1.1 A brief introduction to Baule

Baule [‘bauleɛ] is a language of the Niger-Congo Kwa family spoken in Ivory-Coast, mostly in the central and southern parts of the country. According to WorldAtlas (2017), the language counts approximately 1.2 million speakers.¹ The language also has several dialectal variants, of which the Kode [kodɛ] dialect will be the object of our discussion. Kode is spoken in the “Northwestern corner of the Baule region” (Timyan 1977:15), that is, the area of Béoumi. Kode shares with Central Baule most of its features. It nevertheless exhibits variation at the level of the lexicon, but also at the syntactic level. Here, we will be using the term Baule whenever the variants do not exhibit any distinction (that is, the data is compatible with both Central Baule and Kode), and we will specifically use the term Kode when the data belongs exclusively to the Baule-Kode variant. In this section, we give a brief description of the main properties of the language.

Baule is a tonal language, with 5 contrasting tones and nasal-oral vowel contrast (see a.o. Timyan 1977; Ahoua 1996; Konan 2019). Typically, the language will exhibit the following contrasting pairs (where the diacritic ʹ indicates high tone, ˉ middle tone, ˋ low tone, ͝ indicates rising tone and ͡falling tone):²

Baule is described as an SVO language (see Timyan 1977, Konan 2019):

The verb bears no agreement or tense marking. In addition to tonal mode marking,³ the verb can also bear the aspectual prefixes a- (for perfective) and su- (imperfective or progressive) and a “factual past” suffix -li, which expresses “an action, process or state which has actually occurred and is a fact of the world.” (Timyan 1977: 229). The suffix -li has an allomorph, realized as the reduplicated final vowel of the verb if it is not sentence final. The default, non-marked aspect yields a habitual, general statement:

The contrast between perfective and factual past is illustrated in (4):

There is no number agreement on the verb either:

Nouns bear no singular/plural marking:

Baule has six personal pronouns, (1,2,3 sg, 1,2,3 pl). They can function as full pronominal subjects (7a, b). They can also co-occur, as a resumptive pronoun, along with the full nominal subject.⁴ Clearly, the presence of the 3^rd pl pronoun me in Kode (be in Central Baule) signals plurality (7c). However, the absence of me/be is not a mark of singular, but only indicates that number is unspecified (7d).

Nouns may be marked for specificity with the marker nĩ, as in (8) below, a property which will be discussed in more detail in this paper:

Adjectives and numerals follow the noun they modify:

The order within the noun phrase, as given in Timyan (1977), is the following:

Note that this order corresponds to one of the few orders observed in Greenberg’s universal 20 (Greenberg 1963) and viewed as one of the most common orders, as discussed at length in Cinque (2005).

The data provided and discussed in this paper has three main sources. First, as mentioned explicitly, the authors rely on the rich and detailed data provided in Timyan (1977). The rest of the data comes in part from corpora recorded, transcribed, annotated by Yannick Konan in various villages in the Béoumi area and available in Konan (2019), and in part from Yannick Konan, who is a native speaker of the Kode variant of Baule, as well as from individual Kode-speaking informants.

1.2 The problem

Baule has two adnominal quantifiers which participate in universal quantification, ngba (glossed as ‘all’) and klwaa (‘every’):

Both ngba and klwaa are post-nominal. As shown by the glosses, they both quantify over the noun but they have different interpretations. While ngba yields the interpretation of ‘the whole of Monday’ or ‘the entire Monday’, klwaa splits the set into its individual atoms. The examples in (12) and (13) further illustrate this distinction:

In (12), ngba follows the noun ba (‘child’). The quantifier bears on the nominal it immediately follows, and the interpretation is that all the recipients (as a group) receive food without any exception.

Similarly, (13) shows that the quantifier klwaa (‘every’) follows the noun ba (‘child’). The interpretation is that every child, taken individually, will receive food from Aya, maybe in different food-giving events. Again, the quantifier bears only on the nominal preceding it. Moreover, the interpretation shows that klwaa operates as an ‘individuator’.

Baule also has various elements expressing existential quantification. The element jwe (‘some’), when it occurs after a nominal expression (noun or pronoun), is interpreted as an indefinite, and may be glossed as ‘some (undefined, unspecific)’:

We observe that jwe follows the noun jaswa (‘man’) and the 3^rd person plural pronoun me (‘they’) respectively in (14a) and (14b).

The problem arises from the observation that in Baule, quantifiers can combine in an unexpected way. First, the universal quantifiers klwaa and ngba may occur together to form some complex quantification. The nominal element precedes the sequence of quantifiers:

Second, the existential jwe can combine with nĩ, a “specifier” or mũ, an “individuator” following Timyan’s (1977) terminology. Again, the sequence follows the nominal element it quantifies over (note that Kode uses a reduced form of the specifier nĩ, transcribed here as ’n):

The combination of various quantifiers produces complex quantification, and raises the following question, which we address in this paper: given the properties of individual quantifiers in Baule and their unexpected combinatorial possibilities, how do we account for their distribution and their contribution to the interpretation of quantified expressions?

In this paper, we propose an approach to quantified nominal expressions which builds on a structure-interpretation mapping. Therefore, any attempt at answering this question inevitably raises another one: if the interpretation of quantifiers is linked to their structural constraints, what is the syntax of the Baule DP?

Section 2 is a brief overview of both the theoretical framework we adopt and the existing literature on Niger-Congo languages in terms of quantification. Section 3 offers a discussion of the structure of the Baule DP, a domain which is, to our knowledge, yet unexplored. With the structure of the DP in mind, we then turn, in Section 4, to universally quantified DPs, and their various combinatorial possibilities. Section 5 looks at existentially quantified DPs. Finally, Section 6 offers a brief summary of our findings and introduces a number of yet unexplored quantifiers in Baule.

2.1 Theoretical framework

2.2 Previous studies on quantification in Niger-Congo languages

While there is a rich literature on the semantics of Niger-Congo languages (see a.o. Anagbogu 1995; Hartmann & Zimmermann 2006; 2007; Korsah 2016; Renans forthcoming), and a growing bulk of studies on their syntactic properties, led by the pioneering work of Aboh (1998; 2004 and sbq), studies on quantification begin to emerge and contribute to our understanding of quantification cross-linguistically. Among the various families, the Bantu languages seem to have drawn the most attention. As shown in Zerbian & Krifka (2008), Bantu languages have few quantifiers, many of the meanings being built up of complex morphosyntactic structures. This is typically the case for the universal quantifier equivalent to every, as opposed to all. However, the huge variety found among the Bantu languages as well as the gaps in documentation necessitate further detailed work on various aspects of quantification (Zerbian & Krifka, 2008). Similarly, Meeussen (1967) for example shows that numerals are not coherent morpho-syntactic classes in Bantu languages. Synchronic data show traces of a former quinary counting system. In addition, diachronic evidence comes from the fact that only the first five numbers can be historically reconstructed for Proto-Bantu.

Other language families have also started to be investigated. Zimmermann (2008) gives a detailed account of quantification in Hausa, a Chadic language, showing that the language has both adnominal and adverbial quantification. Crucially, Zimmermann observes that Hausa has two universal quantifiers, koo+wh, whose meaning corresponds to English ‘every/each’, and duk(a), ‘all’. They contrast in terms of syntactic distribution: koo+wh occurs before the NP, duk(a) before or after; they show difference in agreement; koo+wh only occurs with indefinite NPs, duk(a) is also fine with definite NPs; duk(a) can enter into a partitive. Zimmermann thus proposes that duk(a) is a modifying element operating on definite DPs (it universally quantifies over a contextually-given set denoted by the DP). Koo+wh on the other hand is a functional head in D combining with bare count NPs. The two quantifiers also behave differently with respect to the distributive/collective dimension: koo+wh is distributive (it cannot occur with inherently collective predicates), it is incompatible with mass nouns and binds singular pronouns. Duk(a) can occur with collective predicates and may bind plural pronoun. We will see that Zimmermann’s observations parallel those provided for Baule.

Several studies also examine the nature of quantifiers and the semantic effects of their interactions. Tamba, Torrence & Zimmermann (2012) give a detailed description of quantification in Wolof (Atlantic Niger-Congo), a language which exhibits variations in the realization of existential and universal quantifiers and in their syntactic distribution.⁷ The existential quantifier bí , and its interactions nó in Akan (Kwa, Niger-Congo) was examined by Amfo (2010),⁸ Dua et al (2020) and Owusu (2020). Renans (2017; 2018) delved into the quantifier system of Ga, a Kwa language, focussing on some exclusive particles, such as too or pɛ (‘also’), as well as on the distribution and interpretation of indefinites.

Finally, existing literature on Baule is only emerging. Although there are a few studies on Baule (see a.o. Carteron 1972; 1975; Creissels & Kouadio 1977; Ahoua 1996; Kouamé & Kouadio 2004), Timyan (1977) is still the study that comes closest to being a comprehensive reference grammar. Timyan is thus one of the first researchers to address the question of numerals and quantifiers in this language. In addition to a list of numerals, she mentions the following quantifiers: ngba (‘all’), kwlaa (kwlakwla) (‘every’), kungba (‘only’), ngumi (‘only’), showing that they systematically follow the noun they quantify over. However, she does not discuss the possibility of combined quantifiers.

2.3 Hypothesis

Assuming a strict syntax-semantics mapping within the syntactic structure of DPs, we hypothesize that Baule complex quantified expressions are structurally constrained and accordingly yield interpretations whose meaning can be computed off the structure and whose scope is predictable, as deriving from other properties of the quantifiers involved. In order to verify our hypothesis, we will consider in turn universally quantified and existential quantified nominal expressions. We will show that both syntactically and semantically, the quantified DPs are constrained and subject to DP-internal structural conditions.

4.1 The semantics of universal quantifiers

Having set the DP scene, we now move on to the particularities of quantifiers. Examples in (11), repeated here, show that Baule has two quantifiers interpreted as ‘universal’.

That they both have universal quantificational force can be shown by the following examples:

While the existential jwe (to be discussed in section 4) can occur in an existential construction, neither ngba nor klwaa are licensed. However, the two quantifiers contribute differently to the interpretation of the sentence, as illustrated by the following contrast:

While nzue ngba (‘all water’) in (34a) is natural, (34b) with nzue klwaa is semantically odd. Indeed, klwaa yields an improbable interpretation in which Aya drank every (quantity) of water. The difference between the two is that while ngba takes the set as a whole, klwaa distributes on the set, identifying its individual items. This is not plausible with a mass noun like ‘water’. Note that a possible reading of (34b) could be coerced if Aya drank every kind of water, suggesting that the distributive interpretation can extend to kinds, if individuals are not available.²⁴

The pairs in (35), (36) again illustrate the contrast:²⁵

The pair in (35) builds on a predicate which has a collective reading only. We observe that a DP with the quantifier ngba is grammatical, whereas the presence of the quantifier klwaa is illicit. In (36), the predicate points to an individual reading, as shaking hands is (preferably) an activity involving individuals in a one-to-one relation. The quantifier klwaa is fine (36a), while (36b), without being ruled out, is judged as odd. This suggests that some contrast along the collective-distributive reading is at play. Consider also the following:

Example (37a) is appropriate in a situation where Abo has several fields (tomatoes, yam, maize) and has farmed every (different) field. (37b), in contrast, describes a situation where Abo has one field which he has farmed (for example tomatoes), but completely.

Another relevant diagnostic comes from binding. Recall that only pronouns are marked for number in Baule. In possessive constructions, the possessor may be expressed by a pronoun only:

We see that klwaa enters singular anaphoric relations, and ngba a plural one.

Finally, combining the quantifier with the classifier ba gives different interpretations:

While ngba will provide a set, which is the sum of the individual grains of rice, the presence of klwaa results in an interpretation where the distribution operates over categories.

Putting all the data together, we observe that klwaa is odd with mass nouns, is disallowed with collective predicates, distributes over individuals or categories and requires a singular bindee. In contrast, ngba is compatible with mass nouns but also with count nouns, is comfortable with collective predicates. It can both function as a plural-marked binder (if its restrictor is a count noun interpreted as plural) or a singular-marked binder (if it quantifies over a singular count noun, with an interpretation of ‘whole’).

Building on previous literature (see Brisson 1998; Matthewson 2013) and using some of Zimmermann (2008)’s diagnostics, we would like to propose that klwaa is a bona fide quantified determiner, (i.e. a function of type < <e,t>, <e,t> t>), which combines with an NP to form a Generalized Quantifier (a set of sets < <e,t>, t>). It also bears all the marks of an inherently distributive quantifier. As for ngba, we propose, along the lines of what is discussed in Brisson (1998) and Morzycki (2012), that it is not an intrinsic quantifier, but a modifier of nominal expressions. However, we will rather exploit part of Renans (2017)’s approach, and consider that ngba is a group forming operator à la Landman (2010).²⁶ Renans suggests that these modifiers are of type <<e,t>, <e,t>>. Our analysis diverges from Renans’ one, though, because ngba does not contain an exclusive feature. Its contribution is, along the lines of Brisson (1998), to “maximize” the set denoted by the NP. We propose that it is the denotation of this modified ‘maximized NP’ which combines with the specificity feature to yield the referential DP (or SpecifP) which we assume is of type e.

Although ngba, as expected from our analysis, seems to favour a default collective reading (see 35, 36), a distributive-like behaviour is not excluded. As attested by example (40) below, the distributive reading may sometimes emerge rather naturally:

Here, the quantified DP bla ngba occurs as the possessor in the possessive DP bla ngba dũmã (‘(the) name of all women’) (see footnote 12 for a description of possessive constructions). An exclusively collective reading is absurd, as it is expected that each woman has her own name (even if it is the same one).²⁷ We conclude that ngba, very much like English all, is underspecified for distributivity (see also Beghelli & Stowell 1997, who propose that English all is “pseudo-distributive”). An elegant solution to this behaviour is proposed in Brisson (1998), with the optional introduction of a D(istributive) operator. When the operator is not present, the reading is a default collective one. The distributive operator enables to distribute over properties (predicates), a case which is compatible with the natural interpretation for (40) above, where a name is associated with every individual woman.

4.2 The syntax of universal quantifiers

We now turn to the syntactic implementation of the proposal above and show how the structure of quantified DPs is in adequation with their interpretation. Recall that quantifiers canonically follow the head noun in Baule, and klwaa and ngba linearly follow the NP they interact with within the quantified phrase. This is a property the language shares with Bantu languages (Fourie, 1992), as well as with Wolof (Tamba et al. 2012). But given the semantic differences between klwaa and ngba discussed in the previous section, we expect their syntactic build up to be different, despite surface similarities.

In order to identify their respective specificities, we consider in more details the distribution of elements within the DP. While Baule does not have a definite determiner, it has a demonstrative, nga, which occurs after the noun and, according to the orders given in Timyan (1977), follows the adjective and the numeral. As in the previous examples, the subject DP can be doubled with an optional resumptive pronoun, suggesting topicalization:²⁸

The demonstrative appears to be incompatible with klwaa and ngba, whether it follows the quantifier or precedes it:²⁹

Numerals, which also follow the noun, are compatible with ngba but not with klwaa:

(44a) refers to four children, taken as a group. As for (44b), the possible interpretation which would distribute over groups of 4 children is not available at all.

Finally, combining the quantifiers with the classifier ba and the numeral confirms the orders and the restrictions discussed above:

Crucially, ngba is compatible with the classifier and the numeral, and gives rise to an interpretation where the set composed of three grains of rice is considered. The order is N-class-Num-ngba. Klwaa is ruled out in the same configuration: while it is compatible with ba (see 39b above), it cannot co-occur with the numeral.

The above facts suggest that the two quantifiers appear in different positions in the structure. Let us start with klwaa.

On the surface, the universal quantifier klwaa (‘every’) occurs to the right of the noun. According to the head-initial analysis proposed in section 3, we expect the NP to raise to a position which c-commands that of the quantifier. Recall that klwaa cannot co-occur with numerals. Either its merging position is in competition with that of numerals, or something blocks the occurrence of numerals in the presence of the quantifier. Given what was discussed above on the properties of the quantifier, we propose that klwaa heads a functional projection QP dominating UnitP:

By a roll up movement, the nP raises to SortP, which in turn raises to the specifier of QP to check the individuating feature, transiting through specUnitP:

Note that both a successive roll-up movement to specQP and a standard XP movement from specUnitP to spec QP are compatible with the data. What is crucial is that the movement to specUnitP accounts for the fact that numerals are not licensed. The quantifier thus forms with the nP a quantified nominal expression of type <<e,t,>,t>.

Turning now to ngba, we have observed that it also follows the noun, suggesting a raising analysis similar to that of klwaa. However, it is compatible with numerals. Although we defended the idea that ngba is semantically a modifier, we also adopt recent views on the syntax of modification (see e.g. Kayne 1994, Cinque 1999) which dispense with the notion of ‘adjunction’ and argue in favour of specialized syntactic positions in the form of functional projections for modifiers. In this view, we expect ngba to occur in a dedicated functional projection, let us call it GroupP (for group forming). We propose the following structure, in which the GroupP projection crucially occurs below the Specif projection.

We need to enrich our structure to provide a landing site for the derivation to produce the relevant order: first, np raises to specSortP, giving N ba. SortP must then raise to a position dominating UnitP whose specifier hosts numerals. As seen above, numerals are compatible with noun phrases modified by ngba and occur linearly after noun+classifier. Finally, ZP raises to spec GroupP, yielding the order N-(ba)-(Num)- ngba:

We remain agnostic as to the nature of this ZP projection. It may correspond to the Foc projection identified by e.g. Svenonius (2007). However, the focal properties of the ‘maximized nominal’ constituent would need to be examined in more detail and are left for future research. Crucially, while klwaa creates a generalized quantificational structure, ngba is an NP modifier of type <<e,t>,<e,t>>. Combining with an NP, it produces an output of type <e,t>. GroupP therefore needs to combine with a Determiner (-like) element (<e,t>, e>) to form a referential e type. We propose that the GroupP further raises to specSpecifP to check its specificity/referentiality feature.

4.3 Combining the quantifiers

As was illustrated in (5) above, Baule can combine the two quantifiers within one nominal expression:

The interpretation of (50) is that ‘I go to the bush absolutely every day, without exception’. As the gloss and the interpretation show, the two quantifiers contribute different meaning components to the DP. Clearly, the use of a second quantifier contributes to the reinforcing, or in Brisson’s terms, the maximization of the (already) universally quantified expression, hence the ‘without exception’ interpretation. The examples below show how the meanings build up in different contexts:

In (51a), the target is on the individuals who possess the chickens. Similarly to example (44a) above, the quantifier klwaa scopes over the pronoun me, resulting in the sequence me klwaa ‘every pro (=‘of every(one of) them’). This possessive DP (headed by the pronoun) is then modified by ngba, yielding the sequence me klwaa ngba. These examples appear to support the idea that the semantic contribution of ngba is precisely that of a “maximizer”, in that is operates on the set of individuals without exceptions. Note that here, ngba does not introduce distributivity, which is already inherently encoded in klwaa, but does not block the distributive reading either. This seems to confirm that it is indeed neutral with respect to distributivity/collectivity. The complex quantified phrase me klwaa ngba is merged as the possessor of akɔ (‘chicken’). (51b) is easily computable along the same line.

Baule also allows the two quantifiers to occur in the reverse order:

Given our discussion on the semantics of these quantifiers, we expect this reverse pattern to produce a significant difference in meaning. As shown by the gloss, this is indeed the case. In (52), ngba takes srã in its scope, yielding the quantified expression all humans (as an unspecified group). This quantified phrase is then further modified by klwaa, which divides the set into its atomic subparts. The presence of kɔ̃ngɔ̃ (‘one-by-one’), a distributed share, confirms that the whole nominal expression is interpreted distributively. In order to show the difference in the meaning which, we contend, is semantic (i.e. truth-conditional), we will consider the following situations:

Consider first (53a):

(53a) is true if Abo has farmed every field of his domain, irrespective of the produce or the size of the field. In other words, he needs to have farmed every field belonging to the set of his fields. It is false if Abo failed to farm one field of his domain. The set of fields has to be exhausted.

Compare with (53b):

(53b) is true if Abo farmed completely every piece of land, namely 100% of the tomato field, 100% of the potato field and 100% of the mango field. It is false if Abo failed to farm his fields completely. What emerges from the contrasts above is that ngba indeed functions as a maximizer, in that it creates readings which ban any exception to the situation. The difference in meaning confirms the different semantic contribution of the two quantifiers. This, we contend, is reflected in their syntax.

4.4 Refining the syntax of Baule universal quantifiers

The combination of two universal quantifiers within the nominal constituent uncovers a more complex structure of quantified DPs than what was previously assumed. One possibility would be to generate the quantifiers in the order in which they appear, Although the distributional pattern seems to suggest that the two quantifiers can occur in any order, the difference in interpretation argues against a pure optionality approach. Given our theoretical approach which rests on the idea that each structure yields a given interpretation (or one interpretation can only be accounted for by one structure), we expect quantifiers to occupy fixed positions in the structure.

In order to account for these, we need to elaborate on the DP structure proposed in section 3 above. Namely, while we have defined a projection (and hence a structural position) for each of the quantifiers under scrutiny, we have, up to now, not investigated their potential relative hierarchy.

Let us first consider the nominal phrase in (53a) above:

We propose the following derivation:

The steps of the derivation are as follows. First, the nP fje raises to SortP, a movement which results in the individuation of the entities denoted by the noun. Then, a snowballing movement raises SortP to XP, and XP to GroupP, giving the order fje ngba. Finally, GroupP moves to spec QP, resulting in the quantified nominal expression fje ngba klwaa. The syntactic derivation proposed here is in perfect adequation with the semantic properties of the quantified expression: recall that ngba, as a modifier, was argued to be of type <<e,t>, <e,t>>. Combining with the nP, it produces an output of type <e,t>. The latter is a perfect input candidate for the universal quantifier klwaa (<e,t>, <<e,t>,t>). Thus, semantic composition and syntactic derivation seem to work hand in hand.

The reverse order, exemplified in (53b), raises interesting challenges. It is possible to derive it, on the basis of the structure proposed for (53a), by building on Cinque’s (2005) proposal that languages resort to two kinds of DP-internal movement: the roll-up movement discussed above, and the cyclic XP movement. A cyclic XP movement along these lines can raise SortP to specGroupP and to spec QP giving the order fje ngba klwaa. However, a major objection to this derivation for (53b) is that it predicts that there is no interpretive difference with (53a). Both end up being QPs headed by klwaa. This seems to go against the data. An alternative proposal, which we offer here rather tentatively, would be to take the role of ngba as a modifier seriously, and consider that it may occur in different positions in the structure. The main drawback of such a proposal is that it does not seem to be able to prevent overgeneration: how do we block such “free” modifiers from surfacing in any position in the DP? On the advantage side, though, the flexible nature of the modifier offers a rather straightforward account of the contrast in interpretation discussed in section 4.3. The derivation we propose is as follows:

The nominal phrase nP raises to specSortP to verify its count feature. By roll-up movement, SortP raises to specUnitP, which raises to specQP, Finally the QP raises to specGroupP, giving the order fje klwaa ngba. Given that ngba functions here as a modifier not of the nominal component, but of the quantified phrase, its semantic type as defined above raises a problem. One possible solution, which was suggested to us by a reviewer and which turns out to be the most viable one, is to consider ngba as a versatile modifier which can accommodate both <e,t> and <<e,t>, t> types.