Dan (South Mande < Southeastern Mande < Mande < Niger-Congo) is a macrolanguage spoken in Côte d’Ivoire, Liberia and Guinea by about 1,600,000 people (2012, my evaluation).2 The total number of Dan dialects is approximately 35, see Figure 1. In Liberia, Dan is known under the name of Gio; in Côte d’Ivoire Dan is most often referred to as Yakuba (Yacouba). Since the 1960s, two distinct written norms have been elaborated in Côte d’Ivoire: Western Dan, based on the Blo dialect, and Eastern Dan, based on the Gwɛɛtaa dialect (sousprefecture of Santa, prefecture of Biancouma, Tonkpi province). There is also a northern variety of Kla-Dan which is different enough from Eastern Dan to be regarded as a separate language.

In the introduction to (Baker & Gondo 2020) it is said that the Dan dialect under study is that of Man (the main administrative center of the Dan country in Côte d’Ivoire): “see Gondo (2016) for the first extended study of the Man dialect, which we investigate here”.3 In fact, in Gondo’s PhD dissertation it is the dialect of Gble or Gblewo (in Dan, Gblʌ̀ʌ̀ or Gblʌ̀ʌ̀wȍ; wȍ means ‘language, dialect’) that is concerned, Bleu Gildas Gondo being a native speaker of this variety. Evidently, Gble data have also been used for the 2020 article. This dialect belongs to Eastern Dan, and is spoken to the west and northwest of Man, at some distance from the city. It is true that all of the Eastern Dan dialects are more or less mutually intelligible, but it is not accurate to refer to Gble as “the dialect of Man”: the dialects whose areas are immediately adjacent to Man are Sleng, Soon and Kaa (Slʌ̋ŋ̏, Sɒ̰̋ɒ̰̏, Kȁȁ), see Figure 1.

It should be mentioned that most of the data for Eastern Dan published so far stem from the Dan-Gwɛɛtaa (Gwɛ̋ɛ̏tàà) dialect, and most of my unpublished data are from this dialect too. Eastern Dan has a written tradition and a codified orthography (some publications are available in the Dan Electronic Library, http://cormand.huma-num.fr/danbiblio/index.php). In what follows, I will operate with the data of the Gwɛɛtaa dialect (unless otherwise specified). Differences between the Gwɛɛtaa and Gble dialects have no impact on the points discussed in this paper; nevertheless, such differences will be mentioned and discussed whenever they may be relevant.

Eastern Dan has 12 to 15 oral and 9 nasal vowels,4 and a vowel ŋ with limited distribution. A remarkable feature of the consonantal system is the presence of two implosive phonemes, ɓ and ɗ, which are realized as [m] and [n] respectively before nasal vowels or after ŋ. In the same way, the sonorants y, w are realized in nasal contexts as [ɲ] and [w̃] respectively. This means that the nasal consonants m, n, ɲ, w̃ have no phonological status in Dan; they will not be represented in my transcription (although m, n appear in examples taken from (Baker & Gondo 2020)).

Eastern Dan has five level tones (extra-high: a̋, high: á, mid: ā, low: à, extra-low: ȁ; the extra-low floating tone is designated by an apostrophe) and three modulated tones (all falling), only one of which is relatively frequent, high-falling: â. Tones bear a great functional load, fulfilling both lexical and grammatical functions (Bearth & Zemp 1967; Vydrin 2016).5

The basic word order in a simple verbal sentence is (S) Aux (DO) V (X), where S is the subject, Aux stands for auxiliaries indexed for the person and number of the subject and expressing TAM and polarity meanings (in the Mandeist tradition these auxiliaries are usually referred to as “predicative markers”), DO is the direct object, V is the verbal predicate, and X is an indirect/oblique object or adjunct. Dan is a null-subject language (the explicit presence of a subject NP is not obligatory, the subject being indexed in the Aux). The presence of a DO makes a verb transitive; if the DO position is void, the verb is intransitive. X can be represented by a postpositional phrase (NP + postposition), an adverb or an NP headed by a declinable noun (see below).

Non-verbal sentences of different types are formed with the copulas ɓɯ̰̏, ɓā, ɗɛ̰̀, ɗɤ́.6 Existential, conjoint and negative imperfective series of auxiliaries (see Table 1) can also serve as copulas in non-verbal sentences and can therefore be regarded as “bifunctional auxiliaries” (Vydrin 2020a: 79).

A noun can be the head of a noun phrase. It can appear in the positions of subject, direct object, or oblique (introduced by a postposition, i.e., as the dependent component of a postpositional phrase); it can also be the head of an attributive nominal construction.

In the noun phrase, a dependent noun precedes the head noun, while an adjective follows the head noun. A determiner follows the NP. Thus, the basic NP structure is [[NP] N [Adj] Det].

A dependent noun can be connected to the head noun directly (ɗɯ̋ zɔ̋ɗi̋ɤ̋ <tree top> ‘top of a tree’) or by means of a postposition (ɗūɤ̄ɤ̄ ɓȁ gbīŋ̄gā <raffia.palm on caterpillar> ‘raffia palm caterpillar’).7 If the possessor is human, it is connected to a relational noun without a postposition (Gbȁtȍ dʌ̄ <Gbato father> ‘Gbato’s father’), and to a non-relational noun by means of the postposition ɓȁ (Gbȁtȍ ɓȁ kʌ̋ʌ̋ <Gbato on hoe> ‘Gbato’s hoe’). In other words, the postposition ɓȁ serves as a marker of dependency between nouns.8 When a possessive construction with a non-relational head appears in post-verbal position, the connecting postposition ɓȁ is most often replaced by another postposition, gɔ̏.

There is a subclass of nouns, so-called declinable nouns. In adjunct function they appear in their oblique case forms without postpositions.9 Six morphological cases can be identified, with a great deal of irregularity in their formation. For more detail, see (Vydrin 2011).10

Adjectives have semi-regular derivative models for plural and intensive meaning based on reduplication, modification of tones, suffixation and transfixation. They can appear in three syntactic functions: a noun modifier in an attributive nominal construction (2), the predicate in a postpositionless construction with a bifunctional auxiliary in copular function (3), the predicate in a postpositional construction (the postposition is ká ‘with’), and with a bifunctional auxiliary in copular function (4).

For a detailed presentation of Eastern Dan adjectives, see (Vydrine 2007; Vydrin 2009; Vydrin 2010a).

Verbs in finite verbal constructions are necessarily preceded by auxiliaries (predicative markers, see Table 1). The 3SG existential, conjoint and subjunctive predicative markers can optionally be omitted, and the same is true of the 2SG imperative predicative marker (I prefer to interpret such cases as involving the use of zero exponents rather than a lack of markers). Predicative markers may merge with adjacent pronouns, conjunctions and determiners.

Apart from the predicative markers, the TAM and polarity meanings are expressed by verbal inflection which may be tonal or segmental. In the neutral aspect construction (Vydrin 2010b; Vydrin 2020b), verbal lexical tones are replaced by the extra-low grammatical tone (5b), cf. the perfect construction, where the verb ɗṵ̄ ‘come’ carries its lexical tone (5a).

This tonal modification is a formal feature distinguishing verbs from adjectives: the adjectives can also appear with the existential series of bifunctional auxiliaries, but keep their lexical tones (3). This formal difference between verbs and adjectives was first described by Piper (1983) for Liberian Dan.

In the conjoint construction (an analogue of the neutral aspect construction used in dependent clauses and after a focalised NP), tonal lowering on the verb takes place; this modification follows a complex pattern which depends on the lexical tone and the segmental structure of the verbal stem, see in detail (Vydrine & Kességbeu 2008: 48–50; Vydrin 2016; Vydrin 2017: 536; Vydrin 2021a: 41–42).

The following means of nominalization are found in Dan-Gwɛɛtaa (see also 3.3, 6.1):

The first three suffixes (-sɯ̏, -ɗɛ̏, -ɗē)11 can also appear in phrasal nominalization constructions (in this case, they are added to a post-verbal adverbial element or PP). In such constructions, the phrasal nominalization is also marked by the grammatical extra-low tone on the verb. On phrasal nominalization see in more detail sections 3.4 and 7.

Nouns, verbs and adjectives are distinguished in Dan by their distribution. There is a limited amount of conversion between these classes, but this is an idiosyncratic phenomenon, rather than a general property. The authors specially mention a complication: nouns and adjectives are used as predicates preceded by the “particle” ȅ, in the same way as verbs (6/BG14a, b, c). Ȅ can optionally be omitted in all these cases (7/BG15a, b, c).

According to the authors, “the difference between verbs and nouns/adjectives in Dan is concealed in this particular environment because of the confluence of two features of the language, neither of which is remarkable in itself: T remains as a separate word in Dan, and Pred is lexically null (also T can be elided in faster/informal speech)” (Baker & Gondo 2020: 9).

The authors say the following about the cross-linguistic differentiation between the main lexical classes (p. 14):

The authors suppose that the fact “that simple verbs combine directly with their core argument, whereas adjectives and nouns can only do so indirectly, with the help of a functional head that may or may not be overt” may explain why nominalized verbs combine directly with their core argument, “whereas nominalized adjectives and nouns can only do so indirectly, with the help of an overt functional head ɓa”.

Another difference between the main parts of speech is mentioned on p. 9:

In Dan, relational and non-relational nouns are formally distinguished in possessive constructions: the latter are connected to their possessors by means of the connector ɓa, e.g. (9/BG2b), and the former require no connector (the two nouns are simply juxtaposed), e.g. (8/BG1a). In both cases, the possessor precedes the head noun.

The syntactic structures of these constructions are interpreted by Baker & Gondo as follows: the inalienable possessor is the complement of a noun (10/BG8a), while in a possessive construction with an alienable head noun, the “possessor particle” ɓa is the head, the possessum NP is its complement, and the possessor is the specifier (11/BG10).

Verbs are productively nominalized by the morpheme –sɯ. According to Baker and Gondo, the event- and fact-denoting deverbal nouns formed with –sɯ follow the mode of relational nouns and are connected to their possessors without ɓa. A transitive verb can be nominalized retaining both of its core arguments: in this case the theme (i.e., the internal argument) is connected with the verbal noun without ɓa, and the agent (the external argument), with ɓa, see (12/BG33a).

Adjectives are regularly nominalized by means of the morpheme -ɗɛ to form names of states, e.g. (BG16a) zɔ̄ɔ̄zɔ̏ɔ̏-ɗɛ ‘foolishness’. Deadjectival nouns follow the model of non-relational nouns, in that they are connected to possessors by means of the connector ɓa, see (BG5a-d).

The authors explain this difference in terms of Baker’s theory of lexical categories, according to which theme arguments are direct arguments of verbs but only indirect arguments of adjectives.

-ɗɛ is also compatible with nominals: it derives abstract nouns (e.g., nʌ́ ‘child’ → nʌ́-ɗɛ́ ‘childhood’, gbɛ̰́ ‘dog’ → gbɛ̰́-ɗɛ́ ‘dog’s behavior’). Baker and Gondo refer to this derivation as “nominalization of nouns” and consider it to be similar to deadjectival nominalization with -ɗɛ, inasmuch as these forms also require the possessive connector ɓa when used as possessum.

Finally, -ɗɛ can derive nouns from verbs, but in this case, according to Baker and Gondo, its role is different: first, the derived form has the meaning of the location of the event (Zȍta̋ gā-ɗɛ̄ ‘place where Zota died’, ɓa̋a̋ kpà-ɗɛ̀ ‘the place where rice is cooked’); second, the theme argument of the derived nominal is not marked with the “possessive particle” ɓa.

Further on, we read (Baker & Gondo 2020: 13): “The affix -sɯ also has uses other than its use in deriving event/fact nouns from verb [sic]. It productively derives adjectives from a root of any category.” This claim is confirmed by the following examples (see 6.1 for my interpretation of the semantics and functions of –sɯ):

According to Baker & Gondo (2020: 16), cross-linguistically, a nominalizing morpheme can combine directly with a lexical head, before the lexical head combines with any of the arguments (or adjunct modifiers) that it would normally appear with. This is a low nominalization. Otherwise, the nominalizing morpheme combines with a verb or adjective after the verb or adjective has already combined with at least some of its arguments or modifiers: this is a high nominalization, or phrasal nominalization.

The low nominalization marker for verbs is –sɯ, and that for adjectives is -ɗɛ. When a deverbal noun forms a possessive construction with the theme of the original verb, it behaves as a relational (inalienable) noun (14/BG3c). A deadjectival noun behaves as a non-relational (alienable) noun and connects to the possessor by means of the possessive marker ɓa (15/BG5d).

Baker & Gondo (2020: 18) explain this difference on the grounds that a verb takes an internal argument, while an adjective takes an external argument; this difference is projected on the syntax of the nominalized forms.

“Nominalization of nouns” (i.e., derivation of abstract nouns with the suffix -ɗɛ) follows the same pattern as the nominalization of adjectives: “they too have no argument that their nominalization can inherit. As a result, they too behave like nonrelational nouns”.

High nominalization, meanwhile, is possible only on verbs. Baker & Gondo assume that deverbal nominalizations as in (14/BG3c) can be interpreted as reflecting either low nominalization, (Műsȍ) – (gā-sɯ̄), or high nominalization, (Műsȍ gā) – (sɯ̄). Besides, there is a nominalization model where the nominalizer –sɯ comes after a post-verbal constituent (an adverb or a postpositional phrase), as in (16/BG37a, f). This model can be only interpreted as high nominalization.

The authors claim that “there is a second version of the nominalizations in (37)12 in which -sɯ does appear affixed directly to the verb root, and the PP, AP or adverb follows -sɯ. The wrinkle is that these postverbal elements are necessarily connected to the nominalized verb by the linking particle ɤ́” (Baker & Gondo 2020: 23), cf. (17/BG39a, d).

The authors claim that in the latter constructions we have low nominalization, and suggest at the same time that “ɤ is probably not a predicative particle”. They represent the syntactic structure of (17/BG39a) as follows.13

Baker & Gondo further note (p. 26) that “the location-denoting nominalizer -ɗɛ (homophonous with the deadjectival nominalizer <…>”, as well as an agentive nominalizer –mɛ can also be markers of high nominalization, in the same way as –sɯ.

At the same time, they show that in those cases where adjectives (in their predicative function) can have adjuncts, the deadjectival nominalizer -ɗɛ can be suffixed only to the adjective, and not the adjunct. This fact confirms the impossibility of the high nominalization for adjectives.

Let us pass now to a critical analysis of the theses of the paper (Baker & Gondo 2020).

According to Baker and Gondo (2020: 11–12), deadjectival nous are derived by means of the suffix -ɗɛ (which can also derive denominal abstract nouns), while deverbal nouns are derived by the suffix –sɯ. They admit that deverbal nouns can also be derived by -ɗɛ, but claim that “the meaning of the derived form is quite different: it does not refer to a state/event/fact/eventuality, but rather to the location at which an event or events of the type named by the verb root takes place”.

In my view, there are three homonymous (or quasi-homonymous) suffixes -ɗɛ and three homonymous suffixes –sɯ in Eastern Dan. In Dan Gwɛɛtaa they all carry the extra-low tone (-ɗɛ̏, -sɯ̏); in the Gble dialect they are toneless, and the tone of the preceding syllable spreads on to the suffix.

The three suffixes -ɗɛ are the following:

It is likely that all three suffixes go back to one and the same noun ɗɛ̏ ‘place’, but in synchrony, it is appropriate to consider the common case suffix as separate one, while the status of the other two (the quality/state and action nominal suffixes) is more debatable: they can be regarded as representing a single transcategorial suffix or two different suffixes (or even three, if we decide that suffixes added to adjectival and nominal stems are distinct).

My first objection to what is said in Baker & Gondo’s article is that although the verbal noun (masdar) derived with -ɗɛ may have the meaning of “location of action” in some contexts (29), elsewhere (depending on the context) it can simply name an action or process (30).

To my mind, examples (29) and (30) represent different stages of grammaticalization of the noun ɗɛ̏ ‘place’. There are also intermediate cases, where both interpretations can be accepted, as in (31).

Therefore, the semantic difference between the verbal nouns derived with –sɯ and those derived with -ɗɛ, at least in some contexts, cannot be explained in terms of semantic opposition between event/fact nominals and location nominals.

And the three suffixes –sɯ are as follows:

Dan Gwɛɛtaa

If we return to the examples cited by Baker and Gondo, we can establish that in nʌ́ dĩ-sɯ ‘hungry child’, -sɯ is the denominal adjectivizer; in gbɔ̄ pɤ̀-sɯ̀ ‘fallen pot’ –sɯ is the gerund suffix (pɤ̀ is a verb ‘fall’); and in mɛ̄ pűű-sɯ̋ ‘white man’ -sɯ is the selectivity suffix (pűű is the adjective ‘white’).

In my view, there is no doubt that all three suffixes -sɯ are different, i.e. we have here homonymy of suffixes, rather than polysemy and/or transcategoriality.

The central topic of Baker & Gondo’s paper is that the deverbal event and fact nominals formed with –sɯ follow the model of relational nouns and are connected to their possessors without ɓa, while the deadjectival nouns of state derived by means of the morpheme -ɗɛ follow the model of non-relational nouns, i.e. they are connected to possessors by means of the connector ɓa.

However, the rule requiring that “the theme argument of an adjectival nominalization <…> must be marked by the possessive particle ɓa, like the possessors of the nonrelational nouns” is not as strict as is claimed by Baker and Gondo. In my data, I find numerous cases of adjectival nominalization where the theme is not marked with ɓȁ, thus partly disproving the paper’s central claim. Here are some examples.

I also find numerous examples in my data where ɓȁ is present, as in (37, 38).

It is rather improbable that the Gble dialect is different in this respect from what I observe in the Gwɛɛtaa dialect. Even if it seems premature to formulate a strict rule regulating the presence or absence of ɓȁ in possessive constructions with deadjectival nouns, some trends can be identified. The first such trend is that a human possessor requires ɓȁ, as in (20),18 with the exception of two nouns derived from quantifiers, gbɛ̋ɗɛ̏ ‘majority’ (from gbɛ́ ‘many, much’) and kpíȉɗɛ̏ ‘overwhelming majority’ (from kpíȉ ‘big’), as in (39).

Another factor may be the pragmatic status of the possessor: it seems that a high pragmatic status of the possessor (as in (37), where the possessor is relativized) may also favour the presence of ɓȁ.

To sum up: contrary to what Baker and Gondo write, possessive constructions with deadjectival nouns may or may not feature the possessive marker ɓȁ, depending on the semantics of the possessor, and probably also its pragmatic status and some other factors.

Let us now consider possessive constructions with deverbal nouns in –sɯ. According to Baker and Gondo:

As for prototypical unergative verbs, the authors claim (p. 21) that such verbs seem to be lacking in Dan, because the meanings in question are expressed by transitive constructions with light verbs: ‘sing’ is ‘song+pick/harvest’, ‘dance’ is ‘dance+do’, ‘sleep’ is ‘sleep+kill’,19 ‘swim’ is ‘water+do’, ‘walk’ is ‘walk+do’ and so on, and the less canonical unergative predicates represented by simple verbs (pè ‘vomit’, wlɤ̀ ‘fly’), when nominalized, are not marked with the possessive ɓa.

Although Baker & Gondo are correct on this point in general, there are some nuances to be taken into account. It is true that “prototypical unergative verbs” are rare in Dan, but still, among the approximately 250 morphologically simple verbs found in the Eastern Dan dictionary (Vydrin 2021b) there are some intransitive verbs which can have agentive subjects. I find examples in my corpus of texts where gerunds of such verbs are connected with their possessors (corresponding to verbal subjects) by means of ɓȁ.

The verb ɗā is labile: it can be intransitive (‘save oneself’) or transitive (‘save someone’). If the intransitive meaning is concerned in (41), ɗā is an unergative verb (the saved person is an active actor), and if the transitive meaning is intended, then the saved person is the theme, which is in overt contradiction with Baker & Gondo’s claims.

The gerund of the verb ɗṵ̄ ‘come’ also appears in my data in possessive constructions without ɓȁ, just as in Baker & Gondo’s paper, cf. (BG3a). It turns out that, at least for some speakers (or in some dialects) of Eastern Dan, ɓȁ is acceptable in such context.

On the other hand, converse examples can be found, although rarely: my language assistants accept, in some cases, a gerund of a transitive verb connected with both arguments without ɓȁ, as in (42), although a variant with a possessive connector, wɯ̄-ɗṵ̏ ɓȁ glɛ̏ŋ̏ wɯ́-sɯ̏…, is also accepted.

Without pretending to have conclusive analysis, I would surmise that here again, the presence or absence of ɓȁ may be conditioned by the agentivity of the possessor (the animals mentioned in (42) are agentive, but less so than humans). Accordingly, ɓȁ is not accepted by the same informant in (43), where the possessor is ɗā ‘rain’, while it is obligatory in (44), where the agent is human.