Mid-vowels and vowel harmony in Civili

This article was downloaded by: [North West University]On: 23 April 2014, At: 03:55Publisher: RoutledgeInforma Ltd Registered in England and Wales Registered Number: 1072954 Registered office: MortimerHouse, 37-41 Mortimer Street, London W1T 3JH, UK

South African Journal of African LanguagesPublication details, including instructions for authors and subscription information:http://www.tandfonline.com/loi/rjal20

Mid-vowels and vowel harmony in CiviliHugues Steve Ndinga-Koumba-Binza aa Centre for Language and Speech Technology , Stellenbosch University , Private Bag X1,Matieland , 7602 , South Africa E-mail:Published online: 24 Oct 2012.

To cite this article: Hugues Steve Ndinga-Koumba-Binza (2006) Mid-vowels and vowel harmony in Civili, South AfricanJournal of African Languages, 26:1, 26-39

To link to this article: http://dx.doi.org/10.1080/02572117.2006.10587267

PLEASE SCROLL DOWN FOR ARTICLE

Taylor & Francis makes every effort to ensure the accuracy of all the information (the “Content”) containedin the publications on our platform. However, Taylor & Francis, our agents, and our licensors make norepresentations or warranties whatsoever as to the accuracy, completeness, or suitability for any purpose ofthe Content. Any opinions and views expressed in this publication are the opinions and views of the authors,and are not the views of or endorsed by Taylor & Francis. The accuracy of the Content should not be reliedupon and should be independently verified with primary sources of information. Taylor and Francis shallnot be liable for any losses, actions, claims, proceedings, demands, costs, expenses, damages, and otherliabilities whatsoever or howsoever caused arising directly or indirectly in connection with, in relation to orarising out of the use of the Content.

This article may be used for research, teaching, and private study purposes. Any substantial or systematicreproduction, redistribution, reselling, loan, sub-licensing, systematic supply, or distribution in anyform to anyone is expressly forbidden. Terms & Conditions of access and use can be found at http://www.tandfonline.com/page/terms-and-conditions

http://www.tandfonline.com/loi/rjal20

http://dx.doi.org/10.1080/02572117.2006.10587267

http://www.tandfonline.com/page/terms-and-conditions

http://www.tandfonline.com/page/terms-and-conditions

26 S.Afr.J.Afr.Lang.,2006, 1

Mid-vowels and vowel harmony in Civili

Hugues Steve Ndinga-Koumba-BinzaCentre for Language and Speech Technology, Stellenbosch University,

Private Bag X1, Matieland, 7602, South [email protected]

This article highlights the phenomenon of vowel harmony in Civili. The distinctive features of vowels are determined in order to analyze the vowel harmony system, and later to explain the distinction between respective problematic vowels in the vowel system. Vowel harmony in Civili is elaborated according to these propositions for vowels’ distinctive features. For vowel harmony in Civili both [+ATR] and [-ATR] are required. On the assumption that vowel harmony is feature filling, a vowel in Civili is unspecified for +ATR/-ATR and receives its value for this feature by spreading.

IntroductionPrevious researchers of the Civili phonological system (Marichelle, 1902; Ndamba, 1977; Blanchon, 1990 and Mabika Mbokou, 1999) noted a certain allophonic relationship between [e] and [ε] on the one hand, and between [o] and [] on the other. The object of this paper is to highlight the phenomenon of vowel harmony in Civili. To this purpose, the relationships between the different vowels concerned are analyzed within an autosegmental framework.

The present study is the continuation of research started in a Master’s dissertation (Ndinga-Koumba-Binza, 2000). Here the problem is re-examined meticulously with contributions from a wide range of new sources in order to confirm or negate previous analyses and conclusions. The results reached affirm the aims of the Masters research and give more clarity on the vowel harmony tendency on the relationships between [e] and [ε], [o] and [].

Dubois et al. understand vowel harmony as:

un phénomène d’assimilation vocalique qui peut jouer sur plusieurs voyelles d’un même mot: le choix d’une ou de plusieurs voyelles dans une position donnée n’est pas libre, mais il est déterminé automatiquement par la présence d’une autre voyelle déterminée. (1994:230)

‘A phenomenon of vowel assimilation that can occur on many vowels of the same word: the choice of one or a number of vowels in a given position is not free, but it is automatically determined by the presence of another determined vowel.’

In other words, it is a phenomenon of vowel assimilation that can concern several vowels in the same word. A determined vowel influences the choice of other vowels for a given place in a word, and/or even a phrase:

[v]owel harmony can be seen as a subclass of long-distance assimilation. Many languages exclude certain combinations of vowels in the word. (Gussenhoven & Jacobs, 1998:191)

Francis Katamba (1989:211) provides a more accurate definition when he states:

[v]owel harmony is a process whereby within a certain designated domain, usually the word, all vowels are required to share one or more phonological properties. The vowels of a language are divided into two mutually exclusive sets and all vowels within a stipulated domain must be, say, either front or back, high or low, rounded or unrounded, etc.

Dow

nloa

ded

by [

Nor

th W

est U

nive

rsity

] at

03:

55 2

3 A

pril

2014

27S.Afr.J.Afr.Lang.,2006, 1

This article acknowledges that vowel harmony is present in Civili and investigates the behaviour of this phenomenon in this language. The main goal here is finally to show how phonological fundaments of the problematic phenomenon can be determined. However, it is important to point out that literature about vowel harmony in Bantu languages is sparse and the treatment of vowel harmony in these languages is in its infancy. Janssens (1999), for instance, does not recognize this phenomenon as structural in Bantu languages.

Civili is spoken in Gabon, mostly in two towns in Nyanga province: Mayumba (3˚23´S, 10˚38´E) and Ndindi (3˚46´S, 11˚10´E). This language, classified by Malcolm Guthrie (1948) as H.12a, is a widely spoken language in the Republic of Congo, in the Democratic Republic of Congo, in the whole enclave of Cabinda and in Angola. All the data analyzed in this paper was collected from native speakers in Mayumba and Libreville (the capital city of Gabon).

Outline of the vowel systemIt is necessary to determine the distinctive features of vowels in order to analyze the vowel harmony system. This article will mostly concern itself with distinctive features in the analyses. It should be noted that

the term distinctive feature is normally understood by linguists to refer to features established within the domain of phonology and – almost universally – to a characterization of speech sounds in phonetic terms. (Brasington, 1994:1042)

Civili has a system of five underlying short vowels (Ndinga-Koumba-Binza, 2000; 2004). The vowel chart in Table 1 displays phonological descriptions of the five underlying short vowels. Only the articulation and aperture natures adapting the standard chart that François Dell (1973:59) proposes are taken into account.

Table 1: Phonological vowel chart of Civili

Articulation

Front Central Back

1st degree i u

2nd degree e o

3rd degree a

Generative phonologies acknowledge that, just as every language sound is definite in the list of its specifications with regard to various pertinent features, the phonemes of a language are characterized by ‘features’ specifications columns’ (Dell, 1973:52, 72). Civili vowels can thus be described with the distinctive features below:

• /i/ is a high vowel: [hi]• /e/ is unspecified vowel• /a/ is a low vowel: [lo]• /o/ a round vowel: [rd]• /u/ is both round and high: [rd, hi].

We will see later that vowels /e/ and /o/ can be underspecified when we regard relations between the phones [e] and [ε], and the phones [o] and []. Note that the vowel system of Civili seems to be structured according to these three features [hi, rd, lo] as shown in Table 2.

Dow

nloa

ded

by [

Nor

th W

est U

nive

rsity

] at

03:

55 2

3 A

pril

2014


Table 2: Distinctive features in Civili vowels

[Front] [Back] [rd]

[hi] i u closed unspecified e o mid [lo] a open

Distinctive features are characterized by a hierarchical structure that is universal. According to Carole Paradis:

[u]n segment est composé principalement d’un nœud de base, auquel sont associés des traits terminaux, un nœud laryngal et un nœud de place. Le nœud de place domine les articulateurs. Dans ce modèle, dont la structure générale est universelle, les articulateurs ont un statut spécial non seulement parce qu’ils sont unaires mais aussi parce que, contrairement aux autres nœuds, ils représentent des paliers indépendants (non ordonnés) les uns par rapport aux autres. (1993: 30)

‘a segment is mainly composed of basic root, to which terminal nodes are associated, a laryngeal knot and a place knot. The place knot dominates the articulators. In this model, of which the general structure is universal, articulators have a special status not only because they are unary but also because, contrary to other knots, they represent tiers independent (randomized) from one another.’1

The hierarchy of distinctive features of Civili vowels can be represented in Figure 1.

Figure 1: Hierarchy of distinctive features

× = timing unit

Cavity nasal oral ( = place) pharyngeal

Front labial coronal dorsal

[rd] [back] [hi] [lo]

From these five vowels, the Civili vowel system works according to phonological processes, the result of which are the following phones, which also present long counterparts and nasalized correspondants in some nasal environments (Ndinga-Koumba-Binza, 2000; 2004).

(1a) [i] [bílílì] ‘lips’(1b) [e] [ndébù] ‘net’(1c) [ε] [fáyεt] ‘tailor’(1d) [a] [bbákl] ‘men’(1e) [] [símbt] ‘stars’

Dow

nloa

ded

by [

Nor

th W

est U

nive

rsity

] at

03:

55 2

3 A

pril

2014


(1f) [o] [líkósì] ‘nape (of the neck)’(1g) [u] [múnù] ‘mouth’

[i] and [u], as we can see in the above data, also appear in their respective devoiced forms [] and [], and [a] in its reduced form transcribed [] (Ndinga-Koumba-Binza, 2000:89-95).

This section focuses on working with only the voiced oral and precise forms noted in Table 3 below. However, every counterpart (long, nasalized devoiced or reduced) of each voiced oral vowel behaves like it with respect to vowel harmony.

Table 3: Voiced oral phones of Civili

Front Central BackClosed i uMid-closed e oMid-open ε

Open a

It must be noted that the phonological identity of vowel units listed in Table 1 comes from minimal contexts of commutation called minimal pairs. Illustrations of the pairs in question follow in example (2).

(2a) /i/ vs /e/ /ndíbù/ ‘collar’ /ndébù/ ‘net’

(2b) /o/ vs /i/ /kúbólà/ ‘to rot’ /kúbílà/ ‘to greet’

(2c) /u/ vs /o/ /kúkúsì/ ‘to lengthen’ /kúkósì/ ‘to put into’

(2d) /a/ vs /o/ /kukátùkà/ ‘to break away’ /kukótùkà/ ‘to wake up’

Only the criterion of pertinence through commutation for identifying these vowels as structural units of Civili is considered (Ndinga-Koumba-Binza, 2000:42-43; 2004). However, bringing out only the distinctiveness of a segment at the conclusion of minimal pairs is not enough to determine its structural nature or underlying form, for this segment can appear from phonological processes in the language. That is so, for instance, in the case of long vowels in Civili (Ndinga-Koumba-Binza, 2000).

In Civili there is no contrast between [ε] and [e], or between [] and [o]. Nevertheless it has been shown in several languages that a segment, which does not enter any commutation, can be confirmed as an underlying unit in a given language for the simple reason that it does not derive from any phonological process of the language, or from any phenomenon of linguistic borrowing. Accordingly, we can thus be tempted to regard |ε| and || as underlying segments since it seems, as will be observed in the following section, that there is no rule accounting for their realization. It will be shown, however, that this cannot be the case.

Vowel harmony is treated within the framework of phonological analysis. Therefore, only phonological analyses display treatments on vowel harmony (Ringen, 1977; Katamba, 1989; Wald, 1992). Furthermore, as Benji Wald (1992:159) remarks, ‘[p]honologically, for most of Bantu languages, a form of vowel harmony applies to certain derivational suffixes: a mid-vowel suffix is selected following a mid-vowel stem.’

Dow

nloa

ded

by [

Nor

th W

est U

nive

rsity

] at

03:

55 2

3 A

pril

2014


This article observes that this phenomenon occurs in Civili and hence needs to deal with phonetic considerations of the language. As mentioned above, phonologically [ε] and [] might be merely variants of /e/ and /o/ respectively. This issue is discussed in the next section. For the rest of this article it is important to regard |ε|, ||, |e| and |o| as being at the same phonemic level; in the phonological view that regards them as various allophones it may not be possible to observe analytically the problematic phenomenon.

Preliminaries to vowel harmony This section will state the impossibility of grouping together as allophones Civili vowels according to rules as suggested by the aforementioned authors. A new analysis of Civili data and a minute observation of their proposed allophonic rules will show that the problematic vowel relationship should rather be examined in a framework of other phonological processes, such as vowel harmony.

Allophonic variation in the phonological context is often described to refer to the relationship between phones grouped together as allophones of a particular phoneme (Katamba, 1989:20; Gussenhoven & Jacobs, 1998:55). So, the term ‘phoneme’ will refer to ‘the segment category that the various allophones are variants of’ (Gussenhoven & Jacobs, 1998:55). In other words, an allophone is every phone that is to be one of the particular realizations of a specific phoneme. Allophones are necessarily context-dependent variants.

It was mentioned earlier that vowels of the 2nd degree of aperture (/e/ and /o/, /ε/ and //), are hard to specify into distinctive features. With regard to the phonetic representations of the vocalic system we can realize that the phones [e] and [o] are respectively near to phones [ε] and [], such as is the case in many languages of the world (Ladefoged, 1993; Ladefoged & Maddieson, 1996). In the specific case of Civili, [e] and [o] are less frequent than [ε] and []. Moreover, the phonic units [e] and [ε] on one hand, and [o] and [] on the other hand do not permute with each other. In other words, they do not appear in shared word-contexts which might permit significative contrasts. That makes both of them allophones of respective phonemes /e/ and /o/ according to the classical principle, as discussed above, and as is shown in Ndamba (1977), Blanchon (1990) and Mabika Mbokou (1999).

However, these allophones are not related to each other by any segmental complementary distribution which might permit an elaboration of an allophonic segmental rule. Since the establishment of Ndinga-Koumba-Binza’s framework (2000), analysis does not amount to a rule that might confirm the allophonic relation of these various segments; an alternative framework needs to be proposed. And yet, unlike her predecessors (Ndamba, 1977; Blanchon, 1990), Mabika Mbokou (1999: 29-30) formalizes the distribution of these phones by stating:

‘[l]e phonème /e/ a une représentation phonétique [ε] dans le contexte où il se trouve en position médiane...; il est [e] partout ailleurs’ ([t]he phoneme /e/ has a phonetic representation [ε] when it is in a median position …; it is [e] everywhere else).

Figure 2: Allophonic relationship between [ε] and [e]

/e/ [ε] / C_C

[e] / Elsewhere

Likewise she says , ‘[l]e phonème /o/, lorsqu’il porte un ton bas s’ouvre et se réalise []’ (ibid.:30) ([t]he phoneme /o/ opens when it bears a low tone and it is realized []).

Dow

nloa

ded

by [

Nor

th W

est U

nive

rsity

] at

03:

55 2

3 A

pril

2014


Figure 3: Allophonic relationship between [o] and []

/o/ [] / C_ C

[o] / Elsewhere

This rule would be sufficient if we were concerned only with phonetic representations for a phonological minimal-pairs analysis. In a quite similar case in Zulu, Poulos and Msimang confirm such formalism seen in Mabika Mbokou when they argue that ‘it has been observed that mid-vowels in these examples are sometimes heard as vowels which are slightly higher on the vowel chart’ (1998:512). They equally conclude that an allophonic situation prevails: ‘[i]n other words, instead of [ε] and [], slightly different vowels are heard and these would be represented on the vowel chart by [e] and [o] respectively.’(ibid.:512). Poulos and Msimang’s reasons for comingto such a conclusion are:

the differences in vowel quality are slight in each case, but the important point to note here is that the substitution of [e] for [ε] and [o] for [] does not in any way affect the meaning of the word. (1998:512)

Unfortunately, this formalization is certainly not confirmed by further observations made about Civili. Indeed, the explanatory values given for this allophonic variation are not very persuasive.

First, the between-consonant position C_C of a vocalic segment cannot influence its quality in so far as the analysis made on this language does not show any close syllable such as CVC (Mabika Mbokou, 1999:39; Ndinga-Koumba-Binza, 2000:66). Indeed, it can be observed that the minimal structure of the Civili syllable is the unique vowel V. In Civili, a syllable can start with a consonant but ends only with a vowel. This means therefore that the Civili syllable is not closed but open, that is, only CV is attested.

Hence, a consonant that follows a vowel in a Civili word cannot influence the vowel that precedes it, because neither is from the same syllable. The vowel belongs to the previous syllable, and the consonant starts the following syllable for there is no coda in the Civili syllable.

All of these observable considerations are seen in examples (3a-h) below, which negate any allophonic relationship between [e] and [ε] and between [o] and [] as elaborated according to the rules illustrated in Figures 2 and 3.

(3a) [kúdéfì] ‘to owe’(3b) [chè:mbì] ‘disorder’(3c) [kúlεf] ‘to swear’(3d) [mε:s] ‘table’(3e) [líbótì] ‘benefaction’(3f) [mbt] ‘stars’(3g) [líb:l] ‘basin’(3h) [lílò:ndù] ‘wharf’

The interaction between tonal system and segmental phenomena In the case of Civili, segmental phenomena often influence the tonal system, and not the other way around (Ndinga-Koumba-Binza, 2000:118-119; 2003). This is not unique to Civili but is a principle observed in all languages (Clements & Goldsmith, 1984; Miti, 2002). For instance, Ndinga-Koumba-Binza (2003) has shown that in Civili

Dow

nloa

ded

by [

Nor

th W

est U

nive

rsity

] at

03:

55 2

3 A

pril

2014


some cases of vowel lengthening provoke a tonal modulation on the lengthened vowel. Therefore it is impossible in this language for a back closed vowel that bears a low tone to open because of the lowness of this tone. It could be speculated that the lowness trait of a tone can contribute towards opening a closed vowel. But while such a phenomenon may occur in certain languages, that is not the case in Civili. Many examples, such as those following, show the back closed vowel [o] and the back open vowel [] in various contexts where each one of them bears a low tone.

(4a) [mbt] ‘stars’(4b) [lílò:ndù] ‘wharf’(4c) [lítòlù] ‘cheek’(4d) [k] ‘snake’(4e) [tíbòfù] ‘chin’

On the other hand, substituting [e] for [ε] and [o] for [] does not in any way affect the meaning of the word,2 but it remains the case that, in this language, the sequences of syllables / – u/ or / – i/, and /ε: – u/ or /ε: – i/, are unworkable in the same word.3 That makes it impossible to find minimal pairs /e/ vs. /ε:/ or /o/ vs. // in contexts where the final vowel is high. It is also important to notice that such substitutions, [e] for [ε:] and [o] for [], are frequent in Civili. However, any substitution of this kind entails automatically the change of the final high vowel (generally [i]) into a low one (only /a/ realized as []). This is shown in examples (5a-c).

(5a) /cósì/ ‘cold’ can be realized both [tó:sì] and [t:s] with the same meaning, but it will never be *[t:sì] nor *[tó:s].

(5b) /mbélì/ ‘knife’ can be realized both [mbé:lì] and [mbε:l] with the same meaning, but it will never be *[mbε:lì] nor *[mbé:l].(5c) /òsì/ ‘honey’ can be realized both [òsì] and [s], but it will never be *[sì] nor *[òs].

It should be noted that this does not mean that the vowel [i] becomes [], but we can see here that it is normally a case of co-occurrence. Certain vowels occur between them and others are not co-occurrent, and cannot both appear in the same word.

Nevertheless, it would be erroneous to commutate segments that do not enter at all in the same environments and, finding the commutation to be impossible, conclude that these segments are allophones. If this procedure can work in analyses of many languages (while commutation remains the classical principle for bringing out distinctive segments), it is not so in all cases of phonic units.

Analyzing the co-occurrences of problematic phones in a context of dissyllabic or polysyllabic nominal stems leads to the observation that Civili refuses any co-occurrence such as CV1CV2 in which V1 is a low intermediate vowel [ε] or [] and V2 a high vowel [i] or [u]. Likewise it is impossible to have the sequences CV1CV2 in which V1 is an intermediate high vowel (/e/ or /o/) and V2 a low one (/a/, generally in its variant form []4). These phenomena are anticipatory: [e] and [o] appear in V1 only if V2 is high, and [ε] and [] appear in V1 only if V2 is low. Table 4 presents the co-occurrences of vowels in Civili for dissyllabic nominal and verbal stems.

Dow

nloa

ded

by [

Nor

th W

est U

nive

rsity

] at

03:

55 2

3 A

pril

2014


Table 4: Co-occurrences of vowels in Civili for dissyllabic and verbal stems

V2 i a u

V1

i + + + e + − + ε − + − a + + + − + − o + − + u + + +

Table 4 shows mid vowels [e] and [o] appearing only with closed vowels [i] and [u]. In other words, they combine only with high vowels as V2 and generally in the final position of the stem.5 However, [i], [u] and [a] are co-occurrent in any position.

From Table 4 it can also be observed that the co-occurrences below are impossible.

• e…a

• ε…i

• …i• o…a

• ε…u

• …u

The appearances that are strictly possible are those that are found in examples (6 a-o).

(6a) e...i [mbé:lì] ‘knife’(6b) o...i [ndo:si] ‘dream’

(6c) a...i [bútálì] ‘iron’(6d) i...i [cílílì] ‘lip’(6e) u...i [nzú:njì] ‘bile’(6f) e...u [cívékù] ‘tsetse fly’(6g) o...u [misobù] ‘earthworms’(6h) a...u [búbà:lù] ‘nastiness’(6i) i...u [nlì:lù] ‘place in a river where there are many fish’(6j) u…u [cínúnù] ‘old man’(6k) ε...a [cε:s] ‘luck’(6l) ...a [líb:l] ‘basin’(6m) a...a [mtá:t] ‘fathers’(6n) i...a [mbíl] ‘kind of fish’(6o) u...a [mbúl] ‘palm wine’

We have just clarified relations, occurrences and contexts of the vowels [e] and [ε], [o] and []. Nevertheless, in some cases the following questions still remain:

Dow

nloa

ded

by [

Nor

th W

est U

nive

rsity

] at

03:

55 2

3 A

pril

2014


• How can the distinction between respective problematic vowels in the vowel system be explained?• Why are [e] and [ε], and [o] and [], not found in the same contexts of appearance?

The hypothesis may be considered of closing mid-open vowels |ε| and || into [e] and [o] respectively in the environment of a closed vowel [i] or [u], according to an assimilation phenomenon. In this case it might be a matter of posing as structural forms the vowels |ε| and ||6 which are realized as closed, [e] and [o] in assimilation on the feature of closeness to the closed vowels |i| and |u|. Such a hypothesis would more easily explain the allophonic relationship of the vowels under consideration here. But while such a rule can be formulated, it will remain unconvincing for two reasons. First, the assimilation phenomenon on which it leans is a long-distance one. Furthermore, this rule neither explains appearances’ distinction of assumed allophones nor explains impossible phonotactic appearances. Therefore it is advisable to think of another phonological process rather than of a mere closing of structural vowels.

Vowel harmonyIt seems appropriate at this point to revisit the specification of vowels into distinctive features. It is assumed that all vowels |i e ε a o u| are in the phonemic level as recommended previously. For this reason, this analysis on the issue of vowel harmony cannot be based on the features chart preliminarily seen in Table 2, because of missing |ε| and ||.

Vowel harmony in Civili is elaborated according to the following propositions for vowels’ distinctive features.

Table 5: Distinctive features of Civili vowels

Vowel Distinctive features

i [hi], [Advanced Tongue Root (ATR)]

u [hi], [ATR], [rd]

e unspecified

o [rd]

ε unspecified

[rd]

a [lo]

From these specifications and from the possibilities of phonotactic appearances observed in example (6), it is possible to divide the vowels of Civili into two subsets: Advanced Tongue Root (ATR) vowels and non-ATR vowels. These subsets suggest a fragmented vowel harmony system based on the ATR feature such as in Figure 4.

Figure 4: Fragmented vowel harmony

+ATR -ATR

i u

e o ε

a

Dow

nloa

ded

by [

Nor

th W

est U

nive

rsity

] at

03:

55 2

3 A

pril

2014


In this fragmented vowel harmony, some explanations about the possible phonotactic appearances observed in example (6) can be found. But, it does not explain the impossible phonotactic appearances observed from Table 4. However, the harmony system is still not explained.

It has been observed from Tables 2 and 5 that |e| and |ε|, and |o| and ||, are distinctively unspecified. We can therefore assume an underspecification that will display the Civili vocalic system as follows:

Figure 5: The Civili vocalic system assuming underspecification

[hi, ATR] i u [hi, rd, ATR]

E O [rd]

a [lo]

/E/ and /O/ are underspecified vowels.

Harmony may be regarded here as a state in which vowels agree in that they all have or lack a particular element. The preceding definition will, following Harry van der Hulst and Jeroen van de Weijer (1995:504), be conceptualized according to the framework of unary component theory:

the unary approach can be regarded as a radical version of radical underspecification theory. Essentially, its claim is that one and the same value is active across languages and that default values are never phonologically active in any part of the phonology. (van der Hulst & van de Weijer, 1995:504)

In this way it might not be acceptable to formulate a vowel harmony according to the ATR feature as previously indicated in Figure 4. However, it could be pointed out with Donca Steriade that:

[t]he behavior of ATR, [high] and [back] is significantly different from that of the features reviewed so far. There exist good examples of processes engaging either value of these features ... then it is impossible to claim ATR is permanently and universally a privative feature with a cross-linguistically constant phonetic implementation. We may suggest that ATR’s privativity is language specific, but this hypothesis leads to a notational variant of the claim that the feature is binary but possesses reversible markedness. ... Let us accept then that the tongue root feature is equipollent. (1995:149)

For vowel harmony in Civili both [+ATR] and [-ATR] are required. The features of ATR are regarded here as features that reinforce, but they are not basic features as are [hi] or [lo]. Without the feature of ATR, harmony in Civili might be defined according to the assumption that would say that in a context of a [hi] vowel, the underspecified is realized mid-closed; and in the context of [lo] vowel it is realized mid-open. This is formulated in Figure 6.

Dow

nloa

ded

by [

Nor

th W

est U

nive

rsity

] at

03:

55 2

3 A

pril

2014


Figure 6: Harmony in Civili without ATR

/E/ [e] / _ [hi] /O/ [o] / _ [hi]

[ε] / _ [lo] [] / _ [lo]

Autosegmentally observed, vowels, according to the preceding formalism, will display as put together according to the features [hi] and [lo]. This may be an object of assimilation.

On the assumption that vowel harmony is feature filling, a vowel in Civili is unspecified for +ATR/-ATR and receives its value for this feature by spreading. It should be noted that vowels that are underspecified for ATR may effectively acquire the feature [hi] and [lo] phonotactically without changing their phonemic value.

It is universally agreed that +ATR implies [high], and -ATR implies [low] as well. Therefore, Figure 6 can be revised into autosegmental configurations that display the spreading of the features’ value onto an underspecified vowel. It is important to note that +ATR is the active spreading value, and -ATR the default value. -ATR is interpreted in the absence of +ATR.

In such a situation harmony in Civili may be defined as a state in which vowels agree with respect to their value for the feature ATR within a domain of the nominal or verbal stem. The following configurations display the spreading feature showing the orientation of the harmonizing phenomenon.

Figure 7: Spreading of the feature active value

(a) ATR (b) ATR

[E i] [O i]

hi hi

[e........i] [o........i]

(c) ATR (d) ATR

[E u] [O u]

hi hi

[e........u] [o........u]

Dow

nloa

ded

by [

Nor

th W

est U

nive

rsity

] at

03:

55 2

3 A

pril

2014


This phonological phenomenon provides the following examples in Civili:

(7a) /lúvèmù/ [lúvè:mù] ‘wind’(7b) /cílesi/ [cilesi] ‘younger’(7c) /libèyì/ [líbèyì] ‘mockery’(7d) /ntókù/ [nthókù] ‘ground’(7e) /sísònì/ [sísònì] ‘shame’

Figure 8: Spreading the feature default value

(a) -ATR (b) -ATR

[E a] [O a]

lo lo

[ε........a] [........a]

Examples (8a-d) illustrate this.

(8a) /kútékà/ [kútεk] ‘to draw’(8b) /nsésà/ [nsε:s] ‘rods broom’(8c) /zòlà/ [zl] ‘hour, time’(8d) /kúlò:sà/ [kúls] ‘to shoot (gun)’

ConclusionA process of vowel harmony in the whole system better explains the relation between specific vowels [e] and [ε], and between [o] and [], and not merely an allophonic regulation. But for this, all the phonic forms should be regarded as structural. That is to say that we admit as underlying forms the structures |i e ε a o u|. Thus vowel harmony is only explained when searching inside of vocalic distinctive features that underspecify |e| and |ε| as /E/, and |o| and || as /O/, for the feature ATR that implies [high]. Finally, mid-vowels harmonized into mid-closed (high) when the feature ATR respectively spreads onto the unspecified segment or mid-open (low) by default interpretation in the absence of the active value of ATR.

This implies that vowel harmony in Civili comes from the underlying structure of the language sound system. Theoretically, though, all of these phonological assumptions leave the vowel harmony phenomenon in Civili merely as a construction without a concrete segmental rule, which might be verifiable through phonetic data.

We really need other kinds of analyses such as those in experimental phonology in order to confirm or negate the underlying character of both |e| and |ε|, and |o| and ||. This might imply that this problem should be analyzed within experimental phonetic procedures; that is to say, within an acoustic-perceptual framework. Furthermore, experimental analyses might answer these questions:

• is vowel harmony confirmed or negated?• if confirmed, what is a proper response to the problem?• if not, what is it and how should research proceed on this issue?

Dow

nloa

ded

by [

Nor

th W

est U

nive

rsity

] at

03:

55 2

3 A

pril

2014


Without providing specific answers, the discussion in this article has provided a basis and a way forward for the analysis of mid-vowels and vowel harmony in Civili.

AcknowledgementsI am thankful to Dr. Myles F. Leitch who brought this issue of Civili mid-vowels to my attention. I am very grateful to Prof. J.C. Roux for his insightful comments and suggestions on an early version of this paper, and for encouraging me to write scientific papers. I also convey thanks to the anonymous referees whose valuable remarks have helped to improve this paper. I, however, take full responsibility for any error or mistake that may be found here.

Notes1. Translations in this article are by Ndinga-Koumba-Binza.

2. This allows Poulos and Msimang (1998:512) to justify allophonic variants of these vowels in Zulu.

3. In Civili only the high vowels /u/ and /i/, and the low vowel /a/ are found in final position of dissyllabic and polysyllabic stems (Ndinga-Koumba-Binza, 2000).

4. Ndinga-Koumba-Binza (2000:94-95) indicates that [a] reduces into [] when it is in an unaccented syllable.

5. In Civili, nominal and verbal stems are dissyllabic (Marichelle, 1902; Ndamba, 1977; Blanchon, 1990; Mabika Mbokou, 1999; Ndinga-Koumba-Binza, 2000).

6. The forms [e] and [] are considered as structural forms (rather than [e] and [o]) because they are less marked and appear in the greatest number of contexts. That is because the most frequent final vowel of a stem is [] in Civili.

ReferencesBlanchon, J.A. 1990. Civili. Revue Gabonaise des Sciences de l’Homme 2:141-142.

Brasington, R. 1994. Distinctive features, in The encyclopedia of language and linguistics (Vol. 2), edited by R.E. Asher. Oxford/New York: Pergamon Press:1042-1050.

Clements, G.N. & J.A. Goldsmith (Eds). 1984. Autosegmental studies in Bantu tone. Dordrecht: Floris.

Dell, F. 1973. Les règles et les sons: introduction à la phonologie générative. Paris: Hermann.

Dubois, J., Giacomo, M., Guespin, L., Marchellesi, C., Marchellesi, J.B. & Mevel, J.P. 1994. Dictionnaire de linguistique et des sciences du langage. Paris: Larousse.

Gussenhoven, C. & Jacobs, H. 1998. Understanding phonology. London: Arnold.

Guthrie, M. 1948. The classification of the Bantu languages. Oxford University: Press for International African Institute.

Janssens, B. 1999. Diffusions des traits phonologiques. Available: http://www.unice.fr/CRI/html/ comparlingafrica/1999. Accessed on 2002/10/01.

Katamba, F. 1989. An introduction to phonology. London: Longman.

Ladefoged, P. 1993. A course in phonetics (3rd Edition). Orlando, Florida: Harcourt Brace College Publishers.

Ladefoged, P. & Maddieson, I. 1996. The sounds of the world’s languages. Oxford: Blackwell.

Marichelle, C. 1902. Dictionnaire Vili-Français. Loango: Imprimerie de la Mission.

Dow

nloa

ded

by [

Nor

th W

est U

nive

rsity

] at

03:

55 2

3 A

pril

2014


Mabika Mbokou, L. 1999. Les phénomènes analogiques en civili : études phonologiques et morphologiques. Unpublished Masters dissertation. Libreville: Université Omar Bongo.

Miti, L.M. 2002. Aspects of Cisenga tonology. Cape Town: The Centre for Advanced Studies of African Society (CASAS).

Ndamba, J. 1977. Syntagme nominal et groupe nominal en vili (H12): langue du Congo. Unpublished Doctoral thesis. Paris: Université Paris 3-Sorbonne Nouvelle.

Ndinga-Koumba-Binza, H.S. 2000. Phonologie du civili de Mayumba: langue bantu du Gabon (H12a). Unpublished Masters dissertation. Libreville: Université Omar Bongo.

Ndinga-Koumba-Binza, H.S. 2003. ‘An aspect of the Civili tonal system: tone formation through vowel lengthening.’ Paper presented at 12th Biennial International Conference of the African Language Association of Southern Africa (ALASA), University of Stellenbosch: 7-11 July 2003.

Ndinga-Koumba-Binza, H.S. 2004. Vowel Duration Issue in Civili. South African Journal of African Languages 24(3):189-201.

Paradis, C. 1993. Phonologie générative multilinéaire, in Tendances actuelles en linguistique générale, edited by J.P. Nespouslous. Neuchâtel/Paris: Delachaux et Niestlé:11-47.

Poulos, G. & Msimang, C.T. 1998. A linguistic analysis of Zulu. Cape Town: Via Afrika.

Ringen, C. 1977. Vowel harmony: implications for the alternation condition, in Phonologica 1976, edited by W.U. Dressler & I.E. Pfeiffer. Innsbrucker Beiträge zur Sprachwissenschaft:127-132.

Steriade, D. 1995. Underspecification and markedness, in The handbook of phonological theory, edited by J.A. Goldsmith. Cambridge, MA: Blackwell Publishers:114-174.

Van der Hulst, H. & van de Weijer, J. 1995. Vowel harmony, in The handbook of phonological theory, edited by J.A. Goldsmith. Cambridge, MA: Blackwell Publishers:495-534.

Wald, B. 1992. Bantu languages, in International Encyclopedia of Linguistics (Vol. 1), edited by W. Bright. Oxford: Oxford University Press:157-160.

Dow

nloa

ded

by [

Nor

th W

est U

nive

rsity

] at

03:

55 2

3 A

pril

2014

Mid-vowels and vowel harmony in Civili

Documents