Navigation – Plan du site
Varia

Transitivity in Bakel Soninke

La transitivité en soninké de Bakel
Транзитивность в сонинке р-на Бакел
Denis Creissels et Anna Marie Diagne
p. 5-38

Résumés

Dans cet article, après avoir présenté les principales particularités du soninké de Bakel dans le domaine de la phonologie et de la morphosyntaxe, nous examinons de façon détaillée les mécanismes morphosyntaxiques liés à la transitivité : alternances de transitivité non marquées morphologiquement et dérivations exprimant une opération sur la valence verbale. Dans ce domaine, les caractéristiques les plus saillantes du soninké sont une dérivation antipassive productive allant de pair avec la rareté des verbes A-labiles, et un phénomène de P-labilité concernant presque exclusivement les lexèmes verbaux terminés par i ou e, en quasi-complémentarité avec une dérivation à fonction médiopassive propre aux verbes terminés par a, o ou u.

Haut de page

Notes de l’auteur

Abbreviations
ANTIP = antipassive
CAUS = causative
CMP = completive
D = determination marker
DEM = demonstrative
DET = detransitivization marker
GER = gerundive
FOC = focalization marker
IMPER = imperative
INTR = intransitive
LOCCOP = locative copula
LOCCOPF = locative copula in focalization context
NEG = negative
OBL = oblique
This gloss is used for a postposition with a variety of uses that cannot be covered in a satisfying way by a more precise term.
PL = plural
POS = positive
REFL = reflexive
SG = singular
SUBJ = subjunctive
TR = transitive

Texte intégral

1. Introduction

  • 1 This article has benefited from the support of the French National Research Agency (ANR) within the (...)

1Soninke (sooninkanqanne), spoken mainly in Mali, Mauritania, Senegal, and The Gambia, belongs to the Soninke-Bozo sub-branch of the western branch of the Mande language family. The only relatively well-documented Soninke variety is that spoken in Kaedi (Mauritania), for which two comprehensive grammars are available (Diagana O.M. (1984 or 1995) and Diagana Y. (1990 or 1994)), as well as a dictionary (Diagana O.M. 2011). In this paper, building on these works, on the analysis of voice in Kaedi Soninke provided by Creissels (1991a), and on Anna Marie Diagne’s work on the phonology and morphology of Bakel Soninke (Diagne (2008)), we describe the morphosyntactic phenomena related to transitivity in the Soninke variety spoken in Bakel (Senegal).1

2The article is organized as follows. In Section 2, we provide basic information on Soninke phonology and morphosyntax, emphasizing the particularities of Bakel Soninke. In Section 3, we present the three valency-changing morphological derivations found in Bakel Soninke. In Section 4, we discuss the classification of verbs as strict transitive, strict intransitive, A-labile, P-labile, and A/P-labile, and the division of transitive verbs into several sub-classes according to the morphological marking of their deagentive and depatientive uses. In Section 5, we discuss the status of Soninke according to the distinction between transitivizing and detransi­tivizing languages proposed by Nichols & al. (2004). Section 6 summarizes the main conclusions.

2. A sketch of Bakel Soninke phonology and grammar

2.1. Segmental phonology

2.1.1 Consonants

3Bakel Soninke has the following inventory of consonantal sounds:

Table 1. Phonetic inventory of Soninke consonants

labials

alveolars

palatals

velars

uvulars

glottal

voiceless plosives

p

t

c

k

q

voiced plosives

b

d

j [ɟ]

g

fricatives

f

s

x [χ]

h

vibrant

r

lateral

l

glides

y [j]

w

nasals

m

n

ñ [ɲ]

ŋ

4There are also geminates, which mostly represent underlying sequences whose elements do not belong to the same morpheme, with |N| as the first or second element of the sequence.

5Sequences with |N| in initial position can be illustrated by the concatenation of first person singular |Ń| with verbs or nouns:

Ń + fàré

‘donkey’

ḿpàré

‘my donkey’

Ń + sí

‘horse’

ñ́cì

‘my horse’

Ń + xánnè

‘voice’

ŋ́qánnè

‘my voice’

Ń + máamè

‘grandparent’

ḿmáamà

‘my grandfather/mother’

Ń + tá

‘foot’

ńtà

‘my foot’

Ń + kà

‘house’

ŋ́kà

‘my house’

6When the initial consonant of the stem is a vibrant or a lateral, the concatenation of |Ń| gives a lateral geminate:

Ń + lébò

‘knife’

ĺlébò

‘my knife’

Ń + rémmè

‘son/daughter’

ĺlémmè

‘my son/daughter’

7True prenasals and nasal geminates are rare in Soninke and come mostly from loans.

8The consonantal systems of Kaedi and Bakel Soninke are very similar, but in stem-initial position, there is a regular correspondence Kaedi h ~ Bakel f:

Bakel

Kaedi

fàré

hàré

‘donkey’

fàabè

hàabè

‘father’

fàté

hàté

‘skin’

máráfà

máráhà

‘gun’

fállè

hállé

‘back’

kófè

kóhè

‘nape of the neck’

fò

hò

‘thing’

9There are however cases in which both dialects have an initial h:

Bakel

Kaedi

hóorè

hóorè

‘noble’

10The concatenation of f and h with |Ń| gives however equally mp, which allows analyzing them as members of the same phonological class:

Bakel

Kaedi

both varieties

fàré

hàré

ḿpàré

‘my donkey’

fàabè

haabe

ḿpàabà

‘my father’

fàté

hàté

ḿpàté

‘my skin’

11The prenasalization of a glide gives rise to a nasal consonant with the same place of articulation as the glide, which allows considering that glides have no consonantal node, and consequently cannot contribute to quantity in­crease:

Ń + yáaxè

ñáaxè

‘my eye’

Ń + wùllé

ŋûllè

‘my dog’

12Preceding a stem beginning with a vowel, |N| surfaces as a velar consonant:

Ń + ótò

ŋótò

‘my car’

Ń + ànjóbè

ŋànjóbè

‘my catfish’

  • 2 On the absence of tone marks when verbs are quoted in isolation, see Section 2.2.

13|N| also appears in non-initial position in the gerundive suffix -NV (where V is an unspecified vowel copying all the features of the last vowel of the verb root). This suffix attaches to an allomorph of verbal lexemes that may end with a consonant or a vowel, and consequently, in this construction, |N| may occur between two vowels (V_V) or between a consonant and a vowel (C_V). In intervocalic position |N| surfaces as the alveolar nasal n, but when |N| is preceded by a consonant, the sequence |CN| surfaces as a geminate consonant:2

  • 3 √ signals an abstract verb root which in other contexts (in particular, in the quotation form of ve (...)

lifi

+ NV

→ lifini

‘sewing’

soxo

+ NV

→ soxono

‘cultivating’

yiga

+ NV

→ yigana

‘eating’

√kat

+ NV

→ katta

‘beating’3

√kan

+ NV

→ kanna

‘fearing’

√bos

+ NV

→ bocco

‘sucking’

√gem

+ NV

→ gemme

‘meeting’

√kar

+ NV

→ kalla

‘dying’

14On the basis of their phonological behavior the consonants of Bakel Soninke can be classified as follows:

Table 2. Phonemic inventory of Soninke consonants

peripheral

alveolar

palatal

posterior

labial

glottal

velar

uvular

voiceless stops

p

t

c

k

q

voiced stops

nasal

m

n/N

ñ

ŋ

lateral

l

plain

b

d

j

g

spirants

non vibrant

f

h

s

x

vibrant

r

glides

y

w

2.1.2. Vowels

15Bakel Soninke has five vowels qualities and a contrast between long and short vowels. The five vowel qualities can be represented as follows:

Table 3. The short vowels of Soninke

anterior

central

posterior

degree 1

i

o

degree 2

e

a

u

16Contrary to consonants, there is no evidence that long vowels result from the concatenation of two adjacent vowels. Concatenations of two underlying vowels at morpheme boundaries do not surface as long vowels, as illustrated by the suffixation of the detransitivization marker -i (see Section 3.2), which amalgamates with the final vowel of non-monosyllabic verb stems without increasing quantity:

garaba

‘split’

+ -i

→ garabe

‘be split’

yonto

‘push’

+ -i

→ yonte

‘be pushed’

17We therefore consider that Soninke has a series of long vowels contrasting with the short ones:

Table 4. The long vowels of Soninke

anterior

central

posterior

degree 1

ii

oo

degree 2

ee

aa

uu

2.1.3. Syllable structure

18Bakel Soninke has four types of syllables: CV, CVV, CVC and CVN. The CV and CVC (with a non-nasal coda) types can be viewed as light syllables on language-specific grounds. As shown by the examples, the non-nasal coda of CVC syllables is always identical with the onset of the following syllable and forms a geminate with it:

CV

CVC

fà.ré

‘donkey’

kít.tè

‘hand’

kí.yé

‘sun’

sòk.kí.xá.yè

‘grass’

kà

‘house’

sík.kò

‘three’

sù.xú.bà.fá.nà

‘daybreak’

hac.ci.te

‘have lunch’

sì.yà.xà.ré

‘mare’

fìl.ló

‘two’

19The two possible types of heavy syllables are CVV and CVN. The nasal coda in CVN syllables may form either a geminate or a prénasal with the following onset:

CVV

CVN

sàa.né

‘star’

rém.mè

‘son/daughter’

bòo.xí

‘tear’

fún.bún.nè

‘abcess’

xá.báa.nè

‘egg’

xú.rún.gò

‘knee’

juu.ra

‘pray’

kón.pè

‘bedroom’

dáa.rú.mè

‘day before yesterday’

xùn.bá.nè

‘tomorrow’

20The homorganic nasal may also appear as a syllabic consonant in the realization of |Ń| (1st person singular), but this occurs only when this morpheme is the first element of an intonational phrase.

Ń + fàabà

→ ḿpàabà

‘my father’

Ń + máamà

→ ḿmáamà

‘my grandfather/mother’

21In non initial position |Ń| appears as ín, as in [mìŋgàrì] ‘I certainly will come’, decomposable as mà ín gà rì (mà ‘certainly’, ín ‘I’, gà ‘subordination marker’, rì ‘come’).

2.2. Tone

22Like Kaedi Soninke, Bakel Soninke has a tone system with two contrasting tones, H(igh) and L(ow), and downsteps analyzable as the trace of floating L tones. Moreover, nouns in isolation show tonal contours similar to those described for Kaedi Soninke, and in both varieties, the tones of nouns in isolation can be analyzed as resulting from the interaction of lexical tones with a grammatical L tone that plays a role in the determination system of nouns. But in other respects, the tone system of Bakel Soninke is very different from that of Kaedi Soninke as described by Diagana (1984) and Diagana (1995). The most striking characteristic of Bakel Soninke is a relatively high degree of variation in the tonal realization of individual words, which sharply contrasts with the relative stability of tonal realizations in Kaedi Soninke. For example, with the determined form of two syllable nouns that can be analyzed as having a HH lexical tone pattern, we observe a free variation between HL and HH  (HH followed by a downstep if the following word begins with a H tone). At this stage of our inquiry, we are not able to exhaustively classify the variations we have observed as free, conditioned by the phonological environment, or conditioned by syntactic structure, although we already have identified cases of free variation, phonological conditioning, and syntactic conditioning.

  • 4 The Soninke sentences we quote are followed by the initials of the consultants with whom they have (...)

23In particular, in addition to the L tones present in the underlying representations of morphemes, a L tone may be assigned to the first syllable of intonational phrases, if the following syllable is not underlyingly L. This explains for example the alternation in the tonal realization of ‘we’ illustrated in Ex. (1).4

(1)

a.

régé

wùró-n

mùumá. (AK)

1pl

dance

night-d

whole

‘We danced the whole night.’

b.

dà

kén

mùgù. (AK)

1pl

tr

dem

hear

‘We have heard that.’

24The question of the lexical tones of verbs is particularly relevant to our topic. In Kaedi Soninke, verbal lexemes divide into five tone classes (H, HL, HLH, LH, and LHL), and all verbs, whatever their inherent tone pattern, take a L contour in particular syntactic configurations – Creissels (1991b). By contrast, in Bakel Soninke, lexical tone contrasts cannot be observed in the quotation form of verbs (in which intransitive verbal lexemes are preceded by nàn, and transitive verbal lexemes are preceded by n’à < nà à, where is the third person singular pronoun). In our recorded data, we have observed three types of possible contours for verbs:

  • a H contour,

  • a HL contour,

  • a L contour observed so far almost exclusively in contexts in which it can equally be interpreted as the realization of an underlying H contour,

but in addition to a very strong predominance of the H contour, at least in the contexts that we have systematically checked, we observe variations between H and HL in the realization of individual verbs that we are not able to explain by a contextual conditioning (either phonological of grammatical). Additional inquiries with a variety of consultants will be necessary before deciding whether the variation is really free, and Bakel Soninke is engaged in a process of loss of lexical tone contrasts between verbs, or perhaps underlying contrasts are neutralized by processes that we have not been able to analyze yet.

25As regards nouns, lexical tone contrasts are unquestionable in quotation, but we have observed in our corpus H realizations of lexically LH nouns suggesting that lexical tone contrasts between nouns are neutralized in more contexts than those that have been recognized as triggering the neutralization of lexical tone contrasts between nouns in Kaedi Soninke. However, we are not in a position to provide a precise description of this phenomenon yet.

26Consequently, in this paper, the verbal lexemes are quoted without tone marks, since the evidence at our disposal does not allow us to posit lexical tones for verbs, and the tones of the nouns and sentences we quote are transcribed as they have been pronounced by our consultants when checking with them the data we present, without any interpretation or ‘regularization’ of the variations. In particular, in addition to the symbols for the H tone (ˊ), the L tone (ˋ), and the downstep (), we use the macron (ˉ) for syllables that are perceived higher than an immediately preceding L syllable but lower than an immediately following H syllable, because it would be premature to decide whether they should be analyzed as the realization of either H or L in the context L_H.

2.3. Clause structure

  • 5 S = subject, O = object, V = verb, X = oblique.

27In Soninke, as in other West Mande languages, verbal predication can be schematized as S (O) V (X).5 The rigid constituent order and the fixed position occupied by so-called predicative markers immediately after the subject are crucial for the recognition of grammatical relations, since there is neither flagging nor indexation of the core syntactic terms S and O. Predicative constructions with two or more terms encoded in the same way as the patient of typical monotransitive verbs (so-called ‘multiple object constructions’) are not possible.

28The predicative markers occupying a fixed position immediately after the subject include aspect-modality-polarity markers such as má ‘completive, negative’, and the locative copula wá (negative ntá ) fulfilling the function of incompletive auxiliary – Ex. (2). With the locative copula used as an incompletive auxiliary, the verb is in the gerundive form, marked by a suffix whose underlying form is ­|NV| (see Section 2.1.1), otherwise it occurs in its bare lexical form.

(2)

a.

Kè

yúgó

má

xàrà. (AK)

dem

man

cmp.neg

study

‘This man did not study.’

b.

mā

fáabá

tù. (AK)

3sg

cmp.neg

refl

father

recognize

‘He did not recognize his father.’

c.

wà

táaxú-nú

dàagó-n

kàmmà. (LD)

1sg

loccop

sit-ger

mat-d

on

‘He will sit on the mat.’

d.

wà

kē

lémúnú

kóróosì-nì

yàxàré-n

dà. (LD)

1sg

loccop

dem

child.pl

watch-ger

woman-d

for

‘He will watch these children for the woman.’

29As discussed by Creissels (To appear 2013) for Mandinka, the recognition of a paradigm of markers occurring immediately after the subject is crucial for the analysis of clauses with just one noun phrase preceding the verb, since it rules out the analysis according to which clauses such as those in (3a) and (3c), in which a bivalent verb is preceded by a unique noun phrase representing the patient-like participant, might have a transitive construction with a null subject.

(3)

a.

Sǐ-n

wá

wárí-ní

xùnbànè. (DB)

horse-d

loccop

see-ger

tomorrow

‘The horse will be seen tomorrow.’

b.

*Ø

Wá

sǐ-n

wárí-ní

xùnbànè.

loccop

horse-d

see-ger

tomorrow

c.

Yìllé-n

kòrì. (DB)

millet-d

sift

‘The millet has been sifted.’

d.

*Ø

Dà

yìllé-n

kòrì.

tr

millet-d

sift

30In this construction of potentially transitive verbs, as illustrated by the contrast between (3a) and the agrammatical sequence (3b), the markers that occur between the subject and the object in the transitive construction are found after the unique noun-phrase preceding the verb, not before it, as it should be the case if this noun phrase occupied the object position in a transitive construction with a null subject. Similarly, the contrast between (3c) and the agrammatical sequence (3d) shows that the transitivity marker dà does not occur in the construction of kori ‘sift’ with just one noun phrase representing the patient, whereas it should be found to the left of the patient phrase if this were a transitive construction with a null subject.

  • 6 This issue is discussed by Cobbinah and Lüpke (2009), who provide a survey of languages with constr (...)

31The analysis of the unique noun phrase preceding the verb in such constructions as a subject, rather than an object in a null-subject construction, is further confirmed by the fact that it does not select the variant ya of the focalization marker, used to focalize phrases in functions other than subject, and can only be focalized by means of the variant yan, used exclusively to focalize subjects. The only possible analysis of clauses in which a potentially transitive verb is preceded by a sole noun phrase representing the patient is that, syntactically, they include no object phrase, and the noun phrase preceding the verb fulfills the subject role, which implies considering them as instances of a morphologically unmarked passive construction. In other words, the valency properties of Soninke verbs such as kori ‘sift’, bayi ‘spread’, dabari ‘repair’ or wari ‘see’ must be analyzed in terms of active / passive lability, a phenomenon common among Mande languages, but relatively rare cross-linguistically, at least in its fully grammaticalized form (that is, without the restrictions and/or aspecto-modal nuances that characterize the use of zero-coded quasi-passives such as English This book sells well).6 However, in Soninke, this phenomenon can be observed almost exclusively with verbs ending with i. An explanation is given further.

32A striking feature of Soninke is the particularly clear-cut distinction between transitive and intransitive predications, due not only to the rigid S (O) V (X) pattern, which excludes ambiguity between the syntactic roles of object and oblique, but also to the fact that three of the grammatical elements occupying a position immediately after the subject are sensitive to the transitive vs. intransitive distinction:

    • 7 This dà is sometimes labeled ‘completive positive marker’, but this label is hardly compatible wit (...)

    in the completive positive, a morpheme dà analyzed here as a transitivity marker is obligatorily found in transitive constructions, but does not occur in the corresponding intransitive constructions – Ex. (4), and this dà also occurs with the same distribution in the imperative plural;7

    • 8 The form labeled here ‘subjunctive’ combines with noun phrases in subject function in uses broadly (...)

    the subjunctive positive is marked by nà in transitive constructions and nàn in intransitive constructions – Ex. (5);8

  • in clauses including a focalized term, the locative copula wá used as a incompletive marker has two variants depending on the transitivity of the construction: Ø in intransitive constructions, and nà (homonymous with the subjunctive positive marker) in transitive constructions – ex. (6).

(4)

a.

Fàŋŋē

ké

káawá

fànè

yìrìgì. (AK)

river

dem

dry_up

early

this_year

‘The river dried up early this year.’

b.

Yàxāré-n

dà

tíyè-n

qóbó

sáxà-n

wà. (AK)

woman-d

tr

meat-d

buy

market-d

obl

‘The woman bought meat at the market.’

c.

Xà

dà

lémíné-n

pàabà! (DB)

imper.pl

tr

child-d

give_assitance

‘Give the child assistance!’

(5)

a.

nàn

táaxù

yítté-n

wùrè. (LD)

3sg

subj.pos.intr

sit

tree-d

under

‘He should sit under the tree.’

b.

Lèmúnú

kú

nà

tíyè-n

yìgà. (AK)

child.pl

dem

subj.pos.tr

meat-d

eat

‘The children should eat meat.’

(6)

a.

Ø

sállí-ní

yà. (LD)

3sg

loccopf.intr

pray-ger

foc

‘He is PRAYING.’

b.

n’

gáagá-ná

yà. (LD)

3sg

loccopf.tr

3sg

sell-ger

foc

‘He is SELLING it.’

3. Valency-changing derivations

3.1. Introductory remarks

33Bakel Soninke has two morphological devices encoding a valency-decreasing operation, and one encoding a valency-increasing operation.

34The detransitivizing marker ­­‑i (glossed DETR) may express a variety of valency-decreasing operations, depending on the verbs it combines with, whereas antipassive ­‑ndi (glossed ANTIP) unambiguously expresses patient demotion. In both cases, the demoted argument is normally not expressed, although expressing it as an oblique is not completely impossible.

35The only valency-increasing operation that can be encoded morphologically is causativization, expressed by the causative suffix ‑ndi.

36Creissels (1991a) argues that in Kaedi Soninke, causative ‑ndi and antipassive ‑ndi have different tonal properties, and can be analyzed as underlyingly ‑ndí (causative) vs. ‑ndì (antipassive). Unfortunately, we cannot confirm this analysis for Bakel Soninke, because of the doubts we have about the possibility of identifying tone contrasts between verb stems in this Soninke variety.

3.2. The detransitivizing marker ‑i

3.2.1. Formal properties of the detransitivizing marker ‑i

37Most verbs that have a transitive stem ending with a, o, or u also have an intransitive stem that can be analyzed as derived from the transitive stem by the addition of a detransitivizing marker whose underlying form is |i|. However, this detransitivizing marker surfaces as a distinct segment ‑yi with monosyllabic stems only:

ka

‘insult’

kayi (intr.) ‘insult unspecified people’

ku

‘give’

kuyi (intr.) ‘give unspecified things’

tu

‘know’

tuyi (intr.) ‘know unspecified things’

38With non-monosyllabic stems, the presence of detransitivizing ‑i is manifested by a change in the last vowel of the stem that can be explained as the result of the amalgamation of an underlying i according to the following rules:

a + i → e (sometimes also i)

o + i → e

u + i → i

39It is therefore possible to consider that the lack of distinct detransitivized forms with verbs ending with e or i follows from the fact that the morphophonological process manifesting the presence of ‑i would apply vacuously to such stems.

3.2.2. Semantic properties of the detransitivizing marker ‑i

40Depending on the individual verbs with which it combines, ‑i may express various detransitivizing operations, but it is not equally productive in all its possible uses.

41Agent demotion is by far the most productive use of the detransitivizing marker ­‑i, with two semantic subtypes, anticausative (the agent is suppressed from argument structure, and the event is presented as occurring spontaneously, as in Ex. (7b)) and passive (the agent is semantically maintained, but it is not expressed, and the subject role is fulfilled by the patient, as in Ex. (8b)).

(7)

a.

Yúgó-n

dà

wùlù-séyínté-n

yòolà. (DB)

man-d

tr

dog-rabid-d

drown

‘The man drowned the rabid dog.’

b.

Lémíné-n

yóolé

fàŋŋé-n

wà. (DB)

child-d

drown.detr

river-d

obl

‘The child drowned in the river.’

(8)

a.

Yàxàré-n

dà

yíllé-n

gòrò. (AK)

woman-d

tr

millet-d

pound

‘The woman pounded the millet.’

b.

Yìllé-n

gòrè. (AK)

millet-d

pound.detr

‘The millet was pounded.’

42The distinction between these two semantic varieties of deagentive derivation (agent-backgrounding and agent-suppressing) is not rigid. With many verbs, both readings are equally available, depending on the context. What seems to be crucial is the semantic distinction between processes easily conceived as occurring spontaneously (such as ‘drown’) and processes that require the intervention of an agent (such as ‘become pounded’).

  • 9 Soninke has two pronouns used productively to express reflexivity: is a long-distance reflexive (...)

43With a few verbs among those that have the ability to combine with the detransitivizing marker ‑i in deagentive function, the same form also has a reflexive or autocausative use:9

boora

‘undress (tr.)’

→ boore

‘undress oneself’ – Ex. (9)

fubba

‘plunge’

→ fubbe

‘dive’

kafu

‘gather (tr.)

→ kafi

‘gather (intr.)’

fuutu

‘stretch (tr.)’

→ fuuti

‘stretch (intr.)

(9)

a.

Yúgó-n

d’

rémmé-n

bòorà. (DB)

man-d

tr

refl

son/daughter-d

undress

‘The man undressed his son.’

b.

Yúgó-n

bòorè. (DB)

man-d

undress.detr

‘The man undressed.’

44The detransitivizing marker ‑i may also have a depatientive function, for which it is in competition with the dedicated antipassive suffix ‑ndi. All the intransitive stems derived by means of ‑i that can be used in depatientive function also have a deagentive (anticausative or passive) use, as illustrated by yige, intransitive form of yiga ‘eat’ – Ex.  10).

(10)

a.

Lèmúnú

kú

dà

tíyè-n

yìgà. (AK)

child.pl

dem.pl

tr

meat-d

eat

‘The children ate the meat.’

b.

Lèmúnú

kú

yìgè. (AK)

child.pl

dem.pl

eat.detr

‘The children ate.’

c.

Tíyè-n

yìgè. (AK)

meat-d

eat.detr

‘The meat was eaten.’

3.3. The antipassive suffix -ndi

45The depatientive function is the only possible function of the antipassive suffix ‑ndi–, Ex. (11), and this suffix is very productive.

(11)

a.

Sámáqqè-n

dà

lémíné-n

qìñì. (AK)

snake-d

tr

child-d

bite

‘The snake bit the child.’

b.

Sámáqqè-n

qíñí-ndì. (AK)

snake-d

bite-antip

‘The snake bit (someone).’

46Bakel Soninke has a handful of transitive verbs that can be used intransitively in their underived form with a subject representing the agent (see Section 4), and a small set of transitive verbs with which the detransitivizing marker ‑i can be used in depatientive function (see Section 3.2), but all transitive verbs that do not belong to one of these two subsets combine with the antipassive marker ‑ndi.

47The antipassive suffix ‑ndi has an allomorph ‑yindi with monosyllabic stems:

ka

‘insult’

→ (antip.) ka-yindi

si

‘shave’

→ (antip.) si-yindi

3.4. The causative suffix ‑ndi

48Causativization, illustrated by Ex. (12), is the only valency-increasing morphological derivation found in Soninke.

(12)

a.

Sǐ-n

bàamì. (DB)

horse-d

gallop

‘The horse galloped.’

b.

Yúgó-n

d’

sí-n

bàamì-ndì. (DB)

man-d

tr

refl

horse-d

gallop-caus

‘The man made his horse gallop.’

49Causative verbs are derived by means of a suffix ‑ndi which has been claimed to be tonally different from antipassive ‑ndi in Kaedi Soninke – Creissels (1991a), but seems to be fully homonymous with it in the Bakel variety.

50As illustrated by Ex. (12) above, causativization by means of the causative suffix ‑ndi is fully productive for strict intransitive verbs (i.e., for verbs that cannot be used transitively in their underived form). By contrast, morphological causativization is only marginally available with a transitive input. Most of the time, consultants asked to causativize transitive clauses or the depatientive version of transitive clauses reject the use of ‑ndi and propose analytical causatives instead.

  • 10 On causativization in typological perspective, see Creissels (2006: chapter 24) and references ther (...)

51There are however a few transitive verbs for which morphological causativization is accepted by our consultants: dabari ‘repair’, faamu ‘understand’, faccari ‘translate’, goro ‘pound’, kara ‘cross’, kari ‘kill’, lifi ‘sew’, mini ‘drink’, saara ‘give birth’, soro ‘cook’, xara ‘learn’, and yiga ‘eat’. No clear semantic generalization emerges from this list, which includes not only transitive verbs belonging to semantic classes known for lending themselves to causativization even in languages that tend to restrict causativization to intransitive verbs (such as ‘eat’ and ‘drink’), but also transitive verbs belonging to semantic classes for which no such tendency is observed cross-linguistically (for example ‘repair’, or ‘pound’).10

52As illustrated by Ex. (13), the object of causative verbs derived from transitive verbs may correspond semantically either to the subject or the object of the transitive verb from which they derive, but if both are expressed, as in (13c), the object of the initial construction is maintained as the object of the causative verb.

(13)

a.

Lémíné-n

dà

tíyè-n

yìgà. (AK)

child-d

tr

meat-d

eat

‘The child ate meat.’

b.

Fàatú

dà

lémínè-n

yìgà-ndì. (DB)

Fatou

tr

child-d

eat-caus

‘Fatou made the child eat.’

c.

Fàatú

dà

tíyè-n

yígá-ndí

lémínè-n

wà. (DB)

Fatou

tr

meat-d

eat-caus

child-d

obl

‘Fatou made the child eat meat.’

53Morphologically, a few verbal lexemes ending with e, such as bire ‘live’, show an irregular change in their last vowel when combined with the causative suffix (birebira-ndi), which may suggest that bire was originally decomposable as *bira + ‑i.

54An irregular causative form is also observed with bogu ‘go/come out’, caus. baga‑ndi, and xara ‘learn’, caus. xaraŋŋundi.

3.5. Antipassivization of causative verbs

55Although the causativization of derived intransitive forms encoding patient demotion is perfectly conceivable semantically (‘a causer makes a causee act on an unspecified patient’), it does not seem possible to express it in Bakel Soninke by means of the concatenation ‑ndi ‘antipassive marker’ + ‑ndi ‘causative marker’, or -i ‘detransitivization marker’ + ‑ndi ‘causative marker’.

56For example, with a verb like yiga ‘eat’, for which the demotion of the patient triggers the use of the derived intransitive form yige, one could imagine the existence of a causative form *yige-ndi expressing ‘make someone eat some unspecified food’, but this form does not exist, and Ex. (13b) above shows that yiga-ndi ‘make eat’ is used to causativize not only constructions in which the patient is expressed, but also constructions with a demoted patient. Similarly, faamu ‘understand’ and soro ‘cook’ occur as faami and sore in intransitive constructions expressing patient demotion, but can only be used causatively as faamu-ndi and soro‑ndi, even if the initial object is not expressed.

57By contrast, derived verb stems with an ending ‑ndindi decomposable as ‑ndi ‘causative’ + ‑ndi ‘antipassive’ are perfectly possible, and the antipassive marker operates on causative verbs in the same way as on non-derived transitive verbs: the meaning of such forms is that a causer manipulates an unspecified causee – Ex. (14).

(14)

a.

Nàa-ní-n

dà

té-n

bònò-ndì. (DB)

cow-pl-d

tr

field-d

become_spoilt-caus

‘The cows caused damage to the field.’

(lit. ‘The cows made the field become spoilt.’)

b.

Nàa-ní-n

bònò-ndì-ndì. (DB)

cow-pl-d

become_spoilt-caus-antip

‘The cows caused damage.’

(lit. ‘The cows made something unspecified become spoilt.’)

3.6. ‘Die’ and ‘kill’

58Cross-linguistically, ‘die’ and ‘kill’ may be expressed by two unrelated verbs, by a single labile verb, or by two verbs related by a valency-changing operation (causative or anticausative).

59In Soninke, the formal resemblance between kara ‘die’ and kari ‘kill’ suggests the existence of a relationship, but kari cannot be analyzed as the causative form or kara ‘die’, since this causative form could only be *kara‑ndi, and kara cannot be analyzed as the anticausative form of kari, since verbs ending with i cannot have a morphologically distinct anticausative form.

60Consequently, in spite of their formal resemblance, kara ‘die’ and kari ‘kill’ must be viewed as two unrelated lexemes in a synchronic analysis of Soninke.

4. Transitivity in the lexicon

4.1. Labile vs. non-labile verbs

61Labile verbs are verbs that can be used in their underived form either transitively or intransitively. The verbs that can be used intransitively with a subject representing the same agent-like participant as the subject of the same verb used transitively are designated as A-labile, and those that can be used intransitively with a subject representing the same patient-like participant as the subject of the same verb used transitively are designated as P-labile. A/P labile verbs (Section 4.7) can be used intransitively with a subject that may correspond to either of the core arguments of the same verb used transitively, and some labile verbs are characterized by a reflexive reading of their intransitive use (Section 4.6).

62As discussed in Creissels (2009), in languages in which the presence or absence of a noun phrase showing the same formal characteristics as the noun phrase representing the patient of typical action verbs is the only formal manifestation of the distinction between transitive and intransitive predication (as in most European languages), the recognition of a class of A-labile verbs (i.e., of verbs used intransitively with a subject) may be problematic, since in such languages, it is not always easy to find criteria distinguishing true intransitive predications from transitive predications in which an unexpressed P is either interpreted as non-specific or identified to some discursively or situationally salient entity. Symmetrically, in languages in which the presence or absence of a noun phrase in A role is the only formal manifestation of the distinction between transitive and intransitive predication (as in Avar and many other Nakh-Daghestanian languages), the recognition of a class of P-labile verbs may be problematic, since in such languages, it is not always easy to find criteria distinguishing true intransitive predications from transitive predications in which an unexpressed A is either interpreted as non-specific or identified to some discursively or situationally salient entity. By contrast, in Mande languages in general, and more particularly in Soninke, due to the proliferation of transitivity marking in this language, it is equally unproblematic to distinguish A-labile and P-labile verbs from strict transitive verbs and strict intransitive verbs:

  • strict transitive verbs in the completive positive or in the imperative plural obligatorily combine with the transitive marker dà, they obligatorily select the nà variant of the subjunctive positive marker, and in constructions including a focalized term, they obligatorily trigger the choice of the nà variant of the locative copula;

  • strict intransitive verbs in the completive positive or in the imperative plural cannot combine with the transitive marker dà, they obligatorily select the nàn variant of the subjunctive positive marker, and in constructions including a focalized term, they obligatorily trigger the choice of the Ø variant of the locative copula;

  • labile verbs in the completive positive or in the imperative plural can be found with or without the transitivity marker dà, they are compatible with both variants (nà and nàn) of the subjunctive positive marker, and they are compatible with both variants (nà and Ø) of the locative copula; the verbs identified as labile by this definition may have an intransitive construction with an agent-like subject, as illustrated by soxo ‘cultivate’ in Ex. (15), or with a patient-like subject, as illustrated by bayi ‘spread’ in Ex. (16)).

(15)

a.

Bàdárà

d’

té-n

còxò. (DB)

Badara

tr

refl

field-d

cultivate

‘Badara has cultivated his field.’

b.

Bàdárà

sòxò. (DB)

Badara

cultivate

‘Badara has cultivated.’

(16)

a.

Yúgó-n

dà

dàagó-n

bàyì. (DB)

man-d

tr

mat-d

spread

‘The man has spread the mat.’

b.

Dàagó-n

bàyì. (DB)

mat-d

spread

‘The mat has been spread.’

63These criteria are particularly useful with compound verbs incorporating a noun fulfilling the same semantic role as the object in the transitive construction, as in Ex. (17b), since otherwise, the incorporated noun could easily be confused with a noun phrase in object role. Note however that, as illustrated by sókkè ‘grass’ in this example, incorporated nouns occur in a form (here: sókkí‑) attested only in combination with adjectives or derivative suffixes, or in compounding. As can be seen from (17b-c), the incorporation of this special form of a noun fulfilling the same semantic role as a noun phrase in the role of object of the transitive construction correlates with the impossibility of introducing the transitive marker dà.

(17)

a.

dà

sókkè-n

dúqqútà. (AK)

3sg

tr

weeds-d

pull_up

‘He pulled up the weeds.’

b.

sókkí-dúqqútà. (AK)

3sg

weeds-pull_up

‘He did the weeding.’

c.

*

dà

sókkí-dúqqútà.

3sg

tr

weeds-pull_up

64In applying the transitivity criteria to individual verbs in order to classify them as strict transitive, strict intransitive, or labile, the only problem is that polysemous verbs may have different transitivity properties in their different meanings. Consequently, they cannot be straightforwardly classified as labile or non-labile; rather, the different (although related) meanings of such verbs must be classified separately – see Section 4.8.

4.2. Strict intransitive verbs

65Verbs such as baami ‘gallop’, bire ‘live’, bono ‘become spoilt’, etc., can only be used intransitively, and must take the causative suffix (see Section 3.4) in order to be used transitively with a causative meaning.

4.3 Strict transitive verbs

66Strict transitive verbs in their non-derived form can only be used transitively, and a derived form must be used in intransitive constructions with a subject corresponding either to the subject or the object of the transitive construction. Two subtypes of strict transitive verbs can be distinguished, those with a single intransitive form, and those with two distinct intransitive forms.

4.3.1. Strict transitive verbs with two distinct intransitive forms

67Most strict transitive verbs have the behavior illustrated by faaba ‘help, rescue’ in Ex. (18), with two derived intransitive forms: an intransitive form whose subject corresponds semantically to the object of the transitive construction, derived by means of the detransitivizing marker ‑i (see Section 3.2), and an intransitive form whose subject corresponds semantically to the subject of the transitive construction, derived by means of the antipassive suffix ‑ndi (see Section 3.3). Depending on the lexical meaning of the verb, the intransitive form derived by means of ‑i may have an anticausative or passive reading.

(18)

a.

Xà

dà

lémíné-n

pàabà! (DB)

imper.pl

tr

child-d

give_assitance

‘Give the child assistance!’

b.

Lémíné

ké

nàn

pàabè

kárámà! (DB)

child

dem

subj.pos.intr

give_assistance.detr

quickly

‘This child should be given assistance quickly!

c.

Tángáanú-n

pàabà-ndì

kárámà. (DB)

fisherman.pl-d

give_assistance-antip

quickly

‘The fishermen provided assistance quickly.’

68The following verbs can be classified as strict transitive verbs with two distinct intransitive forms:

banbu / banbi / banbundi

‘carry on the back’

boora / boore / boorandi

‘undress’

buruta / burute / burutandi

‘reduce the quantity of food’

buruxa / buruxe / buruxandi

‘erase, exterminate’

buuda / buude / buudundi

‘overcook’

cucufu / cucufi / cucufindi

‘sprinkle’

danca / dance / dancandi

‘provide with an additional gift’

danpu / danpi / danpundi

‘trample’

duqquta / duqqute / duqqutandi

‘pull up’

fata / fate / fatandi

‘separate, harvest, wean’

fubba / fubbe / fubbandi

‘plunge’

garaba / garabe / garabandi

‘split, crack’

gunu/ guni / gunundi

‘fold’

jaga / jage / jagandi

‘dig’

juura / juure / juurandi

‘pray on a grave’

kafu / kafi / kafundi

‘add, gather’

kaga / kage / kagandi

‘scratch’

kara / kare / karandi

‘break’

katu / kati /katundi

‘hit’

kippa / kippe / kippandi

‘turn over’

langa / lange / langandi

‘curse’

mugu / mugi / mugundi

‘hear’

nafa / nafe / nafandi

‘do a service’

raga / rage / ragandi

‘seize’

saada / saade / saadandi

‘heap up’

safa / safe / safandi

‘write’

saga / sage / sagandi

‘clear out (ground)’

sagara / sagare / sagarandi

‘pick up’

sagata / sagate / sagatandi

‘join’

sella / selle / sellandi

‘sweep’

soxo / soxe / soxondi

‘tattoo, embroider’

taga / tage / tagandi

‘build’

tulu / tuli / tulundi

‘plait (hair)’

waara / waare / waarandi

‘open wide’

wara / ware / warandi

‘release’

wundu / wundi / wundundi

‘waken’

wutu / wuti / wutundi

‘take’

xamu / xami / xamundi

‘milk’

xoso / xose / xosondi

‘break, defeat’

yiriga / yitige / yirigandi

‘count’

yonto / yonte / yontondi

‘push’

yoola / yoole / yoolandi

‘drown’, etc.

4.3.2. Strict transitive verbs with a single intransitive form

69The verbs that have the behavior illustrated by soro ‘cook’ in Ex. (19) constitute a minor subclass of strict transitive verbs. They have a single intransitive form, derived by means of the detransitivizing marker ‑i (see Section 3.2), whose subject may correspond semantically either to the subject or the object of the transitive construction.

(19)

a.

Àmí

dà

tíyé-n

yà

sòrò

lènkì. (DB)

Amy

tr

meat-d

foc

cook

today

‘It is meat that Amy has cooked today.’

b.

Tíyé-n

còrè. (DB)

meat-d

cook.detr

‘The meat has been cooked.’

c.

Àmí

sóré

fànè. (DB)

Amy

cook.detr

early

‘Amy did the cooking early.’

70This behavior is shared by the following verbs:

faamu / faami / faami

‘understand’

gaaga / gaage / gaage

‘sell’

joppa / joppe / joppe

‘begin’

kuu / kuyi / kuyi

‘give’

saara / saare / saare

‘give birth’

xobo / xobe / xobe

‘buy’

yaara / yaare / yaare

‘glean’

yiga / yige/ yige

‘eat’

71We also have recorded goro / gore / gore ‘pound’ with one of our consultants, but there seem to be variations in the intransitive forms of this verb.

4.4. A-labile verbs

72A-labile verbs are verbs that can be used in their underived form either transitively, or intransitively with a subject representing the same agent-like participant as the subject of the same verb used transitively, but must undergo a detransitivizing derivation in order to be used intransitively with a subject representing the same patient-like participant as the object of the transitive construction. This behavior can be illustrated by soxo ‘cultivate’ – Ex. (20).

(20)

a.

Bàdárà

d’

té-n

còxò. (DB)

Badara

tr

refl

field-d

cultivate

‘Badara has cultivated his field.’

b.

Bàdáràc

sòxò. (DB)

Badara

cultivate

‘Badara has cultivated.’

c.

Tě-n

còxè. (DB)

field-d

cultivate.detr

‘The field has been cultivated.’

73Xara ‘learn’ is the only other verb in our corpus showing this behavior.

4.5. P-labile verbs

74P-labile verbs are verbs that can be used in their underived form either transitively, or intransitively with a subject representing the same patient-like participant as the object of the same verb used transitively, but must undergo a detransitivizing derivation in order to be used intransitively with a subject representing the same agent-like participant as the subject of the transitive construction. In all cases, the intransitive form is derived by means of the antipassive suffix ‑ndi (see 3.3).

75This behavior can been illustrated by bayi ‘spread’ – Ex. (21).

(21)

a.

Yúgó-n

dà

dàagó-n

bàyì. (DB)

man-d

tr

mat-d

spread

‘The man has spread the mat.’

b.

Dàagó-n

bàyì. (DB)

mat-d

spread

‘The mat has been spread.’

c.

Yúgó-n

báyí-ndí

kànbérà. (DB)

man-d

spread-antip

yard

‘The man has spread things in the yard.’

76Semantically, two varieties of P-lability can be distinguished: causative / anticausative lability, if the subject of the intransitive construction represents a participant undergoing the same process as the object of the transitive construction, but not necessarily as the result of the action of an agent, and active / passive lability, if the intransitive construction implies the participation of an unexpressed agent triggering the process undergone by the referent of the subject.

77Cross-linguistically, causative / anticausative lability is extremely common. By contrast, as already mentionned in Section 2.3, active / passive lability is relatively rare, at least in its most grammaticalized form, but Mande languages constitute an exception to this generalization. Mandinka illustrates the extreme case of a language with no strict transitive verb, a very restricted class of A-labile verbs, and limited possibilities of causative / anticausative lability, but in which all the verbs that have a transitive use can be used intransitively in their underived form, without any restriction related to TAM or to semantic features of the participants, with a subject interpreted as undergoing the action of an unexpressed participant that would be encoded as the subject in the transitive construction of the same verb – Creissels (To appear 2013).

78Active / passive lability is found in Soninke too, but in Soninke, contrary to Mandinka, both subtypes of P-lability are restricted to a subset of the verbs that can be used transitively. It is striking that the vast majority of P-labile verbs end with i, and conversely, all the verbs in our corpus that end with ‑i and can be used transitively are P-labile, which raises the question whether this is really P-lability, or perhaps rather vacuous detransitivization, since Soninke has a detransitivizing suffix ‑i. This point will be discussed further in Section 4.9.

79There is no rigid distinction between P-labile verbs with an agent-triggered process reading or a spontaneous process reading of their intransitive construction. Rather, the choice seems to depend on the semantic distinction between processes easily conceived as occurring spontaneously (such as ‘burn’) and processes that require the intervention of an agent (such as ‘shave’).

80Our corpus includes the following P-labile verbs:

anniya / anniyandi

‘make a wish while performing a sacrifice’

araayi / araayindi

‘advise’

bayi / bayindi

‘spread’

betexi / betexindi

‘soil’

boosi / boosindi

‘take out’

bosi / bosindi

‘suck’

booti / bootindi

‘extract’

buyi / buyindi

‘burn’

daabi / daabindi

‘massage’

daari / daarindi

‘coat’

dabari / dabarindi

‘repair’

danbi / danbindi

‘praise’

danpi / danpindi

‘fold’

daro / darondi

‘honor’

faccari / faccarindi

‘translate’

fesi / fesindi

‘hang out’

firi / firindi

‘precede, outdistance’

ganci / gancindi

‘announce’

guuji / guujindi

‘pour’

jaabi / jaabindi

‘answer’

jaarabi / jaarabindi

‘tempt’

jurumi / jurumindi

‘heap up’

kari / karindi

‘kill’

kori / korindi

‘sift’

koroosi / koroosindi

‘watch’

korosi / korosindi

‘shell’

lifi / lifindi

‘sew’

loggi / loggindi

‘hang’

lori / lorindi

‘transplant’

nooni / noonindi

‘delimit’

ñaagari / ñaagarindi

‘decorate’

sanqi / sanqindi

‘scatter’

sembe / sembendi

‘pass on’

sembe / sembendi

‘lean’

sii / siyindi

‘shave’

somi / somindi

‘aggravate (wound)’

wanqi / wanqindi

‘wash’

wuñi / wuñindi

‘open’

xiyili / xiyilindi

‘drag’

4.6. Reflexive lability

81Wanqi ‘wash’, already mentioned in Section 4.5 as a P-labile verb, is the only verb in our corpus that can also be used intransitively with a reflexive meaning in its underived form – Ex. (22).

(22)

a.

Yàxàré-n

dà

yíráamú-n

wànqì. (DB)

woman-d

tr

cloth.pl-d

wash

‘The woman washed the clothes.’

b.

Yàxàré-n

wànqì-ndì. (DB)

woman-d

wash-antip

‘The woman did the washing.’

c.

Yìràamú-n

wànqì. (DB)

cloth.pl-d

wash

‘The clothes were washed.’

d.

Yàxàré-n

wànqì. (DB)

woman-d

wash

‘The woman washed (herself).’

4.7. A/P-labile verbs

82A/P labile verbs are verbs that can be used in their underived form transitively, intransitively with a subject corresponding semantically to the subject of the transitive construction, and intransitively with a subject corresponding semantically to the object of the transitive construction. We have found three verbs with this behavior: mini ‘drink’, salli ‘pray’, and wari ‘see’ – Ex. (23).

(23)

a.

Yúgó-n

dà

sǐ-n

wàrì. (DB)

man-d

tr

horse-d

see

‘The man saw the horse.’

b.

Sǐ-n

wàrì. (DB)

horse-d

see

‘The horse was seen.’

c.

wá

wàrì-ni. (DB)

3sg

loccop

see-ger

‘He is clairvoyant.’ (lit. ‘He sees’)

4.8. Lability and polysemy

83Soninke also has verbs which in their non-derived form can be used intransitively or transitively, but also have a causative form. However, as illustrated in Ex. (24) by bogu (with the irregular causative form baga‑ndi), at least within the limits of the data we have been able to gather, such verbs are polysemous verbs whose transitivity properties depend on the precise meaning they express. Consequently, they do not really contradict the classification put forward in this section, and do not really constitute a separate type. For example, there is an obvious semantic link between bogu ‘go out’ and bogu ‘distinguish’, but bogu ‘go out’ has the behavior of an intransitive verb from which a causative form baga-ndi ‘take out’ can be derived, whereas bogu ‘distinguish’ is transitive.

(24)

a.

Yúgó-n

bògù. (DB)

man-d

go_out

‘The man went out.’

b.

Yúgó-n

dà

làbó-n

bàgà-ndì. (DB)

man-d

tr

knife-d

go_out-caus

‘The man took the knife out.’

c.

Yúgó-n

dà

lémínè-n

bógú

sòró-n

wà. (DB)

man-d

tr

child-d

distinguish

person.pl-d

obl

‘The man distinguished the child from the crowd.’

4.9. Summary: deagentive and depatientive uses of transitive verbs

84The verbs that have the ability to be used transitively in their underived form (i.e., that are not strict intransitive verbs) can be divided into sub-classes according to their ability to be used intransitively with a subject representing either the agent-like or the patient-like participant.

85The deagentive use of transitive verbs may be morphologically unmarked, or marked by the addition of the detransitivizing marker ‑i, whereas the depatientive use may be morphologically unmarked, marked by the addition of the detransitivization marker ‑i, or marked by the addition of the antipassive marker ‑ndi, which gives six logical possibilities:

deagentive

depatientive

(a)

unmarked

unmarked

(b)

unmarked

marked by ‑i

(c)

unmarked

marked by ‑ndi

(d)

marked by ‑i

unmarked

(e)

marked by ‑i

marked by ‑i

(f)

marked by ‑i

marked by ‑ndi

86The only unattested pattern is (d), with the depatientive use morphologically unmarked and the deagentive use morphologically marked. However, the remaining five patterns are not evenly distributed. Three of them are found with a limited number of verbal lexemes:

  • pattern (a), illustrated above by wari ‘see’ – ex. (23), in which both the deagentive and the depatientive uses are morphologically unmarked, is very rare (three verbs in our corpus) ;

  • pattern (b), illustrated above by soxo ‘cultivate’ – ex. (20), in which the depatientive use is unmarked, and the deagentive use is marked by the detransitivization marker ‑i, is also very rare (two verbs in our corpus) ;

  • pattern (e), illustrated above by yiga ‘eat’ – ex. (10) – and soro ‘cook’ – Ex. (19), in which both the deagentive and depatientive uses are marked by the detransitivization marker ‑i, is less rare than patterns (a) and (b), but the number of verbs in our corpus that attest this pattern does not exceed ten.

87The two productive patterns are:

  • pattern (c), illustrated above by bayi ‘spread’ – ex. (21), in which the deagentive use is unmarked, and the depatientive use is marked by the dedicated antipassive marker –ndi ;

  • pattern (f), illustrated above by faaba ‘rescue’ – ex. (18), in which the deagentive use is marked by the detransitivization marker ‑i, and the depatientive use is marked by the dedicated antipassive marker –ndi ;

88The two productive patterns have in common the use of a dedicated antipassive form in depatientive function, they differ in that the deagentive use is morphologically marked in pattern (f), but not in pattern (c). A possible conclusion is therefore that the tendency to overtly mark the depatientive use of Soninke verbs by means of a dedicated antipassive marker is much stronger than the tendency to use a distinct form in deagentive function.

89There is however another possible analysis of this situation, based on the observation that the two productive patterns are in nearly complementary distribution: more than 90% of the verbs following pattern (c) (with the same form in transitive and deagentive use) end with i, whereas there is no verb ending with i among those following pattern (f).

90Consequently, it can be argued that, underlyingly, Soninke has an equally strong tendency to overtly mark the deagentive and the depatientive uses of transitive verbs, but in the case of the deagentive use, which can only be marked by means of the detransitivizing marker ‑i, the manifestation of this tendency is limited by phonological factors.

91According to this analysis, there is just one productive pattern, the one in which the deagentive use is marked by ‑i and the depatientive use is marked by ‑ndi, and the apparent productivity of the pattern with the deagentive use unmarked and the depatientive use marked by ‑ndi is simply due to the fact that the phonological process distinguishing bare verb stems from verb stems modified by the detransitivization marker applies vacuously to stems ending with i or e.

5. Soninke as a transitivizing language

92In this section, we analyze the competition between morphological causativization and anticausativization in the organization of the verbal lexicon of Soninke. We use the sample of 18 verb pairs proposed by Nichols & al. (2004) to evaluate the preference languages may have for transitivization or detransitivization in the organization of the lexicon, in order to characterize the situation of Soninke in terms that will enhance comparability with that of other languages.

93Each pair consists of a ‘plain’ member and an ‘induced’ member, and the induced member differs semantically from the plain one by the addition of a participant in the semantic role of causer. In the first nine pairs (the ‘animate’ pairs), the plain verb typically expresses a predication about a human or animate, whereas in the following nine pairs (the ‘inanimate’ pairs), the plain verb typically expresses a predication about an inanimate.

94Here is the list of the eighteen verb pairs with their Soninke equivalents:

Plain

Induced

1

‘laugh’

‘make laugh, amuse’

soyi

soyindi

2

‘die’

‘kill’

kara

kari

3

‘sit’

‘make sit, seat’

taaxu

taaxundi

4

‘eat’

‘feed, give food’

yiga

yigandi

5

‘learn’

‘teach’

xara

xaraŋŋundi

6

‘see’

‘show’

wari

koyi

7

‘be/become angry’

‘anger, make angry’

butu

butundi

8

‘fear, be afraid’

‘frighten, scare’

kanu

kanundi

9

‘hide, go into hiding’

‘hide, put into hiding’

muxu

muxundi

10

‘(come to) boil’

‘(bring to) boil’

wari

warindi

11

‘burn, catch fire’

‘burn, set fire’

buyi

buyi

12

‘break (intr.)’

‘break (tr.)’

kare

kara

13

‘open (intr.)’

‘open (tr.)’

wuñi

wuñi

14

‘dry’

‘make dry’

kaawa

kaawandi

15

‘be/become straight’

‘straighten, make straight’

teleŋo

teleŋondi

16

‘hang (intr.)’

‘hang (tr.)’

loggi

loggi

17

‘turn over (intr.)’

‘turn over (tr.)’

kippe

kippa

18

‘fall’

‘drop, let fall’

xenu

xenundi

95The Soninke verb pairs divide into four of the nine types of correspondences considered in Nichols & al. (2004):

  • ‘augmented’ (the induced verb is formally derived), attested by eleven verb pairs: 1, 3, 4, 5, 7, 8, 9, 10, 14, 15, 18;

  • ‘reduced’ (the plain verb is formally derived), attested by two verb pairs: 12, 17;

    • 11 ‘Ambitransitive’ as used by Nichols & al. (2004) is synonymous with P-labile as used in this articl (...)

    ‘ambitransitive’ (no formal derivation), attested by three verb pairs: 11, 13, 16;11

    • 12 To be precise, this is a case of ‘partial suppletion’, since the two verbs seem to be cognate, but (...)

    ‘suppletion’ (the two members of the pair have different roots), attested by two pairs, 2,12 6.

96The augmented correspondences clearly predominate (11 verb pairs out of 18), and Soninke can therefore be characterized as a predominantly transitivizing language. However, the four types of correspondences attested in Soninke are not evenly distributed between animate and inanimate verb pairs, as can be seen from the following chart, in which each pair is represented by its plain member:

augmented

reduced

ambitransitive

suppletion

animate

soyi ‘laugh’

taaxu ‘sit’

yiga ‘eat’

xara ‘learn’

butu ‘become angry’

kanu ‘fear’

muxu ‘hide’

kara ‘die’

wari ‘see’

inanimate

wari ‘boil’

kaawa ‘dry’

teleŋo ‘become straight’

xenu ‘fall’

kare ‘break’

kippe ‘turn over’

buyi ‘burn’

loggi ‘hang’

wuñi ‘open’

97It is striking that:

  • the predominance of the augmented correspondence is very strong among animate pairs, but this correspondence is found in four of the nine inanimate pairs only;

  • the only animate pairs that are not of the augmented type belong to the suppletion type;

  • reduced and ambitransitive correspondences (which in Soninke are in quasi-complementary distribution depending on the final vowel of the verb root) are found exclusively among inanimate pairs.

98The obvious conclusion is that, in Soninke, the preference for causativization as the formal device relating the members of plain / induced verb pairs is counterbalanced, in the case of inanimate pairs, by the preferred lexicalization principle stated by Nichols & al. (2004: 172-73) in the following terms:

“Other things being equal, LANGUAGES TEND TO LEXICALIZE ACTIONS OR STATES OF HUMANS (OR ANIMATES MORE GENERALLY) AS PRIMARY; actions or states of inanimates are less preferred [...] Consequently, for animate verbs there is a preference for treating the plain verb as the basic form and deriving the induced one [...] For inanimate verbs there is the opposite preference for treating the induced verb as basic and deriving the plain one.”

6. Conclusion

99On the basis of an inquiry carried out with speakers of the Bakel variety of Soninke, we have brought out the following features as characterizing the morphosyntactic phenomena related to transitivity in this Soninke variety:

  • the distinction between transitive and intransitive verbal predication is particularly clear-cut, since it is overtly marked in the completive positive, in the imperative plural, in the subjunctive positive, and when the locative copula is used as an incompletive auxiliary in clauses including a focalized term;

  • A-lability is exceptional, and the use of the antipassive suffix ‑ndi constitutes the most common depatientivization strategy;

  • P-lability constitutes the rule for transitive verbs ending with i or e, but is exceptional for verbs ending with other vowels;

  • there is no formal distinction between anticausativization and passivization: P-labile verbs used intransitively as well as intransitive verbs derived by means of the detransitivization marker ‑i can equally encode more or less spontaneous processes and processes that cannot be conceived without the intervention of an agent;

  • according to the typological grid proposed by Nichols & al. (2004), there is a strong preference for transitivization with the verb pairs they characterize as ‘animate’, but not with the ‘inanimate’ ones.

100Finally, we would like to emphasize that these conclusions are probably not limited to the Soninke variety spoken in Bakel, since whenever we found in the published sources on Kaedi Soninke examples of use of verbs corresponding to those we have in our corpus, we observed no discrepancy with our own data.

Haut de page

Bibliographie

Cobbinah Alexander and Lüpke Friederike, 2009, Not cut to fit – zero coded passives in African languages. In: M. Brenzinger and A.-M. Fehn (eds.). Proceedings of the 6th World Congress of African Linguistics. Cologne: Köppe, pp. 153-165.

Creissels Denis, 1991a, La voix en soninké. Mandenkan 23, pp. 1-24.

Creissels Denis, 1991b, Remarques sur le système tonal soninké. Linguistique africaine 6, pp. 129-134.

Creissels Denis, 2006, Syntaxe générale, une introduction typologique, vol. 2. Paris : Hermès.

Creissels Denis, 2009, P-lability and radical P-alignment. 19th International Symposium on Theoretical et Applied Linguistics, Workshop ‘Typology of Labile Verbs: Focus on Diachrony’. Aristotle University of Thessaloniki, April 3–5, 2009. http://www.deniscreissels.fr/public/Creissels-P-lability.pdf

Creissels Denis, 2013, Valency properties of Mandinka verbs. In: Bernard Comrie and Andrej Malchukov (eds.), Valency classes: a comparative handbook.

Diagana Ousmane Moussa, 1984, Le parler soninké de Kaédi, syntaxe et sens. Habilitation thesis. Paris Descartes University.

Diagana Ousmane Moussa, 1995, La langue soninkée: morphosyntaxe et sens. Paris: L’Harmattan.

Diagana Ousmane Moussa, 2011, Dictionnaire soninké-français (Mauritanie). Paris: Karthala.

Diagana Yacouba, 1990, Éléments de grammaire du soninké. PhD thesis. Paris: INALCO.

Diagana Yacouba, 1994, Éléments de grammaire du soninké. Paris: Association Linguistique Africaine.

Diagne Anna Marie, 2008, Phonologie et morphologie du soninké. VDM Verlag.

Geniušienė Emma, 1987, The typology of reflexives. Walter de Gruyter.

Lüpke Friederike, 2007, Vanishing voice – the morphologically zero-coded passive of Jalonke. Linguistische Berichte, Sonderheft 14, pp. 173-191.

Nichols Johanna, Peterson David, and Barnes Jonathan, 2004, Transitivizing and detransitivizing languages. Linguistic Typology 8, pp. 149–211.

Haut de page

Notes

1 This article has benefited from the support of the French National Research Agency (ANR) within the frame of the ‘Sénélangues’ project (ANR-09-BLAN-0326). We also wish to express our thanks to our consultants, Diaman Bathily, Ladji Dianifaba, and Almamy Konaté.

2 On the absence of tone marks when verbs are quoted in isolation, see Section 2.2.

3 √ signals an abstract verb root which in other contexts (in particular, in the quotation form of verbal lexemes) is augmented by the addition of a final vowel: katu, kanu, bosi, gemu, kara.

4 The Soninke sentences we quote are followed by the initials of the consultants with whom they have been recorded. The list of the abbreviations used in the glosses is given at the end of the article.

5 S = subject, O = object, V = verb, X = oblique.

6 This issue is discussed by Cobbinah and Lüpke (2009), who provide a survey of languages with constructions analyzable as zero-coded passives that depart more or less from canonical passives in other respects too, and analyze Manding languages as illustrating the extreme case of zero-coded passives that in all other respects would qualify as canonical passives. See also Lüpke (2007) on the zero-coded passives of Jalonke, and Creissels (To appear 2013) on the zero-coded passives of Mandinka.

7 This dà is sometimes labeled ‘completive positive marker’, but this label is hardly compatible with its use in the imperative plural. Alternatively, given its position, it could be analyzed as an ergative postposition or accusative preposition with a restricted distribution. We prefer the label ‘transitivity marker’ as more neutral, since we are not aware of any decisive evidence for recognizing this dà as forming a phrase with either the subject or the object. This decision is supported by the existence of other instances of transitivity marking that could not be explained as involving (allomorphs of) such an adposition.

8 The form labeled here ‘subjunctive’ combines with noun phrases in subject function in uses broadly similar to those fulfilled by forms traditionally labeled ‘subjunctives’ in grammars of European languages, but it is also found without an overt subject in uses broadly similar to those of European infinitives (it is in particular used as the quotation form of verbs).

9 Soninke has two pronouns used productively to express reflexivity: is a long-distance reflexive used in logophoric contexts, and as a reflexive possessive (as in (9a)), whereas dú is a local reflexive used for object or oblique reflexivization. The term ‘autocausative’ is taken from Geniušienė (1987).

10 On causativization in typological perspective, see Creissels (2006: chapter 24) and references therein.

11 ‘Ambitransitive’ as used by Nichols & al. (2004) is synonymous with P-labile as used in this article.

12 To be precise, this is a case of ‘partial suppletion’, since the two verbs seem to be cognate, but cannot be related by any morphological process attested elsewhere in the language – see Section 3.6.

Haut de page

Pour citer cet article

Référence papier

Denis Creissels et Anna Marie Diagne, « Transitivity in Bakel Soninke », Mandenkan, 50 | 2013, 5-38.

Référence électronique

Denis Creissels et Anna Marie Diagne, « Transitivity in Bakel Soninke », Mandenkan [En ligne], 50 | 2013, mis en ligne le 01 décembre 2013, consulté le 28 juillet 2017. URL : http://mandenkan.revues.org/211 ; DOI : 10.4000/mandenkan.211

Haut de page

Auteurs

Denis Creissels

Université de Lyon
Denis.Creissels@univ-lyon2.fr

Articles du même auteur

Anna Marie Diagne

IFAN (Dakar)
amdiagne@gmail.com

Haut de page

Droits d’auteur

Licence Creative Commons
Les contenus de Mandenkan sont mis à disposition selon les termes de la Licence Creative Commons Attribution - Pas d’Utilisation Commerciale - Partage dans les Mêmes Conditions 4.0 International.

Haut de page
  • Logo Llacan – Langage, langues et cultures d’Afrique noire
  • Logo Search | ERIH PLUS | NSD
  • Revues.org