Negative Indefinite Constructions in Bantu: ‘Nobody’

Abstract

This paper presents a first typology of negative indefinites in Bantu languages. The lack of interest in expressions of ‘nobody’ in Bantu languages is connected with the idea that they merely involve a generic noun for ‘person’ and main clause negation. Our study, based on a dataset of 85 Bantu languages, shows that exponents of ‘nobody’ can be quite diverse, either resembling exponents of ‘somebody’ or differing from them through the use of pragmatically strong modifiers like scalar additive operators or explicit quantifiers like the numeral ‘one’. We also look at the constructions in which exponents of ‘nobody’ occur and observe that next to canonical SVO constructions, two construction types are used which are known to express information structure in Bantu languages: inversion constructions and existential cleft constructions. Inversion constructions are shown to be especially frequent with intransitive clauses but we also find object inversion for either ‘somebody’ or ‘nobody’ but not for both in the same language. Existential left constructions, on the other hand, are shown to be especially frequent with transitive clauses and, more importantly, they are more frequently used for expressions of ‘nobody’ than for expressions of ‘somebody’ in line with recent observations for South-East Asian and Mixtec languages.

1. Introduction

While there are a number of typological or at least cross-linguistic studies on various aspects of Bantu standard and non-standard negation (e.g., Devos et al., 2010; Devos & van der Auwera, 2013; Devos & Van Olmen, 2013; Bernander et al., 2022, 2023), negative indefinites, as expressed by English nobody, have not received much attention. This is partially due to the fact that dedicated pronominal elements with negative polarity such as anybody in (1)a, negative concord such as ain’t … nobody in (1)b, or inherent negativity such as nobody in (1)c are rare in Bantu (Devos, 2025, p. 306).

(1)

Negative indefinites in English

• I haven’t seen anybody
• I ain’t seen nobody.
• I have seen nobody.

In this respect, Bantu is not different from African languages, more generally. Haspelmath (1997, p. 241), when summarizing the worldwide distribution of indefinite pronoun types, states that African languages have generic-noun-based indefinites. In a worldwide typological study of negative indefiniteness Van Alsenoy used a sample of 179 languages and ‘[a]ll 22 African languages use polarity-neutral noun phrases to express negated indefinites’ (Van Alsenoy, 2014, p. 215; see also Zerbian & Krifka, 2008, pp. 393–394; Van Alsenoy, 2011).

Our starting point is a section in Devos (2025, pp. 306–307), which is a survey article on Bantu negation. Devos (ibidem) suggests that for negative indefiniteness the Bantu languages have two main strategies, canonical (S)VO in (2) and existentialization in (3), and in both one finds polarity-neutral generic nouns in combination with negated verb forms.

(2)

Ruwund (L531, Devos, 2025, p. 307, based on Nash, 1992, p. 284)

nì-men-àp	mu-ntu
sm1sg-see.prs-neg	12-person
‘I see nobody.’ (lit. ‘I don’t see a person.’)

(3)

Kanincin (L53A, Devos, 2025, p. 307, Michael Tshibanda Kasombo, pers. comm.)

kw-ììkil	mû-ncw	wàà-bul-àŋ	mw-aàn
sm17-not_be	1-person	sm1.rel.pst-hit-prf	1-child
‘Nobody has hit the child.’ (lit. ‘There is no person who hit the child.’)

Devos (ibidem) formulates three additional hypotheses: (i) the ‘existential cleft’ strategy, as in (3), is the preferred one when the indefinite is in the subject position; (ii) when the first strategy is used for subjects, the construction tends to contain an additional marker of indefiniteness; and (iii) generic nouns drop the preprefixal nominal marker, called ‘augment’, for the expression of ‘nobody’. The last hypothesis, of course, only applies in those Bantu languages that have a productive +/− augment alternation, an example of which is given in (4).

(4)

Kuria (JE43, Mary Zakaria Charwi, pers. comm.)

a.	o-moo-nto	a-roche	u-m-nyarobheri	w-ane
	aug-1-person	sm1-see.pfv	aug-1-neighbor	1-poss1sg
	‘Someone saw my neighbor.’
b.	moo-nto	a-ta-roche	u-m-nyarobheri	w-ane
	1-person	sm1-neg-see.pfv	aug-1-neighbor	1-poss1sg
	‘Nobody saw my neighbor.’

The brevity of these remarks calls for more work. So does the idea that Bantu languages avail themselves of existential constructions. This is interesting for two reasons: (i) there is recent general work on existential constructions in the Bantu languages (Devos & Bernander, 2022; Bernander et al., 2024), so this begs the question of whether there is any link between the general use of existential constructions in Bantu and the specific use for negative indefiniteness; and (ii) there is recent work on the use of existential constructions for negative indefiniteness in other languages, in particular, in languages of Mainland Southeast Asia (van der Auwera et al., 2023) and in the Mixtec languages (van der Auwera, 2023), so one wonders whether the use of existential negative constructions in these languages resembles what we find in Bantu.

In this paper, we will extend the brief account found in Devos (2025) and to characterize what the Bantu languages do for ‘nobody’, we will also study what they do for ‘somebody’. For this purpose, we established a database of 85 Bantu languages. This database is not itself a sample; i.e., it is not representative and has a bias toward Eastern Bantu languages, which is why we also define a sample. In Section 2, we characterize both the dataset and the sample. We then describe the data that we tried to collect and explain the shortcomings of our dataset. In Section 3, we examine the noun phrase exponents of expressions of ‘nobody’ and ‘somebody’, and in Section 4, we describe the different construction types. Correlations between exponent types, construction types, negation, and transitivity are explored in Section 5, which also briefly looks at the distribution of the different patterns. Section 6 is the conclusion.

2. Languages and Constructions

Our dataset contains 85 Bantu languages. They are listed in Appendix A, together with the sources and with the names of the informants and/or colleagues who helped us. To make the data more representative, we took the phylogenetic classification of the Bantu languages in Grollemund et al. (2015) and included only 4 languages for each of the 5 major branches. These four languages were chosen based on the quality of the available data. To these 20 languages, we added 4 Grassfields Bantu languages, also added in Appendix A, giving us a sample of 24 languages (cf. Section 5.4 and Tables 4 and 5). These are the languages that are underlined in Appendix A. We started out by perusing grammars and articles for relevant information. In most, if not all cases, the information was incomplete, and in 51 of the languages, we were able to add information through a questionnaire. The questionnaire was very simple. We wanted to find out how Bantu speakers say what English speakers do with nobody, as contrasted with somebody, dependent on whether they occur in the subject or object position, and for the subject position, whether the verb is transitive or intransitive like either arrive or ‘locative-existential’ be.

(5)

• Nobody saw my neighbor.
• Nobody arrived.
• Nobody was there.
• Somebody saw my neighbor.
• Somebody arrived.
• Somebody was there.
• My neighbor saw somebody.
• My neighbor saw nobody.

That the subject versus non-subject is an important cross-linguistic parameter is clear from the account in Devos (2025), from which we started. For non-subjects, we limit the work to (patient) objects. We also look at intransitives to see whether they involve different construction types from transitives. Of the two intransitive constructions in the questionnaire, one is a locational one (5)c,f, allowing us to see whether languages use an existential–locational construction with ‘somebody/nobody’ or do something different still. We made it clear that we wanted natural sentences, not just translations, and that we were interested in all natural renderings, not just one. Of the 85 languages in our dataset, 49 languages have natural renderings for all 8 sentences. Overall, our dataset is restricted in several ways. First, we do not have complete information for all languages, and the information we do have is sometimes lacking in detail. Not all of the respondents were linguists, and for some languages, we only have written replies to our questionnaire, which lack tonal information and phonetic detail. Although necessarily incomplete, we do include these data, hoping that this will incite further detailed research. Next, we have more and better information on Eastern Bantu languages; hence, the creation of a smaller more representative sample. Third, although we were able to capture some intralingual variations in the expression of ‘somebody’ and ‘nobody’, we most probably still missed a lot. Moreover, we did not perform an in-depth analysis of the attested variation. For now, we assume that the different translations given are all (equally) natural renderings of the sentences in the questionnaire. Finally, our study of negative indefinite constructions is limited to declarative main clauses. As hinted at in Section 5.3, non-declarative clauses may show different patterns.

We classify the constructions along two parameters. One concerns the noun phrase exponent of the entity that is asserted to exist or not exist. In (2), this is muntu, and in (3), it is mûncw, and we characterize them both as generic nouns for ‘person’. The other parameter concerns the construction in which this exponent appears. For (2), we characterize the construction as an (S)VO3 pattern, and for (3) as an existential pattern and more precisely an existential cleft construction. We will now describe and illustrate the various types, first for the exponents (Section 3), and then for the constructions (Section 4).

3. The First Parameter: Exponents of ‘Somebody’ and ‘Nobody’

To characterize the exponent of ‘somebody’ and ‘nobody’, we distinguish between three main types: (1) the exponent is absent, (2) the exponent is the generic noun for ‘person’, and (3) the exponent includes an adnominal modifier.

3.1. No Exponent

The first type is a bit paradoxical—there is simply no exponent, at least not in the shape of a noun phrase (this exponent type corresponds to absent in Table 1, Table 2, Table 3, Table 4 and Table 5). There are two points to make about the absence of the exponent: first, it is exclusively linked to the existentialization pattern (see Section 4.2) and, second, in most cases, the generic noun for ‘person’ can be added.

(6)

Luganda (JE15, Deo Kawalya pers. comm. & Betty Ayugi pers. comm.)

a.	wa-li-wo	e-y-a-tuuse
	sm16-be-loc16	rel-sm1-pst-arrive.pfv
	‘Somebody arrived.’ (lit. ‘There is who arrived.’)
b.	wa-li-wo	o-mu-ntu	e-y-a-gy-a
	sm16-be-loc16	aug-1-person	rel-sm1-pst-come-fv
	‘Somebody came.’

3.2. The Exponent Is the Bare Generic Noun for ‘Person’

The exponent can be the bare generic noun for ‘person’. In the majority of languages, this is a reflex of *mʊ̀-ntʊ̀ ‘person’ (Bastin et al., 2002). However, in the Central-Western Bantu languages Liko and Budu, either this reflex or another noun, also translated as ‘person’, may fill the slot. In Liko and especially Budu, there is a clear preference to use the second (innovated?) noun to render ‘somebody’ and ‘nobody’ alike.

(7)

• Liko (D201, Gerrit de Wit pers. comm.)

  mʊtʊ́ ‘person’ & mʊmbánzʊ́ ‘person’
  mʊ-lángá	á-mɩn-ǎ-ndɩ	mʊ-mbánzʊ́
  1-boy	sm1.pst-see-fv-pst	1-person
  ‘The boy saw somebody.’

• Budu (D332, Asangama, 1983, p. 219; David Kopa wa Kopa pers. comm.)

  moto ‘person’ & moombi ‘person’
  mo-mbi	óná-kɔmɔ́	tídŏ
  1-person	sm1.pst.see-neg	father
  ‘Nobody saw my father.’

Another potentially more telling variation concerns the presence or absence of an augment on the generic noun. It should be noted that we discuss the augment in this Section even though it is a cross-cutting characteristic: it is not just restricted to bare generic nouns but can also occur on generic nouns with an adnominal modifier or on the adnominal modifier itself if the latter is used independently. In Bantu studies, the term ‘augment’ is used to refer to a bound element that precedes the nominal prefix of nouns (or independently used adnominal modifiers) without affecting “the class assignment of the noun, nor its lexical meaning, nor the syntactic position it can occupy” (Van de Velde, 2019, p. 248). It is difficult to characterize the use of the augment in individual Bantu languages, let alone across Bantu (de Blois, 1970; Van de Velde, 2019; Halpert, 2025). Halpert (2025) makes a distinction between languages with a default augmentless pattern and languages with a default augmented pattern. In languages with a default augmentless pattern, it is the presence of the augment that needs to be explained (Halpert, 2025, pp. 170–171). In the North-Western Bantu language Eton, the noun appears with an augment in a restricted set of syntactic environments, more precisely, when the noun is followed by a restrictive relative clause or by a demonstrative (Van de Velde, 2017, p. 55). In some Eastern Bantu languages like Kagulu, on the other hand, the augment is never obligatory. When it appears, it is said to help convey specificity, topicality, or deixis (Petzell, 2003, p. 10). The question now is how these languages deal with exponents of ‘somebody’ or ‘nobody’ as far as the augment is concerned. As could be expected, generic nouns expressing ‘somebody’ or ‘nobody’ are typically augmentless in languages with a default augmentless pattern. An example from Kagulu is given in (8). The augmentless generic noun expresses ‘somebody’ in the subject position.

(8)

Kagulu (G12, Malin Petzell & Sauli Lengoliga pers. comm.)

mu-nhu	yu-monga	ha-kengil-a
1-person	1-other	pst-sm1.pst.arrive-fv
‘Somebody arrived.’

Augmented generic nouns occur in two languages with a default augmentless pattern in our dataset, i.e., the Eastern Bantu languages Vwanji and Kinga. In Vwanji, the noun phrase referring to ‘somebody’ includes an augment when it functions as a subject, both pre-verbally and post-verbally.

(9)

Vwanji (G66, Helen Eaton & Ahimidiwe Mahali pers. comm.)

ʉ-muu-nhu	jʉ-monga	a-lis-ile/a-lis-ile ʉ-muunhu jʉ-monga
aug-1-person	1-other	sm1-arrive-pfv
‘Somebody arrived.’

In Kinga, for which we have data from four varieties, some varieties allow the presence of the augment when the post-verbal generic noun expressing ‘somebody/nobody’ is indexed on the verb either as a class 1 object marker or as a class 1 subject marker. When ‘somebody’ functions as the object, co-indexed on the verb by a class 1 object marker, the augment is present in all four varieties. Interestingly, for ‘nobody’, the presence of the augment is also attested under these conditions, be it only in two varieties, and in each case, the generic noun is followed by a restrictive relative clause that could trigger the presence of the augment (as explained for Eton above). The examples below come from two different varieties. In (10), both ‘somebody’ and ‘nobody’ have an augment when functioning as the object. The generic noun is indexed on the verb by the class 1 object marker and in the case of ‘nobody’ followed by a restrictive relative clause. In (11), only the generic noun referring to ‘somebody’ has an augment. When ‘nobody’ is expressed, the augment is absent. Note that the generic noun is not followed by a restrictive relative clause in this case.

(10)

Kinga (G65, Enock Mbiling’i pers. comm.)

a.	u-jilani	v-ango	a-ka-m-bwene	u-mu-nu
	aug-1a.neighbor	1-poss1sg	sm1-pst-om1-see.pfv	aug-1-person
	‘My neighbor saw somebody’
b.	u-jilani	v-ango	na-ka-m-bon-e
	aug-1a.neighbor	1-poss1sg	neg.sm1-pst-om1-see-neg.pst
	u-mu-nu	uvya-vets-ag-a
	aug-1-person	rel1.sm1.pst-be-ipfv-fv
	‘My neighbor saw nobody’

(11)

Kinga (G65, Enock Mbiling’i pers. comm.)

a.	u-jirani	v-ango	a-ka-m-bwene	u-mu-nu
	aug-1a.neighbor	1-poss1sg	sm1-pst-om1-see.pfv	aug-1-person
	‘My neighbor saw somebody’
b.	u-jirani	v-ango	na-ka-m-bon-e
	aug-1a.neighbor	1-poss1sg	neg.sm1-pst-om1-see-neg.pst
	mu-nu
	1-person
	‘My neighbor saw nobody’

In Kinga ‘somebody/nobody’ functioning as the subject almost always occupies the post-verbal slot. It either follows a copula in an existential cleft construction (Section 4.2), or it follows the main verb in a subject inversion construction (Section 4.3). If the subject marker on the verb is in a locative class, the augment is always absent (12). However, if the subject marker agrees with the post-verbal subject, the augment is present with ‘somebody’ (12)b but absent with ‘nobody’ (12)c.

(12)

Kinga (G65, Enock Mbiling’i pers. comm.)

a.	ku-li	mu-nu	a-sikh-e
	sm17-cop	1-person	sm1-arrive-pfv
	‘Somebody was there.’
b.	a-kha-le-po	o-mu-nu	pala
	sm1-pst-cop-loc16	aug-1-person	16.dem
	‘Somebody was there.’
c.	na-kha-le-po	mu-nu
	neg.sm1-pst-cop-loc16	1-person
	‘Nobody was there.’

In sum, in languages with a default augmentless pattern generic nouns expressing ‘somebody/nobody’ are most likely to occur without an augment. Exceptions can occur and they mostly involve expressions of ‘somebody’ indexed on the verb by a class 1 subject marker or object marker.

Let us now turn to languages with a default augmented pattern. In these languages, it is the absence of the augment that needs to be explained (Halpert, 2025, pp. 172–178). Augment drop, whether optional or obligatory, typically happens in the post-verbal domain and tends to be triggered by negation, indefiniteness, and/or focus (Halpert, 2025, p. 178). In the Eastern Bantu language Luganda, any noun in the scope of negation will drop the augment (Hyman & Katamba, 1993; van der Wal & Namyalo, 2016). Luganda and several more or less closely related interlacustrine languages in our dataset show a contrast between augmented generic nouns expressing ‘somebody’ and augmentless generic nouns expressing ‘nobody’ when they function as the object in post-verbal position, as seen in (13) for Ishenyi. This contrast does not suggest a differential treatment of specific indefinites versus negative indefinites in terms of the augment but simply adheres to a grammatical rule. Hyman and Katamba (1993) show that definite nouns in Luganda also lose their augment when they are in the scope of negation.

(13)

Ishenyi (JE45, Rasmus Bernander field notes)

a.	o-mw-ibhatani	one	na-rooche	o-mo-nto
	aug-1-neighbor	1.poss1sg	foc.sm1-see.pfv	aug-1-person
	‘My neighbor saw somebody.’
b.	o-mw-ibhatani	one	ta-rooche	mo-nto
	aug-1-neighbor	1.poss1sg	neg.sm1-see.pfv	1-person
	‘My neighbor saw nobody.’

Similarly, the contrast between augmented ‘somebody’ and augmentless ‘nobody’ in some languages is due to the specific construction in which they occur. In Kirundi, another interlacustrine Bantu language, the augmented generic noun is used to express ‘somebody’ in the subject position, whereas an augmentless generic noun is used for ‘nobody’ functioning as the subject. This is a direct result of the type of construction the noun occurs in. ‘Somebody’ occupies the pre-verbal slot, as in (14), whereas ‘nobody’ is introduced by a negative existential copula, as in (14). Halpert (2025, p. 172) confirms that augmentless nouns are often required after the copula (but see the Simbiti data in (18)) and that the pre-verbal position typically requires the presence of the augment.

(14)

Kirundi (JD62, Pascal Tuyubahe pers. comm.)

a.	u-muu-ntu	ya-rá-boon-ye		u-mu-báanyi	w-aanje
	aug-1-person	sm1.rem.pst-disj-see-pfv		aug-1-neighbor	1-poss1sg
	‘Somebody saw my neighbor.’
b.	nta	muu-ntu	ya-bóon-ye
	cop.neg	1-person	sm1.rem.pst-see-rel.pfv
	u-mu-báanyi	w-aanje
	aug-1-neighbor	1-poss1sg
	‘Nobody saw my neighbour.’

However, in two languages, the contrast between augmented ‘somebody’ and augmentless ‘nobody’ cannot be attributed to negation nor to a constructional difference. In Kuria, a generic noun occurs in the pre-verbal position to express either ‘somebody’ or ‘nobody’ functioning as the subject. Whereas ‘somebody’ has an augment, ‘nobody’ is augmentless. It should be noted that a subject different from ‘nobody’ does not drop the augment in a negative clause, as seen in (15)c.

(15)

Kuria (JE43, Mary Zakaria Charwi pers. comm.)

a.	o-moo-nto	a-roche	u-mu-nyarobheri	w-ane
	aug-1-person	sm1-see.pfv	aug-1-neighbor	1-poss1sg
	‘Somebody has seen my neighbor.’
b.	moo-nto	a-ta-roche	u-mu-nyarobheri	w-ane
	1-person	sm1-neg-see.pfv	aug-1-neighbor	1-poss1sg
	‘Nobody has seen my neighbor.’
c.	u-mu-nyarobheri	w-ane	a-ta-roche	moo-nto
	aug-1-neighbor	1-poss1sg	sm1-neg-see.pfv	1-person
	‘My neighbor saw nobody.’

This pattern is partially attested in Ngoreme. Whereas ‘somebody’ is expressed by a specific indefinite noun (Section 3.3.1), which never has an augment, ‘nobody’ involves the augmentless generic noun in the pre-verbal position.

(16)

Ngoreme (JE401, Rasmus Bernander fieldnotes)

a.	ŋáni	a-aní	há
	1a.certain	sm1-prs.cop	16.dist.dem
	‘Somebody is there.’
b.	mó-nto	tá-á-t͡ʃ-iɾe
	1-person	neg.sm1-pst-come-pfv
	‘Nobody came.’

This is remarkable, as augmentless subjects appear to be extremely rare in pre-verbal position, at least as far as languages with a default augmented pattern are concerned (Halpert, 2025; Schneider-Zioga, 2007 for Kinande; van der Wal & Namyalo, 2016 for Luganda). Halpert (2025, p. 175) shows that in Kinande, for example, a generic noun expressing ‘somebody’ can be augmentless in the post-verbal position but never in the pre-verbal position.

(17)

Kinande (JD42, Halpert, 2025, p. 175)

a.	o-mu-ndu	a-ma-gong-a
	aug-1-person	sm1-prs-knock-fv
	‘Someone’s knocking./The person is knocking.’
b.	*mu-ndu	a-ma-gong-a
	1-person	sm1-prs-knock-fv
	intended: ‘Someone’s knocking.’
c.	ha-ma-gonga	mu-ndu
	sm16-prs-knock	1-person
	‘Someone’s knocking.’
d.	ha-ma-gonga	o-mu-ndu
	sm16-prs-knock	aug-1-person
	‘A/The person is knocking.’

Considering the data from Kuria and Ngoreme, it thus seems to be the case that the ban on pre-verbal augmentless nouns can be lifted with pre-verbal ‘nobody’. What is more, Kuria adheres to the typology of Becker (2022, pp. 264–269), following which the augment is an ‘inclusive-specific’ article, i.e., an article that may have both a definite and a specific indefinite function, as with ‘somebody’ in (15), but not a non-specific one. ‘Nobody’ is non-specific, so there cannot be an augment/inclusive-specific article, as in (15). However, as with most attempts at giving a cross-linguistic generalization of the function of the augment, counterexamples can be found. In Simbiti, for example, the augment is only dropped if ‘nobody’ is a post-verbal object. When ‘nobody’ is introduced by a negative existential copula, the generic noun is augmented.

(18)

Simbiti (JE431, Johnny Walker, pers. comm.)

a.	o-mo-menyani	w-aane	ti-yaa-rööshë	mö-ntö	hë
	aug-1-neighbor	1-poss1sg	neg-sm1.pst-see.pfv	1-person	neg
	‘My neighbor saw nobody.’

b.	taaho	ö-mö-ntö	ono	yaa-hik-irë	hë
	neg.loc.cop	aug-1-person	rel1	sm1.pst-arrive-pfv	neg
	‘Nobody arrived.’

In summary, we cannot uniformly characterize the way in which Bantu languages deal with exponents of ‘somebody’ or ‘nobody’ as far as the augment is concerned. The only generalization that can be made is that exponents of ‘nobody’ are more likely to drop the augment (or exclude it) than exponents of ‘somebody’ and that languages of the default augmented pattern can even do so in pre-verbal position, against a strong cross-Bantu tendency.

3.3. More Complex Exponents

We now come to syntactically complex exponents of ‘somebody’ or ‘nobody’. They may include an adnominal modifier. Adnominal modifiers normally take a nominal or pronominal prefix of class 1 in agreement with the head noun, i.e., the generic noun for ‘person’. The latter can often be omitted, in which case the adnominal modifier is used independently. Modifying nouns are also attested. They do not agree with the head noun, which may again be omitted. In one case, the generic noun is modified by a restrictive relative clause. Next, we also encounter non-nominal invariables, mostly scalar additive operators (following the typology proposed by Gast & van der Auwera, 2011), but also other focus markers and the interrogative pronoun ‘who’. The expression of ‘nobody’ often includes a combination of these markers. The discussion below is structured as follows: we start with modifiers shared by exponents of both ‘somebody’ and ‘nobody’ (Section 3.3.1) before turning to modifiers and invariables specific to the expression of ‘nobody’ (Section 3.3.2).

3.3.1. Indefiniteness Markers

Modifiers shared by expressions of ‘somebody’ and ‘nobody’ typically express indefiniteness. Our dataset shows several markers expressing notions on a scale between specific indefiniteness and non-specific indefiniteness. We start with specific indefinites (subsumed under indefinites in Table 1, Table 2, Table 3, Table 4 and Table 5). Many Bantu languages have a specific indefinite referring to a certain person, often said to be a known person of whom the speaker has forgotten the name or whom he does not want to name. The translation equivalents given by Laman (1936, p. 267) for the specific indefinite noun kingándi attested in Yombe and possibly other Kongo varieties illustrate the typical meaning range of such specific indefinites:

(19)

Laman (1936, p. 267)

• quelqu’un, un certain, un tel, qui est ainsi et ainsi, personne dont on a oublié le nom
• ou qu’on ne veut pas nommer
• [somebody, a certain, such a, so and so, person whose name one has forgotten
• or whom one does not wish to name]

Specific indefinites can be used to express ‘someone’. They are normally excluded from negative contexts in our dataset, but some specific indefinites also pop up in expressions of ‘nobody’, which suggests that they are evolving into non-specific indefinites. Many specific indefinites in our dataset function as modifying nouns. They either occur in combination with the generic noun for ‘person’ or they are used independently. The specific indefinites nyarebe and nyabhorebhe from the interlacustrine languages Gusii and Simbiti are cases in point. They both seem to involve nominalizations of the adnominal modifier rebe/lebe, attested in several more or less related languages with specific indefinite meanings, as seen in (20).

(20)

rebe/lebe

Kinande JD42	lebe/rebe	‘certain one, certain unnamed thing, so and so’	(Fraas, 1961, pp. 61, 273, 318; Kambale, 2006, pp. 71, 297; Mutaka & Kavutirwaki, 2012, pp. 208, 359)
Hunde JD51	lèbe	‘certain, a certain one’	(Kaji, 1992, p. 168)
Havu JD52	lebe	‘a certain one, designating a precise person or thing which one does not name’	(Aramazani, 1985, p. 208)
Tembo JD531	rébè	‘certain, some’	(Kaji, 1986, p. 328)

What is more, the pre-stem element nya- is known in Gusii (Cammenga, 2002, p. 201) and in Bantu more generally (Schadeberg & Bostoen, 2019, p. 195), to refer to personal names and common nouns. It thus seems quite certain that nyabhorebhe and nyarebe are, at least in origin, specific indefinites.4 In Simbiti nyabhorebhe is used in combination with the generic noun for person (21) to express ‘somebody’ functioning as the subject or the object. It is not attested in expressions of ‘nobody’.

(21)

Simbiti (JE431, Johnny Walker pers. comm.)

a.	ö-mö-ntö	nyabhorebhe	yaa-rööshë	o-mo-menyani
	aug-1-person	1a.certain	sm1.pst-see.pfv	aug-1-neighbor
	w-aane
	1-poss1sg
	‘Someone saw my neighbor.’

b.	o-mo-menyani	w-aane	yaa-rööshë	ö-mö-ntö	nyabhorebhe
	aug-1-neighbor	1-poss1sg	sm1.pst-see.pfv	aug-1-person	1a.certain
	‘My neighbor saw somebody.’

In Gusii, nyarebe is used independently to express ‘somebody’ functioning as the subject or the object (22). It cannot be used as a subject in negative sentences but does occur as an object expressing ‘nobody’, suggesting that it might be developing into a non-specific indefinite marker.

(22)

Gusii (JE42, Enoch Matundura pers. comm.)

a.	o-mo-amate	one	na-roche	nyarebe
	aug-1-neighbor	1.poss1sg	foc.sm1.pst-see.pfv	1a.certain
	‘My neighbor saw somebody.’
b.	o-mo-amate	one	ta-roch-i	nyarebe
	aug-1-neighbor	1.poss1sg	neg.sm1.pst-see.pfv-neg	1a.certain
	‘My neighbor saw nobody.’

The Swahili specific indefinite fulani ‘a certain one, certain’, a loan from Arabic (Sacleux, 1939, p. 229), is used in Swahili like nyabhorebhe is in Simbiti. It may combine with the generic noun for ‘person’ and is not attested in expressions of ‘nobody’. The same goes for the specific indefinite ŋáni in Ngoreme, which is used to express ‘somebody’ but not ‘nobody’ (16). A different distribution is found for kingándi in Yombe, which, judging from the translation equivalents given by Laman (1936, p. 267, cf. (19)), is a specific indefinite but it is used for expressions of both ‘somebody’ and ‘nobody’.

(23)

Yombe (H16c, Hilde Goes & Abel Massiala pers. comm.)

a.	ki-ngandi	be	ku-iz-a
	7-certain_person	sm1.rem.pst	np15-arrive-fv
	‘Somebody arrived.’
b.	ki-ngandi	ka-sa	ku-iz-a	kó
	7-certain_person	sm1-neg.rem.pst	np15-arrive-fv	neg
	‘Nobody arrived.’

A last enigmatic case involves the restrictive relative clause uvyavetsaga. It occurs in Kinga and together with the preceding generic noun its meaning was described by our language consultant as ‘a concerned person’. Surprisingly, this seemingly specific indefinite meaning is only attested for the expression of ‘nobody’ in the object position, as seen in (10), repeated here as (24).

(24)

Kinga (G65, Enock Mbiling’i pers. comm.)

u-jilani	v-ango	na-ka-m-bon-e
aug-1a.neighbor	1-poss1sg	neg.sm1-pst-om1-see-neg.pst
u-mu-nu	uvya-vets-ag-a
aug-1-person	rel1.sm1.pst-be-ipfv-fv
‘My neighbor saw nobody’

Adnominal modifiers either historically or synchronically related to the numeral ‘one’ may be involved in the expression of ‘somebody’ and ‘nobody’. Especially with ‘somebody’, the use of the numeral ‘one’ involves the grammaticalization of this numeral into an indefinite marker. We distinguish three patterns. The first pattern (classified as indefinite in Table 1, Table 2, Table 3, Table 4 and Table 5) involves indefinite markers diachronically related to the numeral ‘one’, which are used for the expression of ‘somebody’ but not for the expression of ‘nobody’. The North-Western Bantu language Tuki shows this pattern. It uses a reflex of the Proto-Bantu numeral ‘one’ (*mòì ‘one’ (Bastin et al., 2002)) for the expression of ‘somebody’ (mo), whereas ‘nobody’ involves the innovated numeral ‘one’ (mwasi).

(25)

Tuki (A601, Edmond Biloa pers. comm.)

a.	mò-tò	omo	a-má-úle
	1-person	1.ind	sm1-rem.pst-arrive
	‘Somebody arrived.’

b.	mò-tò	o-mwasí	a-tá-éná	mòtò	odzo	nakongo
	1.person	1-one	sm1-neg.rem.pst-see	1.person	who	near
	na	ngoto
	with	me
	‘Nobody saw my neighbor.’

The second pattern (classified as numeral ‘one’/indefinite in Table 1, Table 2, Table 3, Table 4 and Table 5) involves a marker diachronically or synchronically related to the numeral ‘one’, which is used for the expression of both ‘somebody’ and ‘nobody’. Crucially, the expression of ‘nobody’, although making use of the same marker, is syntactically more complex, typically involving a scalar additive operator (and thus literally expressing ‘(not) even one person’). In Kako, the expression of ‘somebody’ involves a form identical to the numeral ‘one’. However, whereas the numeral regularly follows the head noun (26)a, the indefinite marker used for the expression of ‘somebody’ precedes it (26)b. For the expression of ‘nobody’, the numeral again follows the head noun. Moreover, it occurs twice with a scalar additive operator in between (26)c.

(26)

Kako (A93, Ernst, 1992, p. 67)

a.	mu-mɔ	wɛtɛ	ta	kwa	si-ɛ	ɓo
	1-person	one	fut	leave	we-and	them
	‘One (1) person will leave with us.’

b.	wɛ̀tɛ	mu-mɔ	kɛ	numɛy
	one	1-person	at	door
	‘There’s someone at the door.’

c.	mu-mɔ	wɛtɛ	nɛ̀	wɛtɛ	tì	nje	na
	1-person	one	sao	one	neg	came	neg
	‘Nobody (not even one person) came.’

In Yombe, the indefinite marker mesi used in expressions of ‘somebody’ is formally related but not identical to the synchronic numeral ‘one’ in the language, which is mosi. Like in Kako, expressions of ‘nobody’ use the same marker but in combination with the scalar additive operator ne (27)b.

(27)

Yombe (H16c, Heidi Goes & Abel Massiala pers. comm.)

a.	mu-tu	mesi	Ø-be-di	vana
	1.person	ind	sm1-pst-cop	there
	‘Somebody was there.’

b.	ne	mu-tu	mesi	ka-sa	va	ba	kó
	sao	1-person	one	sm1-neg.rem.pst	there	be	neg
	‘Nobody was there.’

The two patterns exemplified above show that whereas grammaticalization of the numeral ‘one’ into an indefinite marker is a trigger for its use in expressions of ‘somebody’, this is not the case for expressions of ‘nobody’. They rather use the innovated numeral ‘one’ or the (old) numeral in a negative quantifying configuration typically expressing ‘(not) even one’.

The third pattern (also classified as numeral ‘one’/indefinite in Table 1, Table 2, Table 3, Table 4 and Table 5) involves a marker that is ambiguous between numeral ‘one’ and an indefinite reading, as with mókó in the Lingala example in (28). This marker is used, without further modifications, for the expression of both ‘somebody’ and ‘nobody’, as seen for Lingala in (28) and for North Boma in (29). We assume that this is possible because both the indefinite and the numeral ‘one’ reading are synchronically available.

(28)

Lingala (C30B, Meeuwis, 2020, p. 140; Joseph Koni Muluwa pers. com.)

a.	mo-to	mókó	a-yá-ákí	ko-luk-a	yó
	1-person	one	sm1an-come-pst1	inf-search-fv	2sg
	‘One person came to look for you.’/‘A certain person came to look for you.’

b.	mo-to	mókó	a-zal-ákí	wáná
	1-person	one	sm1an-be-pst1	there
	‘Someone was there.’

c.	mo-to	mókó	té	a-zal-ákí	wáná
	1-person	one	neg	sm1an-be-pst1	there
	‘Nobody was there.’

(29)

North Boma (B82, Sara Pacchiarotti & Guylen Bonkako, pers. comm.)

a.	tà:rá	á-mɔ́:n-ì	mɔ̀:-rɔ̀	í-mɔ́rì
	1a.father	sm1-see-pst	1-person	9-one
	‘Father saw somebody.’
b.	tà:rá	á-mɔ́:n-ì	mɔ̀:-rɔ̀	í-mɔ́rì	kɔ́
	1a.father	sm1-see-pst	1-person	9-one	neg
	‘Father saw nobody.’

Another adnominal modifier used in expressions of ‘somebody’ and ‘nobody’ is the indefinite distributive marker ‘(an)other’ (referred to as ‘(an)other’/indefinite in Table 1, Table 2, Table 3, Table 4 and Table 5). It should be noted that in some Bantu languages ‘other’, especially in the sense of ‘(an)other of the same kind’, is derived from the numeral ‘one’ (Kadima, 1967, pp. 26–27).5 In Tsonga, for example, ŋwana ‘other’ (30), is said to be composed of the numeral one ŋwe and a diminutive suffix -ana (Junod, 1896, p. 83; cited by Kadima, 1967, p. 27). Sitoe (1996, p. 173) gives the translation equivalents ‘(an)other’ and ‘certain’. The adnominal modifiers discussed here are all synchronically attested with the meaning ‘other’ and are often cognate with regional reconstructions of ‘other’ proposed by Kadima (1967). They are either used alone, as in Tsonga, or in combination with the generic noun, as in Kagulu. In both languages ‘other’ only occurs in expressions of ‘somebody’. In Gusii, on the other hand, ‘other’ is restricted to expressions of ‘nobody’.

(30)

Tsonga (S53, Benito Trollip pers. comm.)

u-ŋwana	u-fik-ile
1-other	sm1-arrive-pfv
‘Somebody arrived.’

(31)

Kagulu (G12, Malin Petzell & Sauli Lengoliga pers. comm.)

mu-nhu	yu-monga	ha-kengil-a
1-person	1-other	pst-sm1.pst.arrive-fv
‘Somebody arrived.’

(32)

Gusii (JE42, Enock Matundura pers. comm.)

o-nde	ta-ra-ik-a
1-other	neg.sm1.pst-rec.pst-arrive-fv
‘Nobody arrived.’

In Chewa, ina ‘other’ can be used to express ‘somebody’ too, but this gives rise to ambiguous readings and the preferred constellation includes a third-person singular possessive, which does not have a possessive reading but expresses something like ‘another of the same kind’ (Peter Msaka pers. comm.).

(33)

Chewa (N31b, Peter Msaka pers. comm.)

wina-wake	a-na-fik-a
1.other-1.poss1	sm1-pst-arrive-fv
‘Somebody arrived.’

3.3.2. Adnominal Modifiers and Invariables Restricted to the Expression of ‘Nobody’

We now turn to a set of invariable and adnominal modifiers that are restricted to expressions of ‘nobody’: scalar additive operators, free choice items, (negative) focus markers, and a single instance of an interrogative pronoun. They are not necessarily negative; i.e., most of them can also be used in affirmative contexts but they never partake in expressions of ‘somebody’.

Conjunctive coordinators in Bantu languages are known to be polyfunctional. They are often used not only as the conjunctive coordinator ‘and’ but also as a comitative preposition ‘with’ and an additive focus marker ‘too, also, as well’ (cf. Andrason, 2019a on Xhosa; Andrason, 2019b on Kituba; Creissels, 2016 on Tswana; Schneider-Zioga, 2015 on Kinande). In some languages, the semantic map of the conjunctive coordinator is broadened still to include a use as a scalar additive operator (sao in Table 1, Table 2, Table 3, Table 4 and Table 5). The latter use is attested in Xhosa (Andrason, 2019a) and also, but less prominently so, in Tswana (Creissels, 2016) and Kituba (Andrason, 2019b). In Xhosa the conjunctive coordinator na is used in both affirmative and negative contexts. Following the typology developed in Gast and van der Auwera (2011), it is used as a ‘BEYOND-operator’ in affirmative contexts. In (34), the local proposition of the scalar additive operator, i.e., a mouse, is pragmatically stronger than all corresponding context propositions. In other words, being afraid of a mouse is unexpected in comparison to being afraid of, e.g., a lion.

(34)

Xhosa (S41, Andrason, 2019a, p. 39)

ne-mpuku	le	i-ya-m-nkywantis-a
sao-9.mouse	dem9	sm9-foc-om1-frighten-fv
‘Even a mouse frightens her.’

The use of na in negative contexts shows that it can be considered an extended BEYOND-operator; i.e., its minimal host proposition (including negation) is pragmatically stronger than all relevant context propositions (Gast & van der Auwera, 2011). Although the local proposition of ‘having a cent’ in (35) is pragmatically weak, the minimal host proposition of ‘not having a cent’ is pragmatically strong.

(35)

Xhosa (S41, Andrason, 2019a, p. 40)

U-sipho	a-ka-na-yo	ne-senti
aug-1a.Sipho	neg-sm1-have-rel9	sao-9.cent
‘Sipho has not even a cent.’

When na is involved in the expression of ‘nobody’, it typically combines with the numeral ‘one’ (indicated as sao & numeral ‘one’ in Table 1, Table 2, Table 3, Table 4 and Table 5), as has already been illustrated in Section 3.3.1 for Yombe (27). Another example comes from the Eastern Bantu language Tharaka, in (36). Nyakyusa, another Eastern Bantu language, shows the same construction but without the generic noun for ‘person’, as seen in (37). Merged forms like na-jʊmo in Nyakyusa are further discussed in Section 4.5.

(36)

Tharaka (E54, Muriungi, 2010, p. 100)

mu-ntû	noa	û-mwe	a-ti-ra-gur-a	î-buku
1-person	sao	1-one	sm1-neg-pst-buy-fv	7-book
‘Nobody bought a book.’

(37)

Nyakyusa (M31, Persohn, 2017, p. 73)

po	na-jʊ-mo	a-ti-kw-i-tuufy-a-mo
then	sao/neg-1-one	sm1-neg-prs-refl-praise-fv=some
‘So no one ever praises oneself …’

Languages in which the scalar additive operator and the generic noun alone can express ‘nobody’ are rare in our dataset, but they do occur. The Central-Western language Babole in (38) is a case in point.

(38)

Babole (C101, Leitch, 1994, p. 205)

à-ka-ɛ́n-í	na	mo-tò
sm1-neg-see-pfv	sao	1-person
‘He didn’t see anybody.’

The combination with the numeral ‘one’ is not only attested with na, but also occurs with other scalar additive operators for which we have less information on their etymology. In the Central-Western Bantu language Liko, gʊtʊ́gʊ seems to be used as an extended BEYOND-operator. In the affirmative sentence in (39), gʊtʊ́gʊ signals that its local proposition ‘that they plunge fish hooks (=to fish)’ is pragmatically strong. The men were not used to fishing but had to try it to obtain food (de Wit, 2015, pp. 275–276). In the negative sentence in (40), the minimal host proposition referring to ‘not listening’ is pragmatically strong. The speaker could have disobeyed the advice, but he did not even listen to it.

(39)

Liko (D201, de Wit, 2015, pp. 275–276)

ɓa-lʊ́kʊ́	ɓa-kɩngɩl-ag-ǎ	ndɩ	gʊtʊ́gʊ
2-man	sm3pl-try-plur-fv	pst3	sao
ɓɛ́	ɓó-lub-ǒg-í	masɩkɩɗángɩ́
comp	sm3pl-plunge-plur-sbjv	6.fish_hook
‘The men even tried to fish.’ (lit. ‘that they plunge fish hooks.’)

(40)

Liko (D201, de Wit, 2015, p. 276)

ná-kʊkan-a	gʊtʊ́gʊ	ndɩ	mʊ-tɩ́wɩ	ka-babǎ	ɩɓʊ́
sm1sg-neg:hear-fv	sao	pst3	3-advice	gen-1a.father	2.pro
na	mamá
with	1a.mother
‘I did not even listen to the advice of my father and my mother.’

Just like with na, gʊtʊ́gʊ combines with the numeral ‘one’ in expressions of ‘nobody’. What is more, the pragmatic strength of its minimal host proposition can be reinforced by the addition of the contrastive focus marker áka.

(41)

Liko (D201, Gerrit de Wit pers. comm.)

gʊtʊ́gʊ	mʊ-mbánzʊ́	ɓe-motí	(áka)	Ø-ká-mɩn-á-gʊ
sao	1.person	1.num-one	foc	sm3sg-neg-see-fv.pst-neg
mʊ-lángá
1-boy_(first one to be circumcised)
‘Not even one person/Nobody saw the first boy to be circumcised.’

Note that in Yombe, an emphatic negative marker with a minimizing effect (mvéla ‘not even a bit’, Laman, 1936, p. 633) can be used together with the scalar additive operator instead of the numeral ‘one’ (cf. also (27)). It can also be the only modifier of the generic noun for ‘person’.

(42)

Yombe (H16c, Heidi Goes & Abel Massiala pers. comm.)

a.	ne	mu-tu	mvela	ka-sa	mon-a	kó	di-sia
	sao	1-person	neg.foc	sm1-neg.rem.pst	see-fv	neg	5-neighbor
	di-ama
	5.poss1sg
	‘Nobody saw my neighbor.’

b.	di-sia	di-ama	ka-sa	mon-a	mu-tu	mvela
	5-neighbor	5-poss1	sm1-neg.rem.pst	see-fv	1-person	neg.foc
	kó
	neg
	‘My neighbor saw nobody.’

In the West-Western language Nzadi the scalar additive operator áta also combines with the numeral ‘one’ for the expression of ‘nobody’. áta is probably a loan from the local lingua franca Kituba and possibly, just like in Swahili, it ultimately derives from the Arabic preposition ‘until’ (Sacleux, 1939, pp. 273–274). Whereas in languages like Kituba (Fehderau, 1992, p. 2) and Swahili, it functions as a ‘universal scalar additive operator’ (as in the typology by Gast & van der Auwera, 2011), it is restricted to negative utterances in Nzadi (Crane et al., 2011, p. 179).

(43)

Nzadi (B865, Crane et al., 2011, p. 179)

atá	muur	ómɔtúk	ko	ya	bɔ
sao	person	one	neg.pst	come	neg
‘No one came.’

Interestingly, in a number of languages, scalar additive operators seem to have developed into ‘scalar negators’, incorporating an expression of negation (Gast & van der Auwera, 2011). As will be further discussed in Section 4.5, scalar negators may occur without clausal negation, as seen in (44), from the South-Western language Nyaneka.

(44)

Nyaneka (R13, da Silva Maia, 1966, p. 377)

na-wike	u-mu-hole
sao/neg-1.one	sm1-om1-love-pfv
‘Nobody loves him.’

We now turn to exponents of ‘nobody’ which include a free choice item. Free choice items in Bantu languages tend to be related to the universal quantifier ‘all’ and/or to the distributive universal quantifier ‘every’ (Zerbian & Krifka, 2008; Osa-Gómez, 2015). Whereas in some cases one and the same form means both ‘any’ and ‘every’, in other cases, the free choice item is clearly derived from the universal quantifier, against Haspelmath’s (1997, p. 156) generalization that ‘every’ cannot develop in ‘any’. In a footnote, he already gives exceptions from Hebrew and Turkish. Bantu languages provide additional counterexamples. In Swahili, for example, the (distributive) universal quantifier ‘all, every’ preceded by a bound pronoun6 has a free choice interpretation. In Chewa, the syntactically complex distributive universal quantifier ‘every’, a relative clause consisting of the copula (li) followed by the universal quantifier (onse ‘all’ preceded by a bound pronoun), may have a free choice interpretation (Zerbian & Krifka, 2008, p. 15; Paas, 2004, p. 13).

(45)

Swahili

• vitabu vyote ‘all the books’
• vitabu vyovyote ‘any books, whichever books’

(46)

Chewa (Paas, 2004, p. 13)

• munthu aliyense ‘every person, any person’

Osa-Gómez (2015) gives a detailed analysis of the free choice item in the Eastern Bantu language Nata. Like in Swahili, it consists of the universal quantifier (in Nata osé ‘all’) preceded by a bound pronoun. It can be used in (non-episodic) affirmative as well as negative utterances and, following Osa-Gómez (2015), its main function in both environments is domain widening. Through the addition of the free choice item, the speaker urges the hearer to consider all possible, even marginal, alternatives rather than just contextually relevant or salient ones. The free choice item in (47), for example, is used when the speaker thinks that the hearer might exclude marginal alternatives like ‘rotten bananas’. Likewise, the free choice item in (47), makes clear that not a single child will be taken, contrary to the assumed expectation of the hearer.

(47)

Nata (JE45, Osa-Gómez, 2015, p. 157)

a.	aa	e-bh-ebhúse	m-bhi-háa-ri	e-bhi-tọọké	bhjobhjoosé
	no	aug-8-baboon	comp-sm8-hab-eat	aug-8-banana	fci8
	‘No, baboons eat any banana.’
b.	ne-te-ku-ghégh-a		mu-aná	wowoosé
	sm1sg-neg-ipfv-pick-fv		1-child	fci1
	‘I won’t take any child.’

Exponents of ‘nobody’ including a free choice item are restricted to post-verbal position in our dataset (but an interesting exception is found in the non-declarative clause, cf. Section 5.3). In Lusoga and Kagulu, the free choice item is derived from the universal quantifiers ena ‘all’ and ose ‘all’, respectively. In Lusoga, the universal quantifier is fully reduplicated, whereas in Kagulu, it is preceded by a bound pronoun, resulting in a partially reduplicated pattern.

(48)

Lusoga (JE45, Osa-Gómez, 2015, p. 157)

n-kaali	kú-kóbél-á	múú-ntú	yéenayééná
sm1sg-not_yet	15-tell-inf	1-person	fci1
‘I have not yet told anybody.’

(49)

Kagulu (G12, Malin Petzell & Sauli Lengoliga pers. comm.)

h-echak-a	mu-nhu	wowose	y-eng-ile
sm16-lack-pst	1-person	fci1	sm1-arrive-pfv.rel
‘Nobody arrived.’

In Chewa (cf. (46)) and Yao, the free choice item is derived from the univerbation of a relative clause involving a copula (li and ana, respectively) and the universal quantifier, both of which agree in class 1. In Chewa, the construction is also used to express ‘every’ (Zerbian & Krifka, 2008, p. 15; Paas, 2004, p. 13). The free choice item combines with the indefinite ‘(an)other’, rather than with the generic noun for ‘person’ to express ‘nobody’, as seen in (51).

(50)

Chewa (N31b, Peter Msaka pers.comm.)

neba	w-anga	sa-na-on-e	w-ina	aliyense
1.neighbor	1-poss1sg	neg.sm1-pst-see-pfv	1-other	fci1
‘My neighbor saw nobody.’

(51)

Yao (P21, Julius Taji pers.comm.)

jilani	jw-angu	nga-na-mw-on-a	mu-ndu	jwanajoose
1a.neighbor	1-poss1sg	sm1sg.neg-pst-om1-see-fv	1-person	fci1
‘My neighbor saw nobody.’

In Southern Sotho, yet another form is attested, involving disjunctive reduplication, not of the universal quantifier ‘all’, but rather of an interrogative pronoun, either ‘who’ (mang kapa mang) or ‘which’, as seen in (52).

(52)

Southern Sotho (S33, Malete, 2001, p. 267)

ha-ho	mo-tho	o-fe	kapa	o-fe
neg-loc.cop	1-person	1-which	or	1-which
eo	b-ana	ba-mo-rat-ang
dem1	2-child	sm2-om1-like-prs
‘The children like nobody. (lit.: There is no person which or which whom the children like.)’

This brings us to the use of the bare interrogative pronoun mang ‘who’ for the expression of ‘nobody’ in Southern Sotho. Markers that are semantically interrogative but can also be used as indefinites have been referred to as ‘ignoratives’ (Karcevski, 1969; see also van der Auwera et al., 2023).

(53)

Southern Sotho (S33, Malete, 2001, p. 341)

ha-ho	mang	ya-tseb-ang
neg.loc-cop	who	sm1.rel-know-prs
	‘Nobody knows. (lit.: There is no who who knows.)’

3.4. Exponent Types: Overview and Discussion

Exponents of ‘somebody’ and ‘nobody’ may be quite similar in Bantu languages: they may consist of a bare generic noun or contain an indefinite marker. When expressions of ‘somebody’ and ‘nobody’ differ, the expression of ‘nobody’ is typically pragmatically stronger. We have seen that expressions of ‘nobody’ are more likely to drop the augment than expressions of ‘somebody’, a feature often linked to focus in Bantu languages. It should be noted that we did not include information on the presence versus absence of the augment in Table 1, Table 2, Table 3, Table 4 and Table 5 as this feature is not always directly linked to the expression of ‘somebody’ or ‘nobody’. As explained in Section 3.2, augment drop may be licensed by negation whether the noun in question is indefinite or not. Next, ‘nobody’, but not ‘somebody’, may involve scalar additive operators and free choice items, both signaling pragmatic strength. Finally, ‘nobody’ often involves a syntactically complex noun phrase including more than one marker like with scalar additive operators expressing ‘(not) even one’, free choice items involving either (partial) reduplication or a relative clause and configurations including an (optional) emphatic focus marker. Table 1, Table 2 and Table 3 include three exponent types that have not yet been discussed in detail: adnominal negative markers derived from a negative copula (negative copula/neg in Table 1, Table 2 and Table 3), an SAO in combination with the numeral ‘one’ (sao & numeral/neg) and the standard negative marker used as an adnominal negative marker (neg). Since they are intertwined with construction types and/or the position of the negator, they are further discussed in Section 4.2, Section 4.5, and Section 4.4, respectively.

Something that has remained underexposed in our discussion so far is the fact that different exponent types may co-exist in one and the same language. The West-Western Bantu language Yombe, for example, has five co-existing strategies for ‘nobody’: indefinite kingándi, the generic noun in combination with the scalar additive operator ne, the generic noun in combination with ne and either the minimizer mvéla or the numeral mesi, and the generic noun in combination with the minimizer mvéla alone (Heidi Goes & Abel Massiala, pers. comm.). Most languages in our dataset do not show this amount of variation but this is an area in need of further research.

4. The Second Parameter: Construction Types

In this Section, we discuss three construction types attested for both ‘somebody’ and ‘nobody’: SV(O) constructions (Section 4.1), existential cleft constructions (Section 4.2), and inversion constructions (Section 4.3). We then take a closer look at the position of the negator in constructions for ‘nobody’ (Section 4.4). In Section 4.5, we suggest that a reduction in the negative existential cleft construction through a grammaticalization process well-known to affect basic clefts may exceptionally give rise to dedicated pronominal expressions of ‘nobody’ with inherent negativity or negative concord (Section 4.5). The findings of this Section are summarized in Section 4.6, where we also point out that just as exponent types, different construction types may co-exist in one and the same language.

4.1. SV(O) Constructions

In Bantu languages, the pre-verbal domain is typically interpreted as non-focal or topical, whereas the Immediate-After-Verb position has a non-topical or focal interpretation, and the basic or canonical word order associated with this configuration is SVO (Bearth, 2003; van der Wal, 2015). We refer to constructions as SV(O) constructions if the exponents of ‘somebody/nobody’ occur in their ‘normal’ pre-verbal S or post-verbal O slots. The examples from the Eastern Bantu language Kuria (54) and the Central-Western language Boa (55) show SVO constructions for ‘somebody/nobody’ in the subject and object positions, respectively.

(54)

Kuria (JE43, Mary Zakaria Charwi pers. comm.)

a.	o-moo-nto	a-roche	u-m-nyarobheri	w-ane
	aug-1-person	sm1-see.pfv	aug-1-neighbor	1-poss1sg
	‘Somebody saw my neighbor.’
b.	moo-nto	a-ta-roche	u-m-nyarobheri	w-ane
	1-person	sm1-neg-see.pfv	aug-1-neighbor	1-poss1sg
	‘Nobody saw my neighbor.’

(55)

Boa (C44, Gerrit de Wit pers. comm.)

a.	pombó	a-mén-i	mo-tó
	1a.young_man	sm1.pst-see-rec.pst	1-person
	‘The boy saw somebody.’
b.	pombó	ká-mén-í	mo-tó
	1a.young_man	neg.sm1.pst-see-rec.pst	1-person
	‘The boy saw nobody.’

Bantu languages like Kuria, and Boa for that matter, allow specific and negative indefinites functioning as the subject to occur early in the sentence. However, the following two construction types show that Bantu languages also have strategies to avoid such a configuration.

4.2. Existential Cleft Constructions

Existential cleft constructions are bi-clausal structures consisting of a matrix, which is an existential clause introducing the exponents of ‘somebody/nobody’, and a relative clause co-indexed with ‘somebody/nobody’. The examples in (56) and (57) from the Eastern Bantu languages Lusoga and Nyamwezi illustrate such ‘existential clefts’. The exponent for ‘somebody’, i.e., the generic noun for ‘person’, is the nominal predicate of an existential clause and the semantic main verb is relativized. As was already hinted at in the introduction and will be further discussed in Section 5.2, existential cleft constructions are especially frequent with expressions of ‘nobody’ and less so ‘somebody’ in the subject position. Below, we first take a look at the characteristics of the matrix, which are shown to correspond to the characteristics of simple existential clauses in Bantu languages. We then look at the whole construction with special attention to indications of signs of a reduction in negative existential cleft constructions.

(56)

Lusoga (JE16, Betty Ayugi pers. comm.)

e-li-yo	o-mu-ntu	e-ya-bonn-a	mu-lirwana
sm19-cop-loc19	aug-1-person	rel1-sm1.pst-see-fv	1-neighbor
w-ange
1-poss1sg
‘Someone saw my neighbor.’ (lit. ‘There is a person who saw my neighbor.’)

(57)

Nyamwezi (F22, Ponsiano Kanijo pers. comm.)

kw-iina	m-nhʊ	waa-lɩ-hoo
sm17-com	1-person	sm1.rel.pst-cop-loc16
‘Somebody was there.’ (lit. ‘There is a person who was there.’)

Recent research on existential clauses in Bantu languages (Devos & Bernander, 2022; Bernander et al., 2024) has shown that they typically display the following three characteristics: locative agreement marking, subject inversion, and the use of a locative copula or a comitative predicate. These characteristics are also attested in the matrix of existential cleft constructions. Locative nouns in Bantu languages are generally derived by the addition of a locative nominal prefix of class 16 *pa-, 17 *kʊ, 18 *mʊ or less typically by the locative nominal prefix of class 23/25 (also referred to as class 19) *ɪ- or the locative suffix -(i)ni (Meeussen, 1967; Grégoire, 1975; Maho, 1999, pp. 204–206; Schadeberg & Samsom, 1994). Locative nouns can induce locative agreement within the noun phrase and on the verb. Locative agreement on the verb can also occur independently to refer to an implicit location or expletively. Locative marking within existential clauses most typically concerns the subject marker slot, as in (57). However, secondary locative marking, often in the form of an enclitic7 and exceptionally a proclitic, is attested as well. This secondary locative marking may combine with a locative subject prefix, as in (56), or may be the only locative marking within the existential clause, as seen in (58), from Ishenyi. The existential predicate has a class 1 subject marker that is co-referential with the generic noun for person as well as a class 16 locative enclitic.

(58)

Ishenyi (JE45, Rasmus Bernander field notes)

ta-re-ho	mo-nto	ono	a-hik-ire
neg.sm1-cop-loc16	1-person	rel1	sm1-arrive-pfv
‘Nobody arrived.’ (lit. ‘There is a person who arrived.’)

A second characteristic of existential clauses is subject inversion; i.e., the entity said to exist follows rather than precedes the existential predicator resulting in a non-canonical VS word order. Based on the typology of Bantu subject inversion proposed in Marten and van der Wal (2014), Devos and Bernander (2022) make a distinction between locative inversion and agreeing inversion in existential clauses. Locative inversion is seen in (56) and (57): the existential predicate has a locative subject marker (used referentially or expletively) instead of a class 1 subject marker agreeing with the post-verbal generic noun. The Makhuwa example in (59), just like the Ishenyi one in (58), displays agreeing inversion: its subject marker is non-locative and agrees with the post-verbal generic noun.

(59)

Makhuwa-Enahara (P31E, Jenneke van der Wal pers. comm.)

khaá-vó	ń-tthú	o-n-ly-á	woó-hí-wuryá
neg.1sm-be.there	1-person	rel1-prs-eat-fv	1.conn.15-neg-drink
‘Nobody drinks without eating.’ (lit. ‘There’s no person who eats without drinking.’)

The third characteristic of existential clauses concerns the verbal element, which typically is either a locative copula, as with li in Lusoga, re in Ishenyi, and vo in Makhuwa, or a comitative predicate, as with ina in Nyamwezi. Both types tend to involve reflexes of Proto-Bantu ‘be’ verbs, either *dɪ̀ or *bá (Bastin et al., 2002). When used as a locative copula in plain locational or existential clauses, the ‘be’ verb may take a locative enclitic. If the two merge and the copula part is eroded, the locative enclitic itself may function as a copula. This is what must have happened in Makhuwa. Likewise, the merger of the copula with the comitative preposition na to express ‘be with, have’ followed by complete or partial erosion of the copula part, or drop of the copula in present tense contexts, has led to inflected forms of the comitative in many Bantu languages (Creissels, 2024). Although negative existential clauses, used for the expression of ‘nobody’, often differ from affirmative existential clauses only in the presence of standard negative marking (Section 4.4), they sometimes involve a specialized predicator typically derived from an inherently negative lexeme (Bernander et al., 2022, pp. 16–22; Bernander et al., 2023, pp. 27–28). In one of the Kinga varieties, for example, the existential cleft construction involves the locative copula li for expressions of ‘somebody’ but uses the inherently negative verb tsila ‘be without’ for expressions of ‘nobody’.

(60)

Kinga (G65, Enock Mbiling’i pers. comm.)

a.	ku-li	mu-nu	a-m-bwene	u-jirani	v-ako
	sm17-cop	1-person	sm1-om1-see.pfv	aug-1a.neighbor	1-poss1sg
	‘Somebody saw my neighbor.’ (lit. ‘There’s a person who saw my neighbor.’)
b.	ku-tsil-a	mu-nu	a-m-bwene	u-jirani
	sm17-be_without-fv	1.person	sm1-om1-see.pfv	aug-1a.neighbor
	v-ako
	1-poss1sg
	‘Nobody saw my neighbor.’ (lit. ‘There lacks a person who saw my neighbor.’)

In all of the examples in this Section, except for the Kinga ones in (60), the existential matrix clause is accompanied by a relativized semantic main verb. Data from another Kinga variety, presented in (61), show the presence of a relativized main verb but the absence, at least in the negative construction, of an existential predicate in the matrix clause.

(61)

Kinga (G65, Enock Mbiling’i pers.comm.)

a.	pwa-le	uva-ka-sikh-e
	sm16.pst-cop	rel1.sm1-pst-arrive-pfv
	‘Someone arrived.’ (lit. ‘There’s who arrived.’)
b.	na-pamoto	uva-ka-m-bwene	u-jilani	v-ane
	sao/neg-one	rel1.sm1-pst-om1-see.pfv	aug-1a.neighbor	1-poss1sg
	‘Nobody saw my neighbor.’ (lit. ‘There isn’t who saw my neighbor.’)

The absence or reduction of the existential predicate (61) and/or the relative clause marking (60) is probably due to a grammaticalization process known to affect basic clefts crosslinguistically (Harris & Campbell, 1995; Heine & Traugott, 1991) and also in Bantu languages (van der Wal et al., 2025). It concerns the development from bi-clausal constructions to mono-clausal ones and is typically manifested by the reduction or loss of the predicator in the matrix clause and by the reduction or loss of relativization in the relative clause. Relative clauses in Bantu languages are often marked tonally, and because some of our data lack tonal information, we are hesitant to say much about reductions within the relative clause. However, we do see evidence of a reduction in the existential predicator. A number of interlacustrine Bantu languages make use of a more or less similar invariable negative copula in the matrix of the existential cleft constructions. They are listed in (62).

(62)

Reduced existential predicators

JD61 Kinyarwanda:	nta	(Zorc & Nibagwire, 2007, p. 300)
JD62 Kirundi:	ńta(a)	(Meeussen, 1959, p. 176)
JD63 Kifuliiru:	ndáá	(van Otterloo, 2011, p. 226)
JE14 Rukiga:	taa	(Kaji, 2023, p. 549)
JE43 Kuria:	ta	Mary Charwi pers. comm.
JE431 Simbiti:	taaho	Johnny Walker pers. comm.

In Kirundi expressions of ‘somebody’ use the SV(O) construction type, whereas expressions of ‘nobody’ involve an existential cleft construction with the invariable negative existential predicator nta.

(63)

Kirundi (JD62, Pascal Tuyubahe pers. comm., Meeussen, 1959, p. 183)

a.	nta	muu-ntu	ya-bóon-ye	u-mu-báanyi	w-aanje
	cop.neg	1-person	sm1.pst-see-pfv.rel	aug-1-neighbor	1-poss1sg
	‘Nobody saw my neighbor.’
b.	ntaa	ma-búye	(a-rí-ho)
	cop.neg	6-stone	(sm6-cop-loc16)
	‘There are no stones.’

Although we do not have historical data on these languages, the existential predicators all seem to be reduced forms of an inflected locative copula involving the negative prefix ta-. The example in (58) from Ishenyi, yet another interlacustrine Bantu language, gives an idea of what this copula could have looked like. Interestingly, in Rukiga and Kuria, we also see reductions in the relative marking. In Rukiga, the affirmative verb following the matrix clause with the invariable copula taa can either be relative (64) a,b or non-relative (64) c, whereas in Kuria it is, at least segmentally, identical to non-relative verbs (65). Still, we need to keep in mind that relativization could be marked tonally.

(64)

Rukiga (JE14, Kaji, 2023, p. 549)

a.	ta:-mu-ntu	o-wa:-yij-a
	cop/neg-1-person	rel1-sm1.pst-come-fv
	‘Nobody came./There’s nobody who came.’
b.	ta:-mu-shéja	o-wá-gyend-ire
	cop/neg-1-man	rel1-sm1.pst-go-pfv
	‘No man has gone./There is no man who has gone.’
c.	ta:-mu-shê:ja	wá:-gyend-ire
	cop/neg-1-man	sm1.pst-go-pfv
	‘No man has gone.’

(65)

Kuria (JE43, Mary Zakaria Charwi pers. comm.)

a.	te-moo-nto	a-hik-ire	hai
	cop/neg.aug-1-person	sm1-arrive-pfv	neg
	‘Nobody arrived.’
b.	o-moo-nto	a-hik-ire
	aug-1-person	sm1-arrive-pfv
	‘Somebody arrived.’

Equally interesting is the fact that these invariable negative copula merge with the following noun. Merged forms like the ones in (65) are reminiscent of dedicated negative indefinites without negative concord. However, their distributional features betray their link with existential cleft constructions. In Rukiga, the merged forms can also be used in the object position, but they need to precede the verb, whether followed by a relative clause or not.

(66)

Rukiga (JE14, Kaji, 2023, p. 549)

a.	ta:-mw-á:na	a-gír-a
	cop/neg-1-child	rel.sm1-have-fv
	‘He has no child./There is no child that he has.’
b.	ta:-mw-â:na	a-gír-a
	cop/neg-1-child	sm1-have-fv
	‘He has no child.’

In Kuria, on the other hand, the merged forms are only attested in the subject position. In the object position, the (bare) generic noun for ‘person’ is used, as in (67), reflecting a general tendency for existential cleft constructions to be a preferred strategy only when ‘nobody’ functions as the subject (cf. also Section 5.2).

(67)

Kuria (JE43, Mary Charwi pers. comm.)

u-mu-nyarobheri	w-ane	a-ta-roch-e	moo-nto
aug-1-neighbor	1.poss1sg	sm1-neg-see-pfv	1-person
‘My neighbor saw nobody.’

Babole has an invariable negative marker nàká, which, judging from its distributional features, could also be linked to a negative existential cleft construction. It can only occur before the verb, and there is no negative concord, as seen in (68).

(68)

Babole (C101, Leitch, 1994, pp. 205, 208)

a.	nàká	ná	dz-úmbá	dí-kwíe-í	lá	bw-èndò
	cop/neg	even	5-thing	sm5-fall-pfv	from	14-sky
	‘Nothing fell from the sky.’
b.	*tò-ɛ́n-í	nàká	ná	mbíendzà
	*sm1pl-see-pfv	cop/neg	even	10.buffalo
	(intended: ‘We saw no buffalos.’)

Nzadi has a negative existential predicator ka, which may or may not combine with the clause-final negative marker bɔ (69)a. The same combination occurs in what looks like an existential cleft construction (69)b, i.e., an existential clause followed by a relative clause, which in Nzadi may be left unmarked (Crane et al., 2011, p. 194). However, ka can also combine with clausal negation (69)c, suggesting a reanalysis as an adnominal negator in combination with clausal negation, i.e., negative concord.

(69)

Nzadi (B865, Crane et al., 2011, pp. 240, 230 & Simon Nsielanga Tukumu pers. comm.)

a.	ka	mwàán	(bɔ)
	cop/neg	child	neg
	‘There’s no child.’
b.	ka	muur	ómɔtúk	(bɔ)	o	lyaa
	cop/neg	person	one	(neg)	pst	cry
	‘Nobody cried.’
c.	ka	muur	ómɔtúk	ko	lyaa	bɔ
	cop/neg	person	one	neg.pst	cry	neg
	‘Nobody cried.’

Nyoro has an adnominal negator that combines with clausal negation (70). We mention it here because the negator is also used as a negative answer particle in combination with locative prefixes (hâ:ha (class 16), kwâ:ha (class 17) ‘no’; Kaji, 2015, p. 532). This is interesting, seeing that Bernander et al. (2022) show that negative answer particles may be derived from dedicated negative existential predicators. The adnominal negator in (70) may thus have an origin in an existential cleft construction.

(70)

Nyoro (JE11, Kaji, 2015, p. 532)

o-mú:-ntu	w-â:ha	ta-iz-írê
aug-1-person	1-neg	neg.sm1-arrive-pfv
‘Nobody has come.’

Finally, we return to the Kinga example in (61), which shows the complete omission of the negative existential predicator. In Section 4.5, we suggest that this omission of the predicator in negative existential cleft constructions is intertwined with another process, whereby a scalar additive operator attached to the numeral ‘one’, napamoto in (61), is reinterpreted as a scalar negator with inherent negativity, rendering the negative predicator redundant.

4.3. Inversion Constructions (Without Clefting)

The third construction type involves inversion without clefting. Inversion is most typically attested when ‘somebody/nobody’ functions as the subject. It then involves subject inversion. As with existential matrix clauses in existential cleft constructions, two main agreement patterns are attested: the verb either agrees with the inverted subject (71) or the verb carries a locative agreement prefix (72), used referentially or expletively. Here as well, we refer to these agreement patterns as ‘agreeing inversion’ and ‘locative inversion’, respectively.8

(71)

Xingoni (N12x, Heidrun Kröger, Mustafa Mussa Binamo; Jacinto Saidine Chauré & Paulino Emede pers.comm.)

a-hik-ite	mu-ndu
sm1-come-pfv	1-person
‘Somebody arrived.’

(72)

Changana (S53, Bento Sitoe pers. comm.)

a-ku-fik-anga	mu-nhu
neg-sm17-arrive-neg.pst	1-person
‘Nobody arrived.’

Secondary locative agreement occurs with both agreeing and locative inversion. Cuwabo and Vwanji display agreeing inversion. The class 1 subject prefix agrees with the inverted subject. Secondary locative agreement occurs in the form of a class 16 locative enclitic and proclitic in Cuwabo (73) and Vwanji (74), respectively. Luganda, on the other hand, has locative inversion. The verb has a class 16 locative subject marker and involves secondary locative agreement in the form of a class 19 locative enclitic (75). In cases of double locative agreement, the subject marker is often used expletively, whereas the locative enclitic has a referential function and shows less restrictions as to the choice of locative class (see, e.g., Devos et al., 2017 on Kirundi).

(73)

Cuwabo (P34, Guérois, 2015, p. 246)

ka-dh-ílé-vó	mu-ttu
neg.sm1-come-pfv-loc16	1-person.pl
‘Nobody came.’

(74)

Vwanji (G66, Helen Eaton pers. comm.)

pala	na-pw-a-lya-l-e	muu-nhu
loc16.dem	neg-loc16-sm1-pst-be-pfv	1-person
‘Nobody was there.’

(75)

Luganda (JE15, Betty Ayugi pers. comm.)

wa-li-yo	o-mu-ntu
sm16.pst-be-loc19	aug-1-person
‘Somebody was there.’

A few languages display object inversion, i.e., ‘somebody’/’nobody’ functioning as the object occurs in pre-verbal rather than post-verbal position. Yombe is a case in point. When ‘somebody’ functions as the subject, the canonical SVO pattern is attested, as in (76). However, when ‘somebody’ functions as the object, it appears in a pre-verbal position resulting in an SOV pattern, seen in (76). The construction in (76)b is also morphologically distinct from the one in (76)a, as it uses a ka-prefix rather than a Ø-prefix as the third-person singular subject marker of class 1.

(76)

Yombe (H16c, Heidi Goes & Abel Massiala pers. comm.)

a.	mu-tu	mesi	Ø-be-mon-a	di-sia	di-ama
	1-person	ind	sm1-pst-see-fv	5-neighbor	5-poss1sg
	‘Somebody saw my neighbor.’
b.	di-sia	di-ama	mu-tu	mesi	ka-be-mon-a
	5-neighbor	5-poss1sg	1-person	ind	sm1-pst-see-fv
	‘My neighbor saw somebody.’

Bostoen and Mundeke (2012) argue that in Mbun and in H16 Kongo languages like Ntandu and Manyanga object focus involves both syntactic marking, putting the object in the ‘Immediate-Before-Verb’ position, and morphological marking, i.e., the selection of the allomorph ka- as the class 1 subject marker. The utterances in (76), thus, show that ‘someone’ in the object position behaves like a focused object, at least in Yombe. Although more research is needed, it seems that this combined syntactic and morphological marking of object focus is restricted to affirmative utterances. In Yombe, ‘nobody’ in the object position does not trigger a marked SOV word order nor special verbal morphology.

(77)

Yombe (H16c, Heidi Goes & Abel Massiala pers. comm.)

di-sia	di-ama	ka-sa	mon-a	ki-ngandi	kó
5-neighbor	5-poss1sg	sm1-neg.rem.pst	see-fv	7-certain_person	neg
‘My neighbor saw nobody.’

Finally, the Central-Western Bantu language Mbochi has SVO word order when ‘somebody’ functions as the object. When ‘nobody’ functions as the object, the generic noun for person occurs between the auxiliary and the main verb. Dryer (2013) suggests that Mbochi has a different word order with negation (i.e., SOVNeg instead of SVO in affirmative clauses). Our data suggest that the word order change is instigated by more than negation alone. When ‘my neighbor’ functions as the object of a negative clause, no word order change occurs, possibly because ‘my neighbor’ contrary to ‘nobody’ is not in focus.

(78)

Mbochi (C25, Guy Kouarata pers. comm.)

a.	obia	la	ngá	á-dz-ee	mo-ro	o-táa	kaá
	1a.neighbor	of	I	sm1-be-pfv	1-person	15-see	neg
	‘My neighbor saw nobody.’
b.	mo-ro	á-dz-e	o-táá	obia	la	ngá	kaá
	1-person	sm1-be-prf	15-see	1a.neighbor	of	I	neg
	‘Nobody saw my neighbor.’

4.4. The Position of the Negator

In what follows, we take a closer look at an important additional parameter in constructions with ‘nobody’, i.e., the position of the negator.

In constructions with an SV(O) pattern, three subtypes can be distinguished: clausal negation, adnominal negation, and both adnominal negation and clausal negation (i.e., negative concord). Negation is overwhelmingly expressed at the clausal level. As is typical for Bantu languages, this mostly implies that the verb includes negative morphology. The Liko example in (41), repeated here as (79), in which the verb has a post-initial and a post-final negative marker, is a case in point.

(79)

Liko (D201, Gerrit de Wit pers. comm.)

gʊtʊ́gʊ	mʊ-mbánzʊ́	ɓe-motí	(áka)	Ø-ká-mɩn-á-gʊ
even	1.person	1.num-one	foc	sm3sg-neg-see-fv.pst-neg
mʊ-lángá
1-boy_(first one to be circumcised)
‘Not even one person/Nobody saw the first boy to be circumcised.’

Negation can also be expressed on the exponent. This pattern is attested in Lingala and Kituba (both lingua franca): they have analytical verb phrase negation through the particles té and vé, respectively, which come at the end of the clause (Meeuwis, 2020, p. 252, Joseph Koni Muluwa pers. comm.). When ‘nobody’ functions as the subject, the negative particle follows the exponent implying that noun phrase negation rather than verb phrase negation is at stake here, as seen for Lingala in (80) and for Kituba in (81). When ‘nobody’ functions as an object, the negative marker follows the exponent of ‘nobody’ immediately and precedes an (optional) adverb, as seen in (80)b.

(80)

Lingala (C30B, Joseph Koni Muluwa pers. comm., Michael Meeuwis pers. comm.)

a.	mo-to	mókó	té	a-món-ákí	tata	na	ngáí
	1-person	one	neg	sm1an-see-pst1	1a.father	of	me
	‘Nobody saw my father.’ (lit. ‘Not one person saw my father.’)

b.	tatá	a-món-ákí	mo-to	mókó	té	na	balabála
	1a.father	sm1an-see-pst1	1-person	one	neg	on	9.street
	‘Father saw nobody on the street.’ (lit. ‘My father saw not one person on the street’)

(81)

Kituba (H10, Joseph Koni Muluwa pers. comm.)

mu-ntu	mósi	vé	ku-vand-aka	páana
1.person	one	neg	15-be-pst	there
‘Nobody was there.’ (lit. ‘Not one person was there.’)

Other instances of negative marking in the vicinity of the exponent without clausal negation can be linked to existential cleft constructions, as was already suggested for the Babole example in (68), repeated here as (82), and will be further discussed in Section 4.5.

(82)

Babole (C101, Leitch, 1994, p. 208)

nàká	ná	dz-úmbá	dí-kwíe-í	lá	bw-èndò
neg.foc	even	5-thing	sm5-fall-pfv	from	14-sky
‘Nothing fell from the sky.’

A third and last pattern concerns the negative marking of the exponent as well as clausal negation, i.e., ‘negative concord’. Examples are found in four languages in our dataset: Barombi, Mokpe, Nzadi, and Nyoro, and also in the Grassfields language Bamum. The North-Western languages Barombi and Mokpe, as well as the Grassfields language Bamum, have a strict negative concord; i.e., the negated exponent combines with clausal negation whether it is in pre-verbal or post-verbal position, as illustrated for Mokpe and Bamum in (83) and (84), respectively.

(83)

Mokpe (A22, Tanda & Neba, 2005, p. 21, Vincent Tanda pers. comm.)

a.	tò	mòtò	à	zrá	ówí	ŋgówà
	neg	person	sm	neg	kill	pig
	‘Nobody killed a pig.’
b.	tò	mòtò	à	zrá	líŋgànì	tò	jómà
	neg	person	sm	neg	likes	neg	thing
	‘Nobody likes anything.’

(84)

Bamum (Grassfields, Nchare, 2012, p. 404)

a.	nʃè-mɯ̀m	pí	mâ	m-fù	ŋì	wá
	neg-person	pst2	neg	ptcp-call	3sg	me
	‘Nobody called me.’

b.	nʃè-mɯ̀m	mâ	ŋ-ɡέr	ì	nʃè-lɛ̀twà
	neg-person	neg	ptcp-have	3sg	neg-book
	‘Nobody has any book.’

For Nyoro, we only have an example with ‘nobody’ in the subject position (70). For Nzadi, it seems that negative marking is more likely with ‘nobody’ in the subject position than with ‘nobody’ in the object position (Simon Nsielanga Tukumu pers. comm.), which confirms the link with an existential cleft construction, suggested in Section 4.2. In an existential cleft construction, we expect negation to be marked in the matrix only, as in (85)a. However, as already mentioned in Section 4.2, the negative predicator seems to have been re-analyzed as an adnominal marker that can combine with clausal negation, i.e., negative concord, as in (85)b. Still, when ‘nobody’ functions as the object, a construction with clausal negation only seems to be preferred, as in (85)c.

(85)

Nzadi (B865, Crane et al., 2011, p. 230, Simon Nsielanga Tukumu pers. com.)

a.	ka	atá	(bɔ)	muur	ómɔtúk	o	ya
	cop/neg	even	neg	person	one	pst	come
	‘Nobody came.’
b.	ka	atá	muur	ómɔtúk	ko	ya	bɔ
	cop/neg	even	person	one	neg.pst	come	neg
	‘Nobody came.’
c.	mí	ko	mɔ́n	àtá	muur	ómɔtúk
	I	neg.pst	see	even	person	one
	‘I saw nobody.’

Existential constructions come in subtypes as well. Negation is either marked on the existential copula or on the exponent. The first subtype is by far the most frequent one. The existential verb either carries negative morphology, as in (86), or it is intrinsically negative, either a full negative verb as in (87), or a copula as in (88)a. It should be noted that we are not entirely sure whether the utterance in (88)a contains a relative clause or whether it exemplifies an SVO construction with negation marked on the exponent. There are two elements in favor of the former analysis: the low-toned class 1 subject prefix ò- is typical for subject relative clauses (Stappers, 1986, p. 24) and the sentence-initial negative marker â is reminiscent of a negative copula used in identificational clauses (Stappers, 1986, p. 31), see (88)b. If this analysis is the right one, North Boma is another example of a language using an invariable negative copula in the existential clause (cf. Section 4.2).

(86)

Swahili (G42, Enock Matundura pers. comm.)

ha-ku-na	(m-tu)	a-li-ye-wasili
neg-sm17-com	(1-person)	sm1-pst-rel1-arrive
‘Nobody arrived.’

(87)

Lusoga (JE16, Betty Ayugi pers. comm.)

e-zir-a	mu-ntu	e-ya-bonna	mu-lirwana	w-ange
sm19-lack-fv	1-person	rel1-sm1-see.pfv	1-neighbor	1-poss1sg
‘Nobody has seen my neighbor.’

(88)

North Boma (B82, Sara Pacchiarotti & Guylen Bonkako pers. comm., Stappers, 1986, p. 31)

a.	â	mɔ̀:-rɔ̀	í-mɔ́rì	ò-mɔ́ːn-ì	tá:r	ì	ŋgàlí	kɔ̀
	neg.cop	1-person	9-one	sm1.rel-see-pst	1a.father	of	me	neg
	‘Nobody saw my father.’
b.	a	bá-kâ:ru	kɔ́
	neg.cop	2-woman	neg
	‘Those are not women.’

Note that the North Boma clause-final negative marker kɔ is hard to interpret in terms of scope. It is normally the second element of discontinuous negative marking, but in relative clauses, the first negative element is absent (Stappers, 1986, p. 45). The same problem arises in Kivunjo, in which the clause-final negative marker pfho is the sole exponent of negation. The available data do not allow us to ascertain whether these clause-final negative markers have scope over the existential matrix clause or over the relative verb. Still, seeing that elsewhere in our database negation is never marked on the relative clause, scope over the matrix clause seems more likely.

(89)

Kivunjo (E622C, Gundelinda Shayo pers. comm.)

ku-wo-re	mu-ndu	a-m-bom-a	mo-mraso	o-ko	pfho
sm17-pst-be	1-person	sm1-om1-see-fv	1-neighbor	1-poss1sg	neg
‘Nobody saw my neighbor.’

The existential matrix clause of a few languages does not have a predicator. As mentioned in Section 4.2, the drop of the existential predicator seems to be intertwined with a process whereby scalar additive operators become reinterpreted as (scalar) negators. As these changes pertain to the construction type, the exponent type, as well as to negation, we return to them in a separate Section (Section 4.5). We have already given one example from the Eastern Bantu language Vwanji. Expressions of ‘nobody’ are rendered through an existential cleft construction consisting of a matrix clause without a predicator and a relative clause. The exponent of ‘nobody’ is the scalar additive operator naa added to the numeral wumo. The scalar additive operator is here re-analyzed as a scalar negator. It is the only negative element in the sentence.

(90)

Nyiha (M23, Helen Eaton pers. comm.)

naa-wu-mo	we	á-fis-ile
sao/neg-1-one	rel1	sm1-arrive-pfv
‘Nobody arrived.’

Interestingly, there is no split for the inversion types. One could imagine the negator to go either on the verb or on the (inverted) exponent or maybe even on both. However, we only find negation on the verb, both with agreeing inversion (91)—whether or not accompanied by a secondary locative agreement (92)—and locative inversion (93).

(91)

Makhuwa (P31E, Jenneke Van der Wal pers. comm.)

kha-w-aálé	n-tthu
neg.sm1-come-pfv	1-person
‘Nobody came.’

(92)

Vwanji (G66, Helen Eaton pers. comm.)

pala	na-pw-a-lya-l-e	muu-nhu
loc16.dem	neg-loc16-sm1-pst-be-pfv	1-person
‘Nobody was there.’

(93)

Changana (S53, Bento Sitoe pers. comm.)

a-ku-fik-anga	mu-nhu
neg-sm17-arrive-neg.pst	1-person
‘Nobody arrived.’

4.5. Scalar Additive Operators, Negative Concord, and Existential Cleft Constructions

In this Section, we take a closer look at expressions of ‘nobody’ that are univerbations of a scalar additive operator and the numeral ‘one’. The forms in question and the languages they appear in are listed in (94). In each case, the scalar additive operator has merged with the numeral ‘one’, which takes a pronominal prefix of class 1. The latter is optional in Chokwe and absent in Kinga.

(94)

‘Nobody’ as an univerbation of a scalar additive operator and ‘one’

Kinga (G65)	napamoto (/na-pamoto/)
Rukiga (JE14)	nomwe (/na-u-mwe/)
Chokwe (K11)	nyi(u)mwe (/nyi(-u)-mwe)
Mbukushu (K333)	noyofotji (/no-yo-fotji/)
Nyiha (M23)	naawumo (/naa-wu-mo/)
Ndali (M301)	naayumo (/naa-yu-mo/)
Nyakyusa (M31)	najʊmo (/na-jʊ-mo/)
Umbundu (R11)	lomwe (/la-u-mwe/)
Nyaneka (R13)	naumwe (/na-u-mwe/)
	nawike (/na-u-ike/)

In the previous Section (Section 4.4), we have seen that constructions with ‘nobody’ are most frequently negated on the main verb in SV(O) and inversion construction types, and on the existential predicator in existential cleft constructions. We also saw that a reduction in the negative existential predicate may give rise to different negation patterns involving adnominal negation with or without negative concord (Section 4.2). As shown in Section 3.3.2, SAOs can combine with generic nouns and/or other syntactic markers, typically the numeral ‘one’, to express ‘nobody’. Exponents of ‘nobody’ including an SAO combine with clausal negation, as seen in (38), repeated here as (95), and in (96).

(95)

Babole (C101, Leitch, 1994, p. 205)

à-ka-ɛ́n-í	na	mo-tò
sm1-neg-see-pfv	sao	1-person
‘He didn’t see anybody.’

(96)

Kanincin (L53A, Michaël Kasombo, pers. comm.)

mu-ntw	aâp	wù-mwîng	kàà-shik-a	pend (SV)
1.person	sao	1-one	neg.sm1.pst-arrive-fv	neg
‘Nobody arrived.’

One can easily imagine that SAOs that are frequently used in negative contexts (especially expressions of ‘not even (one)’) become reinterpreted as scalar negators—a standard pathway for a language to acquire negative concord. With this in mind, we will now investigate how the pronominal forms listed in (94) pattern with negation. Do they combine with clausal negation, and could this then be interpreted as negative concord or do other patterns occur?

Four different patterns can be distinguished. First, in Rukiga, nomwe occurs in an existential cleft construction. It is introduced by a negative existential verb and does not have inherent negative meaning. We do not know whether nomwe can also be used as an object.

(97)

Rukiga (JE14, Taylor, 1985, p. 59)

ti-ha-ri-ho	n-o-mwe	o-kor-a	ekyo
neg-sm16-cop-loc16	soa-1-one	sm1.rel-do-prs.fv	that
‘Nobody does that.’

Second, in Kinga, Mbukushu, Umbundu, and Nyaneka, the merged forms do not combine with clausal negation. Rather, they behave like dedicated negative indefinites with inherent negativity. When taking a closer look at their distribution, some particularities are observed. In Kinga, napamoto only occurs in the subject position, where it is followed by a relative clause (61), never in the object position where a different exponent is used, which combines with clausal negation, for which see (24). In Umbundu (98), Nyaneka (99), and Mbukushu (100), lomwe, naumwe/nawike, and noyofotji, respectively, can fulfill both functions but only in a pre-verbal position. This distributional restriction seems to be dictated by an existential cleft configuration, whether or not the relative clause is reduced, which cannot be deduced from the available data.

(98)

Umbundu (R11, Keiling, 1937, p. 25)

a.	l-o-mwe	wey-a
	sao/neg-1-one	sm1.come-pst
	‘Nobody came.’
b.	la-vi-mwe	va-ling-a
	sao/neg-8-one	sm2-do-pst
	‘They didn’t do anything.’

(99)

Nyaneca (R13, da Silva Maia, 1966, p. 377)

a.	na-w-ike	u-mu-hol-e
	sao/neg-1-one	sm1-om1-love-pfv
	‘Nobody loves him.’
b.	na-u-mwe	ndyi-lwete
	sao/neg-1-one	sm1sg-see.pfv
	‘I see nobody.’

(100)

Mbukushu (K333, Fisch, 1998, p. 134)

a.	no-y-ofotji	temba	mu-nu	gho-ku-pir-a	mu-randu
	sao/neg-9-one	also	1-person	conn1-15-have-fv	3-blame
	‘No one is blameless.’
b.	no-th-ofoji	temba	thi-yama	tu-na-mon-o
	sao/neg-9-one	also	9-animal	sm1pl-prs-see-pfv
	‘We saw not a single head of game.’

Next, in Nyiha and Ndali, naawumo and naayumo behave like dedicated negative indefinites with non-strict concord (van der Auwera & Gianollo, 2025, pp. 26–32). When used as subjects, they head a matrix clause without an existential predicator and are followed by a relative clause, as illustrated for Nyiha in (101). When used as objects, however, they follow a negative main verb, for which see (101) also from Nyiha. Their use is less intimately connected with the existential cleft construction.

(101)

Nyiha (M23, Helen Eaton & Damas Mwashitete pers. comm.)

a.	naa-wu-mo	we	á-mu-vhweny-i	u-mu-palamani	w-ane
	sao/neg-1-one	rel1	sm1-om1-see-pst	aug-1-neighbor	1-poss1sg
	‘Nobody saw my neighbor.’
b.	u-mu-palamani	w-aane	a-tá-mu-vhweny-i	naa-wu-mo
	aug-1-neighbor	1-poss1sg	sm1-neg-om1-see-pst	sao/neg-1-one
	‘My neighbor saw nobody.’

A third pattern is attested in Nyakyusa; najʊmo or related negative indefinites always combine with clausal negation, whether in the pre-verbal (37) or post-verbal position (102). They could thus be said to have strict negative concord. However, it could also be the case that in Nyakyusa, the SAO has not been re-analyzed as a scalar negator and that we simply have the predominant main clause negation pattern here.

(102)

Nyakyusa (M31, Persohn, 2017, p. 48)

ne	n-ga-job-a	na-si-mo
1sg	sm1sg-neg-speak-fv	sao/neg-10-one
‘I haven’t said anything.’

Chokwe, finally, seems to be in between the second and suggestive third type. We do not have data on its post-verbal use, but pre-verbally niumwe can have negative concord (103)b, or not (103)a.

(103)

Chokwe (K11, Mukanda wa Zambi, 1970)

a.	ni-u-mwe	w-anyingik-a	mwe	mw-ana	shina	tata
	sao/neg-1-one	sm1-know-fv	?	1-child	except	1a.father
	‘No one knows the son except the father.’
b.	mu-tu	ni-u-mwe	k-echi	k-evw-a	li-ji	li-enyi
	1-person	sao-1-one	sm1-neg.cop	sm1-hear-fv	5-voice	5-poss1
	mu	ma-jila
	18	6-street
	‘No one will hear his voice in the street.’

In summary, it could be hypothesized that the SAO became reinterpreted as a scalar negator and that this triggered the grammaticalization toward a mono-clausal structure: the existential verb, no longer necessary for the expression of negation, could easily be dropped. The merged SAO-‘one’ forms, freed from the existential negative verb, either behave like dedicated negative indefinites with inherent negativity or like dedicated negative indefinites with non-strict or strict negative concord.

4.6. Construction Types: Overview and Discussion

Exponents of ‘somebody’ and ‘nobody’ may occur in their canonical pre-verbal subject or post-verbal object slot. However, since pre-verbal subjects in Bantu languages are generally interpreted as non-focal if not topical, strategies avoiding ‘somebody/nobody’ in a pre-verbal position occur in many languages. Next to SV(O) constructions, we identified existential cleft constructions and inversion constructions. Clefting and word order permutations are well-known strategies for information structure expressions in Bantu languages (van der Wal et al., 2025).

Existential cleft constructions are bi-clausal constructions consisting of an existential clause introducing ‘somebody/nobody’ and a relative clause. The existential matrix has all the characteristics of Bantu existential clauses, i.e., the presence of locative marking, locative or agreeing inversion, and the use of a locative copula or comitative predicator. Basic or it-clefts are known to be used for exhaustive focus in Bantu languages (van der Wal et al., 2025, p. 22). However, less is known about the pragmatic correlates of existential clefts in Bantu languages. They are mainly claimed to be used as presentationals and, more specifically, to introduce a new entity into the discourse (Devos & Bernander, 2022, pp. 583, 588–591; Lafkioui et al., 2021 on Kirundi).

Word order permutations (without clefting) mostly involve subject inversion, with ‘somebody/nobody’ functioning as the subject appearing after the verb in a slot dedicated to focus in many Bantu languages, i.e., the Immediate-After-Verb focus position (Kerr et al., 2023, pp. 9–10). We also identified cases of ‘object inversion’ in two DRC Bantu languages. In Yombe and Mbochi, ‘somebody’ and ‘nobody’, respectively, occur before the (semantic) main verb when functioning as the object. In the area where these languages are spoken, the dedicated focus position is not IAV but rather IBV, i.e., the Immediate-Before-Verb position (Kerr et al., 2023, pp. 9–10).

It is thus clear that languages that do not use an SV(O) construction, treat exponents of ‘somebody’ and/or ‘nobody’ like they treat focused constituents. However, this does not imply that this is not the case in languages making use of an SV(O) construction. For one, exponents of ‘somebody/nobody’ functioning as objects occur in the IAV focus position in these languages. Moreover, some (zones M and N) Eastern Bantu languages are known to allow pre-verbal focus (Kerr et al., 2023, pp. 10–12). Western Serengeti languages (zone JE), on the other hand, allow detopicalized constituents to occur in a preverbal position (Nicolle, 2015; Aunio et al., 2019; Bernander & Laine, 2020; Devos & Bernander, 2022). In other languages, the use of an SV(O) construction with ‘somebody/nobody’ functioning as the subject is probably due to a rather fixed word order. In Section 5.4, we will investigate whether fixed word orders correlate with heavier syntactic marking of the exponent.

In constructions with ‘nobody’, negation is predominantly marked on the verb. Another, exceptional, pattern involves the negation of the exponent with or without negative concord. We have suggested that some adnominal negative markers are derived from SAO or negative copula and that the change into adnominal negative markers is concomitant with a change from existential cleft into mono-clausal structure.

Before we look for correlations between construction types and exponent types and between construction types and features like negation and transitivity (Section 5), it should be noted that just as is the case with exponent strategies (Section 3.4), construction strategies can compete with each other. In transitive clauses, SVO and existential cleft constructions tend to compete, as in Luganda (104) and Bulu (105), for ‘somebody’ and ‘nobody’, respectively, whereas, in intransitive ones, the same two strategies can compete, as seen in Bulu, but optionality more frequently involves inversion and existentialization, as in Xhosa (107), or inversion and SV, as in Kanyok (108).

(104)

Luganda (JE15, Betty Ayugi pers. comm.)

a.	o-mu-ntu	ya-lab-a	mu-lirwaana	w-ange (SVO)
	aug-1-person	sm1.pst-saw-fv	1-neighbor	1-poss1sg
	‘Somebody saw my neighbor.’
b.	wa-li-wo	e-ya-lab-a	mu-lirwaana	w-ange (existential)
	sm16-cop-loc16	rel-sm1.pst-saw-fv	1-neighbor	1-poss1sg
	‘Somebody saw my neighbor.’

(105)

Bulu (A74, Ebanga, 2019, pp. 53, 19)

a.	môd	a-à-á-b-ya	ki	yənə	Zamba (SVO)
	1.person	sm1-neg-rec.pst-be-prf	neg	see	God
	‘Nobody has seen God.’

b.	tə-kə	môt	Ø-éziŋ	a-a-yiane	duma	ya	dz-am	ə-tə (exist.)
	cop-neg	1.person	1-one	sm1-prs-deserve	honor	of	4-affair	4-dem
	‘Nobody deserves the honor of this affair.’

(106)

Bulu (A74, Ebanga, 2019, p. 45)

a.	môt	Ø-éziŋ	a-s-kə		zôsô (SVO)
	1.person	1-one	sm1-cop.neg-neg		right
	‘Nobody is right.’
b.	tə-kə	môt	Ø-éziŋ	a-n	zôsô (existential)
	cop-neg	1.person	1-one	sm1-cop	right
	‘Nobody is right.’

(107)

Xhosa (S41, Carstens & Mletshe, 2016, p. 782)

a.	a-ku-kho	m-ntu	o-fik-ile-yo (existential)
	neg-sm17-cop	1-person	rel1-arrive-disj-rel
	‘Nobody arrived.’
b.	a-ku-fik-anga	m-ntu (locative inversion)
	neg-sm17-arrive-neg.pst	1-person
	‘Nobody arrived.’

(108)

Kanyok (L32, Michael Kasombo pers. comm.)

a.	ka-kw-àkà-dy-oh	bènd	muut	nànsh	wu-mwǐh (locative inversion)
	neg-sm17-pst-cop-neg	neg	1.person	sao	1-one
	‘Nobody was there.’

b.	muut	nànsh	wu-mwǐh	kà-àka-di-oh	bènd (SV)
	1.person	sao	1-one	neg.sm1-pst-cop-neg	neg
	‘Nobody was there.’

5. Correlations

Now that we know what the exponents can look like and in what type of constructions they occur, we can investigate how types of exponents and types of constructions correlate with each other, with negation (Section 5.1), and with transitivity (Section 5.2). In Section 5.3, we take a brief look at ‘nobody’ in one type of non-declarative clause. In Section 5.4, finally, we discuss the distribution of exponents and constructions in our dataset versus our sample.

Types of exponents and types of constructions are combined in two sets of parameters in Table 1, Table 2, Table 3, Table 4 and Table 5 (given below). The vertical and horizontal axes of each table represent, respectively, variations as to the type of exponent and type of construction. Table 1, Table 2 and Table 3 (this Section) concern the languages in our dataset. They plot languages with respect to the way they express ‘somebody’ and ‘nobody’ in three environments: functioning as the object and functioning as the subject of a transitive and intransitive verb. Bolded languages concern expressions of ‘nobody’.

Table 4 and Table 5 (in Section 5.4) plot the languages of our sample following the way they express, respectively, ‘somebody’ and ‘nobody’ functioning as the subject of a transitive clause. We use different typographic conventions for the different branches of the phylogenetic tree in Grollemund et al. (2015): North-Western languages are underlined, Central-Western languages are bolded, West-Western languages are in italics, South-Western languages are in italics as well as bolded, and Eastern languages are left unmarked. Grassfield languages are unmarked and preceded by ‘GR’.

Table 1. ‘Somebody/Nobody’ object in the dataset.

	canonical VO	existential cleft	object inversion
absent		JD62, JE15, K402, S42
		JE15, K402, P21, S33
generic noun for ‘person’	A22, A41, A601, A71, A801, A93, C25, C322, C44, D201, D332, D54, E55, E622C, E73, F22, G32, G33, G37, G62, G65, H10, JD42, JD62, JD63, JE15, JE16, JE251, JE405, JE43, JE45I, N11, N12x, N13, P312, P31E, P34, R21, S33, S42, S53, S54	E622C, JD61, JD62, JE15, K402, S42	H16c
	A22, A41, A601, A73a, A93, A74, B52, C25, C322, C44, D201, D332, D54, E54, E55, E73, F22, G32, G33, G37, G42, G62, JD42, JE15, JE16, JE405, JE43, JE431, JE45i, JE45n, N11, N12x, P312, P31E, P34, S33, S41, S42, S407, S53, S53(Tsonga), S54	B865, E622C, JE14, JD61, JD62, S33, S42
indefinite	G42, JE42, JE43, M23, M301, R11 S32, S407		H16c
	H16c, G65, JE42
numeral ‘one’/indefinite	B82, B865, C30B, D201, H10, L32, L53, L53A, M31, P21, S33
	B82, K12, R11
‘(an)other’/indefinite	G12, G66, K402, N31b, S42, S53(Tsonga)
	G66
scalar additive operator (sao)	A801, C101
sao & numeral ‘one’	B865, G66, H10, K333, L32, L53, L53A, M31, M23, M301
free choiceitem	E622C, G12, G42, G65, G66, JE16, JE22, P21	S33
‘(an)other’ & free choice item	N31b
ignorative		S33
minimizer	H16c
sao/neg & numeral ‘one’		K333, R11, R13
negative copula/neg (& sao)	B865	JE14
neg	C30B, H10

Table 1. ‘Somebody/Nobody’ object in the dataset.

	canonical VO	existential cleft	object inversion
absent		JD62, JE15, K402, S42
		JE15, K402, P21, S33
generic noun for ‘person’	A22, A41, A601, A71, A801, A93, C25, C322, C44, D201, D332, D54, E55, E622C, E73, F22, G32, G33, G37, G62, G65, H10, JD42, JD62, JD63, JE15, JE16, JE251, JE405, JE43, JE45I, N11, N12x, N13, P312, P31E, P34, R21, S33, S42, S53, S54	E622C, JD61, JD62, JE15, K402, S42	H16c
	A22, A41, A601, A73a, A93, A74, B52, C25, C322, C44, D201, D332, D54, E54, E55, E73, F22, G32, G33, G37, G42, G62, JD42, JE15, JE16, JE405, JE43, JE431, JE45i, JE45n, N11, N12x, P312, P31E, P34, S33, S41, S42, S407, S53, S53(Tsonga), S54	B865, E622C, JE14, JD61, JD62, S33, S42
indefinite	G42, JE42, JE43, M23, M301, R11 S32, S407		H16c
	H16c, G65, JE42
numeral ‘one’/indefinite	B82, B865, C30B, D201, H10, L32, L53, L53A, M31, P21, S33
	B82, K12, R11
‘(an)other’/indefinite	G12, G66, K402, N31b, S42, S53(Tsonga)
	G66
scalar additive operator (sao)	A801, C101
sao & numeral ‘one’	B865, G66, H10, K333, L32, L53, L53A, M31, M23, M301
free choiceitem	E622C, G12, G42, G65, G66, JE16, JE22, P21	S33
‘(an)other’ & free choice item	N31b
ignorative		S33
minimizer	H16c
sao/neg & numeral ‘one’		K333, R11, R13
negative copula/neg (& sao)	B865	JE14
neg	C30B, H10

Table 2. ‘Somebody/Nobody’ transitive subject in the dataset.

	canonical SVO	existential cleft	loc inv.	agr inv.
absent		G65, JE15, JE16, K402, P31E, P34, S42, S53, S54
		G33, G37, G66, JD62, JE14, JE15, JE16, K402, N12x, N31b, P21, P312, P31E, P34, R13, R21, S33, S42, S53, S53(Tsonga), S54
generic noun for ‘person’	A22, A41, A74, A801, B52, B85, C25, C322, C44, D332, E73, H10, JD42, JD62, JD63, JE14, JE15, JE16, JE405, JE43, JE45I, JE45N, K402, N11, N12x, N13, P21, P312	E622C, F22, G32, G33, G37, G42, G62, G65, H21, JD61, JD62, JE15, JE16, S33, S41, S407	M60	P312
	A71, A73a, A74, B52, B85, C25, C322, C44, D332, H10, JE43, JE45n	A74, E55, E622C, E73, F22, G12, G32, G33, G37, G62, G65, H21, JD42, JD61, JD62, JE14, JE15, JE16, JE405, JE43, JE431, JE45i, M60, N11, P31E, P34, R21, S32, S41, S407	S41
indefinite	A601, B865, G42, H16c, JE431, K11, M23, M301, R11	B865, R21
	H16c
numeral ‘one’/indefinite	B82, C30B, C44, D201, H10, L32, L53, L53A, M31, S33
	A601, A74, C322, C44
‘(an)other’/indefinite	G12, G66, JE42, S32, S53(Tsonga)
	JE42
sao	H16c
sao & numeral ‘one’ (& focus)	H10, B865, D201, E54, K12, L32, L53, L53A, M31	JD63, JE14
sao & minimizer	H16c
free choice item		G42, S33
ignorative		S33
sao/neg	A22, A41
sao/neg & numeral ‘one’	K11, R11, R13	K333, G65, M23, M301
negative copula/neg (& sao & numeral ‘one’)	B865	B865, JE14, JE43
neg	C30B, H10

Table 2. ‘Somebody/Nobody’ transitive subject in the dataset.

	canonical SVO	existential cleft	loc inv.	agr inv.
absent		G65, JE15, JE16, K402, P31E, P34, S42, S53, S54
		G33, G37, G66, JD62, JE14, JE15, JE16, K402, N12x, N31b, P21, P312, P31E, P34, R13, R21, S33, S42, S53, S53(Tsonga), S54
generic noun for ‘person’	A22, A41, A74, A801, B52, B85, C25, C322, C44, D332, E73, H10, JD42, JD62, JD63, JE14, JE15, JE16, JE405, JE43, JE45I, JE45N, K402, N11, N12x, N13, P21, P312	E622C, F22, G32, G33, G37, G42, G62, G65, H21, JD61, JD62, JE15, JE16, S33, S41, S407	M60	P312
	A71, A73a, A74, B52, B85, C25, C322, C44, D332, H10, JE43, JE45n	A74, E55, E622C, E73, F22, G12, G32, G33, G37, G62, G65, H21, JD42, JD61, JD62, JE14, JE15, JE16, JE405, JE43, JE431, JE45i, M60, N11, P31E, P34, R21, S32, S41, S407	S41
indefinite	A601, B865, G42, H16c, JE431, K11, M23, M301, R11	B865, R21
	H16c
numeral ‘one’/indefinite	B82, C30B, C44, D201, H10, L32, L53, L53A, M31, S33
	A601, A74, C322, C44
‘(an)other’/indefinite	G12, G66, JE42, S32, S53(Tsonga)
	JE42
sao	H16c
sao & numeral ‘one’ (& focus)	H10, B865, D201, E54, K12, L32, L53, L53A, M31	JD63, JE14
sao & minimizer	H16c
free choice item		G42, S33
ignorative		S33
sao/neg	A22, A41
sao/neg & numeral ‘one’	K11, R11, R13	K333, G65, M23, M301
negative copula/neg (& sao & numeral ‘one’)	B865	B865, JE14, JE43
neg	C30B, H10

Table 3. ‘Somebody/Nobody’ intransitive subject in the dataset.

	canonical SV	existential cleft	loc inv	agr inv
absent		G65, JE15, JE16, K402, S42
		E73, JD62, JE14, JE15, JE16, K402, N31b, P21, P312, S33, S53(Tsonga)
generic noun for ‘person’	A22, A41, B85, C25, C322, C44, D332, E73, H10, JD42, JD62, JD63, JE15, JE16, JE405, JE43, JE45i, JE45n, N11, N12x, N13, P34, S41	E622C, F22, G33, G37, G42, G62, G65, JD61, JD62, JE15, JE16, K402, S42, S33, S407	A22, D54, G32, G33, JD42, JD62, JE11, JE12, JE14, JE16, K12, K14, M60, R21, S32, S41, S53, S54	A41, G32, G65, G66, JE251, K333, N12x, N13, N31b, P312, P31E, P34
	A601, A71, A801, B52, B85, C25, C322, C44, D201, D332, G66, H10, JE401, JE43	E622C, F22, G37, G42, G62, G65, H21, JD61, JD62, JD63, JE14, JE15, JE16, JE405, JE43, JE431, JE45, P34, S32, S33, S41, S407	A41, D54, E55, G32, G33, G37, JE11, JE12, K14, M60, N31b, R13, S41, S42, S53, S54	A41, E55, JE14, N11, N12x, P31E
indefinite	A601, A93, G42, H16c, JE401, JE42, JE431, K11, M23, M301, R11		R13	R21
	H16c
numeral ‘one’/indefinite	B82, C30B, C44, D201, G66, H10, K11, L32, L53, L53A, M31, P21, R11, S33	B82
	A74, B865, C322, C44, D201	A74, B82
‘(an)other’/indefinite	G12, N31b, S53(Tsonga)
	JE42
sao	A93, C101, H16c		C322	P31E
sao & numeral ‘one’	B865, H10, D201, E54, L32, L53, L53A		C322, L32, L53, L53A
sao & minimizer	H16c
sao & (an)other	G66			G66
free choice item		G12, G42, S33
ignorative		S33
sao/neg	A22, A41
sao/neg & numeral ‘one’	K11, K12, R11, R13	K333, M23, M301
negative copula/neg (& sao & numeral ‘one’)	B865, C101, JE11, JE14, JE43	B865, JE14, JE43
neg (&numeral ‘one’)	H10, C30B

Table 3. ‘Somebody/Nobody’ intransitive subject in the dataset.

	canonical SV	existential cleft	loc inv	agr inv
absent		G65, JE15, JE16, K402, S42
		E73, JD62, JE14, JE15, JE16, K402, N31b, P21, P312, S33, S53(Tsonga)
generic noun for ‘person’	A22, A41, B85, C25, C322, C44, D332, E73, H10, JD42, JD62, JD63, JE15, JE16, JE405, JE43, JE45i, JE45n, N11, N12x, N13, P34, S41	E622C, F22, G33, G37, G42, G62, G65, JD61, JD62, JE15, JE16, K402, S42, S33, S407	A22, D54, G32, G33, JD42, JD62, JE11, JE12, JE14, JE16, K12, K14, M60, R21, S32, S41, S53, S54	A41, G32, G65, G66, JE251, K333, N12x, N13, N31b, P312, P31E, P34
	A601, A71, A801, B52, B85, C25, C322, C44, D201, D332, G66, H10, JE401, JE43	E622C, F22, G37, G42, G62, G65, H21, JD61, JD62, JD63, JE14, JE15, JE16, JE405, JE43, JE431, JE45, P34, S32, S33, S41, S407	A41, D54, E55, G32, G33, G37, JE11, JE12, K14, M60, N31b, R13, S41, S42, S53, S54	A41, E55, JE14, N11, N12x, P31E
indefinite	A601, A93, G42, H16c, JE401, JE42, JE431, K11, M23, M301, R11		R13	R21
	H16c
numeral ‘one’/indefinite	B82, C30B, C44, D201, G66, H10, K11, L32, L53, L53A, M31, P21, R11, S33	B82
	A74, B865, C322, C44, D201	A74, B82
‘(an)other’/indefinite	G12, N31b, S53(Tsonga)
	JE42
sao	A93, C101, H16c		C322	P31E
sao & numeral ‘one’	B865, H10, D201, E54, L32, L53, L53A		C322, L32, L53, L53A
sao & minimizer	H16c
sao & (an)other	G66			G66
free choice item		G12, G42, S33
ignorative		S33
sao/neg	A22, A41
sao/neg & numeral ‘one’	K11, K12, R11, R13	K333, M23, M301
negative copula/neg (& sao & numeral ‘one’)	B865, C101, JE11, JE14, JE43	B865, JE14, JE43
neg (&numeral ‘one’)	H10, C30B

5.1. ‘Nobody’ Versus ‘Somebody’

As shown in Section 3, negation entails more variation as to types of exponents with some syntactic markers being specific to expressions of ‘nobody’. We remarked that adnominal modifiers specific to the expression of ‘nobody’ tend to be pragmatically stronger. Another observation is that they typically involve explicit quantifiers like the numeral ‘one’ or the (distributive) universal qualifiers ‘all’ and ‘every’. We noted that whereas the use of the numeral ‘one’ for expressions of ‘somebody’ is typically triggered by its grammaticalization into an indefinite marker, this is not the case for expressions of ‘nobody’, which rather appeal to the (original) cardinal semantics.

This does not imply that syntactic marking as such occurs more frequently with exponents of ‘nobody’ than with exponents of ‘somebody’. Syntactic marking for expressions of ‘somebody’ occurs in 26, 24, and 28 languages (Table 1, Table 2, and Table 3, respectively), whereas syntactic marking for expressions of ‘nobody’ occurs in 27, 27, and 27 languages, respectively. Syntactic marking is also well-spread throughout the Bantu domain for the expression of both ‘somebody’ and ‘nobody’. In each case, it occurs with languages from all five branches of the phylogenetic classification.

However, the likelihood of a language having competing exponent strategies is bigger with ‘nobody’ than with ‘somebody’. Examples of languages having a choice between two to four exponent strategies occur for ‘nobody’ in all three Tables (i.e., G65 Kinga, G66 Vwanji, H16c Yombe, and S33 Sesotho in Table 1; C44 Boa, D201 Liko, H16c Yombe, and S33 Sesotho in Table 2; B865 Nzadi, C322 Zamba, H16c Yombe, and S33 Sesotho in Table 3), whereas only one language (i.e., K11 Chokwe), in Table 5, shows a choice between two different exponent strategies for ‘somebody’.

Another difference is that syntactic marking is less restricted to the SV(O) construction type with ‘nobody’ than with ‘somebody’. Some exceptions notwithstanding, all languages with exponents of ‘somebody’ including syntactic marking co-occur with the SV(O) pattern, whereas exponents of ‘nobody’ including syntactic marking, although still showing a slight preference for the SV(O) pattern, show more diversification construction-wise.

Expressions of ‘somebody’ have the same type of constructions as expressions of ‘nobody’. There is no marked difference in the choice of construction type between ‘somebody’ and ‘nobody’ functioning as the object (Table 1). However, when ‘somebody or ‘nobody’ functions as the subject, there is a clear correlation between expressions of ‘nobody’ and a preference for the existential cleft construction type. In transitive and intransitive clauses, the existential cleft construction type respectively occurs in almost and more than twice as many languages for the expression of ‘nobody’ (43 languages in Table 2 and 35 languages in Table 3) as for the expression of ‘somebody’ (24 languages in Table 2 and 14 languages in Table 3). The data allow the implication that if a Bantu language can use an existential cleft construction for the subject ‘somebody’, it can also use it for the subject ‘nobody’. This is in line with what is found for South-East Asian and Mixtec languages, and we can appeal to the same explanation, viz., a version of the traditional ‘Given before New’ principle (van der Auwera, 2023; van der Auwera et al., 2023): starting to say something about someone who does not even exist is a radical breach of the principle, and it is useful to have the negative existential flag this explicitly. An additional possible explanation lies in the semantics and functional correlates of existential clefts in Bantu languages. As was mentioned in Section 4.6, existential clefts have received little attention in Bantu languages. In line with the general literature on there-clefts (e.g., Lambrecht, 2001), they are assumed to be presentational clefts whose main function is to introduce new entities into a discourse. However, Davidse et al. (2022), in a detailed and comprehensive account of different types of there-clefts in English, show that there-clefts can be either presentational or specificational. Specificational there-clefts specify a value (the NP in the matrix clause) for a variable (the relative clause). The most common type of specificational there-cleft is a quantifying one that quantifies instances of the value. Importantly, negative quantification is a frequent subtype of quantifying there-clefts because “the semantic option that there are no instances qualifying for the variable is an important one”. (Davidse et al., 2022, p. 78). Davidse et al. (2022, p. 75) give the following example as a typical profile of specificational there-clefts: ‘There was not one of them there could justify the disparity between the way Britain’s being treated and the treatment of the members of the Six’.

Expressions of ‘nobody’ thus seem to be especially suited for the matrix of a there-cleft. Expressions of ‘somebody’, on the other hand, typically involve indefinite markers that, in combination with an existential cleft, tend to trigger a presentational reading (Davidse et al., 2022, p. 79), which is more than ‘somebody’ caters for. In sum, if Bantu existential clefts are like English there-clefts, then they are especially suited for expressions of ‘nobody’ but less so for expressions of ‘somebody’. Specific characteristics of existential clefts, although in need of further research in Bantu languages, together with the ‘Given before New’ principle, help to explain the pattern of selected construction strategies in, for example, Digo. In the subject position, an SVO construction is used for ‘somebody’ (109)a, whereas ‘nobody’ uses an existential cleft (109)b. In the object position, where the Given naturally occurs before the New, both ‘somebody’ and ‘nobody’ make use of an SVO construction (109)c,d.

(109)

Digo (E73, Steve Nicolle pers. comm.)

a.	mu-tu	wa-ni-on-a
	1-person	sm1.pst-om1sg-see-fv
	‘Somebody saw me.’
b.	ta-pha-na	mu-tu	a-ri-ye-ni-on-a
	neg-loc16-com	1-person	sm1-pst-rm1-om1sg-see-fv
	‘Nobody saw me.’
c.	ná-on-a	mu-tu
	sm1sg.pst-see-fv	1-person
	‘I saw somebody.’
d.	si-on-ere	mu-tu
	neg.sm1sg-see-neg.pst	1-person
	‘I saw nobody.’

Another finding is that in many languages (34 languages in Table 3) the locational intransitive clause with ‘nobody’ (i.e., ‘Nobody was there.’) does not simply use an existential clause but rather an existential cleft. Kagulu, for example, has an existential clause involving locative inversion and the negative verb ichaka ‘lack, be without’, as seen in (110)a. However, it uses an existential cleft for ‘nobody’, as seen in (110)b.

(110)

Kagulu (G12, Petzell, 2008, p. 167, Malin Petzell & Sauli Lengoliga pers. comm.)

a.	kw-ichak-a	wa-nhu
	sm17-lack-fv	2-person
	‘There are no people.’ (lit. ‘It (the place) lacks people.’)
b.	hechak-a	mu-nhu	wowose	yow-ile	bahaya
	sm16.lack-fv	1-person	fci1	sm1.rel.be-pfv	here
	‘Nobody is here.’ (lit. ‘Here lacks any person who is here.’)

As argued in Devos and Bernander (2022, p. 597), following Creissels (2019, p. 10), existential clauses (as in (110)a mainly flag a difference in perspectivization (from ground to figure, rather than from figure to ground) and are only indirectly related to information structure. The use of the negative existential cleft with ‘nobody’ emphasizes that there is not one person that qualifies for the variable (i.e., being here).

5.2. Syntactic Roles and Transitivity

The attested types of exponents do not vary much depending on whether ‘somebody’/‘nobody’ is used as a subject or as an object or depending on whether the clause is transitive or intransitive. The specific position and whether it is pre-verbal or post-verbal might be more important. It was observed that exponents of ‘nobody’ including a free choice item only occur post-verbally, i.e., in the dedicated IAV focus position. Interestingly, the apparent ban on pre-verbal ‘nobody’ when including a free choice item is lifted when the main verb is in the subjunctive. This is further discussed in Section 5.3.

Construction types do show a strong correlation with syntactic roles and transitivity. When ‘somebody’/‘nobody’ functions as the object, VO is by far the most common construction type (cf. Table 1). When functioning as the subject, however, existential cleft constructions become more important, especially for the expression of ‘nobody’ (cf. Section 5.1). Intransitive clauses allow for still more diversification as they also make more frequent use of inversion constructions (cf. Table 3). The higher incidence of inversion constructions in intransitive clauses is in line with the implicational hierarchy of predicate types in inversion constructions given in Marten and van der Wal’s (2014) paper on the typology of subject inversion in Bantu languages. Following this hierarchy, inversion will not occur with transitive verbs if it does not also occur with copula and intransitive verbs.

Let us now also take a closer look at the two types of intransitive clauses our questionnaire catered for, i.e., locational intransitives (‘somebody/nobody was there’) and unaccusative intransitives with the verb ‘arrive’ (‘somebody/nobody arrived’). Although we do not always have data for both types and we merge the types in Table 3, we do observe that locative inversion is more easily allowed with locational intransitives, a tendency that also adheres to Marten and van der Wal’s (2014) implicational hierarchy in which copulas come before unaccusatives. For some languages, we see locative inversion in the equivalent of ‘somebody/nobody was there’ but agreeing inversion or the non-inverted SV type in the equivalent of ‘somebody/nobody arrived’. Kwere (111) illustrates the former alternation and Kanincin the latter. In Kanincin, locative existentials use either locative inversion or SV, whereas the unaccusative intransitive only uses SV (112).

(111)

Kwere (G32, Aron Zahran, pers. comm.)

a.	ha-kal-a	ha-bule	mu-nhu (loc. inv.)
	sm16-be.pst-fv	sm16-neg.exist	1-person
	‘There was nobody there.’
b.	chi-kal-a	chi-gonsel-a	bali	ha-fik-ile	mu-nhu (agr. inv.)
	sm1pl-be.pst-fv	sm1pl-wait-fv	but	neg.sm1-arrive-pfv	1-person
	‘We waited but nobody came.’

(112)

Kanincin (L53A, Michaël Kasombo, pers. comm.)

a.	kwà-ding	pènd	mu-ntw	aâp	wù-mwîng (loc. inv.)
	sm17.pst-cop.pst	neg	1-person	sao	1-one
	‘Nobody was there.’

b.	mu-ntw	aâp	wù-mwîng	kàà-shik-a	pend (SV)
	1.person	sao	1-one	neg.sm1.pst-arrive-fv	neg
	‘Nobody arrived.’

5.3. ‘Nobody’ and the Subjunctive

This Section takes a preliminary look at the expression of ‘nobody’ in marked clause types, and more specifically at ‘nobody’ in combination with a verb in the negative subjunctive. The Bantu subjunctive is used for prohibition and (deontic) modality in main clauses (Nurse & Devos, 2019). Although our questionnaire only asked for ‘somebody/nobody’ in main clauses, we came across some data suggesting that languages using the existential cleft construction with ‘nobody’ revert to an SV(O) construction when the verb is in the negative subjunctive. A search for translation equivalents of ‘nobody’ in the Swahili New Testament shows the omnipresence of the existential construction (113)a,b, except when the verb is in the subjunctive, in which case there is a clear preference for an SV(O) (113)c–e word order. The exponent of ‘nobody’ may be the bare generic noun (113)e or the generic noun with the free choice marker (113)c,d. This is also exceptional as we noted that expressions of ‘nobody’ including a free choice item show a strong preference for the post-verbal slot (Section 5.2).

(113)

Swahili (G42, New testament in Swahili (Zanzibar), 1921)

a.	ha-ku-na	a-m-ju-a-ye	Mw-ana	ila
	neg-sm17-com	sm1-om1-know-fv-rel1	1-child	except
	Baba
	9.father
	‘Nobody knows the Son, except for the Father.’

b.	kwa_sababu	ha-ku-na	m-tu	a-li-ye-tu-ajiri
	because	neg-sm17-com	1-person	sm1-pst-rel1-om1pl-hire
	‘Because nobody hired us.’

c.	basi,	sasa	m-tu	yeyote	a-si-ni-sumbu-e	tena
	so	now	1-person	fci1	sm1-neg-om1sg-harass-sbjv	again
	‘Furthermore, let no one be harassing me.’

d.	tangu	leo	hata	mi-lele	m-tu	yeyote
	from	9.today	until	4-forever	1-person	fci1
	a-si-l-e		ma-tunda		kw-ako
	sm1-neg-eat-sbjv		6-fruit		17-poss2sg
	‘No one may eat fruit from you; no longer, forever.’

e.	basi,	m-tu	a-si-ji-vun-i-e	wa-tu
	so	1-person	sm1-neg-refl-boast-appl-sbjv	2-person
	‘So, let no one boast in men.’

For now, our data mainly come from Swahili and closely related languages like Kwere and Zalamo, both of which use the free choice marker and an SV(O) word order in combination with a verb in the negative subjunctive, as seen in (114), whereas they use existential cleft constructions or inversion constructions in (unmarked) main clauses. Further research will indicate whether this is a widespread feature in Bantu and whether the subjunctive triggers changes in default focus positions.

(114)

Kwere (G32, Zahran, 2023, p. 76)

yoyose	seke-ya-swel-e	hishima
fci1	proh-sm1-lack-sbjv	9.respect
‘Nobody should lack respect.’

5.4. The Distribution of Types of Exponents and Types of Constructions

In this Section, we look at the distribution types of exponents and types of constructions for ‘somebody/nobody’ functioning as the subject of a transitive clause. In our dataset, the number of languages using the bare generic noun as the exponent of ‘somebody/nobody’ almost doubles the number of languages involving syntactic marking (cf. 42 versus 24 languages for ‘somebody’ and 41 versus 27 languages for ‘nobody’, Table 2). The high number of bare generic nouns is tuned down in our sample (Table 4 and Table 5), where the bare generic noun and syntactic marking are evenly distributed. Bare generic noun exponents occur in all five branches. They are not attested in Grassfields for ‘somebody’ but do occur for ‘nobody’. Syntactic marking occurs in all five branches and in Grassfields for ‘somebody’ and ‘nobody’ alike.

In our dataset, ‘somebody’ and ‘nobody’ differ greatly as to construction type, at least as the subject of a transitive clause (cf. Section 5.1). For ‘somebody’, the number of languages taking an SVO construction is double the number of languages using an existential cleft construction (cf. 50 versus 24 languages in Table 2). For ‘nobody’, the scales are reversed, with fewer languages opting for an SVO than for an existential cleft construction (cf. 32 versus 43 languages in Table 2). In our sample as well, existential cleft constructions are better represented with ‘nobody’ than with ‘somebody’. However, there is no reversal of scales as SVO constructions still outnumber existential cleft constructions with both ‘somebody’ and ‘nobody’ (cf. 22 versus 3 languages in Table 4 and 18 versus 7 languages in Table 5). The higher number of existential cleft constructions in our dataset is a consequence of the bias toward Eastern Bantu languages. In our sample (Table 4 and Table 5), SVO constructions are found in all five branches and in Grassfields for both ‘somebody’ and ‘nobody’. Existential cleft constructions are less well distributed. They occur in Eastern Bantu, West-Western, and South-Western Bantu for ‘somebody’ and additionally in North-Western Bantu for ‘nobody’. Grassfields and Central-Western languages in our sample do not have existential cleft constructions. When returning to our dataset, we see that existential cleft constructions are entirely absent from North-Western Bantu and West-Western Bantu, except for the North-Western Bantu language Bulu. In this language, an existential cleft construction can be used as an alternative to an SVO construction for expressions of ‘nobody’, as seen in (105), repeated here as (115).

(115)

Bulu (A74, Ebanga, 2019, pp. 53, 19)

a.	môd	a-à-á-b-ya	ki	yənə	Zamba (SVO)
	1.person	sm1-neg-rec.pst-be-prf	neg	see	God
	‘Nobody has seen God.’

b.	tə-kə	môt	Ø-éziŋ	a-a-yiane	duma	ya	dz-am	ə-tə (exist.)
	cop-neg	1.person	1-one	sm1-prs-deserve	honor	of	4-affair	4-dem
	‘Nobody deserves the honor of this affair.’

The near-absence of existential cleft constructions in North-Western and Central-Western Bantu languages does not come as a surprise seeing that these languages typically do not have dedicated existential constructions (Devos & Bernander, 2022). Rather, they show a complete identity between plain locational and existential constructions. These languages are characterized by a rather fixed word order, which might have been triggered by the loss or reduction of the locative agreement system, which plays an important role in inversion constructions. Be that as it may, languages without existential constructions do not have an existential cleft construction as a potential strategy for expressions of ‘nobody’.

A follow-up question is whether languages with more fixed word orders rely more heavily on syntactic marking of the exponent for the expression of ‘somebody/nobody’. Our dataset does not show evidence of such a compensation. Bare generic nouns are just as frequent as more complex expressions of ‘somebody/nobody’ in North-Western and Central-Western Bantu languages.

Table 4. ‘Somebody’ transitive subject in the sample.

	canonical SVO	existential
absent		K402, JE16
generic noun for ‘person’	A22, A41, A74, C25, C44, D332, B52, K402, JE16, N11	F22, JE16
indefinite	GR_Bafut, GR_Awing, GR_Babungo, GR_Bamum, A601, B865, H16c, K11, M23	B865
numeral ‘one’/indefinite	C44, D201, B82, L32, L53

Table 4. ‘Somebody’ transitive subject in the sample.

	canonical SVO	existential
absent		K402, JE16
generic noun for ‘person’	A22, A41, A74, C25, C44, D332, B52, K402, JE16, N11	F22, JE16
indefinite	GR_Bafut, GR_Awing, GR_Babungo, GR_Bamum, A601, B865, H16c, K11, M23	B865
numeral ‘one’/indefinite	C44, D201, B82, L32, L53

Table 5. ‘Nobody’ transitive subject in the sample.

	canonical SVO	existential
absent		K402, JE16
generic noun for ‘person’	GR_Awing, A74, C25, C44, D332, B52	A74, F22, JE16, N11
indefinite	GR_Bafut, GR_Awing, H16c
numeral ‘one’/indefinite	A601, A74, C44
sao	GR_Babungo, H16c
sao & numeral ‘one’ (& focus)	D201, B865, H16cL32, L53
sao/neg	A22, A41
sao/neg & numeral ‘one’	K11	M23
negative copula/neg (& sao & numeral ‘one’)	B865	B865
neg	GR_Awing, GR_Bamum

Table 5. ‘Nobody’ transitive subject in the sample.

	canonical SVO	existential
absent		K402, JE16
generic noun for ‘person’	GR_Awing, A74, C25, C44, D332, B52	A74, F22, JE16, N11
indefinite	GR_Bafut, GR_Awing, H16c
numeral ‘one’/indefinite	A601, A74, C44
sao	GR_Babungo, H16c
sao & numeral ‘one’ (& focus)	D201, B865, H16cL32, L53
sao/neg	A22, A41
sao/neg & numeral ‘one’	K11	M23
negative copula/neg (& sao & numeral ‘one’)	B865	B865
neg	GR_Awing, GR_Bamum

6. Conclusions

This study started out from the observation in the typological literature, as well as in the Bantu literature, that Bantu does not have dedicated expressions for ‘nobody’. Bantu languages were believed to simply use a generic noun in combination with clausal negation. Still, some preliminary investigations showed that ‘nobody’, when functioning as the subject, favors an existential cleft construction. This was interesting for two reasons: (1) research on East Asian and Mixtec languages showed a similar preference and (2) recent research into existentials in Bantu languages has given us a detailed picture of Bantu existential constructions. Hence, despite the apparent unexceptional nature of Bantu indefinites, we embarked upon a study of negative indefinites in comparison with specific indefinites, in a dataset of 85 Bantu languages.

As for the exponents of ‘nobody’ and ‘somebody’, we found that although they share some expressions, the bare generic noun for person being the most obvious one, we observed differences and could point to interesting details. The expression of ‘nobody’ typically involves pragmatically heavier exponents containing, for example, scalar additive operators or free choice items. Concerning the latter, we confirmed a doubt put forward by Haspelmath (1997) about the unidirectional nature of the semantic change of ‘any’ into ‘every’. Several cases of ‘any’ clearly derived from ‘every’ or ‘all’ are attested in Bantu languages. Regarding the semantics of the exponents of ‘nobody’, it was observed that they often involve explicit quantifiers. Interestingly, in this respect, the use of the numeral ‘one’ for expressions of ‘somebody’ tends to be triggered by its change into an indefinite marker, whereas expressions of ‘nobody’ typically appeal to the (original) cardinal semantics, often in combinations like ‘not even one’. We also observed that ‘nobody’ can be expressed by an augmentless generic noun in the pre-verbal position, which is a very unusual slot for augmentless nouns throughout the Bantu domain. Future studies on Bantu augments in specific languages could thus benefit from a systematic investigation of ‘nobody’. Last but not least, dedicated negative indefinites have been shown to occur in Bantu languages. They involve a univerbation of the scalar additive operator na and the numeral ‘one’, and display varying negation patterns. Some have inherent negativity, in which case they are closely related to existential cleft constructions. Non-strict and strict negative concords also occur.

Turning to our findings on construction types, we could confirm the initial hypothesis that existential cleft constructions are more frequent with exponents of ‘nobody’ than with exponents of ‘somebody’. Just like South-East Asian and Mixtec languages, Bantu languages allow the implication that if existentialization is possible with specific indefinites, then it is also possible with negative indefinites. Existential cleft constructions are most likely to occur when ‘nobody’ is the subject of a transitive clause, less likely in an intransitive clause, and least likely when ‘nobody’ is the object. In intransitive clauses, inversion competes with existentialization. The fact that inversion is more likely to occur in intransitive clauses is in line with the implication that inversion only occurs with transitive verbs if it also occurs with intransitive verbs. However, the existential cleft construction is not just an alternative to inversion in transitive clauses. It also frequently occurs with locational clauses instead of a simple existential clause. One would think that the locational clause, ‘Nobody is there’, could simply be rendered by an existential clause (‘There is not a person there’). Although this is indeed sufficient in some languages, many languages use an existential cleft construction (‘There is not a person who is there’). So, existential cleft constructions are not only used to flag a radical breach of the Given before the New Principle, but their use for ‘nobody’ also confirms the importance of recognizing quantifying there-clefts (Davidse et al., 2022). This type of specificational there-cleft is especially suited for negative quantification, and specifically to emphasize that there is no one qualifying for the variable expressed through the relative clause. There-clefts with indefinite markers tend to have a presentational interpretation, which is not a necessary meaning of ‘someone’. They are thus less well suited for the expression of ‘somebody’.

Although we have not yet performed a detailed examination of the geographical distribution of types of exponents and types of constructions, our study confirms the (near-)absence of existential constructions in North-Western and West-Western Bantu languages, possibly due to more fixed word orders in these languages. Exponents of ‘somebody/nobody’ occur in their canonical pre-verbal S and post-verbal O slots, and contrary to expectations, these languages do not seem to compensate for the lack of marked construction types by a heavier marking of the exponent. We also found interesting variations within languages that are known to have an IBV focus position. In Yombe, ‘somebody’ functioning as an object occurs immediately before the verb but ‘nobody’ occurs after the verb. In Mbochi, on the other hand, the reverse pattern is attested.

Finally, this study has shown that an interesting field of further research, next to obtaining complete and reliable data for all the languages of the dataset (and more), is to see how ‘nobody’ behaves in non-declarative clauses. When the verb is in the subjunctive in Swahili and closely related languages, ‘nobody’ functioning as the subject occurs in its canonical pre-verbal slot, and it can be accompanied by a free choice item, which in our dataset, pertaining to declarative clauses, was systematically avoided.