulative
– Cumulative sg vs. noncumulative pl
– Noncumulative with specific gender (G) and number (N) markers
– Noncumulative
– No gender marking on nouns
– No gender
– No data
• Syncretism of gender across number
– Yes
– No
– No data
220
B.3 Number systems
• Language ID = ISO code
• Type of number system in terms of obligatoriness
– Obligatory number
– General number
– No number
– No data
• Number values
– Singular vs.
– Singular vs.
– Singular vs.
– Singular vs.
– No number
– No data
plural
plural vs. dual
plural vs. dual vs. trial
plural vs. dual vs. paucal
• Number of number-indexing targets
– One indexing target
– Two indexing targets
– Three indexing targets
– Four or more indexing targets
– No number indexation
– No data
B.4 Evaluative morphology
• Language ID = ISO code
• Evaluative distinctions
–
–
–
–
–
Diminutive and augmentatives
Only diminutives
Only augmentatives
None
No data
• Type of diminutive markers
– Suffix
– Prefix
– Infix
– Stem change
– Clitic
– Analytical
– Other (reduplication, phonosymbolic patterns etc...)
221
–
–
–
–
–
Different diminutive markers on different nouns
Co-occurring diminutive markers
Different diminutives on different nouns + co-occurring diminutives
No diminutives
No data
• Type of augmentative markers
– Suffix
– Prefix
– Infix
– Stem change
– Clitic
– Analytical
– Other (reduplication, phonosymbolic patterns etc...)
– Different augmentative markers on different nouns
– Co-occurring augmentative markers
– Different augmentatives on different nouns + co-occurring augmentatives
– No augmentatives
– No data
• Relations among gender and evaluation
– Present
– Absent
– No gender
– No data
• Number of diminutive genders
– One
– More than one
– None
– No data
• Number of augmentative genders
– One
– More than one
– None
– No data
• Type of gender shift in sex-based gender systems
– M → F = Dim. and F → M = Aug.
– M → F = Dim.
– F → M = Aug.
– M/F → N = Dim.
– None
– No data
222
B.5 Gender complexity
• Language ID = Iso code
• Number of gender values (GV)
– Two = 0
– Three = 1/3
– Four = 2/3
– Five or more = 1
– No data = NA
• Nature of assignment rules (AR)
– Semantic assignment = 0
– Semantic and formal assignment = 1
– No data = NA
• Number of indexing targets (IND)
– One = 0
– Two = 1/3
– Three = 2/3
– Four or more = 1
– No data = NA
• Cumulative exponence of gender and number (CUM)
– Noncumulative = 0
– Partially cumulative = 1/2
– Cumulative = 1
– No data = NA
• Manipulation of gender assignment triggered by number/countability (M1)
– Absent = 0
– Present = 1
– No data = NA
• Manipulation of gender assignment triggered by size (M2)
– Absent = 0
– Present = 1
– No data = NA
223
C Gender- and number-indexing targets in
the languages of the sample
Table C.1: Number of gender- and number-indexing targets in the languages of the sample
G. indexes
N. indexes
No. of
lngs.
%
One
Genealogical groups
One
4
4%
Two
Two
15
15%
Bantu (2/23)
Chadic (1/8)
Cushitic (3/13)
Khoe-Kwadi (4/5)
Mel (1/3)
South Omotic (1/1)
Ta-Ne-Omotic (3/4)
Three
Three
25
25%
Bantu (3/23)
Berber (5/6)
Chadic (1/8)
Cushitic (7/13)
Eastern Nilotic (1/3)
North-Central
Atlantic
(1/7)
Sandawe (1/1)
Semitic (5/7)
Ta-Ne-Omotic (1/4)
Four or more
Four or more
34
34%
Bantu (17/23)
Berber (1/6)
Chadic (2/8)
Cushitic (1/13)
Eastern Nilotic (2/3)
Hadza (1/1)
Kwa (1/3)
Mel (1/3)
North-Central
Atlantic
(5/7)
Semitic (2/7)
Tuu (1/1)
Chadic (1/8)
Khoe-Kwadi (1/5)
Kxa (1/1)
North-Central
Atlantic
(1/7)
225
Table C.1: (continued)
Total
226
G. indexes
N. indexes
No. of
lngs.
%
Three
Two
1
1%
Dizoid (1/1)
Three
One
1
1%
Cushitic (1/13)
Two
Three
1
1%
Cushitic (1/13)
One
Two
1
1%
Chadic (1/8)
No gender
One
6
6%
Igboid (1/1)
Kwa (1/3)
Mande (4/4)
No gender
Two
4
4%
Chadic (2/8)
Defoid (1/1)
Kwa (1/3)
No gender
No
number
indexation
6
6%
Western Nilotic (6/6)
No data
No data
2
2%
Bantu (1/23)
Mel (1/3)
100
100%
Genealogical groups
D Examples of lexical plurals in four
Cushitic languages
Table D.1: Lexical plurals in Iraqw (based on Mous 2008)
Semantic groupings
Nouns
Abstract nouns
aldafiri ‘interest, something returned with what was borrowed’; diidaa ‘boasting pride’; fayda ‘profit’; huwaa ‘burden’; iilo ‘weight, load’;
loeemaa ‘truth’; qatsuwa ‘heroic success (in hunting or war)’; slaaèareri
‘aroma’; tsaxwa ‘danger’; waaqooda ‘hypochrisy’
Activities involving multiple
actions/participants
aai ‘journey’; da’ri ‘witchcraft’; gila ‘quarrel, fight’; ibyaa ‘pointless
activity with the hands’; waayaa ‘work of different kind, not heavy,
routine’
Animate and inanimate collectives
haywa ‘term to address children’; kumbeeri ‘women accompanying
the bride’s mother during the wedding cerimony’; wa‘ree ‘boys and
girls accompanying the bride’; makay ‘animals’; laqaya ‘thorns’;
yakawaa/hikwaa ‘cattle’; hurwa’i ‘bad maize grains’; baynu ‘pigs’;
maanda ‘Nyiramba, Bantu (land and people)’; afi ‘scraping of stiff
porridge at the sides of the pot’
Body parts
afeetlo ‘waist, loin’; duunga ‘nose’; gitsee’a ‘forehead, face’; gwe’edo
‘buttocks’; hayso ‘tail, penis’; xxatli ‘afterbirth, placenta of an animal,
trees’; xaxardu ‘palate’; da’awa ‘chest’
Diseases
kuuko ‘mumps’
Location
alu ‘behind, reverse’; baray ‘down (on a slope), low, inside’; dimbé ‘side,
far, separate, different’; tsee’a ‘outside’
Miscellaneous
ayla ‘song improvised for the occasion’
Masses
da’ata ‘red of blood’; duwa ‘milk from plant’; ilwa ‘milk’; hinqeereeri
‘saliva’; tsnuqaa ‘saliva as blessing, gifts in the form of money to newly
wed’; dara’ma ‘roasted meat and intenstines for the skinners’; fu’naay
‘meat (for eating)’; kund’i ‘bundle to carry, bale’; ma’ay ‘water’ slaèoo
‘mucus’
Internally complex objects
dara’ma ‘roasted meat and intestines for the skinners’
Places
maraay ‘houses’; doori ‘sky, heaven’; irqwá da’áw ‘montaneous area
southeast of Mbulu’; sièú ‘far land’; yaamu ‘earth, world below’; uwa
‘west’
Periods of time
ameetleemu ‘midday’; aymadu ‘lunch time’; amsi ‘midnight, night’;
axweeso ‘evening, night’; baloqa ‘day after tomorrow’; buhaaree ‘rainy
season’; de’ma ‘time’; ki’ima ‘turn, time, coming back’; matlo ‘tomorrow’; xweeraa ‘night’; tsiindo ‘evening (before dark)’
227
Table D.2: Lexical plurals in Borana-Arsi-Guji Oromo (based on Andrzejewski 1960: 70)
Semantic groupings
Nouns
Body parts
fúnnaan ‘nose(s)’; áfaan ‘mouth, mouths’; ı́lkaan ‘tooth, teeth’;
tS’ı́âaan ‘penis, penes’
Mass nouns
ı́mimmaan ‘tear, tears’, údaan ‘faeces’; fı́ntS’aan ‘urine’; fôon ‘meat’;
bı́saan ‘water’
Animate and inanimate collectives
lôon ‘cattle’; indüaan ‘lice’ (or louse); mı́ãaan ‘a berry, a grain, a fruit,
or berries, grains, fruit’; mı́tSiraan ‘stones, stone’; séep’an ‘leather
straps’; búusan ‘Pleiades’
Periods of time
halkán ‘night’
Table D.3: Lexical plurals in Rendille (based on Oomen 1981: 51)
Semantic groupings
Nouns
Body parts
marát ‘brains’; sombób ‘lungs’
Animate collective
biná ‘wild animals’; èolá ‘domesticated animals’
Mass nouns
èaanú ‘milk’; bicé ‘water’; oncabá ‘maize’; sonxór ‘sugar’; basbás ‘different kinds of milk mixed’; dúubát ‘fog’; banáy ‘light’
Miscellaneous
anxád ‘lightning’
Table D.4: Lexical plurals in Baiso (based on Corbett & Hayward 1987: 9)
Semantic groupings
Nouns
Body parts
ilkoo ‘tooth, teeth’; kalaljaa ‘kidneys’; luk
ff k
ff aa ‘foot, feet, leg(s)’; ilffoo
‘eye(s)’; ogorroo ‘hair’; moo ‘hips, lumber region’
Animate collective
saé ‘cattle’
Mass nouns
eenoo ‘milk’; soo ‘meat’; udú ‘faeces’
Objects coming in pairs
keferoo ‘sandals’
228
E Singular and plural suffixes in Mabaan,
Dinka and Luwo
Table E.1: Singular markers in Mabaan, Dinka and Luwo (based on Storch 2005: 382)
Mabaan
Dinka50
-(C)2̀
V3-
Luwo
Semantics
long, one-dimensional, dominant
objects
V:
-w
-gOn
derogative concepts
-t2̀
-V:[‘]
HL
-O
¨
-Ò
¨
singulative concepts
-n
-V:-V3-
general singular
2̀
-V3-o
-a
-O
¨
-g2̀
-à
-V:-
abstract concepts
-k
abstract concept
-O
¨
spherical, round, small, objects;
mass items; specialized people
[‘]
-w
-V2-i
fast moving objects
Ò
-o
-a
locatives, domestic objects
-ù / Nù
-V2-
-U
¨
shape + possession, body, spatial
orientation
-a
soft, circular objects
-V3[‘]
50
-N2̀
[‘]
-w
-l
àn
” /ñàn
”
-w
part of a larger unit
-(C)in
”
-l
animacy, mass
The cardinal numbers that appear next to the vocalic symbol V refer to the degrees of vowel alternation
in Dinka. For a more detailed discussion on the grammatical functions of vowel alternation in Dinka,
see Storch (2005: 165-166).
229
Table E.2: Plural markers in Mabaan, Dinka and Luwo (based on Storch 2005: 385)
Mabaan
Dinka
Luwo
Semantics
-k2̀
-k
-V2-V3-V:[‘]
-k2̀
general plural
-(C)ln
”
-N
-V2-V3-V:[‘]
-V-F+X -E
-VNÉ
general plural
án
”
È
round, mass, small
-kù
Ì
body, space
semantically unspecified
L
-t”(àn
230
-t”
-t”
semantically unspecified
F Complexity scores for the individual
features in the metric
Table F.1 shows how the languages with gender in the sample scored with respect to
the features of the complexity metric presented in chapter 7. Unlike in table 7.3, where
the GCSs are rounded up to numbers with two decimal places, unrounded figures are
provided in table F.1. The data are ordered alphabetically based on the ISO codes of the
sampled languages. See table A.2 for the correspondent language names. The following
abbreviations are used in the headings of the table:
GV = Number of gender values
AR = Nature of assignment rules
IND = Number of gender-indexing targets
CUM = Cumulation
M1 = Manipulation of gender assignment triggered by number/countability
M2 = Manipulation of gender assignment triggered by size
GCS = Gender Complexity Score
Table F.1: Complexity scores
ISO
GV
AR
IND
CUM
M1
M2
GCS
ahg
amh
arb
ary
awn
baz
bcq
bej
bem
bjg
bip
bqj
bsw
bvx
cgg
cou
dal
0
0
0
0
0
1
1/3
0
1
1
0
1
0
1
1
1
0
1
1
1
1
1
1
1
1
1
1
0
1
1
1
1
1
0
2/3
1
2/3
2/3
1/3
2/3
1/3
1
1
1
1
1
1
1
1
0
1
1
1
1
1/2
1
1
0
1
0
0
1
1
0
1
0
0
1
1
0
1
0
1
1
1
0
1
0
0
1
1
1
1
1
1
0
1
0
0
1
1
0
0.533333333
0.666666667
0.611111111
0.611111111
0.555555556
0.944444445
0.611111111
0.5
1
1
0.222222222
1
0.361111111
0.777777778
1
0.833333333
0.277777778
1
1/3
1
2/3
2/3
1
1
2/3
231
Table F.1: (continued)
232
ISO
GV
AR
IND
CUM
M1
M2
GCS
dim
dsh
eto
ffm
gax
gid
gdl
gol
hau
heb
hnh
hts
irk
jbn
kab
kdj
kik
kki
kqy
ksf
kss
ktb
ktz
kwz
lea
lin
lln
lol
mas
mcp
mdx
mdy
mkf
mlt
naq
nhr
ndg
nmn
nso
nya
pip
rel
sad
shy
sna
snf
snw
som
srr
ssw
0
0
1
1
0
0
0
1
0
0
1/3
0
0
0
0
1/3
1
1
0
1
1
0
2/3
0
1
1
0
1
0
1
0
0
0
0
1/3
1/3
1
1
1
1
0
0
0
0
1
1
1
0
1
1
0
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1/3
1/3
1
1
2/3
1
1/3
1/3
2/3
2/3
1/3
2/3
2/3
2/3
2/3
1
1
1
2/3
2/3
1
1/2
1
1
1
1
1
1
1
1
1
1
1/2
1/2
1/2
1
1
1
1/2
1
1
1/2
1/2
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1/2
1
1
1/2
1
1
1
1/2
1
1
0
0
1
1
0
0
0
0
0
0
0
0
0
1
1
0
1
1
0
1
1
1
0
1
0.388888889
0.472222222
0.833333333
1
0.533333333
0.5
0.466666667
0.666666667
0.611111111
0.444444445
0.533333333
0.611111111
0.433333333
0.694444445
0.694444445
0.722222222
1
1
0.527777778
0.777777778
0.75
0.361111111
0.361111111
0.25
1
0.555555556
0.466666667
1
0.5
1
0.444444445
0.555555556
0.6
0.833333333
0.611111111
0.611111111
1
1
0.833333333
1
0.125
0.466666667
0.611111111
0.75
1
0.666666667
0.833333333
0.433333333
1
0.833333333
1
1
1
1
0
1
0
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
2/3
0
0
1
1/3
1/3
1
1
1
2/3
1/3
1
1
1/3
1/3
1
1
1
1
0
1/3
2/3
1
1
0
1
2/3
1
1
0
0
0
1
0
0
1
0
1
0
0
0
1
0
0
1
1
1
0
0
0
1
1
0
0
0
1
0
1
0
1
1
1
1
1
1
1
0
0
0
0
1
0
1
1
1
1
1
1
1
1
1
0
1
1
1
1
1
1
1
0
Table F.1: (continued)
ISO
GV
AR
IND
CUM
M1
M2
GCS
sur
swa
taq
tem
tig
toi
trg
tsb
tsn
tuv
tzm
ven
wal
wol
xuu
zen
zul
0
1
0
1
0
1
0
0
1
1/3
0
1
0
1
0
0
1
0
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
0
1
2/3
1
2/3
1
2/3
2/3
1
2/3
2/3
1
1/3
2/3
1/3
2/3
1
1/2
1
1/2
1
1
1
1
1
1
1
1/2
1
1
1
1
1/2
1
0
1
1
0
1
1
0
1
0
0
1
1
1
1
0
0
0
1
1
1
1
0.083333333
1
0.694444445
1
0.611111111
1
0.533333333
0.444444445
0.833333333
0.833333333
0.694444445
1
0.555555556
0.777777778
0.466666667
0.694444445
0.833333333
0
0
1
1
1
1
1
1
0
233
Sammanfattning på svenska
Francesca Di Garbo
Gender and its interaction with number and evaluative morphology: An intra- and intergenealogical typological survey of Africa
(Översättning från engelska av Eva Lindström)
Denna avhandling undersöker morfosyntaktisk och semantisk interaktion i genussystem med numerus och evaluativ markering (diminutiva och augmentativa konstruktioner).51 Detta görs genom en kombination av tvärspråklig forskning och djupstudier
av besläktade språk inom en kontinent, nämligen Afrika. Diakroniska och synkroniska
infallsvinklar på interaktionen tas i beaktande. Begreppet interaktion används ganska
löst i detta arbete. Faktum är, särskilt i afrikanska språk, att genus och numerus å ena
sidan, och genus och evaluativ morfologi å den andra, kan vara så tätt sammanflätade
att de inte bara interagerar utan visar tecken på att ha smält samman.
Avhandlingen centreras kring tre huvudsakliga forskningsfrågor:
(1) Interaktion mellan genus och numerus
– Har genus och numerus samma markering? Hur fördelar sig genusdistinktioner över numerusdistinktioner?
– Konkurrerar indexeringsmönstren för genus och numerus med varandra distributionsmässigt?
– Har genus och numerus samma relevans för substantiv?
– Interagerar genus och numerus semantiskt?
(2) Interaktion mellan genus och evaluativ morfologi:
– Kan storlek vara en faktor i genustilldelning?
– Hur interagerar diminutiva och augmentativa genus med de övriga genusdistinktionerna i ett språk?
(3) Interaktion mellan genus och grammatisk komplexitet:
– Är det möjligt att mäta den grammatiska komplexiteten hos genussystem?
– Kan ett sådant komplexitetsmått redogöra för interaktion mellan genus och
andra grammatiska domäner?
51
Begreppet genus används i denna avhandling på samma sätt som det oftast gör i typologisk litteratur,
där genus och nominalklasser ses som ett och samma fenomen, och genus används som överordnad
term.
235
Summary in Swedish
– Vilken roll spelar sådan interaktion för den sammantagna komplexiteten hos
ett genussystem?
Avhandlingen består av åtta kapitel. Kapitel 1 utgörs av en kort presentation av de
huvudsakliga undersökningstemana och en översikt över avhandlingens struktur.
Den teoretiska bakgrunden diskuteras i kapitel 2. I kapitlets första del presenteras
de tre undersökta domänerna (genus, numerus och evaluativ morfologi) var för sig, med
fokus på deras funktionella och semantiska egenskaper, strukturella uttryck och typologiska fördelning. I kapitlets andra del ges en översikt över tidigare litteratur om interaktion mellan genus och numerus, och genus och evaluativ morfologi, liksom begreppet
absolut komplexitet och dess relevans för genussystem. Följande nyckelbegrepp introduceras i kapitlet:
Indexering används i denna avhandling i stället för kongruens, såsom det definieras av
Corbett (2006). Det står för de grammatiska strategier som ett språk använder sig
av (1) för att uttrycka lexikala och grammatiska egenskaper hos substantiv, eller (2)
för att uttrycka semantiska egenskaper hos np-referenter, i båda fallen oavsett om
en syntaktisk antecedent eller nominalt huvudord finns overt uttryckt i kontexten.
Termen indexeringsmottagare (indexing target) används om konstituenter i kontexten vars böjningsmorfologi uttrycker genus och numerus (t.ex. adjektiv, determinerare, verb, pronomen osv.). Indexeringsutdelare (indexation trigger) används
för de konstituenter som aktiverar användningen av ett visst indexeringsmönster
(t.ex. substantiv och pronomen) inom en viss kontext.
Exponens definieras som betydelsemängden per morfologisk enhet. Exponens är kumulativ när betydelser härrörande till åtminstone två grammatiska domäner realiseras
i ett och samma morfem. Genus och numerus har ofta kumulativ exponens.
Synkretism definieras som en typ av paradigmatisk asymmetri där vissa grammatiska
distinktioner neutraliseras eller reduceras givet vissa betingande faktorer. Det
är typologiskt vanligt att genusdistinktioner neutraliseras eller reduceras när ett
substantiv inte står i singularis.
Manipulerbar vs. strikt genustilldelning: genustilldelning är manipulerbar när substantiv kan tilldelas mer än ett genus, beroende på hur np-referenten betraktas. Genustilldelning är strikt när varje substantiv konsekvent tillhör ett enda genus.
Absolut komplexitet definieras som antalet distinktioner inom en grammatisk domän,
eller längden på dess beskrivning.
Studiens urvalsmetodologi och hur data samlats in och behandlats presenteras i kapitel 3. Avhandlingen bygger på ett urval om 100 afrikanska språk, varav 84 med genus
och 16 utan genus. Urvalet utformades med avsikten att kunna göra jämförelser både
inom och mellan språkfamiljer, och består därför av flera delurval inom skilda genetiska
grupper (t.ex. bantu, kushitiska och berber).
236
Sammanfattning på svenska
Urvalet är snedvridet iåtminstone två avseenden: släktskapsmässigt och bibliografiskt.
Det kan ses som släktskapsmässigt snedvridet för att det bygger på relationer mellan
undergrupper av inbördes besläktade språk inom en avgränsad del av världen. Det är
bibliografiskt snedvridet då urvalet inom varje språkfamilj gjordes på basis av tillgänglig
litteratur snarare än med matematiska metoder, även om alla undergrenar är representerade. Dessa två typer av snedvridningar inverkar på analysen såtillvida att urvalet
inte kan användas för att göra statistiskt grundade förutsägelser om dominerande typologiska mönster, varken inom eller utanför den afrikanska makroregionen. Denna studie
syftar heller inte till att göra statistiska förutsägelser. Målet är i stället att redogöra för
förekomst, stabilitet och fördelning av grammatiska fenomen som rör interaktionen mellan genus och numerus, och mellan genus och evaluativ morfologi. Jag har utgått ifrån
att dessa mål bäst kan uppnås genom att se på ett flertal grupper av inbrdes besläktade
språk inom en enda världsdel. För att strukturera data från de olika språken i urvalet
skapades en relationell databas. Vad gäller släktskapsklassificeringen använder sig denna
studie av Glottolog (Nordhoff et al. 2013).
Kapitel 4 ger en översikt över de system för genus, numerus och evaluativ morfologi som påträffades i de undersökta språken. De parametrar som använts för var
och en av de tre domänerna i klassificeringen av databasen presenteras och diskuteras.
Frekvensfördelningen för de värden som hör ihop med de respektive parametrarna ges
också i kapitlet.
Interaktion mellan genus och numerus diskuteras i kapitel 5. Kapitlet har fyra delar. I den första delen undersöker jag exponensmönster hos genus och numerus, samt
numerusbetingad synkretism i genusmarkering. Exponens och synkretism ses ofta som
två grundläggande dimensioner i interaktionen mellan genus och numerus. Såvitt jag
känner till utgör dock denna avhandling den första systematiska genomgången av exponensmönster och synkretism för genus och numerus som grundar sig på ett så pass
stort urval av språk. Kumulativ exponens, där genus och numerus uttrycks i en och
samma form, är extremt vanlig i de undersökta språken, och det gäller även synkretism.
Eftersom inte alla språk med kumulativ exponens har synkretism, men nästan alla språk
med synkretism har kumulativ exponens, framlägger jag hypotesen att numerusbetingad
genussynkretism förutsätter kumulativ markering av dessa två grammatiska domäner.
Jag menar vidare att kumulativitet bör ses som en faktor som ökar den absoluta komplexiteten i genussystem, såtillvida att det inför brus överensstämmelsen mellan form och
betydelse. Slutligen argumenterar jag för att mina resultat rörande kumulativitet och
synkretism kan förstås som möjliga indikatorer på att det finns en relevanshierarki bland
nominala grammatiska särdrag. I överensstämmelse med Greenberg (1963a), Carstairs
(1987), Carstairs & Stemberger (1988), and (Vafaeian 2013) - som alla studerade olika
grammatiska fenomen och använde sig av olika dataset - visar dessa resultat att om det
finns en nominal relevanshierarki så skulle numerus vara den högst rankade kategorin.
I kapitlets andra del fokuserar jag på de konkurrerande mönster för genus- och numerusindexering som förekommer inom ett relativt litet antal språk inom det större
urvalet. Dessa språk hör alla till det afroasiatiska fylumet och uppvisar ett tudelat,
237
Summary in Swedish
könsbaserat genussystem. De har dessutom en delning av indexeringsmönstren på substantiv i pluralis. Vissa substantiv i pluralis aktiverar samma indexering som antingen
maskulinum singularis eller femininum singularis, medan andra substantiv i pluralis ger
upphov till indexering som skiljer sig både från maskulinum och femininum singularis.
Eftersom det senare indexeringsmönstret bara används för att uttrycka egenskapen pluralis kallar jag det Specialiserad pluralindexering (Dedicated Plural Indexation, DPI).
Jag föreslår vidare benämningen språk med delat pluralindexeringssystem (languages
with split plural indexation) för språk av denna typ. Jag visar hur användningen av
DPI kan begränsas antingen av animacitet (t.ex. i modern standardarabiska), eller av
lexikal pluralitet hos huvudordet (kushitiska).
En fallstudie utfördes på de kushitiska språken inom det större urvalet för att kartlägga
hur olika språk inom gruppen beter sig vad gäller användandet av DPI. I litteraturen om
kushitiska språk beskrivs språk med relativt begränsad användning av DPI traditionellt
som att de har en tredje indexeringsklass, alltså ett tredje genus, dvs. pluralis förutom
maskulinum och femininum (se Mous 2008 för en översikt). Detta synsätt har mött stark
kritik i mer typologiskt inriktad litteratur (se bl.a. Corbett 2000; Corbett & Hayward
1987). I dessa studier har dock DPI ofta setts bara som ett undantag, utan något försök
till förklaring av dess säregenhet (ett undantag är Corbett 2012 som försöker förklara
genus- och numerusindexering i Baiso genom att i stället fokusera på den roll som lexikal
pluralitet spelar). Min studie visar att en övergripande alternativ förklaring till detta
länge debatterade fenomen blir möjlig om man ser till hur DPI fördelar sig över olika
språk inom familjen och undersöker de faktorer som påverkar dess förekomst. I kushitiska språk med delad pluralindexering är DPI bara möjlig om indexeringsutdelaren har
en inneboende koppling till pluralitet. Detta innebär inte att DPI är ett morfosyntaktiskt uttryck för genus. Det visar bara att två domäner kan konkurrera med varandra:
en obligatoriskt uttryckt genom indexering (genus), och den andra potentiellt uttryckt
genom indexering (numerus). Jag visar att arbetsfördelningen dem emellan görs upp på
olika sätt i olika kushitiska språk. I de flesta språken indexeras nominal pluralitet alltid,
antingen den är morfologisk eller lexikal; i andra språk indexeras bara lexikal pluralitet.
I kapitlets tredje del diskuterar jag korrelationer mellan förekomst av genus och typ
av numerusmarkering. Jag visar hur framväxten av genomgripande indexeringssystem i
språken i urvalet alltid förutsätter att det finns både genus och numerus, och att språken
i denna studie tenderar att uppvisa samma antal indexeringsmottagare för genus som
för numerus.
I den fjärde och sista delen av kapitlet visar jag att genusbyte i språk med manipulerbar genustilldelning kan användas för att uttrycka räknebarhet. Detta fenomen har en
väldigt sned fördelning i undersökningens språk, i och med att det bara påträffas inom
familjerna bantu, berber, nord-centrala atlantiska och semitiska, samt i ett östnilotiskt
språk.
I kapitel 6 undersöker jag morfosyntaktiska och semantiska interaktioner mellan
genus och den morfologiska kodningen av diminutiver och augmentativer. Kapitlet syftar särskilt till att undersöka (1) hur vanligt det är att storlek fungerar som ett genus för
sig, och (2) hur olika typer av genussystem (könsbaserade kontra icke-könsbaserade) och
238
Sammanfattning på svenska
strategier för genustilldelning (strikt kontra manipulerbart) ger förutsättningarna för de
mönster för interaktion mellan genus och evaluativ morfologi som kunnat observeras.
Genom detta kapitel bidrar denna avhandling med den första större typologiska redogörelsen för hur markering av storlek (‘stor’ eller ‘liten’) kan inlemmas i ett språks
genussystem. Resultaten tyder på att arten av interaktion mellan genus och evaluativ
morfologi varierar enligt vilken typ av genussystem ett språk har. De relevanta variablerna är: (1) typ av genussystem (könsbaserat kontra icke-könsbaserat), (2) antalet
distinktioner inom systemet, och (3) graden av manipulerbarhet i genustilldelning.
Två typer av interaktion mellan genus och evaluativ morfologi förekommer i undersökningsspråken. Språk med stora genussystem har typiskt specialiserade evaluativa
genus som används för att uttrycka diminutiva och augmentativa betydelser. I språk
med mindre och könsbaserade genussystem finns å andra sidan inga specialiserade evaluativa genus, utan maskulinum, femininum eller neutrum kan anvn̈das för att uttrycka
diminutiva och augmentativa betydelser när ett ord används med ett annat genus än
det brukar ha. Båda typerna har bred spridning i mina undersökningsspråk, även över
språkfamiljsgränser. I vissa fall är det möjligt att uppskatta deras stabilitet, såväl som
delar av deras framväxt diakroniskt. Data visar också att även om storlek kan vara
ett produktivt kriterium för substantivklassificering är det aldrig centralt framträdande,
varken synkront eller diakront, i de två typer av genussystem som undersökningsspråken
företer. Slutligen begränsas möjligheten att uttrycka evaluativa betydelser genom genusbyte i språk med könsbaserat genus av np-referentens animacitet. För animata substantiv, där könstillhörighet är den mest framträdande egenskapen, används genusbyte
bara för att uttrycka biologiska genusskillnader. När det gäller inanimata substantiv är
könstillhörighet irrelevant och genusbyte används för att uttrycka skillnader i storlek.
Kapitel 7 syftar till att mäta den effekt som interaktioner mellan genus och numerus, och mellan genus och evaluativ morfologi, har på ett genussystems komplexitet.
Resultaten tyder på att ju mer interaktiv en grammatisk domän är, desto högre är komplexiteten. Jag använder begreppet grammatisk komplexitet i dess absoluta bemärkelse,
dvs. som en objektiv egenskap hos en grammatisk domän. Måttet på absolut komplexitet kan grunda sig antingen på antalet delar som ingår i ett system, eller längden på
systemets beskrivning. Det bör noteras att mått på absolut komplexitet bara kan vara
lokala, i meningen att de bör gälla specifika grammatiska domäner snarare än hela språk.
Miestamo (2008) och Sinnemäki (2011) lägger fram två grundläggande principer som
allmänna riktlinjer för utveckling av komplexitetsmått inom vilken grammatisk domän
det än vara månde:
Principen om färre distinktioner. Ju högre antalet grammatikaliserade distinktioner är, desto komplexare är domänen.
Principen om en betydelse-en form. Den minst komplexa domänen är den där det
råder ett-till-ett förhållande mellan betydelse och form.
I kapitel 7 argumenterar jag för att dessa två principer inte räcker för att redogöra för
hela den vidd av fenomen som kan vara relevanta för studiet av språklig komplexitet.
239
Summary in Swedish
Jag menar att en tredje princip bör läggas till för att hantera interaktioner mellan
grammatiska domäner. Jag kallar denna princip Principen om oberoende.
Principen om oberoende. Den minst komplexa domänen är den vars beskrivning
är oberoende av semantiska och funktionella egenskaper hos andra domäner.
Att utveckla mått för att avgöra den absoluta komplexiteten hos grammatiska domäner
är ett relativt nytt område inom språktypologi. Vad gäller genus har det behandlats i
banbrytande arbete av Audring (2014). Författaren föreslår tre dimensioner längs vilka
genuskomplexitet kan berk̈nas, men ger ingen metod för att överföra dimensionernas
variabler till mätbara värden. I kapitel 7 utvecklar jag Audrings modell för genuskomplexitet, med två huvudsakliga syften:
(1) att redogöra för de morfosyntaktiska och semantiska interaktioner mellan genus och
numerus och mellan genus och evaluativ morfologi som utforskas i avhandlingen.
(2) att omvandla de värden som hör ihop med de respektive dimensionerna till tal.
Det komplexitetsmått som utarbetades för att uppnå dessa syften består av sex komponenter: GV (antal genusdistinktioner), AR (antal regler för genustilldelning), IND
(antal indexeringsmottagare för genus), CUM (förekomst av kumulation av genus och
numerus), M1 (manipulering av genustilldelning som styrs av numerus/räknebarhet),
M2 (manipulering av genustilldelning som styrs av storlek). För att omvandla värdena
för var och en av parametrarna till siffror använder jag en metod utvecklad av Parkvall
(2008).
Resultaten av uträkningarna pekar på: (1) att komplexitetsgraden är knuten till
språkfamilj, (2) att språk med samma komplexitetsgrad kan uppnå den på olika sätt,
(3) att vissa av parametrarna i beräkningsmetoden kan interagera implikationellt med
varandra (som att strikt semantisk genustilldelning bara färekommer i språk med små
genussystem), och (4) att vissa av måttets komponenter (i synnerhet GV, IND och M1)
korrelerar starkt med varandra och förefaller ha en större inverkan på den sammantagna komplexitetsgraden i undersökningsspråken. Slutligen visar resultaten att språk i
mitt urval som har genus oftast har relativt komplexa genussystem medan genussystem
med låg komplexitetsgrad är sällsynta. Om detta urval kan anses representativt för den
afrikanska makroregionen så visar resultaten att genus är ett komplext drag i afrikanska
språk.
Kapitel 8 ger en sammanfattning av avhandlingen, utvärderar den metodologi som
använts, och skisserar möjliga teman för framtida forskning. Behov av vidare forskning identifieras särskilt för fyra områden: nominala relevanshierarkier; system med delad pluralindexering; genustilldelning och dess manipulering; och absolut komplexitet i
genussystem. Morfosyntaktiska och semantiska interaktioner mellan enskilda grammatiska drag har utforskats grundligt inom domänen verb, med resultatet att drag som
tempus, aspekt, modus och evidentialitet numera betraktas som en sammanhängande
helhet, både på deskriptiv och teoretisk nivå. De resultat som redovisas i denna avhandling visar på några av de fördelar som kan nås genom att utvidga detta angreppssätt
240
Sammanfattning på svenska
till den nominala domänen och se på nominala drag i termer av deras morfosyntaktiska,
semantiska och funktionella interaktion, snarare än som isolerade grammatiska domäner.
241
Bibliography
Acquaviva, Paolo. 2008. Lexical Plurals: A morphosemantic approach. Oxford: Oxford
University Press.
Agbedor, Paul & Yvonne Agbetsoamedo. forthcoming. Ewe. In Nicola Grandi & Lı́via
Körtvélyessy (eds.), Edinburgh handbook of evaluative morphology, Edinburgh University Press.
Agbetsoamedo, Yvonne. 2014. Noun classes in SElEE. The Journal of West African
Languages 41. 95–124.
Agbetsoamedo, Yvonne & Francesca Di Garbo. forthcoming. Selee. In Nicola Grandi &
Lı́via Körtvélyessy (eds.), Edinburgh handbook of evaluative morphology, Edinburgh
University Press.
Aikhenvald, Alexandra. 2003. Classifiers: A typology of noun categorization devices.
Oxford: Oxford University Press.
Aikhenvald, Alexandra. 2012. Round women and long men: Shape, size, and the meanings of gender in New Guinea and beyond. Anthropological Linguistics 54. 33–86.
Ajı́bóyè, O
. ládı́pò.. 2010. Plural strategies in Yorùbá. Studies in African Linguistics 39.
141–181.
Allan, Keith. 1977. Classifiers. Language 53. 285–311.
Allan, Keith. 1980. Nouns and countability. Language 56. 541–567.
Amha, Azeb. 2001. The Maale language. Leiden: Research School of Asian, African and
Amerindian Studies.
Amha, Azeb. 2012. Omotic. In Zygmunt Frajzyngier & Erin Shay (eds.), The Afroasiatic
languages, 423–504. Cambridge: Cambridge University Press.
Andrzejewski, B. W. 1960. The categories of number in noun forms in the Borana dialect
of Galla. Africa 30. 62–75.
Appah, Clement K. I. & Nana Aba Appiah Amfo. 2011. The morphopragmatics of the
diminutive morpheme (-ba/-wa) in Akan. Lexis 6. 85–102. [Special Issue: Diminutives
and augmentatives in the languages of the world].
Appleyard, D. L. 1975. A descriptive outline of Kemant. Language 38. 316–350.
243
BIBLIOGRAPHY
Audring, Jenny. 2009. Gender assignment and gender agreement: Evidence from
pronominal gender languages. Morphology 18. 93–116.
Audring, Jenny. 2014. Gender as a complex feature. Language Sciences 43. 5–17. [Special
issue: Exploring grammatical gender].
Baerman, Matthew, Dunstan Brown & Greville Corbett. 2005. The syntax-morphology
interface: A study of syncretism. Cambridge: Cambridge University Press.
Bakker, Dik. 2011. Language sampling. In Jae Jung Song (ed.), Handbook of linguistic
typology, 100–127. Oxford: Oxford University Press.
Bamgbose, Ayo. 1966. A grammar of Yoruba. Cambridge: Cambridge University Press.
Barlow, Michael. 1992. A situated theory of agreement. New York: Routledge.
Barlow, Ruffell. 1951. Studies in Kikuyu grammar and idiom. Edimburgh: William
Blackwood.
Bauer, Laurie. 1996. No phonetic iconicity in evaluative morphology. Studia Linguistica
50. 189–206.
Bauer, Laurie. 1997. Evaluative morphology: In search of universals. Studies in Language
21. 533–575.
Bauer, Laurie. 2002. What you can do with derivational morphology. In O. E. Pfeiffer
& M. Voeikova (eds.), Morphology 2000: Selected papers from the 9th Morphology
Meeting, Vienna 24-28 February 2000, 37–48. Amsterdam: John Benjamins.
Beachy, Marvin Dean. 2005. An overview of Central Dizin phonology and morphology.
Arlington: The University of Texas at Arlington dissertation.
Beguinot, Francesco. 1942. Il berbero di Nefûsi di Fassâto. Roma: Istituto per l’Oriente.
"
Belnap, Kirk R. 1999. A new perspective on the history of Arabic variation in marking
agreement with plural heads. Folia Linguistica 33. 169–185.
Belnap, Kirk R. & Osama Shabaneh. 1994. Variable agreement and non-human plurals in Classical and Modern Standard Arabic. In Ellen Broselow, Mushira Eid
& John McCarthy (eds.), Perspective on Arabic linguistics IV, 245–262. Amsterdam/Philadelphia: John Benjamins.
Bickel, Balthasar & Johanna Nichols. 2007. Inflectional morphology. In Timothy Shopen
(ed.), Language typology and syntactic despription, vol. 3: Grammatical categories and
the lexicon, 169–240. Cambridge University Press.
Bokamba, E. 2009. The spread of Lingala as a lingua franca in the Congo Basin. In
Fiona McLaughlin (ed.), The languages of urban Africa, 50–70. London: Continuum.
244
BIBLIOGRAPHY
Bokamba, Eyamba. 1977. The impact of multilingualism on language structures: the
case of Central Africa. Anthropological Linguistics 19. 181–202.
Bolozky, Shmuel. 1994. On the formation of diminutives in Modern Hebrew morphology.
Hebrew Studies 35. 47–62.
Borg, Albert & Marie Azzopardi-Alexander. 1997. Maltese. London and New York:
Routledge.
Botne, Robert. 2003. Lega (Beya dialect) (D25). In Derek Nurse & Gérard Philippson
(eds.), The Bantu languages, 422–449. London: Routledge.
Breedveld, J. O. 1995. Form and meaning in Fulfulde: A morphological study of Maasinankoore. Leiden: Research School CNWS.
Bybee, Joan. 1985. Morphology. Amsterdam: John Benjamins.
Bybee, Joan. 1994. The grammaticization of zero: Asymmetries in tense and aspect
systems. In William Pagliuca (ed.), Perspectives on grammaticalization, 235–254.
Amsterdam: John Benjamins.
Carrell, Patricia. 1970. A transformational grammar of Igbo. Cambridge: Cambridge
University Press.
Carstairs, Andrew. 1984. Outlines of a constraint on syncretism. Folia Linguistica 18.
73–85.
Carstairs, Andrew. 1987. Allomorphy in inflection. London: Croom Helm.
Carstairs, Andrew & Joseph Paul Stemberger. 1988. A processing constraint on inflectional homonymy. Linguistics 26. 601–617.
Carter, Hazel. 2002. An outline of Chitonga grammar. Lousaka: Bookwords Publishers.
Childs, T. 1995. A grammar of Kisi. Berlin: Mouton de Gruyter.
Childs, Tucker. 1983. Noun class affix renewal in Southern West Atlantic. In Jonathan
Kaye, Hilda Koopman, Dominique Sportiche & Andr Dugas (eds.), Current approaches
to African linguistics II, 17–29. Dodrecht: Mouton de Gruyter; Foris Publications.
Clements, G. N. & Annie Rialland. 2008. Africa as a phonological area. In Bernd Heine
& Derek Nurse (eds.), A linguistic geography of Africa, 36–85. Cambridge: Cambridge
University Press.
Coffin, Edna A. & Shmuel Bolozky. 2005. A reference grammar of Modern Hebrew.
Cambridge: Cambridge University Press.
Cole, Desmond T. 1955. An introduction to Tswana grammar. Cape Town: Longmans
Green.
245
BIBLIOGRAPHY
Comrie, Bernard. 1975. Polite plurals and predicate agreement. Language 51. 406–18.
Contini-Morava, Ellen. 1997. Noun classification in Swahili. In Robert K. Herbert (ed.),
African linguistics at the crossroads, 599–628. Cologne: Rüdiger Köppe Verlag.
Contini-Morava, Ellen. 2000. Noun class as number in Swahili. In Hellen ContiniMorava & Yishai Tobin (eds.), Between grammar and lexicon, 1–30. Amsterdam: John
Benjamins.
Corbett, Greville. 1991. Gender. Cambridge: Cambridge University Press.
Corbett, Greville. 1998. Prototypical inflection: implications for typology. In Geert Booij
& Jaap van Marle (eds.), Yearbook of morphology 1998, 1–22. Dordrecht: Kluwer.
Corbett, Greville. 2000. Number. Cambridge: Cambridge University Press.
Corbett, Greville. 2006. Agreement. Cambridge: Cambridge University Press.
Corbett, Greville. 2012. Features. Cambridge: Cambridge University Press.
Corbett, Greville. 2013a. Gender typology. In Greville Corbett (ed.), The expression of
gender, 87–130. Berlin: Mouton de Gruyter.
Corbett, Greville. 2013b. Number of genders. In Matthew Dryer & Martin Haspelmath
(eds.), The world atlas of language structures online, Max Planck Digital Library,
chapter 30. Available online at: http://wals.info/chapter/30. Accessed on 2014-02-14.
Corbett, Greville. 2013c. Sex-based and non-sex-based gender systems. In Matthew
Dryer & Martin Haspelmath (eds.), The world atlas of language structures online, Max
Planck Digital Library, chapter 31a. Available online at: http://wals.info/chapter/31.
Accessed on 2014-02-14.
Corbett, Greville. 2013d. Systems of gender assignment. In Matthew S. Dryer
& Martin Haspelmath (eds.), The world atlas of language structures online,
Leipzig: Max Planck Institute for Evolutionary Anthropology. Available online at:
http://wals.info/chapter/32. Accessed on 2014-02-14.
Corbett, Greville & Richard J. Hayward. 1987. Gender and number in Bayso. Lingua
72. 195–222.
Corbett, Greville & Alfred Mtenje. 1987. Gender agreement in Chichewa. Studies in
African Linguistics 72. 195–222.
Creissels, Denis. 1999. Origin et evolution des diminutifs et augmentatifs dans quelques
langues africaines. Silexicales 2. 29–35.
Creissels, Denis, Gerrit J. Dimmendaal, Zygmut Frajzyngier & Christa König. 2008.
Africa as a morphosyntactic area. In Bernd Heine & Derek Nurse (eds.), A linguistic
geography of Africa, 86–150. Cambridge: Cambridge University Press.
246
BIBLIOGRAPHY
Crisma, Paola, Lutz Marten & Rint Sybesma. 2011. The point of Bantu, Chinese and
Romance nominal classification. Italian Journal of Linguistics 23. 251–299.
Cristofaro, Sonia. 2012. Typological universals in synchrony and diachrony: The evolution of number marking. Paper presented at the 45th meeting of the Societas Linguistica Europea, Stockholm, September 2012.
Croft, William. 1994. Semantic universals in classifier systems. Word 45. 145–171.
Croft, William. 2000. Explaining language change: An evolutionary approach. London:
Longman.
Croft, William. 2001. Radical construction grammar. Oxford: Oxford University Press.
Croft, William. 2003. Typology and universals. Cambridge: Cambridge University Press.
Croft, William. 2013. Agreement as anaphora, anaphora as coreference. In Dik Bakker &
Martin Haspelmath (eds.), Languages across boundaries: studies in memory of Anna
Siwierska, 107–129. Berlin: Mouton de Gruyter.
Dahl, Östen. 1990. Standard Average European as an exotic language. In Johannes
Bechert, Giuliano Bernini & Claude Buridant (eds.), Towards a typology of European
languages, 3–8. Berlin: Mouton de Gruyter.
Dahl, Östen. 2000a. Animacy and the notion of semantic gender. In Barbara Unterbeck, Matti Rissanen, Tettu Nevalainen & Mirja Saari (eds.), Gender in grammar and
cognition, 99–115. Berlin: Mouton de Gruyter.
Dahl, Östen. 2000b. Elementary gender distinctions. In Barbara Unterbeck, Matti
Rissanen, Terttu Nevalaien & Mirja Saari (eds.), Gender in grammar and cognition,
577–593. Berlin: Mouton de Gruyter.
Dahl, Östen. 2004. The growth and maintenance of linguistic complexity. Amsterdam:
John Benjamins.
Dahl, Östen. 2006. Diminutives and augmentatives. In Keith Brown (ed.), Encyclopedia
of language and linguistics, 594–595. Oxford: Elsevier.
Dahl, Östen. 2011. Grammaticalization and linguistic complexity. In Heiko Narrog &
Bernd Heine (eds.), The Oxford handbook of grammaticalization, 153–162. Oxford:
Oxford University Press.
Dahl, Östen & Maria Koptjevskaja-Tamm. 2001. Kinship in grammar. In Irene Baron,
Michael Herslund & Finn Sørensen (eds.), Dimensions of possession, 201–225. Amsterdam: John Benjamins.
Dammel, Antje & Sebastian Kürschner. 2008. Complexity in nominal plural allormorphy.
In Matti Miestamo, Kaius Sinnemäki & Fred Karlsson (eds.), Language complexity:
Typology, contact, change, 243–262. Amsterdam: John Benjamins.
247
BIBLIOGRAPHY
Daniel, Michael & Edith Moravcsik. 2013. The associative plural. In Matthew S.
Dryer & Martin Haspelmath (eds.), The world atlas of language structures online,
Leipzig: Max Planck Institute for Evolutionary Anthropology. Available online at:
http://wals.info/chapter/36. Accessed on 2014-02-14.
De Mauro, Tullio, Federico Mancini, Massimo Vedovelli & Miriam Voghera. 1993. Lessico
di frequenza dell’italiano parlato. Milano: Etatslibri.
Delafosse, Maurice. 1901. Essai de manuel pratique de la langue mande ou mandingue.
Paris: Ernest Leroux.
Denny, Peter & A. Chet Creider. 1976. The semantics of noun classes in Proto-Bantu.
Studies in African Linguistics 7. 1–30.
Desmuth, Katherine. 2000. Bantu noun class systems: loanword and acquisition evidence
of semantic productivity. In Senft Gunther (ed.), Systems of noun classification, 270–
92. Cambridge: Cambridge University Press.
Dickens, Patrick. 1992.
Namibia.
Ju|’hoan grammar. ms.
Nyae Development Foundation of
Diessel, Holger. 1999a. Demonstratives: Form, function and grammaticalization. Amsterdam: John Benjamins.
Diessel, Holger. 1999b. The morphosyntax of demonstratives in synchrony and diachrony.
Linguistic Typology 3. 1–49.
Dimmendaal, Gerrit. 1983. The Turkana language. Dordrecht: Foris Publications.
Dimmendaal, Gerrit. 2000. Number marking and noun categorization in Nilo-Saharan
languages. Anthropological Linguistics 42. 214–261.
Dimmendaal, Gerrit. 2008. Language ecology and linguistic diversity on the African
continent. Language and Linguistics Compass 2. 840–858.
Dixon, Robert M. W. 1982. Where have all the adjectives gone? And other essays in
semantics and syntax. Berlin: Mouton de Gruyter.
Dixon, Robert M. W. 2004. Adjective classes in typological perspective. In Adjective
classes: A cross-linguistic typology, 1–45. Oxford: Oxford University Press.
Dressler, Wolfgang & Lavinia Merlini Barbaresi. 1994. Morphopragmatics: Diminutives
and intensifiers in Italian, German, and other languages. Berlin: Mouton de Gruyter.
Dryer, Matthew. 1989a. Large linguistic areas and language sampling. Studies in Language 13. 257–292.
Dryer, Matthew. 1989b. Plural words. Linguistics 27. 865–895.
248
BIBLIOGRAPHY
Dryer, Matthew & Martin Haspelmath (eds.). 2013. The world atlas of language structures online. Leipzig: Max Planck Institute for Evolutionary Anthropology. Available
online at http://wals.info, Accessed on 2014-02-14.
Dryer, Matthew S. 1992. The Greenbergian word order correlations. Language 68.
81–138.
Dryer, Matthew S. 2013. Coding of nominal plurality. In Matthew S. Dryer
& Martin Haspelmath (eds.), The world atlas of language structures online,
Leipzig: Max Planck Institute for Evolutionary Anthropology. Available online at:
http://wals.info/chapter/33. Accessed on 2014-02-14.
Eaton, Helen. 2010. A Sandawe grammar. Dallas: SIL International.
Edenmyr, Niklas. 2004. The semantics of Hadza gender assignment: a few notes from
the field. Africa & Asia 4. 3–19.
Fachner, Regine. 1990. The noun phrase in Gola. Afrikanistische Arbeitspapiere 23.
5–52.
Ferry, Marie-Paule & Konstantin Pozdniakov. 2001. Dialectique du régulier et de
l’irrégulier: le système de classes nominales dans le groupe tenda des langues atlantiques. In Robert Nicolaı̈ (ed.), Leçons d’Afrique: Filiations, ruptures et reconstitution
de langues. Un hommage à Gabriel Manessy, 153–167. Paris: Peeters.
Fleming, Harold. 1969. The classification of West Cushitic within Hamito-Semitic. In
Daniel McCall, Norman Bennet & Jeffrey Butler (eds.), Eastern African history, 3 –
27. New York: Praeger.
Foley, W. A. & R. Van Valin. 1984. Functional syntax and universal grammar. Cambridge: Cambridge University Press.
Fortune, G. 1955. An analytic grammar of Shona. London: Longmans Green.
Frajzyngier, Zygmunt. 1989. A grammar of Pero. Berlin: Dietrich Reimer Verlag.
Frajzyngier, Zygmunt. 1993. A grammar of Mupun. Berlin: Dietrich Reimer Verlag.
Frajzyngier, Zygmunt. 1997. Grammaticalization of number: From demonstratives to
nominal and verbal plural. Linguistic Typology 1. 193–242.
Frajzyngier, Zygmunt. 2001. A grammar of Lele. Standford, CA: CSLI Publications:
Center for the Study of Language and Information.
Frajzyngier, Zygmunt. 2008. A grammar of Gidar. Frankfurt am Main: Peter Lang.
Frajzyngier, Zygmunt & Eric Johnston. 2005. A grammar of Mina. Berlin: Mouton de
Gruyter.
249
BIBLIOGRAPHY
Frajzyngier, Zygmunt & Erin Shay. 2002. A grammar of Hdi. Berlin: Mouton de Gruyter.
Garbell, Irene. 1965. The Jewish Neo-Aramaic dialect of Persian Azerbaijan. Berlin:
Mouton de Gruyter.
Gil, David. 1996. Maltese ‘collective nouns’: A typological perspective. Rivista di
Linguistica 8. 53–87.
Gil, David. 2013. Numeral classifiers. In Matthew Dryer & Martin Haspelmath (eds.),
The world atlas of language structures online, Max Planck Digital Library, chapter
55. Available online at: http://wals.info/chapter/55. Accessed on 2014-02-14.
Givón, Talmy. 1971. Some historical changes in the noun class system of Bantu. In C. W.
Kim & S. Herbert (eds.), Papers in African linguistics, 33–54. Alberta: Linguistic
Research Inc.
Glinert, Lewis. 1994. Modern Hebrew: An essential grammar. London and New York:
Routledge.
Grandi, Nicola. 2001. Su alcune presunte anomalie della morfologia valutativa: Il rapporto con il genere ed il numero. Archivio Glottologico Italiano 1. 25–56.
Grandi, Nicola. 2002. Morfologie in contatto: Le costruzioni valutative nelle lingue del
Mediterraneo. Milano: Franco Angeli.
Grandi, Nicola. 2011. Renewal and innovation in the emergence of Indo-European evaluative morphology. Lexis 6. 5–25. [Special Issue: Diminutives and augmentatives in
the languages of the world].
Grandi, Nicola. forthcoming-a. Berber languages. In Nicola Grandi & Lı́via Körtvélyessy
(eds.), Edinburgh handbook of evaluative morphology, Edinburgh University Press.
Grandi, Nicola. forthcoming-b. Evaluative morphology and number/gender. In Nicola
Grandi & Lı́via Körtvélyessy (eds.), Handbook of evaluative morphology, Edinburgh
University Press.
de la Grasserie, Raoul. 1898. La catégorie psychologique de la classification: révélée par
la langage. Revue de Philolosophique de la France et de l’Étranger 45. 594–624.
Green, M. M. & Rev. G. E. Igwe. 1963. A descriptive grammar of Igbo. Berlin: AkademieVerlag. KB.
Greenberg, Joseph. 1959. Africa as a linguistic area. In William Bascom & Melville
Herskovits (eds.), Continuity and change in African cultures, 15–27. Chicago: Chicago
University Press.
Greenberg, Joseph. 1963a. The languages of Africa. Bloomington: Indiana University.
250
BIBLIOGRAPHY
Greenberg, Joseph. 1963b. Some universals of grammar with particular reference to the
order of meaningful elements. In Joseph H. Greenberg (ed.), Universals of language,
73–113. Cambridge, MA: MIT Press.
Greenberg, Joseph. 1978. How does a language acquire gender markers? In Joseph
Greenberg, Charles Ferguson & Edith Moravcisk (eds.), Universals of human language,
vol. 3: Word structure, 47–92. Stanford: Standford University Press.
Greenberg, Joseph. 1980. Circumfixes and typological change. In Elisabeth Traugott
(ed.), Papers from the 4th International Conference on Historical Linguistics, 233–241.
Amsterdam and Philadelphia: John Benjamins.
Greenberg, Joseph. 1983. Some areal characteristics of African languages. In Ivan Dihoff (ed.), Current approaches to African linguistics 1, 3–22. Dordecht/Cinnaminson:
Foris.
Gregová, Renátá, Lı́via Körtvélyessy & Július Zimmermann. 2010. Phonetic iconicity in
the evaluative morphology of a sample of Indo-European, Niger-Congo and Austronesian languages. Word Structure 3. 156–180.
Guarisma, Gladys. 2003. KpāÜ. In Derek Nurse & Gérard Philippson (eds.), The Bantu
languages, 307–334. London: Routledge.
Gudmundson, Anna. 2012. L’accordo nell’italiano parlato da apprendenti universitari
svedesi: Uno studio sull’acquisizione del numero e del genere in una prospettiva funzionalista. Stockholm: Stockholm University dissertation.
Güldemann, Tom. 1998. The Kalahari Basin as an object of areal typology: A first
approach. Khoisan Phorum, Working Papers 3. Köln: Institut für Afrikanistik,
Universität zu Köln.
Güldemann, Tom. 1999. Head-initial meets head-final: nominal suffixes in eastern and
southern Bantu from a historical perspective. Studies in African Linguistics 29. 49–91.
Güldemann, Tom. 2000. Noun categorization systems in Non-Khoe lineages of Khoisan.
Afrikanistische Arbeitspapiere 63. 5–33.
Güldemann, Tom. 2004. Reconstruction through ‘de-construction’: the marking of person, gender and number in the Khoe family and Kwadi. Diachronica 21. 251–306.
Güldemann, Tom. 2008. The Macro-Sudan belt: Towards identifying a linguistic area
in northern sub-Saharan Africa. In Bernd Heine & Derek Nurse (eds.), A linguistic
geography of Africa, 151–185. Cambridge: Cambridge University Press.
Güldemann, Tom. 2010. Sprachraum and geography: Linguistic macro-areas in Africa.
In Alfred Lameli, Roland Kehrein & Stefan Rabanus (eds.), Language and space: An
international handbook of linguistic variation, vol. 2: Language mapping, 561–585,
Maps 2901–2914. Berlin: Mouton de Gruyter.
251
BIBLIOGRAPHY
Güldemann, Tom & Rainer Vossen. 2000. Khoisan. In Bernd Heine & Derek Nurse (eds.),
African languages: An introduction, 99–122. Cambridge: Cambridge University Press.
Guthrie, Malcom. 1966. Grammaire et dictionnaire de lingala. Westmead: Greg Press
Limited.
Hagman, Roy S. 1977. Nama Hottentot grammar. Bloomington: Indiana University.
Hämeen-Anttila, Jaakko. 2000. Grammatical gender and its development in Classical
Arabic. In Barbara Unterbeck, Matti Rissanen, Tettu Nevalainen & Mirja Saari (eds.),
Gender in grammar and cognition, 595–608. Berlin: Mouton de Gruyter.
Harrell, Frank E. 2001. Regression modeling strategies. New York: Springer.
Harrell, Richard S. 1965. A short reference grammar of Moroccan Arabic. Washington
D.C.: Georgetown University Press.
Haspelmath, Martin. 2001. The European linguistic area: Standard Average European.
In Martin Haspelmath, Ekkehard König, Wulf Oesterreicher & Wolfgang Raible (eds.),
Language typology and language universals: An international handbook. (Handbücher
zur Sprach- und kommonikationwissenshaft), 1492–1510. Berlin: Mouton de Gruyter.
Haspelmath, Martin. 2006. Against markedness (and what to replace it with). Journal
of Linguistics 42. 25–70.
Haspelmath, Martin. 2007. Pre-established categories don’t exist: Consequences for
language description and typology. Linguistic Typology 11. 119–132.
Haspelmath, Martin. 2009. An empirical test of the agglutination hypothesis. In Sergio
Scalise, Francesca Masini, Elisabetta Magni & Antonietta Bisetto (eds.), Universals
of language today, 13–29. Dordrecht: Springer.
Haspelmath, Martin. 2012.
Continent-wide typology: a recent trend within
diversity linguistics.
In Diversity linguistics comment, Available online at
http://dlc.hypotheses.org/340.
Haspelmath, Martin. 2013. Occurrence of nominal plurality. In Matthew Dryer & Martin
Haspelmath (eds.), The world atlas of language structures online, Max Planck Digital
Library, Chapter 34. Available online at: http://wals.info/chapter/34. Accessed on
2014-02-14.
Heath, Jeffrey. 2005. A grammar of Tamashek (Tuareg of Mali). Mouton de Gruyter.
Heath, Teresa. 2003. Makaa (A 83). In Derek Nurse & Gerard Philippson (eds.), The
Bantu languages, 335–348. London: Routledge.
Heine, Bernd. 1970. Status and use of African lingua francas. München: Welt Forum
Verlag.
252
BIBLIOGRAPHY
Heine, Bernd. 1975. Language typology and corvengence areas in Africa. Linguistics
144. 27–47.
Heine, Bernd. 1976. A typology of African languages based on the order of meaningful
elements Kölner Beiträge zur Afrikanistik 4. Berlin: Dietrich Reimer Verlag.
Heine, Bernd. 1982. African noun class systems. In H. Seiler & C. Lehmann (eds.), Apprehension: Das sprachliche Erfassen von Gegenständen, 189–216. Tübingen: Gunter
Narr Verlag.
Heine, Bernd. 1997. Grammaticalization chains across languages: an example from
Khoisan. Paper presented at the Symposium on Grammaticalization, Houston, March
1997.
Heine, Bernd. 1999. The {Ani: Grammatical Notes and Texts. Khoisan Phorum 11.
Universität zu Köln, Institut für Afrikanistik.
Heine, Bernd & Christa König. 2011. The !Xun language: a dialect grammar of Northern
Khoisan. Cologne: Rüdiger Köppe.
Heine, Bernd & Christa König. 2013. Northern Khoesan: !Xun. In Rainer Vossen (ed.),
The Khoesan languages, 293–325. London and New York: Routledge.
Heine, Bernd & Tania Kuteva. 2006. The changing languages of Europe. Oxford: Oxford
University Press.
Heine, Bernd & Derek Nurse. 2000. Introduction. In Bernd Heine & Derek Nurse (eds.),
African languages: An introduction, 1–10. Cambridge: Cambridge University Press.
Heine, Bernd & Derek Nurse. 2008. Introduction. In Bernd Heine & Derek Nurse (eds.),
A linguistic geography of Africa, 1–14. Cambridge: Cambridge University Press.
Heine, Bernd & Rainer Vossen. 1983. On the origin of gender in Eastern Nilotic. In
Rainer Vossen & Marianne Becchaus-Gerst (eds.), Nilotic studies: Proceedings of the
international symposium on languages and history of the Nilotic peoples, Cologne,
January 4-6, 1982, 245–268. Köln: Dietrich Reimer Verlag.
Hetzron, Robert. 1978. The nominal system of Awngi (Southern Agaw). Bulletin of the
School of Oriental and African Studies, University of London 41. 121–141.
Hockett, Charles F. 1958. A course in modern linguistics. New York: Macmillan.
Hulstaert, Gustaaf. 1965. Grammaire du lomongo. deuxième partie: Morphologie. Tervuren: Musée Royale de l’Afrique Centrale.
Iemmolo, Giorgio. 2011. Towards a typological study of Differential Object Marking and
Differential Object Indexation. Pavia: Università degli studi di Pavia dissertation.
Jackendoff, Ray. 1991. Parts and boundaries. Cognition 41. 9–45.
253
BIBLIOGRAPHY
Jespersen, Otto. 1922. Language: Its nature, development and origin. London: Allen
and Unwin.
Jurafsky, Daniel. 1996. Universal tendencies in the semantics of the diminutive. Language
72. 533–578.
Katamba, Francis. 2003. Bantu nominal morphology. In Derek Nurse & Gérard Philippson (eds.), The Bantu languages, 103–120. London: Routledge.
Kavari, Jekura U. & Lutz Marten. 2009. Multiple noun class prefixes in Otjiherero.
In Peter Austin, Oliver Bond, Monik Charette, David Nathan & Peter Sells (eds.),
Proceeding of Conference on Language Documentation & Linguistic Theory 2, London:
SOAS.
Khan, Geoffrey. 2008. Jewish Neo-Aramaic dialect of Urmi. Piscataway, NJ: Gorgias
Press.
Kibrik, Aleksander. 1992. Defective paradigms: number in Daghestanian. In EUROTYP
working papers, Theme 7, vol. 16, Strassbourg: ESF.
Kilian-Hatz, Christa & Bernd Heine. 2010. On nominal gender marking in Khwe.
Khoisan Phorum Working Paper 1. 2–26. Revised version of 1997.
Koptjevskaja-Tamm, M. 2004. Mass and collection. In Geert E. Booij, Christian
Lehmann, Joachim Mugdan, Stavros Skopeteas & E. Booij (eds.), Morphology: A
handbook on inflection and word formation, 1016–1031. Berlin: Mouton de Gruyter.
Koptjevskaja-Tamm, Maria. 2002. Number (book review). Journal of pragmatics 32.
1911–1919.
Koptjevskaja-Tamm, Maria. 2011. Language contact. In Jae-Jung Song (ed.), The
Oxford handbook of linguistic typology, 568–590. Oxford: Oxford University Press.
Koptjevskaja-Tamm, Maria & Bernhard Wälchli. 2001. The Circum-Baltic languages:
An areal-typological approach. In Östen Dahl & Maria Koptjevskaja-Tamm (eds.),
Circum-Baltic languages, vol. 2: Grammar and typology, 615–750. Amsterdam: John
Benjamins.
Körtvélyessy, Lı́via. 2011. A cross-linguistic research into phonetic iconicity. Lexis 6.
27–39. [Special Issue: Diminutives and augmentatives in the languages of the world].
Körtvélyessy, Lı́via. 2012. Evaluative morphology from cross-linguistic perspective. In
Paper presented at the conference on “Universals and Typology in Word-Formation
II”, Košice, .
Körtvélyessy, Lı́via & Pavel Stekauer. 2011. Introduction. Lexis 6. 3. [Special Issue:
Diminutives and augmentatives in the languages of the world].
254
BIBLIOGRAPHY
Kossman. 2014.
Derivational gender in Moroccan Berber: examples from Ayt
Seghrushen. STUF – Language Typology and Universals 67. 21–33.
Kusters, Wouter. 2003. Linguistic complexity: The influence of social change on verbal
inflections. Utrecht: LOT: University of Leiden dissertation.
Kusters, Wouter. 2008. Complexity in linguistic theory, language learning and language
change. In Matti Miestamo, Kaius Sinnemäki & Fred Karlsson (eds.), Language complexity: Typology, contact, change, 3–22. Amsterdam: John Benjamins.
Lacan, Ph. 1942. Français-soussou soussou-français. Bordeaux: Librairie d’Amérique
et d’Orient.
Lakoff, George. 1987. Women, fire and dangerous things. Chicago: University of Chicago
Press.
Lamberti, Marcello & Roberto Sottile. 1997. The Wolaytta language. Köln: Köppe. MI.
Langacker, Ronald. 1987a. Nouns and verbs. Language 63. 53–94.
Langacker, Ronald W. 1987b. Foundations of cognitive grammar, vol. 1: Theoretical
prerequisites. Standford: Standford University Press.
Lass, Roger. 1990. How to do things with junk: Exaptation in language evolution.
Journal of Linguistics 26. 79–102.
Leitch, Myles. 2003. Babole (C 101). In Derek Nurse & Gérard Philippson (eds.), The
Bantu languages, 392–421. London: Routledge.
Leslau, Wolf. 1945. Grammatical sketches in Tigré (North Ethiopic): Dialect of Mensa.
Journal of the American Oriental Society 65. 164–203.
Leslau, Wolf. 1948. Supplementary observations on Tigré grammar. Journal of the
American Oriental Society 68. 127–139.
Leslau, Wolf. 1995. Reference grammar of Amharic. Wiesbaden: Harassowitz.
Lewis, M. P., Gary F. Simons & Charles D. Fenning (eds.). 2013. Ethnologue: Languages of the world, seventeenth edition. Dallas: SIL International. Online version:
http://www.ethnologue.com.
Lojenga, Constance Kutsch. 2003. Bila (D 32). In Derek Nurse & Gérard Philippson
(eds.), The Bantu languages, 450–474. London: Routledge.
Luraghi, Silvia. 2011. The origin of the Proto-Indo-European gender system: Typological
considerations. Folia Linguistica 45. 435–464.
Lyons, John. 1968. Introduction to theoretical linguistics. Cambridge: Cambridge University Press.
255
BIBLIOGRAPHY
Maho, Jouni. 1999. A comparative study of Bantu noun classes. Göteborg: Orientalia
et Africana Gothoburgensia dissertation. Acta universitatis gothoburgensia.
Maho, Jouni. 2003. Remarks on a few “polyplural” classes in Bantu. Africa & Asia 3.
161–184.
Marten, Lutz. 2010. The great siSwati locative shift. In Anne Breitbarth, Christopher
Lucas, Sheila Watts & David Willis (eds.), Continuity and change in grammar, 249–
267. Amsterdam: Benjamins.
Matasovič, Ranko. 2004. Gender in Indo-European. Heidelberg: Winter.
Matisoff, James A. 1992. The mother of all morphemes: Augmentatives and diminutives
in areal and universal perspective. In Martha Ratliff & Eric Shiller (eds.), Papers from
the first annual meeting of the Southeast Asian linguistic society, 293–249. Arizona
State University, Program for Southeast Asian Studies.
McGregor, William. 2003. The nothing that is, the zero that isn’t. Studia Linguistica
57. 75–119.
McLaughlin, Fiona. 1992. Noun classification in Seereer-Siin. Austin: The University
of Texas dissertation.
McLaughlin, Fiona. 1997. Noun classification in Wolof: When affixes are not renewed.
Studies in African Linguistics 26. 1–28.
McLaughlin, Fiona (ed.). 2009. The languages of urban Africa. London: Continuum.
McWhorter, John H. 2001. The world’s simplest grammars are creole grammars. Linguistic Typology 5. 125–166.
Meeussen, A. E. 1959. Essai de grammaire rundi. Tervuren: Musée Royal du Congo
Belge.
Mel’ čuck, Igor & Élie Bakiza. 1997. Les classes nominales en kirundi. Bulletine de la
société de linguistique de Paris 82. 283–341.
Mettouchi, Amina. 2000. La “t” n’est-il qu’une marque de féminin en berbère (kabyle)?
Faits de Langue 14. 217–225.
Miestamo, Matti. 2008. Grammatical complexity in a cross-linguistic perspective. In
Matti Miestamo, Kaius Sinnemäki & Fred Karlsson (eds.), Language complexity: Typology, contact, change, 23–41. Amsterdam: John Benjamins.
Miestamo, Matti, Kaius Sinnemäki & Fred Karlsson (eds.). 2008. Language complexity:
Typology, contact, changes. Amesterdam: John Benjamins.
Mithun, Marianne. 1988. Lexical categories and the evolution of verbal number. In
Noonan Michael & Michael Hammond (eds.), Theoretical morphology, 211 – 233. New
York: Academic Press.
256
BIBLIOGRAPHY
Moravcsik, Edith. 2003. A semantic analysis of associative plurals. Studies in Language
27. 469–503.
Moscati, S. 1964. An introduction to the comparative grammar of the Semitic languages.
Wiesbaden: Otto Harrasowitz.
Mous, Maarten. 1993. A grammar of Iraqw. Hamburg: Buske.
Mous, Maarten. 2003a. Nen (A 44). In Derek Nurse & Gérard Philippson (eds.), The
Bantu languages, 283–306. London: Routledge.
Mous, Maarten. 2008. Number as an exponent of gender in Cushitic. In Zygmunt
Frajzyngier & Erin Shay (eds.), Interaction of morphology and syntax: Case studies
in Afroasiatic, 137–160. Amsterdam: John Benjamins.
Mous, Marteen. 2012. Cushitic. In Zygmunt Frajzyngier & Erin Shay (eds.), The
Afroasiatic languages, 342–422. Cambridge: Cambridge University Press.
Mous, Marten. 2003b. Loss of linguistic diversity in Africa. In Mark Janse & Sijmen Tol
(eds.), Language death and language maintenance: Theoretical, practical and descriptive approaches, 157–170. Amsterdam: John Benjamins.
Mufwene, Salikoko. 2003. Contact languages in the Bantu area. In Derek Nurse &
Gérard Philippson (eds.), The Bantu languages, 195–208. London: Routledge.
Myachina, E. N. 1981. The Swahili language. London: Routledge & Kegan Paul.
Newman, Paul. 1990. Nominal and verbal plurality in Chadic. Dordrecht: Foris Publications.
Newman, Paul. 2000. The Hausa language: an encyclopedic reference grammar. New
Haven: Yale University Press.
Nichols, Johanna. 1971. Diminutive consonant symbolism in Western North America.
Language 47. 826–848.
Nichols, Johanna. 1992. Linguistic diversity in space and time. Chicago: University of
Chicago Press.
Nicolas, Francis. 1953. La langue berbère de Mauritanie. Dakar: IFAN.
Nordhoff, Sebastian, Harald Hammarström, Robert Forkel & Martin Haspelmath. 2013.
Glottolog 2.2. Max Planck Institute for Evolutionary Anthropology. Available online
at http://glottolog.org, Accessed on 2014-08-19.
Novelli, Bruno. 1985. A grammar of the Karimojong language. Berlin: Drietrich Reimer
Verlag.
257
BIBLIOGRAPHY
Nurse, Derek & Gérard Philippson. 2003. Towards a historical classification of the Bantu
languages. In Derek Nurse & Gérard Philippson (eds.), The Bantu languages, 164–181.
London: Routledge.
Oomen, Antoinette. 1981. Gender and plurality in Rendille. Afroasiatic linguistics 8.
85–75.
Osam, Emmanuel Kweku. 1993. The loss of the noun class system in Akan. Acta
Linguistica Hafniensia: International Journal of Linguistics 26. 81–106.
Parkvall, Mikael. 2008. The simplicity of creoles in a cross-linguistic perspective. In Matti
Miestamo, Kaius Sinnemäki & Fred Karlsson (eds.), Language complexity: Typology,
contact, change, 265–285. Amsterdam: John Benjamins.
Payne, Doris L. 1998. Maasai gender in typological perspective. Studies in African
Linguistics 27. 159–175.
Payne, Thomas. 1997. Describing morpho-syntax: A guide for field linguists. Cambridge:
Cambridge University Press.
Penchoen, Thomas. 1973a. Etude sintaxique d’un parler berbère (Ait-Frah de l’Aurès).
Napoli: Centro di studi magrebini.
Penchoen, Thomas G. 1973b. Tamazight of the Ayt Ndhir. Los Angeles: Undena Publications.
Petzell, Malin. 2008. The Kagulu language of Tanzania: Grammar, texts and vocabulary.
Köln: Rüdiger Köppe Verlag.
Plank, Frans. 1999. Split morphology: How agglutination and flexion mix. Linguistic
Typology 3. 279–340.
Plank, Frans & Elena Filimonova. 2009. The universals archive. Konstanz: Universitát
Konstanz. Available online at http://typo.uni-konstanz.de/archive/intro/index.
Plank, Frans & Wolfgang Schellinger. 1997. The uneven distribution of genders over
numbers: Greenberg Nos. 37 and 45. Linguistic Typology 1. 53–101.
Poulos, George. 1990. A linguistic analysis of Venda. Pretoria: Via Afrika Limited.
Poulos, George & Sonja Bosh. 1997. Zulu. München: Lincom Europa.
Poulos, George & Louis J. Louwrens. 1994. A linguistic analysis of Northern Sotho.
Pretoria: Via Afrika Limited.
Pozdniakov, Konstantin. 2010. La classification nominale: à la croisée des paradigmes. In
Franck Floricic (ed.), Essais de typologie et de linguistique générale: Mélanges offerts
à Denis Creissels, 87–105. Paris: ENS Éditions.
258
BIBLIOGRAPHY
Pozdniakov, Konstantin. 2013. Protolanguage and prototype: a “proto-letter” and a
“proto-spirit” in noun classes in Niger-Congo. Histoire, épistémologie, langage 35.
65–82.
Quirk, Randolph, Sidney Greenbaum, Geoffrey Leech & Jan Svartvik. 1985. A comprehensive grammar of the English language. London and New York: Longman.
Rapold, Christian J. 2006. Towards a grammar of Benchnon. Leiden: University of
Leiden dissertation.
Reid, Nicholas. 1990. Ngan’gityemerri: A language of the Daly River region, northern
territory of Australia. Canberra: Australian National University dissertation.
Reid, Nicholas. 1997. Class and classifier in Ngan’gityemerri. In Mark Harvey & Nicholas
Reid (eds.), Nominal classification in aboriginal Australia, 165–225. Amsterdam: John
Benjamins.
Roper, E. M. 1928. Tu BedawiE: An elementary handbook for the use of Sudan governament officials. Hertford: Stephen Austin and Sons, LTD.
Ryding, Karin. 2005. A reference grammar of Modern Standard Arabic. New York:
Cambridge University Press.
Saeed, John I. 1987. Somali reference grammar. Wheaton: Dunwoody Press.
Sagna, Serge. 2008. Formal and semantic proprties of the Gujjolay Eegima (a.k.a Banjal)
nominal classification system. London: University of London – School of Oriental and
African Studies dissertation.
Sagna, Serge. 2011. Semantic categorizations in the Gújjolaay Eegima collectives and
distributives. In Oliver Bond, Peter K. Austin, David Nathan & Marten Lutz (eds.),
Proceedings of Conference on Language Documentation and Linguistic Theory 3, 237–
246. London: Department of Linguistics, SOAS.
Sagna, Serge. 2012. Physical properties and culture-specific factors as principles of
semantic categorization of the Gújjolaay Eegima noun class systems. Cognitive Linguistics 23. 129–163.
van Sanbeek, Jan Cornelius. 1955. A Bemba grammar. Cape Town: Longmans, Green.
Santos, Rosine Sachot. 1996. Le mey. Paris: Universite de la Sorbonne Nouvelle dissertation.
Sapir, Edward. 1951. A study in phonetic symbolism. In David G. Mandelbaum (ed.),
Selected writings of Edward Sapir in language, culture and personality, 61–72. Berkeley
and Los Angeles: University of California Press.
Sasse, Hans-Jrgen. 1974. Notes on the structure of Galab. Bulletin of the School of
Oriental and African Studies, University of London 37. 407–438.
259
BIBLIOGRAPHY
Savà, Graziano. 2005. A grammar of Ts’amakko. Köln: Rüdiger Köppe Verlag.
Savà, Graziano. 2007. Interaction between gender and number in Cushitic: The case
of Ts’amakko. In Marco Moriggi (ed.), XII incontro italiano di linguistica camitosemitica (afroasiatica), 203–212. Rubbettino: Soveria Mannelli.
Schuh, Russel G. 1989. Gender and number in Miya. In Zygmunt Frajzyngier (ed.),
Current progress in Chadic linguistics, 171–181. Amsterdam: John Benjamins.
Segerer, Guillaume. 2002. La langue bijogo de bubaque (guiné bissau). Paris, Louvain:
Peeters.
Segerer, Guillaume. 2010. ‘Isolates’ in ‘Atlantic’. Paper presented at the workshop on
Language Isolates in Africa (Lyon, December 3-4, 2010).
Serzisko, Fritz. 1992. Collective and transnumeral nouns in Somali. In Adam M. Hussein
& Charles L. Geshekter (eds.), Proceedings of the first international congress of Somali
studies, 513–525. Atlanta: Scholar Press.
Seyoum, Mulugeta. 2008. A grammar of Dime. Leiden: Netherlands Graduate School
of Linguistics dissertation.
Shadeberg, Tilo. 2003. Derivation. In Derek Nurse & Gérard Philippson (eds.), The
Bantu languages, 71–89. London: Routledge.
Shirtz, Shahar & Doris Payne. 2012. Gendered prefixes in Maa. Paper presented at the
45th meeting of the Societas Linguistica Europea, Stockholm, September 2012.
Siewierska, Anna. 1999. From anaphoric pronoun to grammatical agreement marker:
why objects don’t make it. Folia Linguistica 33. 225–251.
Siewierska, Anna. 2004. Person. Cambridge: Cambridge University Press.
Sinnemäki, Kaius. 2011. Language universals and linguistic complexity. University of
Helsinki: General Linguistics, Department of Modern Languages dissertation.
Smith-Stark, Cedric. 1974. The plurality split. Chicago Linguistic Society 10. 657–671.
Soukka, Maria. 2000. A descriptive grammar of Noon. München: Lincom Europa.
Storch, Anne. 2005. The noun morphology of Western Nilotic. Köln: Rüdiger Köppe
Verlag.
Storch, Anne. 2007. Grammatical change and emblematic features in Western Nilotic. In
Doris L. Payne & Jaime Pen̂a (eds.), Selected proceedings of the 37th annual conference
on African linguistics, 1–15. Somerville, MA: Cascadilla Proceedings Project.
Ström, Eva-Marie. 2011. The increasing importance of animacy in the agreement system
of Ndengeleko and other Southern Coastal Bantu languages. In Jan-Olof Svantesson,
Niclas Burenhult, Arthur Holmer, Anastasia Karlsson & Håkan Lundström (eds.),
Language Documentation and Description, Vol. 10, 265–287. SOAS.
260
BIBLIOGRAPHY
Stroomer, Harry. 1995. A grammar of Boraama Oromo. Köln: Rüdiger Köppe Verlag.
Stump, Gregory. 1993. How peculiar is evaluative morphology? Journal of Linguistics
29. 1–36.
Taljard, P.C., J.N. Khumalo & E.S Bosh S. 1991. Handbook of Siswati. Pretoria: J. L.
Schaik.
Talmy, Leonard. 2000a. Towards a cognitive semantics, vol. 1: Typology and process in
concept structuring. Cambridge, Massachussets: The MIT Press.
Talmy, Leonard. 2000b. Towards a cognitive semantics, vol. 2: Concept structuring
systems. Cambridge, Massachussets: The MIT Press.
Taylor, Charles. 1985. Nkore-Kiga. Sidney: Croom Helm.
Teketal, Hirut Woldemariam. 2004. The grammar of Haro with comparative notes on
the Ometo linguistic group. Adis Abab: Addis Ababab University dissertation.
Tendeng, Odile. 2000. Le gusiilay: un essai de systématisation. une contribution à l’étude
du jóola. Freiburg: Albert-Ludwigs-Universität zu Freiburg dissertation.
Thompson, Kathrina D. & Antonia F. Schleicher. 2001. Swahili’s learner’s reference
grammar. Madison, WI: NALCR Press.
Tosco, Mauro. 1991. A grammatical sketch of Dahalo. Hamburg: Helmut Buske.
Tosco, Mauro. 2001. The Dhaasanac language. Köln: Rüdiger Köppe Verlag.
Traill, Anthony. 1994. A !Xóõ dictionary. Köln: Köppe.
Travélé, Moussa. 1955. Petite manuel franç-bambara. Librairie Orientaliste Paul Geuthner.
Treis, Yvonne. 2008. A grammar of Kambaata. Cologne: Rüdiger Köppe.
Treis, Yvonne. 2011. Number in Kambaata. Paper presented at the Workshop on “Number in Africa and beyond”, Cologne, 27-30 Sep.
Treis, Yvonne. 2014. Number in Kambaata. In Anne Storch & Gerrit Dimmendaal (eds.),
Number: Constructions and semantics. Case studies from Africa, India, Amazonia
and Oceania., 111–133. Amsterdam: John Benjamins.
Trudgill, Peter. 1999. Language contact and the function of linguistic gender. Poznań
Studies in Contemporary Linguistics 35. 133–152.
Ultan, Russell. 1978. Size-sound symbolism. In Joseph Greenberg (ed.), Universals of
human language, vol. 2: Phonology, 525–568. Stanford: Stanford University Press.
Vafaeian, Ghazaleh. 2013. Typology of nominal and adjectival suppletion. STUF –
Language Typology and Universals 66. 112–140.
261
BIBLIOGRAPHY
Van de Velde, Mark. 2006. Multifunctional agreement patterns in Bantu and the possibility of genderless nouns. Linguistic Typology 10. 183–221.
Van de Velde, Mark. 2008. A grammar of Eton. Berlin: Mouton de Gruyter.
Veselinova, Ljuba. 2006. Suppletion in verb paradigms. Amsterdam: John Benjamins.
Veselinova, Ljuba. 2013. Verbal number and suppletion. In Matthew S. Dryer
& Martin Haspelmath (eds.), The world atlas of language structures online,
Leipzig: Max Planck Institute for Evolutionary Anthropology. Available online at:
http://wals.info/chapter/80. Accessed on 2014-02-14.
Veselinova, Ljuba. forthcoming. Sampling procedures. In WSK dictionary on theories and methods in linguistics, Online publication under the auspices of Mouton de
Gruyter.
Vossen, Rainer. 1986. Some observations on nominal gender in Naro. In F. Rottland (ed.),
Festschrift zum 60. Geburtstag von Carl F. Hoffmann, 373–391. Hamburg: Buske.
Wälchli, Bernhard. 2005. Co-compounds and natural coordination. Oxford: Oxford
University Press.
Wälchli, Bernhard. 2012. Grammaticalization clines in space: Zooming in on synchronic
traces of diffusion processes. In Björn Wlemer, Bernhard Wälchli & Björn Hansen
(eds.), Grammatical replication and borrowability in language contact, 233–272. Berlin:
Mouton de Gruyter.
Wald, Benji V. 1975. Animate concord in northeast coastal Bantu: Its linguistic and
social implications as a case of grammatical convergence. Studies in African Linguistics
6. 267–314.
Watkins, Mark Hanna. 1937. A grammar of Chichewa, a Bantu language of British
Central Africa. Language 13. 9–158.
Welmers, William. 1973. African language structures. Berkeley: University of California
Press.
Westermann, Diedrich. 1930. A study of the Ewe language. London: Oxford University
Press.
Whorf, Benjamin. 1941. The relation of habitual thought and behavior to languages. In
Leslie Spier, Irving A. Hallowell & Stanley S. Newman (eds.), Language, culture, and
personality: Essays in memory of Edward Sapir, 75–93. Menasha, Wisconsin: Stanley
S. Newman.
Wierzbicka, Anna. 1984. Diminutives and dispreciatives: semantic representation for
derivational categories. Quaderni di semantica 5. 123–130.
262
BIBLIOGRAPHY
Wilson, W.A. A. 1989. Atlantic. In John Bendor-Samuel & Rhonda L. Hartell (eds.), The
Niger-Congo languages: A classification and decsription of Africa’s largest language
family, 81–104. University Press of America.
Wilson, William A. A. 1961. An outline of the Temne language. London: School of
Oriental and African Studies.
263
ISBN 978-91-7447-952-2
Department of Linguistics
Was this manual useful for you? yes no
Thank you for your participation!

* Your assessment is very important for improving the work of artificial intelligence, which forms the content of this project

Related manuals

Download PDF

advertisement