DOI: 10.

26346/1120-2726-150

Vowel elision and reduction in Bambara

Valentin Vydrin
Institute for Linguistic Studies, Russian Academy of Sciences, St. Petersburg, Russia
<[email protected]>

The goal of this study is to test instrumentally the hypothesis that Bambara
disyllabic feet are distributed into three types. The results of the study can be
summarised as follows:
– reduction and elision of a short V1 in disyllabic feet is phonetic, rather than
phonological, and can be explained by phonotactics. Therefore, disyllabic
feet with a short first vowel form just one type;
– V1 length, although phonologically relevant, displays some instability
between speakers;
– in a disyllabic foot (at least when its boundaries coincide with word bounda-
ries), length characteristics are in complementary distribution: if the first
vowel is short, the second is long, and if the first vowel is long, the second is
short. This phenomenon can be defined as ‘foot isochrony’;
– if the first vowel of a disyllabic foot is short, the duration of the second
vowel depends on the position of the foot within the word: word-finally it is
long, otherwise it is short;
– the difference between disyllabic feet types in Bambara can be exhaustively
described by means of the length of the first vowel; there seems to be no
need to postulate the existence of stress.

Keywords: vowel elision, featural foot, syllable weight, Bambara.

1. General information about Bambara

Bambara (also Bamana, Bamanankan < Manding < Western

Mande < Mande < Niger-Congo) is spoken mainly in Mali by some 4
million L1 speakers and by a further 10 to 12 million L2 speakers.

2. Basics of Bambara phonology

2.1. Phonemic inventory

The Bambara vowel system is triangular with three degrees of aper-
ture. It consists of three series of vowels: short oral /i, e, ɛ, a, ɔ, o, u/,
long oral /iː, eː, ɛː, aː, ɔː, oː, uː/, short nasal (ĩ, ẽ, ɛ̃, ã, ɔ̃, õ, ũ).1
The Bambara consonants are represented in Table 1.

Italian Journal of Linguistics, 32.1 (2020), p. 103-124 (received July 2019)

Valentin Vydrin

Place Labials Alveolars Palatals Velars Labialised velar

Manner
Prenasalised voiceless mp nt ɲʧ ŋk
stops
Prenasalised voiced mb nd ɲʤ ŋg
stops
Prenasalised fricatives mf ns / ns
Voiceless stops p t ʧ k
Voiced stops b/b d / (d) ʤ g/g (gw)
Voiceless fricatives f/f s/s (ʃ) h
Voiced fricatives (v) z
Oral sonorants w / (w) l/l j/j
Trill r
Nasal sonorants m/m n/n ɲ/ɲ ŋ/ŋ

Table 1. Bambara consonants.2

The consonants appearing in foot-initial position are given in regu-

lar characters, and those occurring in foot-internal position are in bold
(after a slash, if a consonant can occur both in foot-initial and foot-inter-
nal positions). Marginal phonemes (contact-induced ones or those with
variable status across dialects) are given in brackets.
Realisations of foot-internal /g/ vary between [k], [g], [ɣ] and [Ø],
depending on the vocalic environment and the dialectal background of
the speaker. There is also, to some extent, free variation. Foot-internal
/ŋ/ can be realised as [ŋ] or [ŋg ~ ŋk], depending on the dialect back-
ground of the speaker. Realisations of /ns/, in both positions, vary [ns ~
nz ~ z]. The phoneme /r/ occurs only in foot-internal position.

2.2. Syllable structure

The canonical syllable type in Bambara is CV. Nasal vowels are
realised as such before a pause, but when followed by a consonant, the
realisation is [-VN], where -N is a nasal sonorant homorganic with the
subsequent consonant, and the vowel is denasalised (at least partly), e.g.
dén [dẽ]́ ‘child’ → dénkɛ [déŋkɛ́] ‘son’ (kɛ̀ ‘male’), dénso [dénsó] ‘womb’
(só ‘house’). Onsetless syllables are rare, appearing in personal pronouns
(à ‘3sg’, í ‘2sg’, etc.), in interjections and at the beginning of some
French and Arabic loans. True closed syllables are very rare and can be

104
 Vowel elision and reduction in Bambara

found only in expressive adverbs (of ideophonic origin) and in semi-

adapted loans.

2.3. Tonal system

There is an abundant literature on the Bambara tonal system (for
surveys covering different interpretations, see i.a. Green 2015: e6-e8).
In the present article, tone is not in focus; but, when necessary, the tonal
facts will be interpreted according to the approach presented in Vydrin
(2016; 2019). This approach is summarised in the following paragraphs.
Bambara has two level tones, low (marked by the grave accent)
and high (marked by the acute accent; a circumflex is used in the pho-
netic transcription for the falling tone, i.e. a combination of H + L). A
tonal domain most often (but not always) coincides with the word and
consists of a tonally dominant syllable (the initial syllable of a domain
whose tone is lexically determined) and, optionally, one or more reces-
sive syllables (whose tones are assigned according to rules). It should
be underlined that the tonally dominant character of the syllable is not
equivalent with the headedness of the foot: these two phenomena are
not isomorphic.
There is downdrift and downstep. Bambara has a tonal morpheme,
an article represented by a floating low tone, which manifests itself in
a tonal modulation on the preceding syllable and in a downstep on the
subsequent high-tone syllable.

3. Featural foot in Bambara

The Bambara foot has been discussed in a number of works (Vydrin

2001; Leben 2002; Leben 2003; Weidman & Rose 2006; Green 2010;
Vydrin 2014; Green 2015; Vydrin 2019: 45-47). My current view is that
the Bambara foot is determined by segmental, rather than prosodic fac-
tors (although the segmentation into feet impacts also prosody). It can
be therefore characterised as a featural foot.3
This approach can be summarised as follows.4 In Bambara, the
foot is a rhythmic unit consisting of one or two syllables. A foot may
be equal to or smaller than a morpheme and synchronically,5 it can-
not include more than one morpheme (i.e., in Bambara, the morpheme
boundary is stronger than the foot boundary). The foot is characterised
by an inner integrity which manifests itself in:
– a discrepancy between the inventories of the foot-initial and the
foot-internal consonants (see Table 1) with foot-initial consonants
tending to be ‘stronger’ and internal consonants ‘weaker’;

105
Valentin Vydrin

– restrictions on vocalic combinations within a foot, with combina-

tions of two identical vowels being most preferred and combina-
tions of mid-open and mid-closed vowels among the combinatory
possibilities least preferred;
– tonal realisations: in trisyllabic monomorphemic words, a high tone
is inserted at the end of a low-toned domain followed by another
low-toned domain, and the span of this high tone coincides with
the foot.6

In Vydrin (2014), I postulated three classes of disyllabic feet in

Bambara:
– trochaic: the first vowel is long;
– iambic: the vowel of the initial syllable is susceptible to elision
(or strong reduction), although it can be restored in slow careful
speech;
– neutral: the first vowel is short but it is not susceptible to reduction.

This classification, however, is not without questions.

First, the distinction between the ‘iambic’ and ‘neutral’ types in
Vydrin (2014) was based on my auditory perception and on the data
from big Bambara dictionaries (Bailleul 2007, Dumestre 2011), and has
never been verified experimentally. However, these dictionaries some-
times contradict each other (see below), which makes it difficult to draw
a clear boundary between the two types.
Second, there are questions also about the trochaic feet. On the one
hand, they may seem to be clearly distinct from the two other types if
we proceed from the fact that vowel length in the foot-initial syllable is
phonologically relevant. On the other hand, the duration of long vowels
is not very stable and varies between speakers. Some words lose their
vowel length more readily (i.e. many speakers pronounce them with
short vowels), while others retain it better.
The current study is an attempt to verify experimentally and, if nec-
essary, to update the hypothesis formulated in Vydrin (2014).

4. The experiment

Originally, my experiment was designed to study the distinction

between ‘potentially neutral’ and ‘potentially iambic’ feet: words with
vocalic elisions and reductions were specifically targeted. A limited num-
ber of ‘potentially trochaic’7 feet were included, as well as some trisyllabic
words, for the sake of comparison. At the analysis stage, the goals of the

106
 Vowel elision and reduction in Bambara

experiment were reformulated, but unfortunately it was too late to collect

further data. Therefore, there are no instances of vowel combinations such
as CeCi or CoCu or of foot-internal -g- in my sample, and the rare trisyl-
labic type (CV)(CVːCV), as in kàsaara ‘accident, catastrophe’, is also absent.
These lacunae should be filled in subsequent experiments, but the results
of the current study still shed some light on the problems discussed.
My phonetic questionnaire consisted of 132 words of different
types, both disyllabic (potentially ‘trochaic’, ‘iambic’, and ‘neutral’) and
trisyllabic. Words were almost exclusively nouns and verbs and were
recorded in June 2016 from two inhabitants (and natives) of the village
of Dugukunna (15 km west from Segu), MT and KK. Both speakers were
males in their 40s at the time of the recording.
In each of the two samples, 45 presumably ‘iambic’, 10 presum-
ably ‘trochaic’, and 45 presumably ‘neutral’ forms (the same words for
both speakers) were selected for the ‘disyllabic sample’. Each word was
pronounced by each speaker three times in isolation, then three times in
phrase-internal position.
Trisyllabic monomorphemic words were represented in the ques-
tionnaire by 27 items of the structure (CVCV)(CV), all of them with
short V1, and 5 items of the type (CV)(CVCV) (this type is much less fre-
quent in Bambara in general).
The words were segmented in Praat, and the duration of the vowels
was calculated. The words were taken only in the context of a carrier
phrase (3 pronunciations for each word). For the nouns, it was typi-
cally the phrase Mùsá bɛ X yé ‘Musa sees X’ (in some cases, where this
particular context was semantically inappropriate, other contexts were
taken, as Án tɛ́ bàlawu fɛ̀ ‘We don’t want a catastrophe’). For the verbs,
these were phrases with an adverbial expression (most often, the adverb
bì ‘today’), in order to avoid a phrase-final position, e.g. Mùsa ká fìsa bì
‘Musa feels better today’.
It was also decided to consider a vowel absent if its duration was
equal to or below 20 ms,8 because such vowels are practically impercepti-
ble to the human ear. Tokens with zero V1 (i.e. with a duration below 20
ms) were calculated separately from a non-zero V1 (more than 20 ms).9

Potentially ‘trochaic’ feet can be divided into three groups,

with respect to the representation of their vowel length in the two major
Bambara dictionaries:
1) those with a long V1 in both Bailleul’s (B) and Dumestre’s (D) dic-
tionaries: bòolo ‘fish Clarotes laticeps’;
2) those with a short V1 in both dictionaries: bòoso ~ bòso ‘Bozo’,10
dóolen ~ dólen ‘fishing hook’, jɛ̀ɛrɛ ~ jɛ̀rɛ ‘to gather together’, jɔ̀ɔrɔ

107
Valentin Vydrin

~ jɔ̀rɔ ‘to worry’;

3) those with variable vowel length: dòolo (B. dòlo ~ dòolo, D. dòlo)
‘star’, fòolo (B., D. fòlo ~ fòolo) ‘goitre’, wéele (B., D. wéle ~ wéele)
‘to call’, bɛ̀ɛsɛ (B. bɛ̀sɛ, D. bɛ̀sɛ ~ bɛ̀ɛsɛ) ‘coquettish’.

Potentially ‘neutral’ feet.

bása ‘agama lizard’, bére ‘stick’, bɔ̀lɔ ‘peg, pole’, bɔ́lɔn ‘lane’, dísi
‘chest’, fála ‘orphan’, fɛ̀rɛ ‘public square’, fúru ‘marriage’, gére ‘horn’,
jàla ‘string’, jéle ‘axe’, jíri ‘tree’, jùru ‘rope’, kìsɛ ‘grain’, kólo ‘bone’, kòlon
‘mortar’, kóro ‘calabash’, kɔ́lɔ ‘shea kernel’, kɔ̀lɔn ‘well (with water)’, kùlu
‘mountain’, mìnan ‘bushbock’, ɲími ‘head louse’, ɲínɛ ‘mouse’, sìna ‘co-
wife’, síra ‘road’, tɛ́nɛ ‘paternal aunt’, wòlo ‘skin’, wɔ̀lɔ ‘francolin’, wúla
‘wilderness’, bísi ‘to press’, dɔ́nɔ ‘to lend’, fàra ‘to add’, fúran ‘to sweep’,
gáran ‘to hobble’, gɛ̀rɛn ‘to evaporate’, kísi ‘to save’, ɲìna ‘to forget’, síran
‘to fear’, sìri ‘to tie’, túnu ‘to err; to disappear’, wólo ‘to give birth’, wúlan,
wúran ‘to skin’, fɔ́lɔ ‘first’, kólo ‘to educate’, kólon ‘used, old’.

Potentially ‘iambic’ feet. In the words of this list, V1 is easily

perceptible as elided or strongly reduced in the pronunciation of at least
one of my speakers. The words are divided in four groups, depending on
their representation in Bailleul’s and Dumestre’s dictionaries:
1) elision is indicated in both dictionaries: bíɲɛ ‘liver’, bùlon ‘ante-
chamber’, dàla ‘lake, pond’, dálan ‘bed’, díɲɛ ‘world’, dɔ̀lɔ ‘beer’,
dúnan ‘foreigner, guest’, fílan ‘age-mate’, fìle ‘flute’, fíyen ‘blind’,
fúla ‘Fulbe’, tìle ‘sun’, bìla ‘to put’, díla ‘to make’, dúlon ‘to hang’, fílɛ
‘to look at’, fìsa ‘be better’, tíla ‘to divide’, tílen ‘to straighten’, bìlen
‘red’, fìla ‘two’;
2) no elision is indicated in either dictionary, but is attested (at least
occasionally) in my data: búran ‘in-law’, fólon ‘ravine’, kíli ‘egg’,
kúma ‘speech’, kùru ‘ball, bowl’, kúrun ‘boat’, mùso ‘woman’, ntúra
‘bull’, túlo ‘ear’, túlon ‘play’, tùma ‘time’, bísan ‘to whip’, fíri ‘to over-
turn’, fìri ‘to cook couscous’, gírin ‘to rush’, mìnɛ ‘to catch’, tára ‘to
stick’, dɔ́rɔn ‘only’;
3) elision is indicated in Dumestre (2011), but not in Bailleul (2007):
fíɲɛ ‘water jar’, fíɲɛ ‘wind’, túlu ‘oil’;
4) elision is indicated in Bailleul (2007), but not in Dumestre (2011):
díli ‘root’, fìli ‘to throw’, fìli ‘to mistake’.

As already stated, the limit between ‘iambic’ and ‘neutral’ feet in

Bambara is vague, and it would be difficult (if not impossible) to for-
mulate a clear criterion for classifying each particular word as belong-
ing to the iambic or neutral type. In this situation, we can say that the

108
 Vowel elision and reduction in Bambara

classification of any particular foot to the ‘iambic’ or ‘neutral’ type (if

such classes are proved to be valid) should be as a result of experimental
study, rather than as a precondition of such a study. In order to avoid a
vicious circle, in the following data analysis all ‘potentially iambic’ and
‘potentially neutral’ feet were grouped together.

Trisyllabic (CVCV)(CV) words: bàlawu ‘catastrophe’, béreke

‘stick’, bìlisi ‘demon’, bóloma ‘intermediary’, bɔ̀rɔkɔ ‘mud’, dɔ́rɔmɛ ‘5
francs’, dálasi ‘5 francs’, dáraja ‘celebrity’, dàraka ‘breakfast’, dùlɔki
‘shirt’, fɔ̀lɔkɔ ‘dust’, fúlumɛ ‘blacksmith’s hammer’, gɛ̀rɛntɛ ‘pression’,
jàlaki ‘guilt’, jɔ̀lɔkɔ ‘chain’, kílisi ‘magic formula’, kùlusi ‘pants’, ntíleku
‘lead (metal)’, ntúloma ‘forked post’, séleke ‘angle’, sìlamɛ ‘Muslim’, sìrimɛ
‘melted butter’, sìrife ‘rasor’, bèleke (mɔ̀gɔ lá) ‘to surprise’, bùluku ‘to dig
over’, búruja ‘to dénigrate’, fàrati ‘to take a risk’, féreke ‘to tangle’, fòroki
‘to chafe’, fùruku ‘to grow angry’, gírinti ‘to belch’, kɔ̀lɔsi ‘to observe’,
mùluku ‘to paralyse’, ɲàraki ‘to waste’, pùruti ‘to snatch’, wùlusi ‘to shell’.
Trisyllabic (CV)(CVCV) words:11 bìsigi ‘image’, fìtinɛ ‘oil lamp’,
fìtinɛ ‘conflict’, mísɛli ‘needle’, mìsiri ‘mosque’.

4.1. ‘Iambic’ and ‘neutral’ feet

The following tables present correlations between different seg-
mental factors and the susceptibility of disyllabic feet to the reduction of
their first vowel.
For each evaluation criterion, the data are represented in a pair
of tables: first, the percentage of the tokens with compete V1 elision
(the ‘elision tables’); second, the number and mean V1 duration in the
remaining tokens (i.e. those which do not undergo V1 elision, the ‘dura-
tion tables’).
In both types of tables, when the numbers of pronunciations by KK
and MT are different, the first number is for KK, and the second one (after
the slash) is for MT. For example, ‘27/30’ in the second column, line two
of Table 2 means that KK pronounced 27 tokens and MT pronounced 30
(i.e., for some reason, 3 tokens were not pronounced by KK).
In the ‘elision tables’, in lines 3 and 4, the number of V1 drops is
given in absolute figures, and in lines 5 and 6, the percentage of drops
in each respective sample. In line 7 (Combined %), the rate of drops in
the whole sample across both speakers is given. The data are arranged in
decreasing order according to the rate of V1 drops.
In the ‘duration tables’, the data are arranged in the order of
increasing mean vowel duration.

109
Valentin Vydrin

Initial consonants.12
Initial t d f b m g k s ɲ j w
consonant
Nr. of 27/30 33 57/60 33 8/9 12 39 12 9 12 15
tokens,
KK/MT
KK n of 11 12 25 12 1 1 5 0 0 0 0
drops
MT n of 24 24 30 11 3 3 6 0 0 0 0
drops
KK % of 41% 36% 44% 36% 12,5% 8% 13% 0% 0% 0% 0%
drops
MT % of 80% 73% 50% 33% 33% 25% 15% 0% 0% 0% 0%
drops
Combined 60,5% 54,5% 47% 34,5% 22,5% 16,5% 14% 0% 0% 0% 0%
% of drops

Table 2. V1 drops for ‘neutral’ and ‘iambic’ feet, by foot-initial consonants.

As we can see, in MT’s pronunciation V1 elision is much more fre-

quent. Otherwise, both speakers display the same trends: feet with ini-
tial dental /t d/ and labial /f b/ obstruents are most liable to V1 elision
(and for MT, dental-initial feet are considerably more liable to exhibit
elision than labial-initial ones). Feet with initial velar or /m/ are less
susceptible to V1 elision, while those with initial palatal and labial sono-
rants /ɲ, w/, as well as the dental fricative /s/ and palatal glide /j/ do
not exhibit it at all.

Initial s f m ɲ t k g b d j w
consonant
Nr. of tokens 12 25/28 8/9 9 16/6 33/32 11/9 21/22 21/9 12 15
KK/MT
KK, ms 33 37 34 42 48 42 47 45 53 59 47
MT, ms 36 38 51 44 38 45 41 46 44 47 65
Combined, ms 34,5 37,5 42,5 43 43 43,5 44 45,5 48,5 53 61

Table 3. Mean V1 duration for ‘neutral’ and ‘iambic’ feet in ms, by foot-initial con-
sonants.

110
 Vowel elision and reduction in Bambara

If the tokens with full elision of V1 are not taken into account, the
fricative initial consonants /s f/ have the shortest realisations of V1, and
the oral sonorants (/j/ and especially /w/), the longest ones. The stops
and nasal sonorants occupy an intermediate position. More precisely,
the situation of /m/ is rather ambiguous: as a sonorant, it tends to be
resistant to V1 reduction, and as a labial, it tends to favour it.
Unlike the ‘elision table’, the ‘duration table’ reveals no clear ten-
dencies (with the exception of the feet with initial /j/ and /w/ in which
V1 is not susceptible to elision and which are realised longer than else-
where). Data for non-elided V1 recorded from both informants are often
contradictory and in most cases, the variability in the duration of such
vowels seems to be rather irrelevant. Therefore, the rate of V1 drops
appears as a more relevant factor than the duration of V1 in the remain-
ing tokens, i.e. after the cases of V1 elision are discarded, remaining feet
behave relatively homogeneously.

Foot-internal consonants.13
Internal consonant ɲ l m j n r s
Nr. of tokens KK/MT 12 117/123 9 3 27 68/69 20/21
KK n 11 51 2 2 1 6 0
MT n 5 71 6 0 7 9 3
KK % 92% 44% 22% 67% 4% 9% 0%
MT % 42% 58% 67% 0% 26% 11% 14%
Combined % 67% 51% 44,5% 33,5% 15% 10% 7%
Table 4. V1 drops for ‘neutral’ and ‘iambic’ feet, by foot-internal consonants.

The results are indicative primarily for the most frequent internal
consonants, /l/ and /r/. The presence of /l/ in foot-internal position
certainly favours V1 elision, while a foot-internal /r/ does so to a much
lesser degree. However, as for other internal consonants, the samples
are not representative enough and words with vocalic combinations of
V1 close vowel and V2 open or semi-open vowel predominate in this
sample. It will be shown below that this combination favours V1 elision
and it can only be stated that any foot-internal consonant represented in
Table 5 allows V1 elision if other favourable conditions (type of initial
consonant; vocalic combination) are satisfied.

111
Valentin Vydrin

Internal consonant j ɲ n r s m l
Nr. of tokens KK/MT 1/3 1/7 26/20 62/60 20/18 7/3 66/52
KK, ms 22 21 38 43 43 40 49
MT, ms 27 41 41 41 45 51 47
Combined, ms 24,5 31 39,5 42 44 45,5 48

Table 5. Mean V1 duration for ‘neutral’ and ‘iambic’ feet in ms, by foot-internal
consonants.

If we discard the rare feet with palatal consonants /j, ɲ/, aver-
age V1 durations in feet with all other internal consonants do not vary
much. Interestingly, the average V1 duration in the feet with internal /l/
exceeds that in the feet with /r/, which does not correlate with the fact
that /l/ facilitates V1 elision, while /r/ does not. It can be said that after
omitting words with V1 elision (and if rare palatal consonants are not
taken into account) the foot-internal consonants do not really influence
the susceptibility of feet to V1 reduction.

Vocalic combinations
In Bambara disyllabic feet, some vowel combinations are impossible,
and among the allowed combinations, some are very frequent, while oth-
ers are rare (see in particular Konaté 1989; Konaté & Vydrine 1989).14
In my sample, all more or less frequent combinations appear.
However, because of the considerable number of combinations, most of
them are represented by a small number of tokens: the highest score is for
the combination i-i, i.e. the type CiCi, which is represented by 12 words
(36 tokens). The low frequency makes it difficult to discover tendencies
concerning vocalic elision or reduction. For this reason, these combina-
tions are grouped into clusters, to make tendencies more visible.15
The following clusters were formed:
1. Combinations with vowels of different aperture but not contrasting
for the feature ‘front/back’ (heterotimbral):
(a) close V1 (i, u), mid V2 (ɛ, e, ɔ, o, ɛ̃, ẽ, ɔ̃, õ), V1 and V2, without
front-back differences (e.g. CiCɛ or CuCo, but not CiCo or CuCɛ); in
Table 6 these are labelled i-E and u-O;16
(b) close V1 (i, u), open V2 (a, ã), labelled i/u-A in Table 6.

2. Combinations with vowels for the same aperture and same front-
back value (i.e. homotimbral: identical or differing only in the
nasality feature):
(c) close vowels: V1 i, u, V2 i, u, ĩ, ũ, labelled i-I, u-U in Table 6.

112
 Vowel elision and reduction in Bambara

(d) open vowels: V1 a, V2 a, ã, labelled a-A in Table 6.

(e) mid vowels: V1 ɛ, e, ɔ, o, V2 ɛ, e, ɔ, o, ɛ̃, ẽ, ɔ̃, õ, labelled E-E, O-O in
Table 6.

Vocalic combinations clusters i-E, u-O i/u-A i-I, u-U a-A E-E, O-O
Nr. of tokens KK/MT 50/54 63 56/60 24 66
KK n 35 21 11 3 3
MT n 35 30 22 8 6
KK % 70% 33% 20% 12,5% 4,5%
MT % 65% 48% 37% 33% 9%
Combined % 67,5% 40,5% 28,5% 23% 7%
Table 6. V1 drops for ‘neutral’ and ‘iambic’ feet, by V1-V2 combination.

Table 6 reveals quite clear tendencies:

– heterotimbral combinations (especially, the type with close V1 and
mid V2, both anterior or both posterior) are more favourable to V1
elision/reduction;
– for homotimbral combinations, those with close vowels are the
most favourable to elision and reduction while those with mid vow-
els are most unfavourable.

Vocalic combinations clusters i-E, u-O i/u-A i-I, u-U E-E, O-O a-A
Nr. of tokens KK/MT 15/19 42/33 45/38 63/60 21/16
KK, ms 35 36 43 47 57
MT, ms 41 42 41 45 45
Combined, ms 38 39 42 46 51
Table 7. Average V1 duration for ‘neutral’ and ‘iambic’ feet in ms, by V1 and V2
combination.

Here too, after removing the feet with V1 elision, tendencies are
unclear. For MT, differences among the types are insignificant. In KK’s
pronunciation, V1 is shorter in heterotimbral combinations, while in the
homotimbral ones the shortest V1s are the close ones and the longest
V1 are the most open ones, possibly a manifestation of a general cross-
linguistic tendency (de Lacy 2007: 27).

113
Valentin Vydrin

Let us also test the hypothesis of correlation between V1 duration

and the orality/nasality of the final vowel, see Tables 8 and 9.

Vocalic combinations clusters V-V V-Ṽ

Nr. of tokens KK/MT 184/192 75
KK n 51 22
MT n 71 30
KK % 28% 29%
MT % 37% 40%
Combined % 32,5% 34,5%

Table 8. V1 drops for ‘neutral’ and ‘iambic’ feet, V-V vs V-Ṽ.

Vocalic combinations clusters V-V V-Ṽ

V1 V2 V1 V2
Nr. of tokens KK/MT 133/121 183/192 53/45 75
KK, ms 44 80 44 102
MT, ms 43 81 44 101
Combined, ms 43,5 80,5 44 101,5
Table 9. Average V1 and V2 duration for ‘neutral’ and ‘iambic’ feet in ms, V-V vs V-Ṽ.

The rate of V1 drops in the types V-V and V-Ṽ is almost the same,
and it can be assumed that the nasality of the final vowel does not influ-
ence the susceptibility of V1 to elision. The V1 lengths in both groups of
vocalic combinations are also practically identical. Therefore, the nasal-
ity/orality of V2 does not influence the V1 duration. On the other hand,
the nasalised V2 turn out to be by 20% longer than non-nasalised ones.

4.2. ‘Trochaic’ feet

As mentioned in 3.1, the contrast of vowel length in Bambara is
pertinent only in the non-final syllable of a disyllabic foot. Even in this
position, long vowels are relatively infrequent and often unstable among
speakers. In my questionnaire, only nine single-foot words are ‘poten-
tially trochaic’, i.e. appear with a long V1 in one or both dictionaries of
reference (or previously attested so in my field data).
The analysis of the recordings has shown that the speaker KK pre-
serves vowel length much better than MT. Out of the 9 words, only one

114
 Vowel elision and reduction in Bambara

(wéle ‘to call’) is pronounced by KK with a short V1, and in the remain-
ing 8 words, V1 is longer than V2. As for MT, he pronounced a relatively
long vowel only in one word (bòoso ‘Bozo’), in 3 other words (dòolo
‘star’, dóolen ‘fishing hook’, bɛ̀ɛsɛ ‘coquettish’) V1 was of the same length
as V2. Elsewhere, V1 was short (although it did not elide).
It is of interest that in some words V1 is always long, and in others
it varies among speakers. According to my field data (exceeding the lim-
its of the current experiment), stability or instability of the long vowel in
V1 position seems to be relatively independent of the dialect.17

4.3. Durational ratio between V1 and V2

Table 10 represents mean durations of V1 and V2 in feet of differ-
ent types (‘iambic’ and ‘neutral’ types are lumped together, as in the
preceding counts; the difference between the number of tokens for V1
and V2 equals the number of V1 drops). For trochaic feet, only words
pronounced with a long vowel (and ‘semi-long’ in MT’s pronunciation)
were included.

Type of foot ‘Trochees’ ‘Neutral’+‘Iambs’

Nr., KK/MT 24/12 186/167 258/267
V1 V2 V1 V2
KK, ms 88 58 44 86
MT, ms 79 82 43 86
Table 10. Average V1 and V2 duration in ms, by foot type.

As we can see, in the ‘neutral/iambic’ feet, even if V1 is not elided,

the second vowel is, on average, twice as long as the first. In the ‘trochaic’
feet, in KK’s pronunciation, V2 is shorter than V1 by approximately 1/3
(and also 1/3 shorter than V2 in ‘neutral/iambic’ feet), while in MT’s
pronunciation, it is roughly equal to the V1. In MT’s idiolect, degradation
of the vowel length contrast proceeds in two directions. First, the vowel
length disappears in less stable words; second, in the words maintaining
vowel length, this feature grows less distinctive, and the contrast between
long and short vowels weakens. However, even in MT’s pronunciation, in
these words, V1 reduction (let alone elision) is never observed.
One may ask: may V2 length be explained by the presence of the
tonal article which manifests itself as tonal modulation at the end of
a noun (cf. 2.3)? To answer this question, let us compare the average
vowel lengths for nouns and verbs, cf. Table 11.

115
Valentin Vydrin

Nouns Verbs
V1 V2 V1 V2
Nr., KK/MT 119/109 160/165 54/46 80/84
KK, ms 46 87 41 86
MT, ms 45 91 39 77
Table 11. Average V1 and V2 duration in ms, ‘iambic’ + ‘neutral’ feet, in nouns
and verbs.

It turns out that in KK’s pronunciation the average duration of both

V1 and V2 in verbs is practically identical to that of V2 in nouns. MT
pronounces both V1 and V2 in verbs a bit shorter than in nouns, but the
ratio between V1 and V2 is roughly the same: V1 is twice as short as V2.
It can be concluded that the difference in V2 and V1 duration in nouns
cannot be explained by grammatical tone.

4.4. Vowel realisations in trisyllabic words

As mentioned at the beginning of section 4, trisyllabic words form
two groups, depending on their segmentation into feet: the more numer-
ous group of the (CVCV)(CV) type, and the rarer type (CV)(CVCV). Mean
durations of vowels for both these groups are represented in Table 12.

(CVCV)(CV) (CV)(CVCV)
V1 V2 V3 V1 V2 V3
Nr., KK/MT 85/91 103/105 103/105 15 9 15
KK 37 51 80 42 50 81
MT 34 56 75 48 51 76
Table 12. Mean V1, V2 and V3 durations in ms in monomorphemic trisyllabic
words.

The following tendencies can be observed (for both speakers):

– in some disyllabic feet, the first vowel elides, cf. bìlisi [blìsî ~ bìlìsî]
‘demon’ (segmentation into feet: (bìli)(si)), fìtinɛ [fìtnɛ̂] ‘oil lamp’
(segmentation into feet: (fì)(tinɛ));
– if words with vowel elision are not taken into account, in both types
of words, the shortest vowel is V1. However, the duration difference
between V1 and V2 in (CVCV)(CV) is larger than in (CV)(CVCV) and
in some realisations, V1 in (CV)(CVCV) can be longer than V2;

116
 Vowel elision and reduction in Bambara

– the word-final vowel (V3) is considerably longer than any other

vowel of the word. The status of the final syllable (the only sylla-
ble of a monosyllabic foot or the final syllable of a disyllabic foot)
seems to have no impact on the duration of its vowel.

4.5. Summary table

KK MT Total/Average
V1 V2 V3 V1 V2 V3 V1 V2 V3
Iamb. + Neut. 186 258 167 267 352 525
CVCV Nr.
duration, ms 44 86 43 86 43,5 86
standard dev., ms 14 27 11 27 12,5 27
Trochees CVCV Nr. 24 24 12 12 36 36
duration, ms 88 58 80 82 85 66
standard dev., ms 12 15 9 17 12 19
(CVCV)(CV) Nr. 85 103 103 91 105 105 176 208 208
duration, ms 37 51 80 34 56 75 36 53 78
standard dev., ms 11 14 24 11 15 18 11 15 21
(CV)(CVCV) Nr. 15 9 15 15 9 15 30 18 30
duration, ms 42 50 81 48 51 76 45 50 78
standard dev., ms 19 14 26 20 14 13 20 14 20
Table 13. Mean duration of vowels in all foot types (both disyllabic and trisyllabic
words).

5. Discussion and conclusions

5.1. ‘Iambic’ vs ‘neutral’ feet

My experiment confirmed that, on the one hand, the elision of a
short V1 is optional. It varies between speakers (and even in the pronun-
ciation of the same speaker), and an elided V1 can always be restored
in careful speech. On the other hand, elision seems to be conditioned by
the segmental composition of a foot, i.e. by phonotactics. The following
factors are of importance:
1) Foot-initial consonants. Bilabial and alveolar stops and fricatives
favour vowel elision; sonorants (both oral and nasal, probably with

117
Valentin Vydrin

the exception of the labial /m/) and palatal stops (or affricates) dis-
favour it, while velar stops (and, probably, the labial sonorant /m/)
occupy an intermediate position.
2) Foot-internal consonants. Sonorants /l/ and probably /j/ and /ɲ/
favour elision of the preceding vowel; the stops /b/ and /g/, the
fricative /f/ and the labiovelar sonorant /w/ are strongly unfavour-
able to it.18 The other foot-internal consonants allow V1 elision pro-
vided that other conditions (foot-initial consonants, vocalic combi-
nations) are met.
3) Vocalic combinations. Heterotimbral combinations with a close V1
favour elision; homotimbral combinations of mid-open vowels dis-
favour it. The homotimbral combinations of closed and open vow-
els occupy an intermediate position.

Susceptibility to V1 elision is determined by the combination of

these three factors. If favourable conditions on all the three parameters
are simultaneously met, the probability of elision is extremely high.
If all three parameters are unfavourable, V1 elision normally does not
occur (in particular, the presence of foot-internal consonants /g, b, f,
ns, ŋ, w/ seems to be sufficient to block V1 elision). If a foot combines
favourable and unfavourable (or ‘intermediary’) factors, the elision var-
ies across and within speakers.
This explanation does not cover cases where, in some feet, reduc-
tion often takes place, and in other feet under identical conditions, it
does not. Here are some examples (among many others).

V1 can elide No elision of V1

dɔ̀lɔ ‘beer’ bɔ̀lɔ ‘stake, peg’
dàla ‘lake, pond’ fála ‘orphan’
fólon ‘ravine’ bólo ‘hand, arm’

The presence of elision in the words of the first column and its
absence in the words of the second column can probably be explained
by a different degree of lexicalisation. On the other hand, the experi-
ment showed that the lack or the presence of V1 elision is not absolute.
We have an entire set of words which are normally pronounced without
elision, but still, elision may occur occasionally (mùso ‘woman’, júfa
‘carrion’, bàra ‘calabash’, etc.), and vice versa (dúnan ‘foreigner’, tùma
‘time’, fìli ‘to mistake’, etc.). The word bólo ‘branch’ is believed to have a
non-eliding V1, however, MT pronounced the word jíribolo ‘branch of a
tree’, contrary to the general tendency, as [jíríbúlô].

118
 Vowel elision and reduction in Bambara

The conclusions are the following:

 the reduction and elision of a short V1 in disyllabic feet is a pho-
netic, rather than phonological, process; it is (at least partially) con-
ditioned by the phonotactics;
 contrary to Vydrin (2014), the distinction between ‘iambic’ and
‘neutral’ feet is phonologically irrelevant. All disyllabic feet with a
short first vowel can be classified into one type;
 V2 in a ‘iambic’ foot which coincides with a disyllabic word is auto-
matically realised as a long vowel. Being automatic and predict-
able, this vowel length cannot be regarded as phonological. This is
confirmed by the fact that in trisyllabic words of the type (CVCV)
(CV), V2 is realised short.

5.2. ‘Iambic’ and ‘trochaic’ feet or ‘light’ and ‘heavy’ feet?

In what concerns the duration of vowels in ‘trochaic’ feet, the
results of my experiment are preliminary (a more thorough experiment
is necessary and is planned). However, some preliminary conclusions
can be drawn.
1) The contrast between ‘iambic’ and ‘trochaic’ feet remains pertinent
in modern Bambara: at least in some words (such as báara ‘work’, míiri ‘to
think’, and, in fact, in many others) the long vowel seems to be indisput-
able for speakers of this language. However, in numerous other words a
long V1 turns out to be more or less unstable, and this durational volatil-
ity seems to be unrelated to dialectal variation. For example, both partici-
pants in my experiment, KK and MT, live on the same street of the same
village, belong to the same age group and are close friends, but display
considerable divergences in the duration of their realisations of long V1).
If we compare Bambara with other languages of the Manding
group, its position can be viewed as intermediary between the extreme
points of a cline. At one extreme, there is Mandinka, where vowel length
is phonologically relevant in all positions, long vowels are very frequent
and their functional load is very high, see (Creissels & Sambou 2013:
21). At the other extreme, there is Guinean Maninka, where the original
vowel length contrast has been practically lost (Diané & Vydrin 2014).19
It seems that the erosion of the vowel length contrast in Bambara (its
irrelevance in foot-final position; its instability in the first syllable) may
be indicative of a further loss of its functional load and a drift towards a
situation such as that seen in Guinean Maninka.
2) An interesting phenomenon discovered through the experiment
is the shortened realisation of V2 in ‘trochaic’ feet.20 It turned out that
in a disyllabic foot (at least when its boundaries coincide with the word

119
Valentin Vydrin

limits), the durational characteristics of V1 and V2 are in inverse rela-

tion. If the first vowel is short, the second is long (a ‘iambic’ foot), and
if the first vowel is long, the second is short (a ‘trochaic’ foot). One can
say that a disyllabic foot tends to maintain its overall duration. This
phenomenon can be defined as foot isochrony; a similar phenomenon is
observed, for example, in Finnic languages (Kuznetsova 2018: 119).
3) The calculation of vowel duration in trisyllabic monomorphemic
words showed that when a disyllabic foot occupies a non-final posi-
tion in a word (i.e. in a structure (CVCV)(CV)), its V2 is still longer
than V1 (approximately by 50%). However, their quantitative differ-
ence is not as great as in a single-foot disyllabic word CVCV (where V2
is approximately twice as long as V1).21 As for a monosyllabic foot in a
non-word-final position, i.e. in structures (CV)(CVCV), its vowel is nor-
mally longer than V2 (i.e. the vowel of the initial syllable of a ‘iambic
foot’), but shorter than V3 (it should be however kept in mind that V2 in
these words is susceptible to elision, while V1 cannot be dropped). This
means that the durational characteristics of vowels are sensitive not only
to foot boundaries but word boundaries are also of relevance. It can be
reformulated as follows: a word-final vowel is, by default, phonetically
long, but it is predictably shortened if it belongs to a disyllabic foot with
an initial heavy syllable.
A direct consequence is that the final vowel of what is hitherto
referred to as the ‘iambic foot’ is not inherently long. It may be either
short or long, depending on the position of the foot in a word.
4) Another conclusion of this study is the following. Contrary to
what was postulated in Vydrin (2014), the difference between the disyl-
labic feet types in Bambara can be exhaustively described in the terms of
the length of the first vowel, there is no need to postulate the existence
of stress.
If we assume that the difference between the two foot types can be
exhaustively described in terms of the quantitative opposition of the first
syllable vowels (and the durational differences in second syllable vowels
are stipulated by the length of V1 and therefore not phonological),22 the
next question is, do we really need to distinguish between iambic and
trochaic feet? In fact, this distinction turns out to be redundant. It seems
enough to distinguish between disyllabic feet with long and short vowels
in the initial syllable. One could designate them ‘heavy’ and ‘light’ feet
respectively.
Therefore, in a language with a featural foot and no stress (like
Bambara), the question of the directionality of the foot turns out to be
superfluous.

120
 Vowel elision and reduction in Bambara

Abbreviations
B, D = Bailleul’s (2007) and Dumestre’s (2011) dictionaries; KK,
MT = names of the informants; V1, V2 = first, second vowel of a (disyl-
labic) foot; V3 = third vowel of a word.

Acknowledgements
The study was supported by the Russian Science Foundation, grant
20-18-00250 ‘Tonal languages of the world: on-line data base and atlas’.
I am thankful to Natalia Kuznetsova for fruitful discussions on numerous
topics dealt with in this study.

Notes
1
In Bambara orthography, vowel nasality is indicated by <n> after the vowel:
mɛ́n /mɛ̃/́ ‘to hear’, sàn /sã/ ̀ ‘to buy’. Vowel length is rendered by a double letter:
míiri /míːrí/ ‘to think’. In this paper, I follow these orthographic conventions.
2
Following the established orthographic tradition for the great majority of West
African languages, in this paper (everywhere, with the exception of Table 1) the voice-
less affricate /ʧ/ is written c, the voiced affricate /ʤ/ as j, and the glide /j/ as y.
3
This term was suggested by Green (2015), although that author’s approach to the
Bambara foot is different from mine.
4
A detailed argumentation for this approach cannot be provided here due to lack
of space.
5
There are cases where ancient bimorphemic words have been re-interpreted as
monomorphemic, and they may function now as consisting of a single foot.
6
It can also be mentioned that a tonal domain can coincide with a foot (mɔ̀gɔ́
‘human being’) or include two or more feet (kóló|gɛ́lɛ́|yá ‘solidity’), and cannot be
shorter than a foot. An exceptional case is represented by a floating low tone which
has no segmental base (and is therefore shorter than a foot).
7
By ‘potentially trochaic’ feet I mean those feet where a long V1 is attested in at
least one of the aforementioned Bambara dictionaries, or those which were at least
sometimes recorded with a long V1 during my previous field work. I use the labels
‘potentially neutral feet’ and ‘potentially iambic feet’ in the same way.
8
I established the threshold of 20 ms when processing the audio files. To my ear,
Bambara vowels slightly longer than 20 ms are still distinguishable. It seems that the
threshold of perceptibility may vary across different languages; so, in English it is
about 30 ms (Lehiste 1977: 256-258), while specialists in Finnic languages put it at
35 ms (Kuznetsova & Verkhodanova 2019: 4).
9
If a speaker pronounced one token of a word with V1 elision and the other two
without elision (or vice versa), these tokens were included in different groups.
10
The name of an ethnic group in Mali.
11
Four out of five words of the type (CV)(CVCV) are Arabic loans; in general, this
type is not very frequent in the original Bambara vocabulary.
12
Words with initial n, y, c (/n, j, ʧ/) have not been included in the questionnaire
which targeted vocalic elision/reduction, as they were presumed to not be suscepti-
ble to V1 reduction.

121
Valentin Vydrin

13
Unfortunately, words with foot-internal consonants /g/, /ŋ/, /ns/, /f/, /b/, /w/
were not included in the questionnaire (it had been presumed that, in such words, V1
is not susceptible to elision), as well as the single-foot words with the rare internal
phoneme /d/.
14
In Konaté & Vydrine (1989), data from the Beledugu dialect of Bambara are ana-
lysed in this respect; in Konaté (1989), both the Beledugu and the Bamako (in fact,
the Standard Bambara) data are presented. There are some divergences between both
dialects but the main tendencies are the same.
15
An anonymous reviewer has objected against lumping together different features
and insisted on a more fine-grained approach. I can answer to this objection that, of
course, in my analysis, each vocalic combination was originally represented separate-
ly, and their conflation was a conscious decision, driven also by space constraints.
16
I use here the convention of designating a group of phonetically similar pho-
nemes by a capital letter. So, ‘E’ stands for e, ɛ, ẽ, ɛ̃; ‘I’ for i, ĩ, etc.
17
This phenomenon is intended to be the subject of a separate experiment.
18
Feet with internal /ns/ [ns ~ z] and /-ŋ-/ [ŋ] ~ [ŋg] ~ [ŋk] were absent from
my sample. In the available dictionaries, their V1 is never marked as reducible,
which is confirmed by my personal experience of work with the Bambara language.
19
To be more precise, in Guinean Maninka a new length contrast emerged as a
result of the elision of velar consonants in intervocalic position (*sìGi > sìi ‘sit down’,
*bùGu > bùu ‘hut’, etc.).
20
One of two informants whose recordings were analysed, KK, displayed this ten-
dency, but not the other one, MT. However, the majority of my other informants
(whose recordings were analysed only preliminarily) pronounce ‘trochaic’ feet in the
same way as KK.
21
We should also not forget that V1 is susceptible to elision; when speaking about
the durational ratio of V1 and V2, only non-elided V1s are meant.
22
In this relation, an interesting point is the length difference between realisations
of oral and nasal vowels in word-final position (see Table 8) and its implications on
the status of vowel nasality in Bambara. However, this question can hardly be solved
without a thorough analysis of realisations of nasal vowels in different positions, and,
in any case, lies outside the scope of the present paper.

Bibliographical references

Bailleul, Charles 2007. Dictionnaire Bambara-Français. 3 rd edition. Bamako:

Donniya.
Creissels, Denis & Sambou, Pierre 2013. Le mandinka. Phonologie, grammaire, tex-
tes. Paris: Karthala.
de Lacy, Paul 2007. Themes in phonology. In Lacy, Paul de (ed.), The Cambridge
handbook of phonology. Cambridge: Cambridge University Press. 5-30.
Diané, Mamadi & Vydrin, Valentin 2014. Segmentnaja fonologija gvine-
jskogo maninka (Сегментная фонология гвинейского манинка)
[Segmental phonology of Guinean Maninka]. In Zheltov, Alexander
(ed.), Antropologija i lingvistika. Materialy peterburgskikh ekspedicij v Afriku
(Антропология и лингвистика. Материалы петербургских экспедиций
в Африку) [Anthropology and linguistics. Materials of field trips to Africa
from St. Petersburg]. St. Petersburg: Muzej antropologii i etnografii RAN
(Kunstkamera). 128-144.

122
 Vowel elision and reduction in Bambara

Dumestre, Gérard 2011. Dictionnaire bambara-français suivi d’un index abrégé

français-bambara. Paris: Karthala.
Green, Christopher R. 2010. Prosodic phonology in Bamana (Bambara): Syllable
complexity, metrical structure, and tone. PhD dissertation. Bloomington:
Indiana University.
Green, Christopher R. 2015. The foot domain in Bambara. Language 91,1.
e1-e26.
Konaté, Adama 1989. Sopostavitel’nyj analiz dialektov beledugu i bamako jazyka
bamana (fonetika, grammatika) (Сопоставительный анализ диалектов
беледугу и бамако языка бамана (фонетика, грамматика)) [Comparative
analysis of the Beledugu and Bamako dialects of the Bamana language (phonol-
ogy and grammar)]. PhD dissertation. Moscow: Institut Jazykoznanija AN
SSSR.
Konaté, Adama & Vydrine, Valentin 1989. Le système phonologique du dialecte
bamana du Bèlèdugu (Village de Diyon). Mandenkan 17. 1-40.
Kuznetsova, Natalia 2018. What Danish and Estonian can tell to modern word-
prosodic typology. In Goedemans, R.; Heinz, J.; van der Hulst, H. (eds.),
The study of word stress and accent: theories, methods and data. Cambridge,
UK: Cambridge University Press. 102-143.
Kuznetsova, Natalia & Verkhodanova, Vasilisa 2019. Phonetic realisation and
phonemic categorisation of the final reduced corner vowels in the Finnic
languages of Ingria. Phonetics 76,1-2. 1-14.
Leben, William R. 2002. Tonal feet. In Gut, Ulrike & Gibbon, Dafydd (eds.),
Proceedings, typology of African prosodic systems (Bielefeld Occasional Papers
in Typology 1). Bielefeld. 27-40.
Leben, William R. 2003. Tonal feet as tonal domains. In Mugane, John (ed.),
Trends in African linguistics 5: Linguistic typology and representation of African
languages. Trenton: AfricaWorld Press. 129-138.
Lehiste, Ilse 1977. Isochrony reconsidered. Journal of phonetics 5. 253-263.
Vydrin, Valentin 2001. Jazyki mande i teorija jazykov slogovogo stroja (Языки
манде и теория языков слогового строя) [Mande languages and the
theory of syllabic languages]. VI mezhdunarodnaja konferencija po jazykam
Dal’nego Vostoka, Jugo-Vostochnoj Azii i Zapadnoj Afriki (25-27 sent’abr’a
2001). Materialy i tezisy dokladov (VI-я международная конференция по
языкам Дальнего Востока, Юго-Восточной Азии и Западной Африки (25-
28 сентября 2001 г.): Материалы и тезисы докладов) [VI International
conference on languages of Extreme Orient, South-Eastern Asia and Western
Africa (September 25-27 2001). Proceedings]. St. Petersburg: Vostochnyj
fakultet SPbGU. 45-53.
Vydrin, Valentin 2014. Ton i akcent v bamana: otsutstvije korrel’acii kak teo-
reticheskij vyzov (Тон и акцент в бамана: отсутствие корреляции как
теоретический вызов) [Tone and accent in Bamana: Lack of correlation
as a theoretical challenge]. In Kolpachkova, Elena (ed.), Jazyki Dal’nego
Vostoka, Jugo-Vostrochnoj Azii i Zapadnoj Afriki. XI Mezhdunarodnaja
konferencija (Sankt-Peterburg, 30 sent’abr’a – 2 okt’abr’a 2014 g.) (Языки
стран Дальнего Востока, Юго-Восточной Азии и Западной Африки:
материалы XI Международной научной конференции (Санкт-Петербург,
30 сентября – 2 октября 2014 г.)) [Languages of the Far East, Southeastern

123
Valentin Vydrin

Asia and West Africa. Acts of the XI International Conference (St. Petersburg,
September 30 – October 2, 2014)]. St. Petersburg: Studia NP-print. 32-38.
Vydrin, Valentin 2016. Tonal inflection in Mande languages: The cases of
Bamana and Dan-Gwɛɛtaa. In Palancar, Enrique L. & Léonard, Jean
Léo (eds.), Tone and Inflection: New facts and new perspectives (Trends in
Linguistics Studies and Monographs 296). De Gruyter Mouton. 83-105.
Vydrin, Valentin 2019. Cours de grammaire bambara. Paris: Presses de l’INALCO.
Weidman, Scott & Rose, Sharon 2006. A Foot-Based Reanalysis of Edge-in Tonal
Phenomena in Bambara. In Baumer, Donald; Montero, David & Scanlon,
Michael (eds.), Proceedings of the 25th West Coast Conference on Formal
Linguistics. Somerville, MA. 426-434.

124