Navigation – Plan du site

AccueilNuméros9(1)Transitivity in Bantu: Event-orie...

Transitivity in Bantu: Event-oriented constructions

Rodrigo Lazaresko Madrid


On considère généralement les langues bantoues comme des langues accusatives, en raison de leur système d’alignement des marques personnelles. Ce type d’analyse est sous-tendu par une vision spécifique de la notion de transitivité, qui repose en particulier sur la distinction sujet versus objet et sur le fait que cette distinction correspond aux rôles sémantiques joués par les actants participant à un événement donné. De fait, certaines constructions attestées en bantou remettent en question cette conception théorique de la transitivité et de la syntaxe, car, dans lesdites constructions, les fonctions grammaticales et les rôles sémantiques ne coïncident pas. Dans cet article, je présente une approche de la transitivité fondée sur les paramètres de Hopper & Thompson (1980) qui supposent une approche graduelle et non discrète de la distinction entre phrases transitives et intransitives. Dans ce cadre explicatif, une construction linguistique donnée peut donc être considérée comme plus ou moins transitive, en fonction de la prise en compte de divers traits sémantiques. Je suggère en outre que, quand des traits autres que la relation sujet/objet revêtent un rôle de premier plan quant à l’implication des actants dans une construction donnée, ladite construction peut être considérée comme régie par l’événement, dans la ligne argumentative de la ‘structure active’ proposé par Klimov (1974) et Wichmann (2007). À la lumière de ce qui vient d’être dit, j’analyse trois constructions attestées en bantou et que l’on pourrait considérer comme régies par l’événement : les constructions inversives, les extensions verbales modifiant la valence, et les verbes prédiquant des qualités. Cette analyse révèle l’avantage présenté par une approche considérant que ces constructions sont régies par l’événement : en effet, la relation entre actants et événements y est alors conçue comme un ensemble de traits morphosémantiques et non comme une suite d’opérations syntaxiques remettant en question les règles établies.

Haut de page

Texte intégral

1. Introduction

1Bantu languages are considered accusative languages, mainly due to their subject agreement system (Blake 2001; König 2008; Dom et al. 2015). However, some constructions that are not expected in this typology occur, and they are of special interest to studies in Bantu syntax. Some examples are inversion constructions, verbal expression of properties, and non-valency changing derivations. In the present article, I suggest that these unusual constructions for accusative languages are due to the expression of transitivity that occurs differently in Bantu when compared to other typologically similar languages. In my proposal, transitivity is seen as a gradient instead of a dichotomy, based on Hopper & Thompsons’ work defining transitivity as “the effectiveness or intensity with which the action is transferred from one participant to another” (1980: 252), so that expressions are analyzed as more transitive or less transitive, and not categorically as transitive or intransitive.

2From a typological perspective, those unusual constructions are found in the set of features Klimov (1974; 1977) and Wichmann (2007) consider to be typical in languages of “active typology” (for Klimov) or “event-oriented” (for Wichmann). The aim of this paper is to explain how constructions that differentiate Bantu languages from other prototypical accusative languages can be related to the active/event-oriented typology and suggest that an active structure might be present in part of the Bantu diachrony, either as reminiscent from earlier stages (more likely) or as more recent innovations.

3The next sections are organized as follows. Section 2 discusses the parameters of transitivity based on Hopper & Thompson (1980) to clarify why transitivity is understood as a continuum instead of a clear-cut category. In Section 3, I present the features attributed to the active typology and suggest that they can be analyzed in terms of an event-oriented grammar (Wichmann 2007). In Section 4, I show some constructions in Bantu that pose challenges to the studies on transitivity and syntax and explain why they are unexpected from a typological point of view, based on their transitivity traits. Finally, in Section 5, I present a summary of the discussion and suggest that diachronic studies on Bantu syntax could improve the explanation on how typologically different constructions might have emerged.

2. Transitivity in the middle

  • 1 Pooth (2016) also defines prototypical transitivity on the cognitive basis of prototypical causati (...)

4In this section, I present Hopper & Thompsons’ (1980) proposal for a more gradient approach to transitivity and sketch an analysis based on a continuum between intransitive constructions and the most transitive construction possible. Therefore, transitivity is seen in this article as a concept reflected in how a construction expresses the level of effectiveness in the realization of an energy transfer event. In other words, the more effectively an agent transfers energy towards a patient, the more transitive the construction responsible for expressing this transfer will be. In this sense, I consider the prototypical transitivity as equivalent of the prototypical causation presented by Lakoff (1987).1 In this prototypical causation, the following features are observed (Lakoff 1987: 54‑55):

  1. There is an agent that does something.
  2. There is a patient that undergoes a change to a new state.
  3. Properties 1 and 2 constitute a single event; they overlap in time and space; the agent comes in contact with the patient.
  4. Part of what the agent does (either the motion or the exercise of will) precedes the change in the patient.
  5. The agent is the energy source; the patient is the energy goal; there is transfer of energy from agent to patient.
  6. There is a single definite agent and a single definite patient.
  7. The agent is human.
  8. a. The agent wills his action.
    b. The agent is in control of his action.
    c. The agent bears primary responsibility for both his action and the change.
  9. The agent uses his hands, body, or some instrument.
  10. The agent is looking at the patient, the change in the patient is perceptible, and the agent perceives the change.

5The proximity between causation and transitivity is clear when one compares the features of causation for Lakoff with the parameters Hopper & Thompson (1980) propose to analyze transitivity. I present the latter in Table 1, with some slight modifications. These modifications were made to broaden the scope of transitivity to expressions that would otherwise be unconsidered, such as nominals. The only difference from their original parameters is that the participants are not labeled as either A or O. Instead, I use “more prominent participant” in lieu of A and “less prominent participant” in lieu of O, and the parameters of affectedness and individuation are considered for both participants, instead of only for O. This table is still in accordance with the assumption that transitivity is not an exclusively verbal feature.

Table 1 — Transitivity parameters (adapted from Hopper & Thompson 1980: 252)

Parameter High transitivity Low transitivity
A. Number of participants 2 or more participants 1 or no participant
B. Kinesis Action Non-action
C. Aspect Telic Atelic
D. Punctuality Punctual Non-punctual
E. Volitionality Volitional Non-volitional
F. Affirmation Affirmative Negative
G. Mode Realis Irrealis
H. Agency High agency Low agency
I. Participant affectedness The less prominent
participant is
totally affected
No participant is
J. Individuation of
Highly individuated Non-individuated

6Transitivity, then, can be represented by a continuum with the less transitive construction on one pole and the most transitive construction on the other. A prototypical transitive sentence such as Robin hit Sam is, then, very transitive, since all the parameters of high transitivity are present (as well as Lakoff’s properties of causation). However, since it is gradual by definition, the level of transitivity could be higher if the construction coded more participants (such as an instrument, and/or a beneficiary), and explicitly specified perfective aspect, punctuality, volitionality, and the affectedness of the patient. Also, transitivity could be lower if the act of hitting in this case was clearly an accident, or if it happened repeatedly in the past. Therefore, the extremes of the transitivity continuum would be, on one side, a construction expressing zero effectiveness in an energy transfer event, i.e., no energy transfer at all, no individuation, and no relation between different entities involved in the event, and, on the other side, the most detailed expression of a very effective energy transfer event, involving all possible participants (including intermediary means of energy transfer) and the result of that transfer. These two poles could be coded as a simple, general, undefined mass noun as the less transitive construction possible on one side, and a multiple-argument finite clause expressing highly individuated participants, with their roles in the action chain explicitly distinguished on the other. A sketch of the transitivity continuum is illustrated in Figure 1.

Figure 1 — Transitivity continuum

Figure 1 — Transitivity continuum

7The area in the middle of the transitivity continuum includes constructions with more balanced parameters, for example, a prototypical transitive sentence closer to the highest transitivity end, and a modified nominal closer to the lowest transitivity end. Considering that middle voice presents two main features, i.e., relative distinguishability among participants and relative elaboration of events (Kemmer 1993), constructions related to the middle domain (such as reflexives and reciprocals) are also in the center region of the transitivity continuum, and not close to the extremes.

  • 2 They are called “agent-oriented” and “theme-oriented” strategies by Langacker (1991; 2008).

8I suggest that there are two main strategies languages follow to navigate along the transitivity continuum, with typological implications. The first strategy privileges the distinctions among participants when describing a more transitive event. It means that either the cause of an event or the element whose state is changed is more prominent. Assuming that the subject is the most prominent element in a clause, the ways of expressing the participants of an event are related to accusative and ergative constructions, respectively.2 The second strategy is oriented to the event itself and the roles performed by the participants in an event are not mapped onto grammatical relations (subject or object) but expressed in the lexicon or with morphosemantic derivations instead. I relate these two strategies to express transitivity to the typologies of “participant-orientation” and “event-orientation”, proposed by Wichmann (2007), which is detailed in the next section.

9In the following sections, I will discuss constructions in Bantu languages that could be considered unexpected under an accusative typology. I also suggest that they are challenging if one considers transitivity as a clear-cut division between transitive and intransitive verbs, and that verbal extensions modify a construction by changing the valency of verbs.

3. Event-oriented grammar

10In this section, I present a brief typological discussion that is helpful to understand how some constructions have emerged (or have remained) in Bantu. I propose to look at these constructions not in comparison with languages construed mainly on an accusative system (participant-oriented), but with those construed on an event-orientation structure instead.

11Event-oriented grammar is a typological concept proposed by Wichmann (2007) in opposition to participant-oriented grammar. Event-oriented languages are those in which the alternation between transitive and intransitive events is not a matter of verbal derivation. Instead, this distinction is made in the lexicon by two different, semantically related forms (one transitive and one intransitive). In participant-oriented languages, “the construal of the relationships among participants takes precedence over the construal of the event and licenses the grammatical roles subject and object, which do not map directly onto the semantic roles agent and patient” (Wichmann 2007: 28). In other words, in participant-oriented languages, the relation between the participants takes precedence over the construal of the event, so the distinction between subject and object is more rigid; on the other hand, in event-oriented grammars, the participants’ construal is dictated by the way the event is conceptualized, and the relation between them depends on the semantics of the predicate. That is why constructions that challenge the subject-object distinction could be analyzed as “event-oriented”, and I will present some examples of these constructions in Bantu.

  • 3 Wichmann (2008) argues in favor of the concept of semantic alignment instead of active alignment, (...)

12The typological distinction between participant-oriented and event-oriented languages is an attempt to group languages of the accusative or ergative systems under one type (participant orientation) and languages that present a semantic alignment pattern (e.g. active-inactive, active-stative, split intransitive) under a different concept (the event-orientation). Wichmann’s proposal is not entirely new, since it is based on Klimov’s (1974) features of an “active structure” that languages with an active-inactive alignment might present.3

  • 4 As mentioned before, that is the reason I consider Wichmann’s concept of “event-orientation” a bet (...)

13According to Klimov’s proposal, the active person-marking alignment is only one of several features a language presents to express the event-orientation. Therefore, there are expressions of the active structure in the lexicon, morphology, and syntax of a language. I consider the features proposed by Klimov not as categorical typological parameters, but as insights to a different perspective on how events can be expressed linguistically without relying so heavily on the subject-object distinction, as accusative and ergative languages prototypically do. I do not assume them as sufficient and necessary conditions for a language to be labeled “active”, but rather as indexes of strategies that languages may use to encode the conceptualization of events.4

14Klimov (1974) lists several features of active languages: lexical features, morphological features, and syntactic features. I will present these in turn and discuss for the Bantu languages whether and how these features are present. According to Klimov (1974: 14‑15), active languages present the following features in the lexicon:

  1. Binary division of nouns between active and inactive.
  2. Lexical opposition of verbs not based on transitivity vs. intransitivity, but on action vs. state verbs. There is also a third lexical group: affective verbs (‘see’, ‘hear’, ‘love’).
  3. Existence of verb pairs: one form exclusively used with active nouns; a different form used with inactive nouns to express a similar meaning.
  4. Absence of verbs of possession.

15Based on these lexical features, transitivity works differently in languages of the active type. The main point is that they lead the analyses of predicates to focus on semantics (activity vs. inactivity) instead of syntax (a clear-cut distinction between transitivity vs. intransitivity). Let us see how these features are present (or not) in the Bantu languages. To my knowledge, despite their rich noun class system, Bantu languages do not split nouns in pairs of active and inactive (Feature 1). The next section will show how transitivity in Bantu is not exactly based on transitivity vs. intransitivity, and how the opposition between activity and inactivity is achieved through suffixation (Parameters 2 and 3). The only lexical feature that is clearly observed in Bantu is Parameter 4, regarding the absence of verbs of possession, which is expressed mainly by non-verbal constructions (Nurse 2008; Stassen 2009; 2013), i.e., when there is no verb ‘have’ but rather ‘be with’.

16As for morphological features, Klimov (1974: 19‑23) suggests that the active structure presents the following:

  1. The morphological category of class is present in the structure of words syntactically connected to nouns.
  2. There are both active and inactive personal affixes in verbal morphology.
  3. Verbal conjugation morphology presents a “mixed” character of class and person.
  4. The category of number in nominal morphology is rarely observed.
  5. The category of case is almost completely absent.
  6. Noun morphology presents a distinction between alienable and inalienable possession.
  7. While the nominal morphology is weakly developed, verbal morphology is well attested.

17The coincidence of some morphological features that Klimov (1974) relates to the active structure and many Bantu languages is stunning. Feature 1 is almost a defining feature of a Bantu language. Noun classes are present in many Bantu languages, and languages with no noun classes are argued to portray traces of them or have a reduced class system (Katamba 2003; Van de Velde 2019). Features 4, the absence of case marking, and 6, richer verbal than nominal morphology, are also clearly present in the Bantu languages (cf. Diercks 2012; Sheehan & Van der Wal 2018), although Feature 3 is not.

18Morphological Features 2 and 3 are the main property of active languages and represent the expression of evident challenge to the transitive vs. intransitive split in verbs. This can be seen in languages such as Guaraní (Tupian), in which there are different paradigms for person marking (1).

(1) Guaraní (Mithun 1991: 511)
a. a‑. I go.’
b. a‑gwerú aí̃na. I am bringing them now.’
c. šé‑rasí̃. I am sick.’
d. šé‑rerahá. ‘It will carry me off.’

19The first-person singular is marked by a‑ in (1a) and (1b), and by šé‑ in (1c) and (1d). The difference between them reflects the different roles performed by the first person in each of the constructions. While a‑ marks higher agency, šé‑ marks the least agentive participant, making (1a) more transitive than (1c), although they both express a single participant. However, both markings are pronominal prefixes that indicate the sole argument of the verb to which they are affixed. As mentioned before, this feature is usually taken as the defining property for an active typology, but the others also show that the active structure is not only related to the person-marking alignment; it is a tendency that languages may present to code the conceptualization of events. The idea of a split in intransitive sentences is possible from the perspective of speakers in participant-oriented languages, i.e., languages for which the subject-object relation usually maps onto the agent-patient distinction. It is worth mentioning the observation made by Mithun & Chafe (1999: 578) that “speakers of Iroquoian languages could say with equal justification that English is a ‘split patient’ language, since English sometimes marks patients as subjects, sometimes as objects”. This observation suggests that, for languages such as the Iroquoian ones, the distinction between subjects and objects does not correlate with a distinction between agents and patients, as usually happens in languages like English. As will become clear in the next section, Bantu languages do not show the same person-marking pattern as observed in Guaraní, but this does not mean that event-orientation is absent from their grammar.

20Finally, at the syntactic level of analysis, the following features are considered by Klimov (1974: 16‑19) as related to active languages:

  1. There is no differentiation in the inventory of parts of the sentence regarding direct and indirect objects.
  2. The mutual syntactic relations are determined by the dominant position of the predicate, while nominals are appositions to the corresponding affixes of the predicate.
  3. Word order is S – O1 – O2 – V for active sentences and S – O1 – V for inactive and affective sentences. It can change depending on the theme of communication.
  4. Verb and noun are mutually differentiated, although they reveal traces of a previous syncretic state.
  5. There is a tendency to polysynthesis.
  6. Verbs vary widely in aspect, but not in time (tense).
  7. Prefixal formation is more prominent than suffixal formation (more prefixal positions).
  8. Active verbs have a higher number of conjugation paradigms.
  9. Voice is absent as a category.

21There is a big debate on the status of object markers in Bantu, both in comparative and language-specific terms (Morimoto 2002; Creissels 2005; Thwala 2006; Riedel 2007; Marten & Kula 2012; Zeller 2012; Zeller & Ngoboka 2015; Van der Wal 2022, among others). Considering it could be related to the Features 1, 2 and 5 from the syntactic features of an event-orientation grammar, it is worth briefly discussing the object markers in Bantu.

22The object marker is a morpheme attached to a verb stem (usually as a prefix) that indexes a less prominent (compared to the most prominent, expressed as the subject) participant of an event, as we see in Kinyarwanda in (2).

  • 5 All glosses in this paper were adapted to ensure consistency across examples from different source (...)
(2) Kinyarwanda (JD61; Zeller & Ngoboka 2015: 208)5
a. A‑ba‑gabo ba‑kuund‑a a‑ka‑zi.
aug‑2‑men 2sm‑like‑fv aug‑12‑work
‘Men like work.’
b. A‑ba‑gabo ba‑ra‑ka‑kuund‑a.
aug‑2‑men 2smdj12om‑like‑fv
‘Men like it (i.e., work).’
c. *A‑ba‑gabo ba‑(ra)ka‑kuund‑a a‑ka‑zi.
aug‑2‑men 2smdj12om‑like‑fv aug‑12‑work
Intended: ‘Men like work.’

23In (2b), the object marker ‑ka‑ means that the object in the sentence belongs to the noun Class 12. In Kinyarwanda, the co-occurrence of the object marker and the noun is not possible, as (2c) shows. In other Bantu languages, such as Chinnima Makonde, it is possible for the object marker to co-occur with the referenced noun, and it raises the question whether it has a clitic pronoun or an agreement affix status.

(3) Chinnima Makonde (P23; Kraal 2005: 235; glosses added by Van der Wal 2022: 47)
a. Tu‑yangata vayéeni. 2.guests
‘We help guests.’
b. Tu‑va‑yangata vayéeni.
1pl.sm2om‑help 2.guests
‘We help the guests.’

24Based only on (3), it is not possible to determine if the OM is a pronoun or an agreement affix (see the aforementioned authors). The discussion on the OM status leaves unclear whether the Feature b —“The mutual syntactic relations are determined by the dominant position of the predicate, while nominals concretize appositions to the corresponding affixes of the predicate”— is observable in Bantu, but if the pronoun/referential hypothesis is right, that could be a trace of event-orientation. Nevertheless, Feature e —“polysynthesis”— is clearly present in Bantu.

  • 6 I acknowledge that this division is debatable (Rugemalira 1991), as one of the reviewers has point (...)

25In constructions with three participants, object marking variation reveals a division between languages considered “symmetric” and languages considered “asymmetric” (as proposed by Bresnan & Moshi 1990).6

26Asymmetrical object marking is found in languages where the two objects behave differently and are also differently referred to by the marker(s) attached to the verb stem. In Swahili, for example, the patient cannot be object-marked in constructions with three participants.

(4) Swahili (G42; Van der Wal 2022: 131)
a. A‑li‑m‑pa kitabu.
‘She gave him a book.’
b. *A‑li‑ki‑pa Juma.
1smpst‑7om‑give 1.Juma
Intended: ‘She gave it to Juma.’

27Example (4) shows that the object marker corresponding to the recipient can be attached to the verb stem, but the object marker referring to the theme cannot. There is, in this case, a correlation between the grammatical relation and the semantic role. In this sense, asymmetrical languages are closer to a participant-oriented grammar than the symmetrical ones, as I will show below.

28In symmetrical object marking, one could consider that there is a split in the object marker expressed in the verb. In other words, the behavior of objects and object markers are not strictly restricted by the semantic roles they are coding. In Zulu, for example, the object marker may shift between encoding the patient and the recipient, as shown in Example (5).

(5) Zulu (S42; Adams 2010: 11)
a. U‑mama u‑ba‑nik‑e in‑cwadi (aba‑ntwana).
1a‑mama 1sm2om‑give‑pfv 9‑book 2‑child
‘Mama gave them a book (the children).’
b. U‑mama u‑yi‑nik‑e aba‑ntwana (in‑cwadi).
1a‑mama 1sm9om‑give‑pfv 2‑child 9‑book
‘Mama gave the children it (a book).’
  • 7 In a Cognitive Grammar framework, “conceptualization” is the process by which people perceive the (...)
  • 8 For a detailed analysis with a more formal approach of such data, see Van der Wal (2022).

29The double possibility presented in (5) suggests that the definition of primary object and secondary object is not clearly established. The possibility of no morphosyntactic differentiation between different object types suggests that the conceptualization of a nominal as a primary or secondary object is mainly a matter of prominence rather than the role performed by the participant in the event encoded by the sentence.7 In this sense, the object marking encoded in the verb indicates which object is the most prominent (i.e., perceived as the figure in a figure-ground relation), regardless of whether it expresses the theme or the recipient.8

30The last syntactic property of the Active Structure (i) —“Voice is absent as a category”— is also observable in Bantu. Although Bantu languages have morphologically marked passives, in the discussion on verbal extensions (Section 4.2), I suggest that constructions such as the passive and the antipassive are construed by a lexical derivation that reduces transitivity in a sentence and not necessarily a syntactic operation regarding voice. That might be the reason why they are both possible in many Bantu languages, and are built by a verbal suffixal extension, in the same fashion as other semantic derivations such as the causative, the applicative, and the neuter.

31As the observations in this section show, there is good reason to treat Bantu languages not primarily as accusative (participant-oriented) but as active (event- oriented). All the parameters presented by Klimov have as a general assumption the fluid relation between subject and object since it is based on a categorical distinction between transitive and intransitive sentences. He summarizes the idea as follows: “the structure of these languages is especially oriented to the rendering, not of subjective-objective relations, which find here only implicit expression, but of relations existing between active and inactive participants” (Klimov 1974: 13). The gradual approach to transitivity proposed by Hopper & Thompson (1980) is, therefore, more appropriate to analyze constructions in which the subject vs. object distinction is not clear. In the following section, I discuss some constructions in Bantu that challenge the transitive-intransitive dichotomy, and I suggest they are an event-oriented part of Bantu grammar.

4. Unusual constructions for an accusative structure

32As mentioned before, Bantu languages are considered accusative because the subject of transitive clauses behaves similarly to subjects of intransitive clauses, and both differ from the object of transitive clauses. This can be seen in (6).

(6) Pagibete (C401; Reeder 2019: 469)
a. Nɛ́‑yɔ́y‑í‑yákà.\pst‑be.tired‑pstnpst
‘I was tired (this morning).’
b. Nɛ‑lék‑à‑gá ò‑kpìà‑kò.\fut‑set.up‑futprog 15a‑net‑15a
‘I am setting up a net.’

33Example (6) shows that the first-person subject marker encodes the subject both when it is the only participant in an event (6a) and when it is the subject in a two-participant event (6b). The subject-object distinction is even clearer when an object marker co-occurs with the subject marker, as Example (7) from another Bantu language shows.

(7) Chichewa (N31; Bresnan & Mchombo 1987: 744)
Njûchi zi‑ná‑‑lúm‑a a‑lenje.
9.bee 9sm‑pst2om‑bite‑fv 2‑hunter
‘The bees bit them, the hunters.’
  • 9 There is a large debate on the status of object markers in Bantu, both in comparative and language (...)

34In (7), there is a clear difference between subject (njûchi ‘bees’) and object (alenje ‘hunters’), which is evidenced by the corresponding class referenced by the prefixes in the verb (Classes 9 and 2, respectively).9 A clause in which the participants are maximally distinct from each other is considered the prototypical transitive sentence (Kittilä 2007; Næss 2007). Such a prototype also reflects high transitivity elements according to the Parameters A (number of participants), E (volitionality), H (agency), and I (participant affectedness), proposed by Hopper & Thompson (1980) and introduced in Section 2.

35However, this clear-cut distinction between subjects and objects is blurred by some constructions that diverge from the transitive prototype and construe two-participant events in less transitive ways. In the following subsections, I discuss inversion constructions, non-valency-changing derivations and verbal expression of properties as unexpected constructions for accusative languages, instead suggesting that they can be better understood as event-oriented.

4.1 Inversion constructions

36Inversion constructions have been extensively studied in the Bantu languages precisely because they “pose a challenge to theoretical analysis as they (appear to) present a mismatch between the syntactic code and semantic roles of an event’s participants”, as observed by Marten (2014: 138). The example in (8b) from Zulu illustrates this type of construction.

(8) Zulu (S42; Zeller 2013: 1108)
a. U‑tshani bu‑mil‑a e‑n‑gadi‑ni.
aug‑14.grass 14sm‑grow‑fv loc‑9‑garden‑loc
‘Grass grows in the garden.’
b. I‑n‑gadi i‑mil‑a u‑tshani.
aug9‑garden 9smgrow‑fv aug‑14.grass
‘In the garden grows grass.’

37Besides the word order (which is not peculiar, if one considers the English translations in the examples above), the interest on inversion constructions is due to the subject agreement that is also “reversed”. In (8a), the subject marker agrees with Class 14 from utshani (grass), but in (8b) the verbal agreement follows Class 9 from ingadi (garden). In both cases, the element performing the event of growing is the same. Besides the inversion presented in (8), in which a locative element triggers the agreement on the subject marker, it is also possible for semantic roles other than locatives to appear in the same pre-verbal position and also trigger the agreement on the subject marker. Marten & Van der Wal (2014) describe inversion constructions involving instruments, patients, and even complement clauses and quotations. Some of these constructions are illustrated by (9), (10), (11), and (12), respectively.

(9) Zulu (S42; Zeller 2012: 135): Instrument inversion
a. U‑mfundi u‑bhal‑a nge‑peni.
1‑1.student 1sm‑write‑fv with‑5.pen
‘The student is writing with a pen.’
b. I‑peni li‑bhal‑a u‑mfundi.
5‑5.pen 5sm‑write‑fv 1‑1.student
‘The student is using a pen to write.’ (lit. ‘A pen is writing the student.’)
  • 10 The numbers referring to the nominal classes were suggested by Marten & Van der Wal (2014: 330‑331) (...)
(10) Kiluguru (G35; Mkude 1974: 133):10 Patient inversion (subject-object reversal)
a. Imw‑ana ka‑tula ici‑ya.
1‑child 1sm‑broke 7‑pot
‘The child broke the pot.’
b. Ici‑ya ci‑tula imw‑ana.
7‑pot 7sm‑broke 1‑child
‘The child broke the pot’ (lit. ‘The pot broke the child.’)
(11) Kinyarwanda (D61; Morimoto 2000: 183): Complement clause inversion
a. Aba‑ana ba‑kuunda gu‑kina.
2‑child 2sm‑like 15‑play
‘The children like to play.’
b. Gu‑kina gu‑kuunda aba‑ana.
15‑play 15sm‑like 2‑child
‘It is the children who like to play.’
(12) siSwati (S43; Ziervogel & Mabuza 1976: 226; as cited in Marten & Van der Wal 2014: 335): Quotative inversion
Kw‑ent‑e njani? ku‑buta unina.
17sm‑happen‑pfv what? 17sm‑ask mother.his
‘What happened? asked his mother.’

38The constructions in (9b), (10b), and (11b) show a relation of agreement between the subject marker and the pre-verbal element, while in (12) the subject marker uses Class 17 as a default agreement (Marten & Van der Wal 2014). None of them presents a subject-object relation mapping to an agent-patient relation, as an accusative sentence usually does. Also, there is no syntactic marking to show this mismatch, as passive constructions show.

39Inversion constructions show that the subject-object distinction in Bantu is not a clear-cut phenomenon, and the main point of the present article is to suggest that a gradual perspective on transitivity might be helpful to understand that there is an event-orientation in Bantu that licences some unusual constructions. I suggest that inversion constructions are a way to reduce transitivity in a sentence, according to the Parameters E, H, I, and O —volitionality, agency, affectedness, and individuation, respectively— from Hopper & Thompson (1980). The comparison between (10a) and (10b), repeated here as (13a) and (13b), respectively, shows how transitivity is reduced by inversion constructions.

(13) Kiluguru (G35; Mkude 1974: 133)
a. Imw‑ana ka‑tula ici‑ya.
1‑child 1sm‑broke 7‑pot
‘The child broke the pot.’
b. Ici‑ya ci‑tula imw‑ana.
7‑pot 7sm‑broke 1‑child
‘The child broke the pot.’ (lit. ‘The pot broke the child.’)

40Example (13a) is a prototypical transitive sentence: the participants are clearly distinguished, there is action (Parameter B), agency (Parameter H), the less prominent participant is affected (Parameter I) and the participants are individuated (Parameter J). The inversion construction in (13b) expresses the same event by a sentence with lower transitivity: the subject marker in the verb indexes the less agentive and less volitional participant. In addition, the less prominent participant is not directly affected by the event. The analysis of which nominal is the subject and which nominal is the object is not so clear for (13b) as it is for (13a). Thus, the sentence in (13b) is further away from the transitive prototype, despite having two participants.

41Along with the inversion constructions, which are strategies to reduce the level of transitivity of a sentence, there are more features present in Bantu languages that could be considered unexpected typologically.

4.2 Non-valency-changing derivations

42The rich verbal morphology in Bantu includes many derivations that express different functions and meanings. Some of them are related to valency increase and others to valency reduction. However, they also occur without any change in the number of participants. Since the number of participants is only one of the parameters Hopper & Thompson (1980) propose for a gradual view on transitivity, I analyze verbal extensions as variations in the level of transitivity in general, which may or may not include a variation of verbal valency. They change the meaning of verbs, and not necessarily how the verbs behave syntactically, which is why I consider these extensions as morphosemantic rather than morphosyntactic phenomena. Non-valency-changing usages of the applicative, causative, neuter, and passive extensions support the idea. In addition, both passives and antipassives are found in Bantu, which is unusual considering the association these constructions have with accusative and ergative systems, respectively, which is why the antipassive construction is also briefly discussed in this paper.

4.2.1 Causative

43The causative extension was reconstructed to Proto-Bantu as *‑i‑ and/or *‑ici‑. The most common understanding of causativization is the addition of an agent (or an external causer) in a verb argument structure (Comrie 1976). The examples in (14) show that the causative extension adds a causer to the event denoted by the verb.

(14) Kikongo (Kizombo) (H10; Fernando 2013: 39‑30)
a. Mvula i‑nok‑ene.
9.rain 9rain‑pst
‘Rain rained.’ (It rained.)
b. Luzolo nok‑es‑e mvula.
1.Luzolo rain‑caus‑pst 9.rain
‘Luzolo caused rain.’

44Fernando (2013) describes the data in (14) as a transformation of the verb ‑nok ‘rain’ from an intransitive form to a transitive one, which would assign objects. However, the increase in the number of participants can also be seen as a consequence of the rise in the transitivity level in the whole clause. In (14), not only are there more participants involved (Parameter A from Hopper and Thompson), but also the agency of A and the affectedness of O are higher (Parameters H and I).

45The examples in (15) show a construction in which the causative extension does not add a causer, but reduces the volition (Parameter H) in the causee element.

(15) Swahili (G42; Dixon 2000: 66)
a. Mwalimu hu‑wa‑som‑esha wanafunzi kurani.
1.teacher hab‑2om‑read‑caus 2.student Koran
‘The teacher teaches the students the Koran.’ (They want to study it.)
b. Mwalimu hu‑wa‑som‑es(h)‑esha wanafunzi wa‑some.
1.teacher hab‑2om‑read‑caus‑caus 2.student Koran
‘The teacher teaches the students the Koran.’ (They do not want to study it.)

46Another piece of evidence for the causative extension to be analyzed as a semantic derivation (and not merely a syntactic one) is the comparison between the expression of causation using the morpheme ‑ish and using a periphrastic construction with the verb tuma (‘make’) in Kinyarwanda (16).

(16) Kinyarwanda (D61; Jerro 2013: 21‑22)
a. Umw‑arimu y‑a‑ndik‑ish‑ije umw‑ana in‑kuru.
1‑teacher 1smpst‑write‑causpfv 1‑child 9‑story
‘The teacher made the child write the story.’
b. Umw‑arimu ya‑tum‑ye umw‑ana ya‑ndik‑a in‑kuru.
1‑teacher 1smpst‑make‑pfv 1‑child 1sm‑write‑ipfv 9‑story
‘The teacher made the child write the story.’

47According to Jerro (Jerro 2013: 22), both sentences in (16) describe the same event involving the same actors. Nevertheless, (16a) is considered to encode a more direct causation than (16b). When expressed by the extension ‑ish, the causation and the event itself are conceptualized as parts of the same event: a scenario for (16a) could be a teacher monitoring the child’s writing while standing by their side. In (16b), there could be a temporal distance between the causation and the writing event —e.g., a teacher inspires a child who writes a story a few years later. The perception of the causation and the writing event as being one single event (when it is construed using a causative extension) is emphasized by the usage of only one verb, with shared tense, aspect, and mood. In this sense, (16a) is more transitive than (16b), since it encodes more participants, perfective aspect, a higher agency of the teacher, and a higher affectedness of the child (Parameters A, C, H, and J, from Hopper & Thompson, respectively).

48A higher transitivity can be expressed in a higher number of participants, but not as a rule. The same is valid for the applicative extension: it changes the verb meaning and increases the transitivity of the construction, not necessarily changing the verb valency.

4.2.2 Applicative

49The applicative extension reconstructed for Proto-Bantu is the morpheme *‑ιd‑. Its main function is to change the status of an adjunct into an argument. In other words, an applicative construction changes an element that is an indirect participant (or not a participant at all) into a direct participant. Example (17) from Ndendeule shows how an applicative construction can express more participants involved in the event of buying:

(17) Ndendeule (N10; Ngonyani 1998: 2)
a. N‑ghɛni a‑ki‑hɛmɛl‑a ngoβo.
1‑guest 1smpst‑buy‑fv 10.cloth
‘The guest bought clothes.’
b. N‑ghɛni a‑ki‑n‑hɛmɛl‑ɛl‑a mw‑ana ngoβo.
1‑guest 1smpst‑1om‑buy‑appfv 1‑child 10.cloth
‘The guest bought the child clothes.’

50Besides the buyer and the bought object, the applied construction in (17b) includes a third participant as a beneficiary. Its status as a core participant is evidenced by the object marker, which agrees with it in class. That would be the only difference between (17a) and (17b), which makes (17b) more transitive than (17a), and, consequently, suggests that the applicative morpheme enhances transitivity. This heightening of the transitivity level is not achieved by a syntactic operation expanding the verb’s argument structure, but rather by changing the transitivity-related semantic features of the construction.

51Examples from Bemba can be illustrative of the semantic effect of the applicative extension (18).

(18) Bemba (M42; Marten 2003: 12)
a. Tu‑ka‑ly‑a.
‘We will eat.’
b. Tu‑ka‑li‑il‑a.
‘We will feast.’ (Idiom.: enjoy)

52The first thing to notice in (18) is that there is no valency change in the applied form of the verb. In both sentences, there is only one participant, the first-person plural. Marten (2003) refers to the phenomenon verified in (18) as “concept strengthening”, identified as a “conceptual-pragmatic function […] irrespective of a change in valency” (Marten 2003: 14). Later, Marten & Mous (2017) developed the pragmatic approach to the applicative extension and reinforced the argument that the “action denoted by the base verb is being carried out in some way remarkably, or in some sense ‘stronger’ than normal” (Marten & Mous 2017: 10). Considering the event-orientation approach, what happens when an applicative suffix is affixed to a verb is that it is conceptualized as more effectively performed, i.e., the transitivity parameters are higher. Example (17) shows how the affectedness and the agency of one participant is higher when the verb has an applicative suffix.

(19) Kinyarwanda (D61; Jerro 2013: 89)
a. Habimana y‑a‑tey‑e Karekezi i‑buye.
Habimana 1smpst‑throw‑pfv Karekezi 5‑rock
‘Habimana threw the rock at Karekezi.’
b. Habimana y‑a‑ter‑ey‑e Karekezi i‑buye.
Habimana 1smpst‑throw‑applpfv Karekezi 5‑rock
‘Habimana threw the rock to Karekezi.’

53The English translation captures the different meanings by changing the preposition. Again, the number of participants is the same in (19a) and (19b), but (19b) is more transitive: while in (19a) Karekezi seems to be a target, unaware of the event he is involved, in (19b) there is a higher participation in the event; it seems Karekezi is expecting to be thrown a rock, maybe as part of a game, or as a result of the rock being used in a construction site. In other words, Karekezi is more “active” in the event denoted by the applied verb, which increases the sentence transitivity even though it keeps the number of participants.

54In the following sections, I will present verbal extensions that reduce transitivity but not necessarily reduce valency, reinforcing the idea that the subject-object relation in this construction is not the main parameter to analyze transitivity.

4.2.3 Passive

55The passive in Bantu has many faces —it means that there are different ways to construe a sentence with a passive-like sense. According to Schadeberg & Bostoen (2019), the reconstructed passive extension (*‑ʊ‑ after C and *‑ibʊ‑ after V) has no evident correlates in Bantu languages from Zones A and B; they also mention that it has been lost in Zone C and that other strategies to express passivity include different extensions, passive participles, and an active verb with a Class 2 subject marker.

56The canonical definition for a passive construction is provided by Siewierska (2013) and presents five parameters:

  1. The passive contrasts with another construction, the active.
  2. The subject of the active corresponds to a non-obligatory oblique phrase of the passive or is not overtly expressed.
  3. The subject of the passive, if there is one, corresponds to the direct object of the active.
  4. The construction is pragmatically restricted relative to the active.
  5. The construction displays some special morphological marking of the verb.

57Interestingly, the Bantu passive construction using the verbal extension is used as a “prototypical example” (20).

(20) Swahili (G42; Ashton 1947: 224; as cited in Siewierska 2013)
a. Hamisi a‑li‑pik‑a chakula.
Hamisi 1smpst‑cook‑fv
‘Hamisi cooked the/some food.’
b. Chakula ki‑lipik‑w‑a (na Hamisi). 7smpst‑cook‑passfv (by Hamisi)
‘The food was cooked (by Hamisi).’

58The transitivity in (20b) is lower than in (20a), since there are many parameters for low transitivity: fewer participants (agent is optional), lower action, lower volition, lower agency (Parameters A, B, E and H). The valency reduction is optional, and even when a second participant is present, it is introduced by a preposition-like particle. Thus, just like there are extensions which enhance transitivity without necessarily changing the number of participants involved in the expressed event, there are also affixes that lower transitivity and do not reduce the number of participants denoted. The alteration in the number of participants is a result of changes in transitivity, not their cause.

59There are constructions with the passive extension that lower the transitivity without affecting the number of participants or the subject-object relation. In Chichewa, for example, we can see that the extension ‑édw‑/‑ídw‑ reduces the subject control over the action (21).

(21) Chichewa (N30; Mchombo 2004: 91)
a. kodz‑a ‘urinate’ kodz‑édw‑a ‘urinate involuntarily’
b. nyel‑a ‘defecate’ nyel‑édw‑a ‘experience involuntary
bowel movement’
c. f‑a ‘die’ f‑édw‑a ‘be in bereavement’
d. uk‑a ‘wake up,
uk‑ídw‑a ‘be sexually aroused’

60Control is not a parameter of transitivity in Hopper & Thompson’s original proposal, but the relation to the “effectiveness or intensity with which is transferred from one participant to another” (1980: 252) is clear. An updated table of parameters could include an additional line (Table 2).

Table 2 — Control as a transitivity parameter

Parameter High transitivity Low transitivity
K. Control Controlled event Uncontrolled event

61It is important to mention that control as a parameter is not an ad-hoc proposal to explain the Chichewa data in (21). Mithun (1991) shows that control is one of the semantic factors (as well as agency, effect, and instigation) to distinguish the active from inactive paradigms in Lakhota, a language with event-orientation. It is clearly a specific parameter among others, which can be seen in a few situations when the event is performed, effected and/or instigated, thus being construed with the active pronoun wa‑, but there is no control from the agent (22).

(22) Lakhota (Siouan Mithun 1991: 516)
blokaska. I hiccough.’ ixa. I smile.’
wapšá. I sneezed.’ waglašnaša. I stutter, stammer.’
waglépa. I vomit.’ awaglašna. I misspoke.’
iyówaya. I yawn.’

62The passive extension is, then, a morpheme that denotes a less controlled or instantiated event. It is more a morphosemantic feature than a voice phenomenon in Bantu that affects the subject-object relation. The absence of voice phenomena is a feature of event-oriented grammars, and this might be the case for the *‑ʊ‑/*‑ibʊ‑ extension in Bantu.

63Passives are prototypically a feature of accusative languages —for which the agent-patient relation tends to be mapped onto a subject-object distinction in order to syntactically express that that mapping was overruled. In an ergative alignment pattern, the overrule of that mapping is marked by antipassive constructions. Interestingly, some studies (Ndayiragije 2006; Bostoen et al. 2015; Dom et al. 2015) suggest that Bantu languages also present an antipassive construction. In the next section, this construction will be discussed as a strategy to lower the transitivity level.

4.2.4 Antipassive

64Antipassives are understood as intransitive constructions derived from transitive sentences by demoting the object. The agent is usually the same in both sentences, but the object is either removed or expressed as an oblique participant. The example in (23) shows the phenomenon in Inuktitut.

(23) Inuktitut (Eskaleut; Bittner & Hale 1996: 36)
a. Juuna‑p Anna kunip‑p‑a‑a.
Juuna‑ergi Annaj kiss‑ind‑[+tr]‑3sgi.3sgj
‘Junna kissed Anna.’
b. Juuna (Anna‑mik) kunis‑si‑v‑u‑q.
Juuna(nom)i (Anna‑ins) kiss‑apind‑[‑tr]‑3sgi
‘Junna kisses/is kissing (Anna).’

65The sentence in (23b) is marked as antipassive and considered an intransitive derivation from (23a). This kind of construction is usually associated with ergative languages, but it can be also present in languages in which ergativity is not predominant.

66In Bantu, the antipassive is developed from the reciprocal/associative extension ‑an‑ (Bostoen et al. 2015; Dom et al. 2015). Some languages have lost the original reciprocal/associative meaning, and others present the ‑an‑ extension as an ambiguous morpheme. In every case, however, the extension functions as a “transitivity reducer”, by affecting the parameters of affectedness, individuation of the less prominent participant and, occasionally, the number of participants.

67Example (24) shows a sentence in which the extension ‑an‑ could not be read as reciprocal. The antipassive reading is possible by comparing the verb ‑yip‑ in Kisongye used with and without the verbal extension. In (24b), there is a second participant, expected due to the causative process denoted by the verb, which is closer to the transitive prototype.

(24) Kisongye (L23; Bostoen et al. 2015: 742)
a. Bà‑mpùlushì abà‑yip‑an‑a bi‑kìle bu‑kùfu.
2‑police 2sm‑kill‑apfv 8‑much 14‑night
‘The police often kill at night.’
b. Bà‑mpùlushì abà‑yip‑a ba‑ngìfi bi‑kìle bu‑kùfu.
2‑police 2sm‑kill‑fv 2‑thief 8‑much 14‑night
‘The police often kill thieves at night.’

68The antipassive extension is, thus, a semantic derivation that reduces the participation of a second participant in events. In (24b), the patient is affected (Parameter I of Hopper & Thompson), while in (24a) there is no patient at all. The conceptualization of the construction is oriented to the realization of the event and not to the subject involved in it.

69An interesting use of the antipassive is found in Cilunda. It is a usage that conveys a property —Kawasha (2003) names it the “potential extension”— of the sole participant, as the examples in (25) show.

(25) Cilunda (L52; Kawasha 2003: 325‑326)
fw‑a ‘die’ > fw‑an‑a ‘be deadly, be dangerous’
dimb‑a ‘deceive’ > dimb‑an‑a ‘be false, deceitful’
sum‑a ‘bite’ > sum‑an‑a ‘be biting, likely to bite’
tuk‑a ‘insult’ > tuk‑an‑a ‘be insulting’
shiñ‑a ‘curse’ > shiñ‑an‑a ‘be one who curses’
zuw‑a ‘be angry’ > zuw‑an‑a ‘be quarrelsome’
shet‑a ‘be drunk’ > shet‑an‑a ‘be intoxicating’
  • 11 Some Bantu languages also present an extension to indicate potentiality in a patientive sense: the (...)

70This usage of the antipassive to express a property of the subject is also found in Lucazi (K13), in the sense of a potentiality of performing an action (Bostoen et al. 2015: 743).11 Since the verbal expression of properties is listed among Klimov’s parameters of an active structure, it is important to discuss it, and that is the main objective of the next section.

4.3 Verbal expression of properties

71There are three strategies to express properties in Bantu: adjectival modification, connective constructions, and verbal constructions. Adjectives in Bantu show agreement in class with the noun they modify, as (26) illustrates:

(26) Kinyarwanda (JD61; Jerro & Wechsler 2015: 158)
a. umw‑ana mu‑to
1‑child 1‑small
‘the small child’
b. aba‑na ba‑to
2‑child 2‑small
‘the small children’
  • 12 The small number of non-derived adjectives is a characteristic attributed to the entire Niger-Cong (...)

72In (26), it is possible to observe that the adjective ‑to (‘small’) in Kinyarwanda has a prefix that agrees with the class of the modified noun: Class 1 in (26a) and Class 2 in (26b). This modification of nouns by adjectives is quite similar to what occurs in participant-oriented languages, such as most Indo-European languages. However, the class of adjectives is quite limited in most Bantu languages. The study of adjectives in African languages carried out by Segerer (2008) shows that, in a sample of 30 Bantu languages, only four have more than 15 adjectives. From these four languages, the largest number of adjectives was found in Shingazidja (G44), which has 26 adjectives.12 The remaining 29 Bantu languages analyzed have closed classes of adjectives with less than 25 items. Therefore, the other strategies to express properties are more used in innovative expressions for noun modification.

73A second strategy to express noun modification in Bantu is the connective construction (Van de Velde 2013). These are constructions in which a noun is modified by a preposition-like particle (usually a variation of the Proto-Bantu reconstructed morpheme *-a) that connects it to a modifier noun, as illustrated in (27) and (28).

(27) Kimbundu (H21; Augusto 2016: 171)
mu‑hatu u‑a nguzu
1‑woman 1‑con strength
‘strong woman’ (lit. ‘woman of strength’)
(28) Lingala (C36; Brisard & Meewis 2009: 28)
Pierre a‑kom‑aka mi‑kandá ya ki‑tóko.
Pierre 1sm‑write‑hab 4‑letter 4.con 14‑beauty
‘Pierre writes beautiful letters.’ (lit. ‘Pierre writes letters of beauty.’)

74Both in (27) and in (28), the particle ‑a (preceded by an agreement prefix) connects a modified noun (mu-hatu and mi-kanda, respectively) to a modifier noun (nguzu and ki-toto, respectively). While a simple adjectival construction presents only one noun modified by conceptualizing the thing it denotes in an evaluation scale (the examples in (26) use a size scale, for instance), the connective construction establishes a relation between two nouns, i.e., a relation between two participants, instead of only one. In other words, an adjectival construction denotes only one thing, which is modified by abstract scales (color, size, dimension, for example). The connective construction, on the other hand, modifies a noun by expressing its relation to another noun.

75The connective constructions are also used to express possession in many Bantu languages, which reinforces the idea that there are two distinct participants involved in this kind of construction, as seen in miguu ya mtoto ‘legs of the child’ from Swahili in (29).

(29) Swahili (G42; Keach & Rochemont 1992: 83)
Miguu ya mtoto a‑li‑funik‑wa.
4.leg 4.con 1.child 1‑pst‑cover‑pass
‘The legs of the child were covered.’

76When compared to the use of simple adjectives, connective constructions are more transitive, since they increase the number of elements involved in predication. They are, however, less transitive than the verbal predication strategy to modify nouns, since the latter grounds the linguistic expression in a temporal domain, increasing the perception of “effectiveness of an event”.

77As mentioned previously, by understanding transitivity as a continuum, I place an exclusively nominal, indefinite, and unbounded expression in the extreme of minimum transitivity, whereas, in the other extreme (that of maximum transitivity), I place a clause with multiple participants, all of them distinct from each other and well-defined, with volitive and agentive action by the term encoded as A (subject). Thus, adjective constructions are presented in a less transitive position than the verbal constructions used to qualify nominals. One example of this type of modification is given in (30).

(30) Lubukusu (E31; Wasike 2018: 335, 337)
a. Wafula a‑la‑saangal‑a.
Wafula 1‑fut‑happy‑fv
‘Wafula will be happy today.’
b. Wafula a‑a‑sangal‑il‑e.
Wafula 1‑prs‑happy‑aspfv
‘Wafula is happy.’
  • 13 In terms of Cognitive Grammar, they are entities of the processual relationship type, rather than (...)

78The Lubukusu examples in (30) show how the meaning translated as ‘happy’ can be expressed through a verbal construction like any other, including the affixes for subject, tense, and aspect. Assuming that transitivity is a property not exclusive to verbs or sentences, I suggest the mentioned features make the verbal expression of properties more transitive when compared to a modification by a simple adjective, since it increases the parameter of participant affectedness. These are indications that the expression refers to something closer to an event than to a reified entity.13

79In sum, while inversion constructions reduce transitivity, the verbal expression of properties enhances it. The “valency-changing” extensions also change the transitivity of Bantu expressions, either to increase it (as in causative and applicative constructions) or to reduce it (in neuter, passive, and antipassive constructions). The main point here is that they express transitivity variation by changing semantic parameters, not only the number of participants involved. As stated before, this reduces the prominence of the subject-object distinction as a parameter to analyze transitivity. Instead, the constructions discussed in this article express an event-orientation in Bantu, by coding changes in transitivity by its semantic features.

5. Summary

80In this article, I have presented two theoretical frameworks to analyze transitivity. The first one is the seminal proposal by Hopper & Thompson (1980) that transitivity is a phenomenon that involves not only verbal valency, but a series of semantic parameters. These parameters capture the variation in the level of transitivity a sentence expresses, creating a continuum of more transitive and less transitive constructions instead of a sharp dichotomy of transitive and intransitive verbs. The second assumption is that a transitivity notion that relies on the subject-object distinction has a typological basis (of participant-oriented grammars) and that, for some languages, transitivity is more influenced by semantic factors such as action, volition, and affectedness, for example. This transitivity strategy is event-oriented, following the terminology proposed by Wichmann (2007).

81Some constructions in Bantu languages have been discussed due to their relation to the subject-object distinction: inversion constructions, four “valency-changing” verbal extensions, and the different strategies to express properties. The causative, applicative, passive, and antipassive extensions were discussed as morphosemantic derivations. They are interesting phenomena in Bantu because they challenge both the subject-object distinction and the relation between grammatical relations and semantic roles. I have suggested that they are related to event-orientation, expressing features of an “active structure” that are not usual in accusative systems. These features are not sufficient and necessary conditions to classify a language as event-oriented, but they are useful to reveal how some constructions express transitivity phenomena without relying on the subject-object distinction.

82In his work on active languages, Klimov (1974) suggests that the active structure was an early stage of an ergative system. Nowadays it is largely accepted that no language is fully construed on a single pattern, especially one regarding person marking alignment. In this article, I have discussed some event-oriented constructions in a group of languages that prototypically have an accusative alignment, so it would be interesting to look at them from a diachronic perspective and see to what extent the event-orientation is a trace of earlier stages of Proto-Bantu or if event-oriented constructions are recent innovations. Studies on tense and aspect might reveal that the event orientation was present in earlier stages of Bantu languages, but the discussion on antipassives presented here shows that some event-oriented features are emerging more recently, so both hypotheses are still valid.


83This study was financed in part by the Coordenação de Aperfeiçoamento de Pessoal de Nível Superior – Brasil (CAPES) – Finance Code 001. I would like to thank the reviewers, as well as Jenneke van der Wal and Evani Viotti for their remarks and suggestions on previous versions of this article. I also would like to thank the audience and the participants from the Bantunivar workshop at WOCAL 10 for their helpful comments. Any shortcomings of the paper remain my own.


84Numbers refer to the noun classes in that language, except when followed by sg = singular or pl = plural. In this case, the numbers refer to the speech person (e.g., 1sg = first person singular). 1a is a subclass in noun Class 1.

ap antipassive
appl applicative
asp aspect
aug augment
caus causative
con connective
dj disjoint form
erg ergative
fut future
fv final vowel
hab habitual
ipfv imperfective
ind indicative
ins instrumental
loc locative
nom nominative
npst non-past
om object marker
pass passive
pfv perfective
pl plural
prog progressive
prs present
pst past
sg singular
sm subject marker
+tr + transitive
‑tr - transitive
Haut de page


Adams, Nikki. 2010. The Zulu ditransitive verb phrase. Chicago, IL: University of Chicago (PhD dissertation).

Ashton, Ethel O. 1947. Swahili grammar, including intonation. London: Longmans, Green & Co.

Augusto, Moisés Alves. 2016. Morfologia contrastiva entre português e kimbundu: obstáculos e suas causas na escrita e ensino da língua portuguesa entre os kimbundu em Angola. São Paulo: Pontifícia Universidade Católica de São Paulo (PhD dissertation).

Bittner, Maria & Ken Hale. 1996. The structural determination of case and agreement. Linguistic Inquiry 27(1). 1‑68.

Blake, Barry. 2001. Case. 2nd ed. Cambridge: Cambridge University Press.

Bostoen, Koen, Sebastian Dom & Guillaume Segerer. 2015. The antipassive in Bantu. Linguistics 53(4). 731‑772.

Bresnan, Joan & Sam A. Mchombo. 1987. Topic, pronoun, and agreement in Chicheŵa. Language 63(4). 741‑782.

Bresnan, Joan & Lioba Moshi. 1990. Object asymmetries in comparative Bantu syntax. Linguistic Inquiry 21(2). 147‑185.

Brisard, Frank & Michael Meewis. 2009. Present and perfect in Bantu: The case of Lingála. Journal of African Languages and Linguistics 30(1). 21‑43.

Comrie, Bernard. 1976. The syntax of causative constructions: Cross-language similarities and divergences. In Masayoshi Shibatani (ed.), The grammar of causative constructions, 259‑312. Leiden: Brill.

Creissels, Denis. 2000. Typology. In Bernd Heine & Derek Nurse (eds.), African languages: An introduction, 232‑258. Cambridge: Cambridge University Press.

Creissels, Denis. 2005. A typology of subject and object markers in African languages. In F. K. Erhard Voeltz (ed.), Studies in African linguistics typology, 43‑70. Amsterdam: John Benjamins.

Diercks, Michael. 2012. Parameterizing case: Evidence from Bantu. Syntax 15(3). 253‑286.

Dixon, R. M. W. 2000. A typology of causatives: Form, syntax and meaning. In R. M. W. Dixon & Alexandra Y. Aikhenvald (eds.), Changing valency: Case studies in transitivity, 30‑83. Cambridge: Cambridge University Press.

Dom, Sebastian. 2014. The neuter in Bantu: A systemic functional analysis. Ghent: Ghent University (M.A. thesis).

Dom, Sebastian, Guillaume Segerer & Koen Bostoen. 2015. Antipassive/associative polysemy in Cilubà (Bantu, L31a): A plurality of relations analysis. Studies in Language 39(2). 354‑385.

Fernando, Mbiavanga. 2013. The causative and anticausative alternation in Kikongo (Kizombo). Stellenbosch: Stellenbosch University (PhD dissertation).

Hopper, Paul J. & Sandra A. Thompson. 1980. Transitivity in grammar and discourse. Language 56(2). 251‑299.

Jerro, Kyle J. 2013. Argument structure and the typology of causatives in Kinyarwanda: Explaining the causative-instrumental syncretism. Austin, TX: University of Texas (M.A. thesis).

Jerro, Kyle J. & Stephen Wechsler. 2015. Person-marked quantifiers in Kinyarwanda. In Jürg Fleischer, Elisabeth Rieken & Paul Widmer (eds.), Agreement from a diachronic perspective, 147‑164. Berlin: De Gruyter Mouton.

Katamba, Francis. 2003. Bantu nominal morphology. In Gérard Philippson & Derek Nurse (eds.), The Bantu languages, 103‑120. London: Routledge.

Kawasha, Boniface Kaumba. 2003. Lunda grammar: A morphosyntactic and semantic analysis. Eugene, OR: University of Oregon (PhD dissertation).

Keach, Camillia N. & Michael Rochemont. 1992. On the syntax of possessor raising in Swahili. Studies in African Linguistics 23(1). 82‑99.

Kemmer, Suzanne. 1993. The middle voice. Amsterdam: John Benjamins.

Kittilä, Seppo. 2007. A typology of tritransitives: Alignment types and motivations. Linguistics 45(3). 453‑508.

Klimov, Grigorij A. 1974. On the character of languages of active typology. Linguistics 12(131). 11‑25.

Klimov, Grigorij A. 1977. Типология языков активного строя. Moscow: Наука.

König, Christa. 2008. Case in Africa. Oxford: Oxford University Press.

Kraal, Pieter J. 2005. A grammar of Makonde (Chinnima, Tanzania). Leiden: Universiteit Leiden (PhD dissertation).

Lakoff, George. 1987. Women, fire, and dangerous things: What categories reveal about the mind. Chicago, IL: University of Chicago Press.

Langacker, Ronald W. 1991. Foundations of cognitive grammar. Vol. 2: Descriptive application. Redwood City, CA: Stanford University Press.

Langacker, Ronald W. 2008. Cognitive grammar: A basic introduction. Oxford: Oxford University Press.

Madrid, Rodrigo Lazaresko. 2023. Considerações tipológicas sobre transitividade em línguas bantu. São Paulo: USP (PhD dissertation).

Marten, Lutz. 2003. The dynamics of Bantu applied verbs: An analysis at the syntax-pragmatics interface. In Kézié K. Lébikaza (ed.), Actes du 3e WOCAL Congrès mondial de linguistique africaine, Lomé 2000, 207‑221. Cologne: Rüdiger Köppe.

Marten, Lutz. 2014. The preverbal position(s) in Bantu inversion constructions: Theoretical and comparative considerations. ZAS Papers in Linguistics 57. 136‑159.

Marten, Lutz & Nancy C. Kula. 2012. Object marking and morphosyntactic variation in Bantu. Southern African Linguistics and Applied Language Studies 30(2). 237‑253.

Marten, Lutz & Maarten Mous. 2017. Valency and expectation in Bantu applicatives. Linguistics Vanguard 3(1).

Marten, Lutz & Jenneke van der Wal. 2014. A typology of Bantu subject inversion. Linguistic Variation 14(2). 318‑368.

Mchombo, Sam. 2004. The syntax of Chichewa. Cambridge: Cambridge University Press.

Mithun, Marianne. 1991. Active/agentive case marking and its motivations. Language 67(3). 510‑546.

Mithun, Marianne & Wallace Chafe. 1999. What are S, A, and O? Studies in Language 23(3). 569‑596.

Mkude, Daniel J. 1974. A study of Kiluguru syntax with special reference to the transformational history of sentences with permuted subject and object. London: University of London (PhD dissertation).

Morimoto, Yukiko. 2000. Discourse configurationality in Bantu morphosyntax. Stanford, CA: Stanford University (PhD dissertation).

Morimoto, Yukiko. 2002. Prominence mismatches and differential object marking in Bantu. In Miriam Butt & Tracy Holloway King (eds.), Proceedings of the LFG ’02 Conference, National Technical University of Athens, 292‑314. Stanford, CA: CSLI.

Næss, Åshild. 2007. Prototypical transitivity. Amsterdam: John Benjamins.

Ndayiragije, Juvénal. 2006. The ergativity parameter: A view from antipassive. In Alan Johns, Diane Massam & Juvénal Ndayiragije (eds.), Ergativity: Emerging issues, 271‑292. Dordrecht: Springer.

Ngonyani, Deo. 1998. Properties of applied objects in Kiswahili and Kindendeule. Studies in African Linguistics 27(1). 67‑95.

Nurse, Derek. 2008. Tense and aspect in Bantu. Oxford: Oxford University Press.

Pooth, Roland. 2016. Causativization by means of transitivizing and detransitivizing morphological strategies. Unpublished ms.

Reeder, JeDene. 2019. Pagibete C401. In Mark Van de Velde, Koen Bostoen, Derek Nurse & Gérard Philippson (eds.), The Bantu languages, 449‑471. 2nd ed. London: Routledge.

Riedel, Kristina. 2007. Object marking in Sambaa. In Bettelou Los & Marjo van Koppen (eds.), Linguistics in the Netherlands 2007, 199‑210. Amsterdam: John Benjamins.

Rugemalira, Josephat M. 1991. What is a symmetrical language? Multiple object constructions in Bantu. In Kathleen Hubbard (ed.), Proceedings of the Seventeenth Annual Meeting of the Berkeley Linguistics Society, February 15‑18, 1991, Special session on African language structures, 200‑209. Berkeley, CA: Berkeley Linguistics Society.

Schadeberg, Thilo C. & Koen Bostoen. 2019. Word formation. In Mark L. O. Van de Velde, Koen Bostoen, Derek Nurse & Gérard Philippson (eds.), The Bantu languages, 172‑203. 2nd ed. London: Routledge.

Segerer, Guillaume. 2008. Closed adjective classes and primary adjectives in African Languages.

Sheehan, Michelle & Jenneke van der Wal. 2018. Nominal licensing in caseless languages. Journal of Linguistics 54(3). 527‑589.

Siewierska, Anna. 2013. Passive constructions. In Matthew S. Dryer & Martin Haspelmath (eds.), The world atlas of language structures online. Leipzig: Max Planck Institute for Evolutionary Anthropology.

Stassen, Leon. 2009. Predicative possession. Oxford: Oxford University Press.

Stassen, Leon. 2013. Predicative possession. In Matthew S. Dryer & Martin Haspelmath (eds.), The world atlas of language structures online. Leipzig: Max Planck Institute for Evolutionary Anthropology.

Thwala, Nhlanhla. 2006. Parameters of variation & complement licensing in Bantu. ZAS Papers in Linguistics 43. 209‑232.

Van de Velde, Mark L. O. 2013. The Bantu connective construction. In Anne Carlier & Jean-Christophe Verstraete (eds.), Genitive case and genitive constructions, 217‑252. Amsterdam: John Benjamins.

Van de Velde, Mark L. O. 2019. Nominal morphology and syntax. In Mark L. O. Van de Velde, Koen Bostoen, Derek Nurse & Gérard Philippson (eds.), The Bantu languages, 237‑269. 2nd ed. London: Routledge.

Wal, Jenneke van der. 2022. A featural typology of Bantu agreement. Oxford: Oxford University Press.

Wasike, Aggrey. 2018. Adjectives in Lubukusu. In Jason Kandybowicz, Travis Major, Harold Torrence & Philip T. Duncan (eds.), African linguistics on the prairie: Selected papers from the 45th Annual Conference on African Linguistics, 325‑339. Berlin: Language Science Press.

Wichmann, Søren. 2007. Valency-reduction in event-oriented languages. In Zarina Estrada Fernandes, Søren Wichmann, Claudine Chamoreau & Albert Álvarez González (eds.), Studies in voice and transitivity (Estudios de voz y transitividad), 33‑51. Munich: Lincom Europa.

Wichmann, Søren. 2008. The study of semantic alignment: Retrospect and state of the art. In Mark Donohue & Søren Wichmann (eds.), The typology of semantic alignment, 3‑23. Oxford: Oxford University Press.

Zeller, Jochen. 2012. Object marking in isiZulu. Southern African Linguistics and Applied Language Studies 30(2). 219‑235.

Zeller, Jochen. 2013. Locative inversion in Bantu and predication. Linguistics 51(6). 1107‑1146.

Zeller, Jochen. 2015. Argument prominence and agreement: Explaining an unexpected object asymmetry in Zulu. Lingua 156. 17‑39.

Zeller, Jochen & Jean Paul Ngoboka. 2015. On parametric variation in Bantu, with particular reference to Kinyarwanda. Transactions of the Philological Society 113(2). 206‑231.

Ziervogel, Dirk & Enos John Mabuza. 1976. A grammar of the Swati language (siSwati). Pretoria: J. L. Van Schaik.

Haut de page


1 Pooth (2016) also defines prototypical transitivity on the cognitive basis of prototypical causation.

2 They are called “agent-oriented” and “theme-oriented” strategies by Langacker (1991; 2008).

3 Wichmann (2008) argues in favor of the concept of semantic alignment instead of active alignment, since not all languages with split intransitives rely exclusively on the active-inactive distinction. That is also the reason I consider “event-orientation” as a more appropriate label than “active structure”. However, when referring to Klimov’s work, I will still use “active structure”.

4 As mentioned before, that is the reason I consider Wichmann’s concept of “event-orientation” a better term than “active structure” to name that set of strategies.

5 All glosses in this paper were adapted to ensure consistency across examples from different sources.

6 I acknowledge that this division is debatable (Rugemalira 1991), as one of the reviewers has pointed out. Nevertheless, recent studies on object marking in Bantu languages use that distinction (Riedel 2007; Zeller 2015; Van der Wal 2022, for instance), and it might be related to the discussion in this paper.

7 In a Cognitive Grammar framework, “conceptualization” is the process by which people perceive the world and make sense of it. Different construals in language reveal different perceptions: e.g., sentences such as the table is under the TV and the TV is on the table reveal two different perceptions of a figure-ground relation (Langacker 1991; 2008).

8 For a detailed analysis with a more formal approach of such data, see Van der Wal (2022).

9 There is a large debate on the status of object markers in Bantu, both in comparative and language-specific terms (Morimoto 2002; Thwala 2006; Creissels 2005; Riedel 2007; Marten & Kula 2012; Zeller 2012; Zeller & Ngoboka 2015, among others). I will not discuss the object marker status in this article, even though the pronoun analysis would reinforce the argument to be developed.

10 The numbers referring to the nominal classes were suggested by Marten & Van der Wal (2014: 330‑331). Mkude has not included them directly in his glosses.

11 Some Bantu languages also present an extension to indicate potentiality in a patientive sense: the neuter ‑ik‑. It is not discussed in the present work for space reasons, but I refer the reader to Dom (2014) for a discussion of the neuter in its different usages (including the potential reading) and to Madrid (2023) for a discussion of it as a transitivity reducer as an event-orientated construction.

12 The small number of non-derived adjectives is a characteristic attributed to the entire Niger-Congo phylum (Creissels 2000).

13 In terms of Cognitive Grammar, they are entities of the processual relationship type, rather than the thing type.

Haut de page

Table des illustrations

Titre Figure 1 — Transitivity continuum
Fichier image/png, 21k
Haut de page

Pour citer cet article

Référence électronique

Rodrigo Lazaresko Madrid, « Transitivity in Bantu: Event-oriented constructions »Linguistique et langues africaines [En ligne], 9(1) | 2023, mis en ligne le 30 juin 2023, consulté le 19 juillet 2024. URL : ; DOI :

Haut de page


Rodrigo Lazaresko Madrid

Universidade de São Paulo (USP)

Haut de page

Droits d’auteur


Le texte seul est utilisable sous licence CC BY-SA 4.0. Les autres éléments (illustrations, fichiers annexes importés) sont « Tous droits réservés », sauf mention contraire.

Haut de page
Rechercher dans OpenEdition Search

Vous allez être redirigé vers OpenEdition Search