1Ginyanga is a North Guang language of the Kwa family.1 It is spoken by a population of 16,500 people in Central Togo and the Volta Region in Ghana (Kalessou 2011). Figure 1 illustrates the Ginyanga-speaking area in Togo.
Figure 1 — Map of the Ginyanga-speaking area in Togo
2Ginyanga manifests a 9-vowel system in which [+ATR] vowels are /i, e, u, o/ while [-ATR] vowels are /ɪ, ɛ, ɔ, ʊ, a/. They form the following harmonic pairs: /i~ɪ, e~ɛ, e~a, u~ʊ, o~ɔ/. The vowel /a/ is neutral and can appear in both [+ATR] and [-ATR] contexts. Ginyanga is a tonal language, with two structural tones. The basic word order is SVO.
3Previous research on Ginyanga includes a brief grammar sketch (Cleal 1976), an encyclopaedia entry (Poidi-Gblem & Kantchoa 2012), a description of segmental phonetics (Kalessou 2011) and a sociolinguistic study (Jordan-Diller & Diller 2008). Ginyanga lexical data also featured in Güldemann & Fiedler’s research on Niger-Congo noun classes among other Guang languages (Güldemann & Fiedler 2019). During 2019-2021 a series of field trips was organised by the Institute of Linguistics, RAS, continued by CNRS. This research was part of Tatiana Mamonova’s master thesis on Ginyanga phonology (Mamonova 2020) at HSE, Moscow, and the authors’ PhDs at IL (Moscow) and CNRS (Paris). This paper draws on and extends the preliminary overview of the Ginyanga noun class system by Pozdniakova (2021). Other relevant research includes an article about the Ginyanga pronominal system (Pozdniakova & Aplonova 2022).
4While the first author’s 2021 paper, published in Russian, was the first work that outlined Ginyanga nominal classification, the present paper proposes a number of significant refinements. First, tone marks have been added, filling in one of the largest gaps in our understanding of the Ginyanga noun class system, as tone is pivotal for African languages morphophonology, lying at the base of many morphosyntactic processes. Secondly, we extend the literature review, i.e. the genealogical and diachronic background. Moreover, this paper is restructured for the sake of clarity. Finally, this paper makes our research accessible to an English-speaking audience.
5The data for the present study comes from grammatical questionnaires and corpus data collected between 2019 and 2024 during fieldwork in Agbandi, Togo (Aplonova & Pozdniakova 2024). The total size of the corpus is about one hour, corresponding approximately to 8K words. All texts are transcribed, glossed and translated to English. Further, a wordlist of approximately 1500 words has been compiled. It was based on the SIL Comparative African Wordlist/SILCAWL (Roberts & Snider 2006) to make the data more accessible for comparative projects. The items are arranged semantically and have English and French glosses.
6One of the distinct features of Niger-Congo languages are noun class systems, although their activity varies across different families. The activity of Kwa noun class systems can be assessed based on four criteria: singular/plural pairings, noun-to-noun derivation (in particular, diminutive formation), verb-to-noun derivation and the diversity of agreement patterns (Konoshenko & Shavarina 2019). The extent to which class markers participate in these processes characterises a system as active or vestigial, although the class/number marking is viewed as decisive. Ginyanga follows the North Guang pattern: class marking alternation is present in all three processes, which is characteristic of an active system.
7The paper is organised as follows. Section 2 outlines the terminology and provides a short overview of the noun class systems in Kwa languages and previous studies of the North Guang branch. In Section 3, the noun class system of Ginyanga is introduced. Its general characteristics are discussed, such as singular-plural pairings and semantic correlates. Section 4 illustrates derivation models that involve class conversion. Section 5 is dedicated to agreement targets such as numerals, demonstratives, verbal prefixes and pronouns. Section 6 compares the Ginyanga noun class system to those of other Kwa languages. Section 7 concludes.
8In the literature on Niger-Congo, the concept of “noun class” is heavily influenced by the Bantu tradition and the assumption of a biunique mapping between nominal form classes, formed by nouns with identical morphological or phonological features (e.g. noun class affixes), and agreement classes, defined on the basis of identical agreement patterns in all contexts. While some Bantu languages (Grinevald & Seifart 2004: 246, 249) with elaborate noun classification systems come close to this prototype, this is rarely the case, even for Bantu. Also, the case of Ginyanga shows that nominal marking and agreement are to be viewed independently. In (1a-b), the words ‘dogs’ and ‘things’ share the noun prefix, therefore belonging to the same nominal form class; however, they have different agreement paradigms. The reverse is also true: in (1c-b), the nouns ‘chief’ and ‘woman’, sharing the agreement class, belong to different nominal form classes.
| (1) |
a. |
ɩ̀-kprɛ́ |
mɛ́-tɔ̀l |
|
|
cl3-dog |
cl3.ipfv-fall |
|
|
‘The dogs fall.’ |
|
b. |
ɩ̀-lá |
ɩ́-tɔ̀l |
|
|
cl8-thing |
cl8.ipfv-fall |
|
|
‘These two things fall.’ |
|
c. |
∅-wúl=lé |
ó-jì |
∅-kìtíkí |
|
|
cl1-chief=art |
cl1.ipfv-eat |
cl1-manioc |
|
|
‘The chief eats manioc.’ |
|
d. |
ɔ̀-cí=yá |
ó-jì |
∅-kìtíkí |
|
|
cl1-woman=art |
cl1.ipfv-eat |
cl1-manioc |
|
|
‘The woman eats manioc.’ |
- 2 We use the term “agreement” for all instances of indexing of the noun class. However, starting fro (...)
9Therefore, we define noun classes in Ginyanga based on agreement,2 in line with the syntactic tradition (Vinogradov 1997: 7–9), referring to a set of nouns that share the agreement pattern in all relevant contexts. For this reason, we do not postulate separate noun classes in cases where lexemes have different noun prefixes but share the same agreement paradigm, which is also in line with the cross-linguistic tradition of defining genders.
10Gender is another frequently debated concept in the Africanist literature, stemming from the same assumption of a one-to-one mapping between nominal form and agreement class. Defined by Corbett as an abstract morphosyntactic feature of nouns reflected in characteristics of associated words (i.e. agreement) in the Africanist tradition, gender is intertwined with the singular-plural pairings of nominal form classes. In particular, both Snider’s and Manessy’s reconstructions of Proto-Guang take into consideration nominal form classes as well as agreement patterns, resulting in perhaps overly complex gender systems (for a discussion, see Güldemann 2018: 128–130).
11The analysis of gender is further complicated by the presence of transnumeral nouns that do not form regular number oppositions and therefore comprise a separate gender in some works (Güldemann & Fiedler 2019). However, in our data this phenomenon seems marginal, so we do not see fit to follow this approach. To avoid confusion, we do not use the term gender and instead simply point out singular/plural noun class pairings (see Section 3.2).
12The term deriflection (portmanteau of “derivation” and “inflection”) was introduced by Güldemann & Fiedler (2019) as the morpho(phono)logical counterpart of genders to address the issue of conflating syntactic agreement between a nominal trigger and its target and morphological and/or phonological regularities of nouns. While we do not discuss gender at this stage of research, in our study we differentiate between noun class based on agreement pattern (referred to as noun class in Table 3) and nominal form classes, cf. Van de Velde (2006) for a similar approach. Table 1 presents the nominal form class system of Ginyanga, and agreement patterns are discussed in depth in the next section. Derivation based on noun class conversion will be discussed in Section 7.
Table 1 — Nominal form (NF) class system of Ginyanga
| NF |
Form |
Number value |
Examples |
| o- |
ɔ~o- |
sg |
ɔ̀-cɩ́ ‘woman’ |
| ø- |
- |
sg |
Ø-wúl ‘chief’ |
| a- |
a- |
pl |
á-sá ‘people’ |
| i- |
ɩ~i- |
sg, pl |
ɩ̀-kprɛ́ ‘dog’ |
| ga- |
ga- |
sg |
gà-tɩ́n ‘house’ |
| gv- |
gɩ~gi~gʋ~gu~gɛ~ge- |
sg |
gù-wú ‘snake’ |
| n- |
n- |
pl |
m-bí ‘children’ |
13Kwa languages are known for their reduced morphology with either completely decayed or residual noun class systems with some exceptions like Ghana-Togo Mountain languages (Aboh 2010: 11), as opposed to the “canonical” Bantu type with robust inflection and elaborate noun class systems (Good 2012: 6). Some scholars do not fully support this claim; for instance, Konoshenko & Shavarina (2019), in their review of the data of 20 Kwa languages with vestigial and active noun class systems, argue that due to the attrition of the elaborate noun class system of “Proto-Kwa”, modern languages present a continuum of complexity in their noun morphology. Guang languages, to which Ginyanga belongs, are among those with fairly well-preserved noun class systems, while the complexity of noun class agreement varies within the group (Güldemann & Fiedler 2019: 122).
14Ginyanga belongs to the northern branch of the Guang group, as first suggested in Painter’s (1967) and Stewart’s (1967) reconstruction of the Proto-Guang vowel system. There are different opinions on the sub-classification of the Guang languages (Williamson & Blench 2000; Simons & Fennig 2018). Simons & Fennig (2018), based on Painter (1967), sub-classify Guang into two language clusters, North Guang and South Guang. Snider (1988) further sub-classifies South Guang into two groups: Hill Guang and Coastal Guang. He also subclassifies North Guang into three branches: Gonja, Oti, and Nkonya.
- 3 “Idiolect” is an original term used in Painter (1967). The likely rationale behind this term was t (...)
15Painter’s (1967) classification of Guang languages is based on the comparison of the lexical data of 25 Guang “idiolects”3. According to him, Ginyanga (Anyanga) is a “decidedly North Guang language” most closely related to Gichode/Achode (75% match in the basic lexicon), Bazanche (72%), Nchumunu (68%), Nawuri (68%), Nkonya (67%), Krachi (65%) and Gonja (lower 60s). Figure 2 presents the classification of Guang languages based on the studies by Painter (1967) and Stewart (1967), while Snider (1988: 153–154) argues that “we know from other linguistic evidence that Gichode and Nawuri represent two different groups”'.
Figure 2 — Modern classification of Guang languages (Hammarström et al. 2022)
16The Proto-Guang noun class system has been independently reconstructed by Manessy (1987) and Snider (1988). Figure 3 shows Manessy’s reconstruction of what he refers to as the “classificational system of Proto-Guang”. Based on 40 nouns from the Swadesh wordlist, he postulates 7 singular prefixes, i.e. *ki-, *ke-, *ka-, *dɩ-, *o-, *e-, *a-, and 5 plural prefixes: *bV-, *e-, *a-, *N-. There is an additional prefix *N- that marks uncountable nouns. Sniders’s reconstruction of Proto-Guang features 4 singular and 4 plural prefixes. He explains the *ke- prefix in Manessy’s work through phonological inaccuracies in the data and does neither support the reconstruction of a singular prefixes *a-, *e- and *dɩ.
Figure 3 — Comparison of Manessy’s and Snider’s reconstructions (from Snider 1988: 160)
17Despite methodological and terminological differences, discussed in Section 2.1, we found Snider’s reconstruction of singular and plural prefixes to be highly relevant for our research on Ginyanga. Table 2 summarises the singular/plural pairings of Proto-Guang noun classes, their corresponding noun prefixes in modern Gichode and Nawuri and the semantic correlates of each group in Sniders’s reconstruction. The capital letters represent allomorphic variation due to [ATR] based vowel harmony and other morphological and/or phonological processes. According to Snider (1988: 141, 153), the partial merger of plural *N- and *I- classes in Proto-Oti Guang has led to the *N- → *I- innovation in Group B prefixes. He suggests that this change occurred in order to equalise the semantic load borne by each noun class prefix. Phonetic innovations are the voicing of the velar stop in *kA- → *gA- and *kI- → *gI- in Proto-Mountain Guang. The change of *bA- to *A- in Group A in most Guang languages is considered a morphological replacement rather than a phonological process (ibid.).
18Gichode and Nawuri belong to the same subgroup as Ginyanga, i.e. Mountain Oti North Guang, the former being more closely related to each other than to Ginyanga, according to Painter’s analysis. As shown in the right column of Figure 3, repeated from Section 3.2, all of Snider’s singular/plural class pairings of Gichode and Nawuri are also attested in Ginyanga, with the addition of a gA-/I- pairing and the I-/A- pattern (the latter has only been observed in two nouns). While we do not support this morphologically-based definition of noun classes (see Section 2.1 for discussion), we find these parallel singular/plural pairings a reassuring sign of linguistic proximity.
Table 2 — Noun class system of Proto-Guang and modern Guang languages (Snider 1988: 153) in relation to Ginyanga singular-plural pairings (groups according to Snider’s classification)
| Proto-Guang |
Gichode Nawuri |
Semantic correlates |
Groups |
Ginyanga pairings |
| *O-/*bA- |
O-/A- |
humans |
A |
ɔ̀-ɲɩ́n ‘man’ – à-ɲɩ́n ‘men’ |
| – |
– |
– |
– |
í-sàfɔ́ ‘foreigner’ – á-sàfɔ́ ‘foreigners’ |
| *O-/*N- |
O-/I- |
unknown |
B |
ɔ̀-kprɛ́ ‘dog’ – ɩ̀-kprɛ́ ‘dogs’ ɔ̀-kplɔ́ ‘table’ – ɩ̀-kplɔ́ ‘tables’ |
| *kA-/*N- |
gA-/N- |
unknown |
C |
gà-bwí ‘bird’ – n-bwí ‘birds’ |
| – |
– |
– |
– |
gá-yɛwá ‘beard’ – ɩ-yɛwa ‘beards’ |
| *kI-/*A- |
gI-/A- |
unknown |
D |
gɩ̀yɛ́ ‘mouse’ – à-yɛ́ ‘mice’ gɩ̀-bwí ‘stone’ – à-bwí ‘stones’ |
| *Ø-/*A-, *Ø-/*I- |
Ø-/A-, Ø-/I- |
animals and plants |
E |
gbɛ̀dɛ́ ‘blacksmith’ – á-gbɛ̀dɛ́ ‘blacksmiths’ mìsín ‘bee’ – í-mìsín ‘bees’ ∅-kìtíkí ‘manioc’ – í-kìtíkí ‘maniocs’ |
| *kI-/*I- |
gI-/I- |
unknown |
F |
gɩ̀-fɛ́ ‘grass’ – ɩ̀-fɛ́ ‘grasses, bush’ |
| *I- |
I- |
mass |
G |
ɩ̀-mɛ́lɩ̀n ‘snot’ |
| *A- |
A- |
abstract |
H |
à-wó ‘cold weather’ |
| *N- |
N- |
(semi-)liquids |
I |
n-cú ‘water’ |
19Snider also concludes that the modern North Guang languages feature a more complex and better preserved noun class system than the South Guang branch.
20As a North Guang language, Ginyanga demonstrates a full-fledged system of noun classes and noun class agreement, comparable with the one that is postulated for Proto-Guang (Snider 1988: 138). There are 8 agreement patterns, based on which we define 8 noun classes. The first class (CL1) is further divided into 3 nominal form classes, depending on the form of the nominal class prefixes: CL1(ɔ-/o-), CL1(ɩ-) and CL1(Ø-) subclass (cf. Table 1 for examples). The CL1(ɩ-) nominal class contains only two words ɩ́-sà ‘person’ and ɩ́-sàfɔ́ ‘stranger’. The second class (CL2) has two nominal form subclasses, i.e. with a prefix a- and with a suffix -áná which is discussed in Section 2.3. Despite different nominal form classes, nouns of CL1 and CL2 have the same agreement pattern.
21CL1, CL4 and CL5 are singular, other classes are plural. Singular/plural pairings are discussed in Section 3.2. Classes are marked by noun prefixes, although a considerable number of CL1 nouns have no overt marking on the noun itself (Ø-nominal form class). Class markers are subject to [+/-ATR]-based vowel harmony, and allomorphs are presented between tildes throughout the paper. In most nouns, the tone of a prefix is predictable via tonal polarisation with the stem, i.e. if the stem starts from H, the prefix is L and vice versa. However, there are a number of nouns where the prefix and the stem have the same tones, e.g. ótú ‘heart’.
22Table 3 gives an overall of all Ginyanga noun classes with their agreement paradigms. In Ginyanga, there is class agreement of cardinal numerals from ‘1’ to ‘6’ (cf. Section 5.1.2), demonstratives (cf. Section 5.1.3), verbal prefixes (cf. Section 5.2) and personal pronouns (cf. Section 5.3).
Table 3 — Ginyanga noun class system in the NP domain, anaphoric pronominal agreement and verbal agreement
23The examples below illustrate the agreement of the CL1 noun kìtíkín ‘manioc’: in (2a-c) agreement is realized with a zero prefix in the NP domain, (2d) exemplifies anaphoric agreement, while (3a-b) shows verbal agreement in the perfective and imperfective.
| (2) |
a. |
∅-kìtíkí |
ɔ̀-kʋ́ |
|
|
cl1-manioc |
cl1-one |
|
|
‘one manioc’ |
|
b. |
∅-kìtíkí |
ɔ̀-bʋ́nʋ̀ |
|
|
cl1-manioc |
cl1-this |
|
|
‘this manioc’ |
|
c. |
∅-kìtíkí |
ɔ̀-bánɩ̀ |
|
|
cl1-manioc |
cl1-that |
|
|
‘that manioc’ |
|
d. |
∅-kìtíkí |
ɔ́-kɩ̀lá |
mà |
ɔ̀-kɩ̀lá |
|
|
cl1-manioc |
cl1-size |
cl1 |
cl1-size |
|
|
‘a manioc’s size, its size’ |
| (3) |
a. |
∅-kìtíkí |
ɔ̀ɔ́-tɔ̀l |
|
|
cl1-manioc |
cl1.pfv-fall |
|
|
‘A manioc fell.’ |
|
b. |
∅-kìtíkí |
ɔ́-tɔ̀l |
|
|
cl1-manioc |
cl1.ipfv-fall |
|
|
‘A manioc falls/is falling.’ |
24Tables 4a and 4b illustrate the distribution of nouns in the corpus according to their class. Table 4a shows the token frequency, i.e. how many occurrences of a word belonging to a noun class were attested. One can see that almost half of the nouns are of CL1 (525 occurrences out of 1128 nouns in total). The second most frequent class is CL5 (271 occurrences, where 78 occurrences correspond to the word gèbí ‘child’). These most frequent classes include nouns that refer to humans and animals in the singular who are the protagonists of the traditional tales that constitute the corpus.
Table 4a — Token frequency of noun classes
| Noun class |
Counts |
% of total |
| CL1 |
525 |
46 |
| CL2 |
57 |
5 |
| CL3 |
8 |
1 |
| CL4 |
125 |
11 |
| CL5 |
271 |
24 |
| CL6 |
29 |
3 |
| CL7 |
54 |
5 |
| CL8 |
59 |
5 |
| Total |
1128 |
100% |
25Table 4b illustrates the type frequency, i.e. how many different nouns of a noun class are attested. CL1 and CL5 are again the most numerous classes, with 55 and 44 unique nouns respectively out of 182.
Table 4b — Type frequency of noun classes
| Noun class |
Counts |
% of total |
| CL1 |
51 |
28 |
| CL2 |
18 |
10 |
| CL3 |
3 |
2 |
| CL4 |
18 |
10 |
| CL5 |
44 |
24 |
| CL6 |
11 |
6 |
| CL7 |
21 |
11 |
| CL8 |
16 |
9 |
| Total |
182 |
100% |
26Overall, token and type frequencies are similar in that CL1 and CL5 are the most frequent, while CL3 and CL6 are the least attested. However, one can notice that the type frequency of the most numerous classes (48% for CL1 vs. 24% for CL5) is more equally distributed in comparison to the token frequency (46% for CL1 vs. 24% for CL5). Moreover, CL4 shows a higher token frequency in comparison to CL2, CL7 and CL8 (24%), which is explained by the extensive use of the words gàɲɩ́n (55 occurrences) and gáfàná (28 occurrences) that mean ‘manner’ or ‘the way of doing’ and are similar in their use to manner adverbs. Note that the type frequency of CL4, CL2, CL7 and CL8 is almost equal.
27In a typological perspeсtive, it is unusual that Ginyanga has less singular than plural classes. Proto-Kwa noun class systems normally follow a different pattern, resulting in one-to-one or one-to-many correspondence between singular and plural classes (Konoshenko & Shavarina 2019). This and the diversity of prefixes in CL1 suggest that CL1 is a result of a merging of the independent Proto-Guang singular classes *O-, *I- and Ø-.
28Table 5 shows the attested singular/plural pairings between Ginyanga noun classes. The last column aligns these pairings with the groups proposed in Snider’s (1988) reconstruction of the Proto-Guang noun system. It does not include the Groups G, H and I because their members do not have singular counterparts, pluralia tantum nouns will be discussed later in this subsection.
Table 5 — Singular/plural pairings of Ginyanga noun classes aligned with groups in Snider’s paper (1988)
29Some animate nouns form plurals with the dedicated plural marker -áná, possibly originating in the associative plural marker -áná, as suggested for other Guang languages (cf. Hansford 1990: 162 for Chumburung). In Ginyanga, this marker can express two meanings: regular plural (4a) for some animate nouns and plants, and associative plural for proper (4b) and animate nouns (4c). In (4a), -áná marks the noun ká ‘wife’, which results in the regular plural meaning. In (4b), the semantics of -áná is ambiguous (it can be interpreted as a regular plural as well as an associative plural), while in (4c) a proper noun Kójó marked by -áná has an associative plural meaning.
| (4) |
a. |
∅-wùl=é |
mà |
káná |
mé-sò |
∅-kìtíkì |
|
|
cl1-chief=art a |
cl1 |
wife:ass.pl |
cl2.ipfv-buy |
cl1-manioc |
|
|
‘The wives of the chief buy manioc.’ |
| a. Articles in Ginyanga are subject to [ATR] harmony and can surface as =wa/=we following a vowel or =a/=e following a consonant (in case the final consonant is -n, it undergoes assimilation). |
|
b. |
à-bál-áná |
ná-cɩ́nà |
|
|
cl2-monkey-ass.pl |
foc.pfv-sit |
|
|
‘There were monkeys…’ (a beginning of a tale) |
|
c. |
Kójáná |
mé-sò |
∅-kìtíkì |
|
|
Kojo:ass.pl |
cl2.ipfv-buy |
cl1-manioc |
|
|
‘Kojo and his associates buy manioc.’ |
30While the -áná suffix would constitute a separate deriflection model in Güldemann & Fiedler’s framework (2019), both groups of -áná nouns (plural and associative plural) follow the CL2 agreement pattern: the verb in both (4a) and (4b) is marked with the prefix mé-, agreeing with the plural human subject (‘chief’s wives’ and ‘Kojo’s associates’).
31Plural forms with -áná are used with kinship terms and, based on speakers’ reports, are preferred to the CL2 correlates (5).
| (5) |
a. |
à-kúl=é |
má-jàŋál |
mɔ̀ |
káná |
|
|
cl2-husband=art |
cl2.pfv-meet |
cl2 |
wife:ass.pl |
|
|
‘The husbands met their wives.’ |
|
b. |
? à-kúl=é |
má-jàŋál |
mɔ̀ |
à-ká |
|
|
cl2-husband=art |
cl2.pfv-meet |
cl2 |
cl2-wife |
|
|
Intended meaning: ‘The husbands met their wives.’ |
- 4 In Güldemann & Fiedler’s framework (2019), such nouns are called “transnumeral” nouns in the sense (...)
32The second group of words that are used with -áná are names of animals and plants that do not form singular/plural pairs;4 they invariably bear the CL2 (animate nouns) or CL7 (inanimate nouns) plural prefix a- (6). Presumably, these nouns refer to singular entities. At this stage of research, we do not know whether they can also designate plural entities without the addition of -áná.
| (6) |
a. |
à-kó cl2-parrot ‘parrot’ > à-káná cl2-parrot:ass.pl ‘parrots’ |
|
b. |
á-bùmbwí kàbú cl2-cattle_egreta ‘cattle egret’ > á-bùmbwí kàbáná cl2-cattle_egret:ass.pl ‘cattle egrets’ |
|
c. |
á-gìmá cl2-chameleon ‘chameleon’ > á-gìmáná cl2-chameleon:ass.pl ‘chameleons’ |
|
d. |
à-yú cl7-garlic ‘garlic’ > à-yáná cl7-garlic:ass.pl ‘garlic (bulbs)’ |
| a. We observed variation in the class marking of á-bùmbwí kàbú ‘cattle egret’. Some speakers use a zero CL1 prefix in singular contexts (bùmbwí kàbú) and a CL2 or CL7 prefix a- for in the plural. |
33Although some Kwa languages, e.g. Lelemi, have preserved elaborate semantic classifications (Allan 1973), most Guang languages have undergone nominal class system attrition, resulting in overlap between various semantic groups in one class (Snider 1988).
34As already mentioned, CL1, CL4 and CL5 are singular and contain both animate and inanimate objects. Other classes are plural. While CL6 has both animate and inanimate nouns, CL2 and CL3 are strictly animate, and CL7 and CL8 are inanimate.
35Kinship terms are assigned to CL1 ɔ~o-/CL2 pairing (ɔ́-ká ‘wife’/a-ka ‘wives’, ó-ɲí ‘mother’), except for ò-kúl ‘husband’/ì-kúl ‘husbands’. Other nouns denoting humans can be found in all three CL1 nominal form subclasses, with plural counterparts in CL2 and CL3 (ɩ́-sàfɔ́ ‘foreigner’/à-sàfɔ́ ‘foreigners’, gbɛ̀dɛ́ ‘blacksmith’/a-gbɛdɛ ‘blacksmiths’, cɔ́cɔ̀ ‘baby’/ɩ̀-cɔ́cɔ̀ ‘babies’).
36Most animal names, including birds, belong to the CL1 Ø-/CL2 pairing (cf. Snider 1988: 148). This confirms the theory that Proto-Guang had a dedicated *Ø-/*A- class, all members of which referred to animals: cágá ‘rat’/a-caga ‘rats’, nárɩ́ ‘bull’/a-narɩ ‘bulls’, pólí ‘eagle’/a-poli ‘eagles’.
37Borrowings from other languages do not get any singular marking and thus end up in the singular CL1 Ø- subclass: kótòkú ‘bag’ (from Ewe kotokú), lòrí ‘car’ (from English lorry), wàcɩ́ ‘bracelet’ (from English watch), làkúmí ‘camel’ (initially from Hausa rakumi). The plural counterparts get CL2/CL3 or CL7/CL8 class marking depending on animacy: ɩ́-wàcɩ́ CL8-bracelet ‘bracelets’, á-làkúmí CL2-camel ‘camels’.
38There are only two CL1 ɩ- class nouns in our data, i.e. ɩ́-sà ‘person’ and ɩ́-sàfɔ́ ‘foreigner’. These nouns have plural pairs in CL2 (cf. Table 5).
39Some terms for body parts belong to CL1 (ó-tú ‘heart’, ɔ̀-ján ‘leg’) and CL4 (gà-mál ‘back’, gà-kán ‘chest’), but mostly they belong to CL5: gɩ̀-bá ‘arm’, gɩ́-mɩ̀lɩ́ ‘knee’. Plants are found in CL5: gɩ̀-lál ‘baobab tree’, gɩ̀-fɛ́ ‘grass’, gɩ́-fàrá ‘leaf’, while fruits and vegetables seem to group in CL1(Ø-): pɛ̀yá ‘avocado’, bɔ́rɔ́dɛ́ ‘plantain’, kìtíkì ‘manioc’. Artefacts are found mainly in CL4 and CL5: gà-cá ‘fishing line’, gà-tá ‘sword’, gɩ̀-gbá ‘door’, gɩ̀-pwál ‘water pot’, although there are exceptions, namely ó-píní ‘pestle’ and ɔ̀-fɛ́ ‘rope’ that belong to CL1.
40Plural CL6 and CL8 contain transnumeral nouns that do not have singular correlates, such as liquids in CL6 (n-cú ‘water’, n-tá ‘wine’), and pluralia tantum nouns in CL8 (ɩ̀-fál ‘field’).
41Table 6a summarises the tendencies of semantic correlations mentioned above. Table 6b attempts a more traditional “gender-style” analysis, but it appears hard to determine semantic correlates of singular/plural pairings apart from the “human” CL1/CL2 pairing. This might be another argument for determining classes in Ginyanga based on agreement patterns and not singular/plural pairings (genders).
Table 6a — Semantic correlates of Ginyanga noun classes
| Noun class |
Number value |
Nominal form class |
Animacy |
Most prominent semantic groups |
| CL1 |
SG |
ɔ~o- |
animate, inanimate |
humans (incl. kinship terms) |
| ɩ- |
animate |
humans (ɩ́sà ‘person’, ɩ́sàfɔ́ ‘foreigner’) |
| Ø- |
animate, inanimate |
animals, fruits and vegetables, borrowings |
| CL2 |
PL |
a- |
animate |
humans (incl. kinship terms) |
| CL3 |
PL |
ɩ~i- |
animate |
some animate nouns, except humans |
| CL4 |
SG |
ga- |
animate, inanimate |
artefacts |
| CL5 |
SG |
gɩ~gi~gʋ~gu~gɛ~ge- |
animate, inanimate |
artefacts, body parts, plants, nominalizations from active verbs |
| CL6 |
PL |
n- |
animate, inanimate |
liquids and substances |
| CL7 |
PL |
a- |
inanimate |
various inanimate nouns |
| CL8 |
PL |
ɩ~i- |
inanimate |
various inanimate nouns |
Table 6b — Semantic correlates of Ginyanga noun classes (singular/plural pairings)
| Noun class |
Singular prefix |
Plural prefix |
Animacy |
Most prominent semantic groups |
| CL1 |
ɔ~o- |
CL2 |
a- |
animate |
humans (incl. kinship terms) |
| CL3 |
ɩ~i- |
animate |
unclear |
| CL8 |
ɩ~i- |
inanimate |
body parts |
| ɩ~i- |
CL2 |
a- |
animate |
humans (ɩ́sà ‘person’, ɩ́sàfɔ́ ‘foreigner’) |
| Ø- |
CL2 |
a- |
animate |
humans (professions), animals |
| CL3 |
ɩ~i- |
animate |
only 2 nouns attested: sɔ̀já ‘soldier’/ɩ-sɔ̀já ‘soldiers’, cɔ́cɔ̀ ‘baby’/ɩ-cɔ́cɔ̀ ‘babies’ |
| CL8 |
ɩ~i- |
inanimate |
fruits and vegetables |
| CL4 |
ga- |
CL6 |
n- |
animate |
unclear |
| CL8 |
ɩ~i- |
inanimate |
unclear |
| CL5 |
gɩ~gi~gʋ~gu~gɛ~ge- |
CL2 |
a- |
animate |
unclear |
| CL7 |
a- |
inanimate |
body parts |
| CL8 |
ɩ~i- |
inanimate |
unclear |
42In African languages, diminutive meaning is often derived from the lexical item ‘child’ (Nikitina 2019). In Ginyanga, however, the word gè-bí ‘child’ has not been fully grammaticalized to a productive diminutive suffix, contrary to Gonja (Nelson et al. 2016: 144), another North Guang language.
43As one can see in (7-9), bi is always [+ATR] and neither harmonises with the stem like the diminutive ‑bi in Gonja (Nelson 2016: 144) nor affects the [ATR] value of [-ATR] stems like in Nawuri (Casali 2002: 144). In Ginyanga, bi is not productive and is used strictly for the formation of animal offspring names (7), different species of a smaller size (8) and singulative-like meanings (9). This process is usually accompanied by the class conversion from CL1 to CL4 or CL5, which is in line with the tendency observed by Konoshenko & Shavarina (2019: 18), who further suggest that CL4-5 *kV- could have been a diminutive class already in Proto-Kwa. We suggest treating these forms given in (7-9) as cases of noun compounding.
| (7) |
a. |
∅-sánɩ́ ‘cl1-sheep’ > gà-sánɩ́-bì cl4-sheep-dim ‘lamb’ |
|
b. |
∅-kúlí ‘cl1-pig’ > gì-kúlí-bì cl5-pig-dim ‘piglet’ |
| (8) |
a. |
∅-sínsín ‘cl1-big ant’ > gì-sínsín-bì ‘ant’ |
|
b. |
∅-kɩ̀nsɩ́l ‘cl1-big rain’ > gɩ́-kɩ̀nsɩ́l-bì ‘small rain/mudfish’ |
| (9) |
a. |
gɩ̀-bá ‘cl5-hand’ > gɩ̀-bá-bì cl5-hand-dim ‘finger’ |
|
b. |
gù-kwádù ‘cl5-banana’ > gù-kwádù-bì cl5-banana-dim ‘one banana; a finger’ (as opposed to the whole banana hand) |
44The actor is derived from a verb by adding a CL1 prefix ɔ- or Ø- and a suffix -bú/bʋ́ (10).
| (10) |
a. |
kínì ‘to refuse’ > ɔ̀-kínì-bú cl1-refuse-act ‘refuser’ |
|
b. |
ká ‘to advise’ > ∅-ká-bú cl1-advise-act ‘counsellor’ |
45Rarely, it involves only a conversion to CL1 without suffixation, except, perhaps, in the most lexicalized cases such as ɔ̀yú ‘thief’ (from yù ‘to steal’).
46Actor to activity derivation is very productive. It involves class conversion to CL5 (11).
| (11) |
a. |
∅-wúl ‘cl1-chief’ > cl5-chief gì-wúl ‘chiefhood’ |
|
b. |
∅-jàmbwí ‘cl1-hunter’ > cl5-hunter gí-jàmbwí ‘hunt’ (noun) |
47Dynamic verbs can be nominalised by adding the CL5 prefix (12), while stative verbs take the CL1 ɔ- prefix and their final vowel is changed to /a/ (13).
| (12) |
a. |
yù ‘to steal’ > gì-yú cl5-steal ‘theft’ |
|
b. |
kà ‘to advise’ > gɩ̀-ká cl5-advice ‘advice’ |
| (13) |
a. |
kílí ‘be huge’ > ɔ̀-kílá cl1-be.huge:art ‘hugeness’ |
|
b. |
pɛ́lɩ́ ‘be ripe/red’ > ɔ̀-pɛ́lá cl1-be.ripe:art ‘ripeness’ |
48Among Niger-Congo languages, the most common class agreement targets in the NP domain are cardinal numerals, adjectives (rarely), demonstratives and indefinite markers (Hepburn-Gray 2016; Konoshenko & Shavarina 2019: 28). However, in North Guang languages, adjectives tend to go unmarked (14).
49A Ginyanga noun phrase is linearly structured as follows:
POSS – N – ADJ – NUM – DEM – ART – RC – Q
50Here POSS stands for a possessor, N for a head noun, while ADJ refers to any qualitative lexeme (in Ginyanga, under certain circumstances these can be adjectives, ordinal numerals, stative verbs and nouns); NUM stands for a cardinal numeral, DEM for a demonstrative, ART for a definiteness marker, RC for a relative clause, and Q for a quantifier. The scheme above is illustrated in (14a-b). These examples were elicited with the intent to demonstrate a maximal NP with all its slots filled, which is by no means obligatory or prevalent in natural language. In both (14a) and (14b), the prefixes on the noun (n-tɩ́n ‘houses’ and í-mùtí ‘oranges’), the numeral (n-ɲɔ́ ‘two’ and ɩ̀-sá ‘three’) and the demonstrative (m-bálɩ̀ ‘these’, ɩ́-bʋ́nʋ̀ ‘those’) coincide. It shows that while qualitative lexemes are unmarked (pòbál ‘new’ and sɩ́n-sɛ́ ‘ripe’) in attributive position, cardinal numerals and demonstratives agree with the head noun.
| (14) |
a. |
n-wù |
∅-wúl=e |
n-tɩ́n |
pòbál |
n-ɲɔ́ |
m-bálɩ̀ |
kpɛ́kpɛ́ |
|
|
1sg.ipfv-see |
cl1-chief=art |
cl6-house |
new |
cl6-two |
cl6-that |
all |
|
|
‘I see both of these two new houses of the chief.’ |
|
b. |
ná-jì |
í-mùtí |
sín-sɛ́ |
ɩ̀-sá |
ɩ́-bʋ́nʋ̀ |
|
|
1sg.pfv-eat |
cl7-orange |
ripe-adj |
cl7-three |
cl7-this |
|
|
fá-sà |
mì=wá |
kɩ́nkɩ́n=mé |
|
|
|
|
2sg.pfv‑give |
1sg=art |
each=art |
|
|
|
|
‘I ate each of those three ripe oranges that you have given me.’ |
- 5 This is the term originally used by Konoshenko & Shavarina (2019: 28).
51The agreement patterns of numerals are summarized in Table 7. Cardinal numerals from ‘1’ to ‘6’ agree with the head noun, numerals from ‘7’ to ‘10’ retain their default noun class marking. Konoshenko & Shavarina (2019: 28) point out that the extent of class “indexation”5 varies greatly in Kwa numerals, so it is not unusual for higher numerals to remain unmarked.
Table 7 — Agreement patterns for numerals from 1 to 10
|
Numeral prefix |
Examples |
| 1 |
2-6 |
7-10 |
| CL1 |
ɔ- |
|
∅-kìtíkì ɔ̀-kʋ́ ‘one manioc’ ɩ́-sàfɔ́ ɔ̀-kʋ́ ‘one foreigner’ ɔ̀-cɩ́ ɔ̀-kʋ́ ‘one woman’ |
| CL2 |
|
a- |
no agreement |
á-sàfó à-sá ‘three foreigners’ á-sàfó gù-dú ‘ten foreigners’ |
| CL3 |
|
ɩ~i- |
no agreement |
ɩ̀-kprɛ́ ɩ̀-sá ‘three dogs’ ɩ̀-kprɛ́ gàkpán ‘two hundred dogs’ |
| CL4 |
ga- |
|
gà-nán g̀a-kʋ́ ‘one scorpion’ |
| CL5 |
gɩ~gi~gʋ~gu~go~gɔ- |
|
gɩ̀-yɛ́ gɔ̀-kʋ́ ‘one mouse’ |
| CL6 |
|
n- |
no agreement |
n-nɔ́ n-sá ‘three mouths’ m-bwí gʋ̀-kwɛ́ ‘eight birds’ |
| CL7 |
|
a- |
no agreement |
à-sʋ́ à-sá ‘three ears’ à-bwí gù-dú ‘ten stones’ |
| CL8 |
|
ɩ~i- |
no agreement |
í-kìtíkì ɩ̀-sá ‘three maniocs’ í-kìtíkì gù-dú ‘ten maniocs’ |
52Composite numbers greater than ‘10’ (‘11’ to ‘19’) have the same agreement pattern as ‘2’-‘6’. In all composite numbers, besides the basic order (noun-numeral), it is possible to put the head noun between the tens and units. In this case, the unit agrees with the noun (15a-b).
| (15) |
a. |
gùdú ɔ̀-kprɛ́ ɔ̀-kwɛ́ ten cl1-dog cl1-eight ‘eighteen dogs’ |
|
b. |
gùdú gè-bí gì-sí ten cl5-child cl5-six ‘sixteen kids’ |
53Numeral ɔ̀-kʋ́ ‘one’ can only be used with the singular (CL1, CL4, CL5) head noun (16a), while the indefinite quantifier ɔ̀-kʋ́ ‘some’ that is originated from the numeral agrees with the plural controllers as well (16b-c).
| (16) |
a. |
gà-bwí gà-kʋ́ cl4-bird cl4-one ‘one bird’ |
|
b. |
m-bwí n-kʋ́ cl6-bird cl6-one ‘some birds’ |
|
c. |
à-wú à-kʋ́ cl7-snake cl7-one ‘some snakes’ |
54Class marking on numerals of a given agreement class is not further affected by animacy in classes that contain both animate (17a) and inanimate subjects (17b).
| (17) |
a. |
m-bwí n-sá cl6-bird cl6-three ‘three birds’ |
|
b. |
n-có n-sá cl6-axe cl6-three ‘three axes’ |
55In distributive (18a-b) contexts, cardinal numerals can be reduplicated, retaining their agreement with the head.
| (18) |
a. |
m-bí=yé |
má-jì |
í-mùtí |
ɩ̀-sá |
ɩ̀-sá |
|
|
cl6-child=art |
cl2.pfv-eat |
cl8-orange |
cl8-three |
cl8-three |
|
|
‘The children have eaten three oranges each.’ |
|
b. |
mɛ́-bà |
à-ɲɔ́ |
à-ɲɔ́ |
gɩ̀-já=wá |
rɔ̀ |
|
|
cl2.ipfv |
cl2-two |
cl2-two |
cl5-market=art |
in |
|
|
‘They go to the market in pairs.’ |
56Ordinal numerals are marked by the suffix -sɛ́. They are similar to other qualitative lexemes in the sense that they do not agree with the controller in attributive position (19).
| (19) |
a. |
gù-jó ɲɔ́-sɛ́ cl5-yam two-adj ‘the second yam’ |
|
b. |
ɔ̀-cɩ́ sònú-sɛ́ cl1-woman seven-adj ‘the seventh woman’ |
57In the majority of Niger-Congo languages, numerals get class marking even when there is no agreement – for example, in counting (Pozdniakov 2018: 15). The default marking of cardinal numerals (when they are used separately from nouns) is illustrated in Table 8.
Table 8 — Ginyanga simple cardinal numerals
| Numeral |
Ginyanga |
Noun class prefix |
| ‘1’ |
ɔ̀-kʋ́ |
CL1 |
| ‘2’ |
ɛ̀-ɲɔ́ |
CL3/CL8a |
| ‘3’ |
ɩ̀-sá |
CL3/CL8 |
| ‘4’ |
ɩ̀-ná |
CL3/CL8 |
| ‘5’ |
ì-nú |
CL3/CL8 |
| ‘6’ |
ì-sí |
CL3/CL8 |
| ‘7’ |
∅-sònú |
CL1 |
| ‘8’ |
gʋ̀-kwɛ́ |
CL5 |
| ‘9’ |
∅-sòŋú |
CL1 |
| ‘10’ |
gù-dúb |
CL5 |
| ‘20’ |
ò-kú |
CL1 |
| ‘100’ |
ì-kwínú |
CL3/CL8 |
| ‘200’ |
gà-kpán |
CL4 |
| ‘1000’ |
n-mànínú |
CL6 |
a. Due to the homonymy of CL3 and CL8 prefixes, it is impossible to determine the class of the numeral in isolation. b. Besides gùdú ‘10’ and gàkpán ‘200’, which seem to be the proper numeral forms, we have elicited the additional lexemes sálá ‘10’ and ŋmán ‘200’. Sálá is used in composite numerals. |
58Other numerals are compounds formed according to the following model: ‘11’ = ‘10’ + ‘1’ gù-dú ɔ̀-kʋ́, 40 = ‘20’ * ‘2’ ì-kú ɛ̀-ɲɔ́, 50 = ‘20’ * ‘2’ + ‘10’ ì-kú ɩ̀-ɲɔ́ sálá, ‘500’ = ‘200’ * ‘2’ + ‘20’ * ‘5’ n-m̀an ɩ̀-ɲɔ́ ì-kú ì-nú and so on. Cross-linguistically, in such cases, the class prefix can be absent in the second formative – otherwise both formatives include coinciding or different class markers (Pozdniakov 2018: 14). Ginyanga follows the second pattern: all elements of the complex numeral are marked and class markers do not coincide. Apart from ‘100’ and ‘1000’, which are derivatives in Ginyanga, the numeric system corresponds to that of other Niger-Congo languages and Pozdniakov’s (Pozdniakov 2018: 126) reconstruction for the Guang branch.
59Ginyanga has two demonstratives that agree in class with the head noun, bálɩ̀ ‘that’ and bʋ́nʋ̀ ‘this’. The only formal difference between the two is that CL1 nouns do not have overt marking on the demonstrative bálɩ̀ (20).
| (20) |
a. |
ɔ̀-cɩ́ ɔ̀-bʋ́nʋ̀ ‘this woman’ |
|
b. |
ɔ̀-cɩ́ ø-bálɩ̀ ‘that woman’ |
60The agreement patterns of demonstratives are listed in Table 9.
Table 9 — Noun class agreement of Ginyanga demonstratives
|
Prefix |
| bʋ́nʋ̀ |
bálɩ̀ |
| CL1 |
ɔ- |
Ø- |
| CL2 |
a- |
a- |
| CL3 |
ɩ- |
ɩ- |
| CL4 |
ga- |
ga- |
| CL5 |
gɩ- |
gɩ- |
| CL6 |
n- |
n- |
| CL7 |
a- |
a- |
| CL8 |
ɩ- |
ɩ- |
61In Ginyanga, both noun class and number-person of the subject are cross-referenced by verbal prefixes. There are two series of verbal prefixes: perfective and imperfective. Perfective and imperfective prefixes are listed in Table 10. Verbal prefixes are subject to [ATR] based vowel harmony: the prefix agrees with the [ATR] value of the stem.
Table 10 — Ginyanga verbal prefixes
|
|
Perfective |
Imperfective |
| Person |
1sg |
ná- |
n- |
| 2sg |
fá- |
fɩ́~fí- |
| 1pl |
á- |
ɛ́~é- |
| 2pl |
fàá- |
fɛ́~fé- |
| Noun class |
1 |
ɔ̀ɔ́-/òó-, á- |
ɔ́~ó-, ɛ́~é- |
| 2 |
má- |
mɛ́~mé- |
| 3 |
má- |
mɛ́~mé- |
| 4 |
gá- |
gɛ́~gé- |
| 5 |
gá- |
gɩ́~gí- |
| 6 |
gá- |
gɩ́~gí- |
| 7 |
á- |
ɛ́~é- |
| 8 |
yá- |
ɩ́~í- |
62In CL1, if a subject is expressed by an overt NP, there are a number of rules of hiatus resolution between the nominal subject and the verbal prefix. In the perfective, the initial syllable of a verb changes from òó to á (21).
| (21) |
Yàwá |
á-bólì |
∅-kútú=wé |
|
Yawa |
cl1.pfv\cntr-break |
cl1-pot=art |
|
‘Yawa broke the pot.’ |
63In the imperfective, if the final vowel of a subject is low, the verbal prefix changes from ó to é (22a) while if the final vowel is high, the prefix changes to í (22b).
| (22) |
a. |
Yàwá |
é-bólì |
∅-kútú=wé |
|
|
Yawa |
cl1.ipfv\cntr-break |
cl1-pot=art |
|
|
‘Yawa breaks the pot.’ |
|
b. |
Kóffí |
í-bólì |
∅-kútú=wé. |
|
|
Koffi |
cl1.ipfv\cntr-break |
cl1-pot=art |
|
|
‘Koffi breaks the pot.’ |
64Table 11 shows the pronominal paradigm, for a detailed discussion see Pozdniakova & Aplonova (2022). In this section, we discuss only those aspects of pronominal paradigms that are relevant for the agreement system, i.e. their anaphoric uses. The possessive construction is discussed in Section 5.3.2, while Section 5.3.3 is dedicated to other contexts with anaphora.
Table 11 — Ginyanga pronouns
|
|
Basic pronoun |
| Person |
1sg |
mì |
| 2sg |
fù |
| 1pl |
àɲí |
| 2pl |
fàɲí |
| Noun class |
1 |
mà |
| wè (in object position)a |
| 2 |
mɔ̀ |
| 3 |
mɔ̀ |
| 4 |
gɔ̀ |
| 5 |
gɔ̀ |
| 6 |
ŋmà |
| 7 |
àwá |
| 8 |
èyá |
| a. In Pozdniakova & Aplonova (2022), we suggested a separate pronominal series for object pronouns (cf. other descriptions of Guang languages, e.g. Segerer (2002) and Sherwood (1982) that follow the same logic). In this study, we use a less redundant classification because the object form is different from the basic pronoun only in CL1. |
65In the Ginyanga possessive construction, the possessor precedes the possessee. The possessive relation is not overtly marked, as seen in (23a). If the possessor is pronominal, it agrees with its antecedent (23b).
| (23) |
a. |
ɔ̀-ɲɩ́n=má |
gà-tɩ́n |
|
|
cl1-man=art |
cl4-house |
|
|
‘the man’s house’ |
|
b. |
mà |
|
gà-tɩ́n |
|
|
cl1 |
|
cl4-house |
|
|
‘his house’ |
66If the possessee is a kinship term (24a), its plural form is derived with the -áná suffix discussed in Section 3.2, even though CL2 class correlates do exist. However, this rule does not apply if the kinship term is used in a general sense or as a vocative, compare ‘the sisters’ in (24b) vs. ‘your sisters’ in (24a, c).
| (24) |
a. |
ná-jàŋál |
fú |
pɩ́cáná |
gì-jó |
rɔ̀ |
|
|
1sg.pfv-meet |
2sg |
sister:ass.pl |
cl5-market |
in |
|
|
‘I met your sisters at the market.’ |
|
b. |
ná-jàŋál |
à-pɩ́cɩ́=á |
gì-jó |
rɔ̀ |
|
|
1sg.pfv-meet |
cl2-sister=art |
cl5-market |
in |
|
|
‘I met the sisters at the market.’ |
|
c. |
*ná-jàŋál |
fú |
à-pɩ́cɩ́=á |
gì-jó |
rɔ̀ |
|
|
1sg.pfv-meet |
2sg |
cl2-sister=art |
cl5-market |
in |
|
|
Expected: ‘I met your sisters at the market.’ |
67With CL1 possessors, kinship terms do not have a noun class prefix and combine with the possessive pronoun even if the possessor is present. In (25a), bí ‘child’ is a kinship term with a CL1 possessor wúl ‘chief’. A possessive pronoun mè that refers to the chief is present despite the nominal possessor. In (25b), a possessee is not a kinship term (gàtɩ́n ‘house’) and a possessive pronoun does not occur.
| (25) |
a. |
ø-wúl=lé |
mè |
bí |
|
|
cl1-chief=art |
cl1 |
child |
|
|
‘the chief’s child’ |
|
b. |
ø-wúl=lé |
gà-tɩ́n |
|
|
cl1-chief=art |
cl4-house |
|
|
‘the chief’s house’ |
68Example below replicates (25) with CL2 possessors. Note that here the semantic property of the possessee does not affect the construction and that the possessive marking is identical in (26a) and (26b).
| (26) |
a. |
a-wúl=lé |
gè-bí |
|
|
cl2-chief=art |
cl5-child |
|
|
‘the chiefs’ child’ |
|
b. |
a-wúl=lé |
gà-tɩ́n |
|
|
cl2-chief=art |
cl4-house |
|
|
‘the chiefs’ house’ |
69The pronominal agreement patterns of Ginyanga are affected by animacy, which is true both in the sense that in agreement “animacy overrides the inherited, more elaborate formal gender assignment” (Güldemann & Fiedler 2019: 126), and in the opposite sense that words belonging to the same nominal form class (sharing a noun prefix) control separate agreement patterns if they differ in animacy. In (27-28), the referents require different object pronouns (mɔ̀ vs. àwá and mɔ̀ vs. èyá, respectively) despite having the same noun prefix (a- in (27) and ɩ-/i- in (28)), while animate antecedents with different noun prefixes in (27a) and (28a) are referred to by the same pronoun.
| (27) |
a. |
n-wù |
à-sáfɔ̀ |
n-wù |
mɔ̀ |
|
|
1sg.ipfv-see |
cl2-foreigner |
1sg.ipfv-see |
cl2 |
|
|
‘I see foreigners, I see them.’ |
|
b. |
n-wù |
à-bɩ́ |
n-wù |
àwá |
|
|
1sg.ipfv-see |
cl7-mountain |
1sg.ipfv-see |
cl7 |
|
|
‘I see mountains, I see them.’ |
| (28) |
a. |
n-wù |
ɩ̀-kprɛ́ |
n-wù |
mɔ̀ |
|
|
1sg.ipfv-see |
cl3-dog |
1sg.ipfv-see |
cl3 |
|
|
‘I see dogs, I see them.’ |
|
b. |
n-wù |
í-kìtíkì |
n-wù |
èyá |
|
|
1sg.ipfv-see |
cl8-manioc |
1sg.ipfv-see |
cl8 |
|
|
‘I see maniocs, I see them.’ |
70In their review of Kwa noun class systems, Konoshenko & Shavarina (2019: 22–26) distinguish three main strategies of class marking that can be used with different agreement targets in the same language. The first strategy is full class indexation, where class markers on the targets correspond to those on the noun controllers, with some discrepancies. The second one is phonological agreement, where targets systematically copy phonological material from class markers on the nouns. The last one is generalised animacy-based indexation driven by the semantic parameter of animacy alone. Diachronically, languages tend to drift from the more elaborate formal marking of the first type towards the generalised semantics-based strategy.
71Ginyanga seems to be in the beginning of this process of noun class attrition (cf. other North Guang languages, e.g. Krachi (Abunya 2018), that are at the more advanced stage of this process, resulting in less rigorous noun class agreement). While numerals, demonstratives and indefinite markers of Ginyanga retain their full class paradigms, mirroring the prefix of the noun (except for CL1), the choice of the marking on predicates and object pronouns is more complicated.
72On the one hand, this is where the semantic criteria come into play: nouns of the animate classes CL2 and CL3 have identical personal pronouns and subject prefixes, despite having different noun prefixes. In contrast, nouns of CL2 and CL7, and CL3 and CL8, while sharing the same nominal form class, require different subject and object markers, very clearly chosen based on their animacy (e.g. mɔ̀ referring to ‘persons’ and àwà referring to ‘teeth’).
73On the other hand, the merger of the personal pronouns of classes CL4 and CL5, and the subject markers of CL5 and CL6, cannot be explained by semantics. In the case of the personal pronoun gɔ̀, one may hypothesise that its functional extension was driven by the phonological similarity with both CL4 and CL5 nominal form classes. The gá-/gɩ́~gí- subject prefix of CL6 is even harder to explain without access to diachronic data. Either way, it is clear that subject and object noun class markers in Ginyanga deviate from the prototypical system where the number of nominal form classes defines the size of the marking paradigms.
74Table 12 aligns the Ginyanga data with the data of other Guang languages from Konoshenko & Shavarina’s review article (2019). It is noteworthy that animacy-based “indexation” prevails in subject and object markers while cardinal numerals tend to have full class “indexation” in languages where they are subject to agreement.
- 6 In Table 12, FC corresponds to full class indexation, NM to number agreement, An to animacy-based (...)
Table 12 — Noun class indexation principles in Guang (adapted from Konoshenko & Shavarina 2019)6
75It is also worth noticing that the distinction between inanimate classes in Ginyanga is preserved better than between animate classes: not only have singular subclasses ɔ~o-, ɩ-, and Ø- (presumably independent in Proto-Kwa) merged in terms of agreement, but both plural animate classes (CL2 and CL3) have the same subject marking and pronoun series, which makes them virtually indistinguishable in these contexts. Note that the plural inanimate classes CL7 and CL8 differ in every agreement context. This is highly unusual for Kwa languages, tending to preserve animate/human over inanimate classes so that the simplification of the noun class system first affects nouns lower in the animacy hierarchy (Konoshenko & Shavarina 2019: 41).
76This paper has featured an overview of the noun class system and agreement of Ginyanga, an understudied language of the Kwa family. Although Kwa languages are generally known for their reduced morphology, the North Guang branch is distinguished by its well-preserved and productive noun class system. This is also the case with Ginyanga: based on agreement, we define 8 noun classes, 3 of which are singular and the others plural. Classes are marked with noun prefixes that are subject to ATR harmony. Agreement classes are further divided into nominal form subclasses that share a agreement paradigm but feature different noun prefixes. Singular/plural pairings and semantic correlates of each class have been discussed; in particular, the role of the semantic parameter of animacy. Class-marking alternation is observed both in singular/plural pairings and derivational processes (diminutive formation), which is characteristic of active noun class systems. The following agreement contexts have been reviewed: class agreement of cardinal numerals, demonstratives, subject-verb and pronominal agreement. Furthermore, this paper has added evidence of a rich class system that could have existed in Proto-Guang.
77The paper was written at the different stages of research first funded by the Russian Science Foundation, then ERC Discourse reporting. It also benefited from the Sorbonne Nouvelle funding for Digital Humanities project as well as from the French ANR and German DFG agencies (CompLETE project). We are grateful to Andrey Shluinsky for his comments and suggestions, and to Svetlana Rubailo-Koudolo as well as Frederic Gnakpo, who helped us get in contact with the Anyanga community. Many thanks to our language assistants, Evagenou Tchala, Yao Amédomé, Emmanuel Kontre and Atchou Edoh, for their help, and to the Anyanga community in Togo for their kindness, patience and support.
| 1sg-3pl |
verb prefixes or anaphoric pronouns |
| =art |
article(clitic) |
| \cntr |
contracted forms |
| act |
actor |
| adj |
adjectivizer |
| ass.pl |
associative plural |
| cl1, 2, 3 etc. |
noun class prefixes or anaphoric pronouns |
| dim |
diminutive suffix |
| foc |
focalized |
| ipfv |
imperfective |
| num |
numeral |
| pfv |
perfective |