Navigation – Plan du site

AccueilNuméros8(1)Moraic tone-bearing units in Kabiye

Moraic tone-bearing units in Kabiye

David Roberts

Résumés

Cet article démontre que la more, plutôt que la syllabe, est l’unité porteuse de ton en kabiyè, une langue gur (gurunsi oriental) du Togo. Après avoir présenté l’inventaire phonémique et un bref aperçu du système tonal, le concept de poids phonologique est exposé, de manière à caractériser avec précision les notions de more et de syllabe telles qu’elles se présentent dans cette langue, ainsi que la relation qu’elles entretiennent. Deux observations plaident en faveur de la more comme unité porteuse de ton : d’une part, les tons modulés sont autorisés sur les syllabes lourdes mais pas sur les légères ; d’autre part, les processus tonals sont sensibles à une unité plus petite que la syllabe, de sorte que le second constituant d’une syllabe lourde opère avec un certain degré d’indépendance à l’égard du premier. Ces observations sont examinées pour quatre phénomènes phonologiques : l’association du schème tonal BHB, l’association des schèmes bitonals, la propagation du ton bas et l’élision vocalique. Deux processus post-lexicaux – le réalignement des contours tonals et le nivellement du schème tonal HBH – constituent des preuves potentielles en faveur de la syllabe en tant qu’unité porteuse de ton mais, même dans ces cas, une explication moraïque est en définitive jugée préférable.

Haut de page

Entrées d’index

Mots-clés :

kabiyè, langues gur, more, syllabe, ton
Haut de page

Texte intégral

1. Introduction

1Several overviews of tone in African languages concur with the importance of identifying whether it is the syllable or the mora that is the tone-bearing unit (TBU) when tackling tone analysis (Clements 2000: 139-144; Odden 1995: 448; Rialland 2003: 105-129; Yip 2002: 74ff.). Gur languages in which it has been claimed that the syllable is the TBU are numerous, including Buli (Schwarz 2001: 121), Dagaare (Anttila & Bodomo 1996), Dagbani (Hyman 1993), Kɔnni (Cahill 2004: 3), Mbelime (Melick 2012a: 104), Vagla (Crouch 1985), Moore, and Lama (Kenstowicz et al. 1988). Gur languages in which it has been claimed that the mora is the TBU, on the other hand, are far less numerous. Apart from Kabiye — the Eastern Gurunsi Gur language of Togo that is the focus of the present study — the only ones I am aware of are Kusaal (Musah 2017) and Gurenɛ (Adongo 2018), although Melick (2012b), in a reanalysis of Mbelime, evokes certain moraic association constraints while still maintaining a syllabic approach. For Kabiye, the claim that the mora is the TBU has not yet been adequately substantiated so it will be helpful to survey the previous research and evaluate the positions of the various contributors.

2Delord’s (1976) pioneering grammar gives considerable space to tonal phenomena, but because it predates the widespread use of autosegmental theory it does not even mention the concept of the TBU and is therefore of limited value to the topic of the present paper. Particularly misleading, as Padayodi (2010: 77) points out, is its interpretation of two-tone patterns on long syllables as contours rather than sequences of level tones, and this is at odds with the position of all ensuing researchers.

3Lébikaza (1999) is the first to tackle the question of the TBU explicitly. For him, short syllables contain one mora, long syllables are composed of two moras, and a morpheme boundary may occur between the latter. All ensuing researchers concur on these points (Kassan 1996: 31f.; Padayodi 2010: 40; Roberts 2016: 137). Lébikaza claims that this is evidence for the mora rather than the syllable as the TBU (1999: 171f.). However, his discussion of tonal association comparing unambiguous and ambiguous forms (1999: 191) is not as persuasive as it might be because it does not compare an alternative syllabic approach. We will return to this issue in detail in section 5.

4Kassan (1996: 31f.) does not explicitly define the TBU. Kassan (2000) mentions it in passing but her prose alternates between references to moraic (2000: 15), and syllabic (2000: 15, 16, 17) tonal associations for no clear reason. Roberts (2003: 95) states that the TBU can be V as well as preconsonantal and final N but without specific reference to syllables or moras. In three more recent analyses, Roberts (2016: 137; 2019: 1; 2021: 11) states more explicitly that the tone-bearing unit is the mora, but with no justification other than citing Lébikaza (1999: 170). Padayodi (2010: 249ff.) concurs with previous researchers in identifying the mora as the TBU but again with no particular justification. The autosegmental representations in all these analyses are consistent with a moraic approach.

5In contrast to the agreement concerning the nature of the TBU, when it comes to the direction of tonal association there is no consensus among researchers. Lébikaza (1999: 204) considers it to be right to left (1999: 204), whereas Padayodi (2010: 256) takes the opposite view. As for Roberts (2016: 148), he considers the obligatory contour principle (OCP) to be inoperative in nouns, thereby obviating the need to stipulate the direction of tonal associations.

6In sum, all the preceding researchers working on Kabiye tonology — apart from Delord (1976) who doesn’t mention the TBU at all — adopt a moraic approach and do so consistently, but at no point do they adequately justify their decision to do so, nor do they contrast their choice with an alternative syllabic analysis and demonstrate why they reject the latter. The novel contribution of the current paper is that it is the first to systematically lay out both possible analyses — moraic and syllabic — in parallel, in order to rigorously contrast and evaluate their relative merits. It concludes that a moraic approach is indeed preferable because it accounts for the tonal facts more lucidly and economically than a syllabic approach.

  • 1 Toolbox is an integrated data management and analysis program (accessed 17 September 2022).

7The analysis is based on 800 nouns and verbs selected from the Kabiye-French dictionary (CLNK 1999). The deselected words (compounds, proper names, borrowings etc.) do not add any facts that contradict the analysis presented in this paper. Ideophones and exclamations were not excluded because their tonal patterning matches the rest of the analysis. Conjugated forms were elicited during the author’s own fieldwork. All data were verified with at least two L1 speakers and stored in Toolbox.1 Throughout the paper, H tone is marked with an acute accent, L tone with a grave accent and non-automatic downstep with a superscript arrow []. The mora and the syllable are represented with the symbols μ and σ respectively. Underlying forms are cited between /forward slashes/, surface realisations between [square brackets] and orthographic data between <chevrons>. The phonemic and phonetic transcriptions in this paper represent tone exhaustively (cf. Padayodi 2010), rather than the widely used convention of leaving one tone unmarked (cf. Lébikaza 1985 and 1999; Kassan 1996 and 2000). This seems preferable in an analysis that needs to clearly show the distinction between syllables and moras, and in a family like Gur where the number of phonemic level tones may vary from language to language.

8The paper is structured as follows. Section 2 presents the segmental phonemic inventory and an overview of the tone system. Section 3 introduces the concept of phonological weight in order to characterise the mora and the syllable. Section 4 presents four domains in which a moraic analysis is deemed to be preferable. Section 5 concedes two domains which may warrant a syllabic approach, but even in these cases a moraic analysis is ultimately judged to be superior. Section 6 concludes with some discussion of the findings.

2. The phonemic inventory

2.1 Segments

9Kabiye has 18 consonant phonemes /p f t d s z ʈ t͡ʃ k k͡p l y w h m n ɲ ŋ/ and nine vowel phonemes /i ɪ e ɛ u ʊ o ɔ a/. All of the latter contrast with their long counterparts, producing both grammatical (1) and lexical (2) contrasts.

(1) a [kì-] sp3sg/c3 ‘it’ [kì-ì-] sp3sg/c3-neg ‘it … not’
b [kɪ̀-] sp3sg/c3 ‘it’ [kɪ̀-ɪ̀-] sp3sg/c3-neg ‘it … not’
c [è-] sp3sg/c1 ‘(s)he’ [è-è-] sp3sg/c1-neg ‘(s)he … not’
d [ὲ-] sp3sg/c1 ‘(s)he’ [ὲ-ὲ-] sp3sg/c1-neg ‘(s)he … not’
e [à-] sp3sg/c7 ‘they’ [à-à] sp3sg/c7-neg ‘they … not’
(2) a [kùzú-ù] st-inf ‘to question’ [kùúzù-ù] st-inf ‘to stress’
b [mát-ʊ̀] rt-c9 ‘writing’ [mátʊ̀-ʊ̀] rt-inf ‘to stammer’
c [kó-ú] rt-inf ‘to shout’ [kòò-ú] rt-c1 ‘niece, nephew’
d [nɔ́] rt[c1] ‘cow’ [nɔ́-ɔ́] rt-c1 ‘mouth’
e [ká-ʊ́] rt-inf ‘to pile up’ [kàà-ʊ́] rt-c3 ‘boulder’
  • 2 The only word in my database in which a long back unrounded vowel is not obviously at the morpheme (...)
  • 3 The five long unrounded vowels are spelled <iɣ, ɩɣ, eɣ, ɛɣ, aɣ> respectively in the standard orthography. It is unfort</i> (...)

10Kabiye also has a series of five long back unrounded vowels /ɯ̘ɯ̘ ɯ̙ɯ̙ ɤɤ ʌʌ ɑɑ/ that only occur at morpheme boundaries,2 either straddling it (3a, 3e) or preceding it (3b, 3c, 3d).3

(3) a [sídɯ̘́-ɯ̘́] rt-ipf ‘is mixing’
b [ɲɯ̙̀ɯ̙̀-ɖέ] rt-c6 ‘piercing tool’
c [k͡pɤ́ɤ̀-ù] rt-inf ‘to pardon’
d [lʌ̀ʌ́-tʊ̀] rt-c9 ‘intelligence’
e [hɑ́-ɑ̀] rt-c5 ‘dog’

11The limited morphological distribution of these vowels suggests that diachronically they are the result of certain morphophonological processes that may legitimately be represented by abstract /VCV/ structures, in which an elided /k/ has left its dorsal trace on the surrounding vowels, a matter beyond the scope of this paper. Synchronically, these five vowels contrast with their front counterparts, their rounded counterparts, and with one another, confirming their phonemic status. Differences in opinion about their exact phonetic quality — specifically whether they are pure vowels or diphthongs — (CLNK 1999: 482; Delord 1976: 21; Lébikaza 1999: 43f., 49-56) present no obstacle for the purposes of the present paper because there is general agreement that these vowels are long and that both elements of the sequence are capable of bearing distinct tones.

2.2 Tones

12Now let us turn to the tone system. Kabiye has two underlying tones, H and L (Lébikaza 1999: 183) and — assuming that the OCP is inoperative (Roberts 2016: 148) — eight underlying tone patterns on non-compound nouns (4).

(4) a HHH /fétí-yé/ rt-c7 ‘toad’
b LLL /kòzòŋ-à/ rt-c5 ‘hare’
c HLL /kpáŋɑ̀-ɑ̀/ rt-c5 ‘donkey’
d HHL /nʊ́mɔ́-ʊ̀/ rt-c3 ‘path’
e LLH /kpèlù-ú/ rt-c3 ‘sparrow hawk’
f LHH /tàtá-ʊ́/ rt-c3 ‘potsherd’
g LHL /kɪ̀lák-ʊ̀/ rt-c3 ‘monitor lizard’
h HLH /nákà-ʊ́/ → [náká-ʊ́] rt-c3 ‘grasshopper’

13Tone conveys meaning in both the lexicon (5) and the grammar (6).

(5) a [k͡pà] rt[imp] ‘sew!’ [k͡pá] rt[imp] ‘catch!’
b [tànàŋ̀] rt[c3] ‘morning’ [tàná-ŋ̀] rt-c4 ‘jokes’
c [kpɤ̀ɤ́-ù] rt-inf ‘to guess’ [kpɤ́ɤ̀-ù] rt-inf ‘to pardon’
(6) a [ὲ-ὲ-tɔ̀-kɪ́] sp3/c1-neg-rt-ipf ‘(s)he is not eating’
b [ὲ-έ-tɔ̀-kɪ̀] sp3/c1-fut-rt-ipf ‘when (s)he will eat’
c /έ-ὲ-tɔ̀-kɪ́/ → [έ-έ-tɔ̀-kɪ́] sp3/c1[cnd]-neg-rt-ipf ‘if (s)he is not eating’

14The transformations in (4h) and (6c) are a foretaste of two post-lexical tonal processes that will provide putative evidence in favour of the syllable as TBU. These will be discussed in detail in sections 5.1 and 5.2 respectively.

3. Definitions

3.1 Defining the mora

15The following analysis will use the theory of phonological weight, as proposed in Hyman’s (1985) seminal book and adopted in various forms in subsequent literature. According to this theory, a syllable weighs more or less according to the number of moras it contains. The mora is defined as a minimal sub-syllabic unit derived from units on a timing tier that mediate, in a non-linear structure, between root nodes and syllable nodes (Broselow 1995: 184; Hayes 1989: 254; Hyman 1985: 9f.). The concept of phonological weight has been useful for explaining a range of phenomena, including syllable structure (van der Hulst & Ritter 1999: 349f.), Japanese vowel length (Labrune 2012), geminate timing (Ham 2001), stress assignment (Watson 2002), compensatory lengthening (Hayes 1989), and — closer to the concerns of the current contribution — the identity of the TBU and the distribution of tone patterns (Odden 1995). In the literature, the terms “weight” and “moraicity” are sometimes used interchangeably. In practice, however, their relationship may vary from language to language and even from process to process within a single language. When this does happen, it is normally because “weight” refers to quantity, stress, and meter, among other things. In the present paper, the term will be employed exclusively with respect to moraicity. In Kabiye, it is not difficult to identify the moras: all vowels and also nasals in preconsonantal and word-final positions fit this description, and they are identical to each other in that each one of these units projects exactly one mora. The only consonants to occupy the coda position are the labial and velar nasals /m ŋ/. Vowels may be preceded by a C onset but this position is weightless (Broselow 1995: 188f.; Broselow, Chen & Huffman 1997; Goedemans 1998; Hayes 1989: 19).

3.2 Defining the syllable

  • 4 In some languages with phonemic stress, it may be necessary to state “below the level of the foot” (...)
  • 5 The standard orthographies of (7d) and (7e) are ɛhaɣaɣ ‘was giving’ and ɛhaɣaɣ-ɩ ‘was giving him/h (...)

16The syllable is defined as a unit of phonological organisation above the level of the mora and below the level of the prosodic word,4 although it may be co-extensive with either of these. A syllable clusters around a single sonority peak contained in the nucleus, following the sonority sequencing principle (SSP) (Broselow 1995: 188f.; Goedemans 1998; Hayes 1989: 19, 258). Identifying Kabiye syllables and their boundaries is more challenging than identifying its moras because the language permits multiple vowel sequences when morphemes are concatenated, up to a maximum of five (7).5

(7) a CV [sú] rt[c3] ‘guineafowl’ frequent
b CVV [ʈòó] rt[c8] ‘night’ frequent
c CVVV [k͡pɤ́ɤ̀-ù] rt-inf ‘to pardon’ less frequent
d CVVVV [hɑ̀ɑ̀-ɑ̀ɑ́] rt-ipa ‘was giving’ infrequent
e CVVVVV [hɑ̀ɑ̀-ɑ̀ɑ́-ɪ́] rt-ipa-op3sg/c1 ‘was giving him/her’ infrequent

17The vowels in trimoraic structures are usually heterogenous as in (7c). A reviewer has questioned whether syllabification of these forms is sensitive to vowel length, but the question does not arise because whenever CVVV contains two vowels, the structure is always CV1V1V2, never *CV1V2V2. Sequences of three homogenous vowels are limited to the infinitive forms of 14 verbs with the underlying structure CVk, e.g. [tʊ̀ʊ́-ʊ̀] ‘to diminish’ (Roberts 2013: 60ff., 140ff.), a few ideophones (e.g. [hὲὲὲ] ‘gently’), one interjection (e.g. [ὲέὲ] ‘yes’), and one arguably onomatopoeic preposition [ɖòòò] ‘since’. Quadrimoraic structures (7d) occur in the imperfective past of 55 verbs with the underlying structure CVk (Roberts 2013: 70ff., 150ff., 230ff.). The addition of a V object pronoun to these forms produces the only quinquemoraic structures (7e).

18A key question that needs to be addressed at the outset, then, is whether Kabiye has syllables that are longer than bimoraic. Odden (2011: 486) reminds us that “[...] nearly all languages with vowel length are treated as having two degrees of length” and goes on to show that most claims of ternary contrast in the literature can actually be traced to something other than length, such as stress assignment. However, a stronger case in favour of three degrees of length has been made for two African languages, Dinka (Andersen 1987; Remijsen & Gilley 2008; Remijsen & Ladd 2008) and Shilluk (Remijsen et al. 2011 and 2019), so it will be advisable to assess that possibility in Kabiye too.

  • 6 A counter-argument is that when V-initial words are pronounced in isolation in Kabiye they are pre (...)

19While it is true that a syllable is defined as having a single sonority peak, not all segmental sequences exhibiting a single peak are necessarily in the same syllable because the SSP operates within syllables, not across syllable boundaries. With this in mind, four phonotactic realities are inconsistent with the putative existence of trimoraic syllables in Kabiye. First, V is a frequently attested word-initial syllable (e.g. [è.g͡bé.lè] ‘maternal uncle’), so sequences of three vowels can legitimately be parsed as bisyllabic (e.g. [k͡pɤ́ɤ̀.ù] ‘to pardon’).6 Second, CVVV never occurs in roots; it is always heteromorphemic (cf. Remijsen & Gilley 2008: 323 for Dinka). Third, CVVV structures never reduplicate in ideophones, unlike CVV (e.g. [láʊ́láʊ́] ‘slender’) and CVN (e.g. [nìm̀nìm̀] ‘absolute silence’). Fourth, it becomes difficult to know where to draw the line. If CVVV is parsed as a single syllable, why would the quadrimoraic (CVVVV) and quinquemoraic (CVVVVV) structures listed in examples (7)d and (7)e not also be considered monosyllabic by the same reckoning? Part of the answer to this question is surely that three moras per syllable is the most that has ever been reliably posited and defended for any language, and even this claim is controversial.

20Another preliminary consideration is the need to establish the phonotactic status of nasals. In Kabiye, they are similar to vowels in five ways (cf. Vydrine & Kességbeu 2008: 7 for Eastern Dan). First, they can occupy slots normally reserved for vowels, both initially (e.g. VCV [ὲzɪ́] ‘as, like’; NCV [ǹzɪ́] rel[c13]) and finally (e.g. CVV [tέʊ̀] ‘rain’; CVN [tέm̀] ‘to finish’). Second, they can be lengthened in the same way as vowels (e.g. VV [ὲὲtὲŋ́] ‘(s)he does not finish’; NN [ŋ̀ŋ̀tὲŋ́] ‘you (sg) do not finish’). On this point, it is legitimate to ask why, if an intervocalic N behaves as an onset, NN is not considered to be an onset-nucleus sequence rather than a branching nucleus. The answer is that NN is always a homogenous, continuous sound, similar to VV. Third, in preconsonantal and final positions nasals always bear tone as vowels do. Fourth, the tonal association patterns of vowels and tone-bearing nasals always match. Fifth, tones on nasals may be different from those of the previous V nuclei.

21However, nasals are not phonotactically identical to vowels in all respects. First, unlike nuclear V, nuclear N is never preceded by an onset (i.e. CN is unattested). Second, unlike intervocalic V, intervocalic N is always an onset (i.e. CVVV and CVNV cannot be syllabified in the same way). Third, CVVN#, a relatively frequent structure, can occur in noun roots (e.g. [kpááŋ́-∅] ‘vestibule-c3’) and CVNN#, though less frequent, occurs in ideophones (e.g. [fáŋ́ŋ́] ‘knowingly’, [lèm̀m̀] ‘vast’). Both of these contexts are monomorphemic, whereas CVVV is always bimorphemic. Fourth, while up to five adjacent Vs may occur, there is an upper limit of two adjacent Ns.

22For all these reasons, the following analysis will be guided and limited by the classic binary feature requirement espoused by early researchers (Chomsky & Halle 1968; Jakobson et al. 1952; Jakobson & Halle 1956) with an upper limit of two moras per syllable, so the more complex structures will be provisionally parsed as CVV.V, CVV.N and CV.NN while acknowledging that this issue would merit further research.

23With these considerations in mind, we are now in a position draw up a definition. A Kabiye syllable has:

24an obligatory nucleus (V, N);

25an optional branching nucleus (CV vs CVV; V vs VV, N vs NN) or an optional coda (CV vs CVN);

26an optional weightless prevocalic onset (CV vs V; CVV vs VV).

27These points can be formally summarised as in Figure 1, where ‘X’ indicates that the slot may be occupied by either V or N.

Figure 1 — Formalisation of Kabiye syllable structure

Figure 1 — Formalisation of Kabiye syllable structure

28This, in turn, leads to a basic distinction between light monomoraic syllables (CV, V, N) and heavy bimoraic syllables (CVV, CVN, VV, NN). Let us look more closely at the longer structures that these generate, since their mora and syllable counts do not match. A sequence of one light syllable followed by one heavy syllable contains three moras (8). Note in passing the downstep in (8d) because this process will provide one of only two pieces of evidence in favour of the syllable as TBU, as will be shown in section 5.2.

29Likewise, a sequence of one heavy syllable followed by one light syllable contains three moras (9).

30And finally, this approach also leads to the following syllabification of the quadrimoraic and quinquemoraic structures already cited (7d, 7e), building syllables left-to-right, with the syllable boundaries matching the morpheme boundaries in these particular cases (10, 11).

31Having characterised the mora and the syllable for Kabiye, we are now in a position to weigh the evidence.

4. Evidence in favour of the mora as TBU

32The evidence in favour of the mora as TBU is based on two fundamental observations. First, if the syllable were the TBU, one might expect tonal contours to be of roughly equal frequency on heavy and light syllables. As it is, in Kabiye they are abundant on the former but prohibited on the latter, contrary to some other Gur languages such as Nateni (Neukom 1995), Mbelime (Neukom 2004) and Lama (Ourso 1989). As Yip (2002: 73) explains:

If the language has both light mono-moraic and heavy bi-moraic syllables, and if these differ in the number of tones they can bear, so that mono-moraic syllables can have only one tone but bi-moraic syllables can have two, then it must be the case that the TBU is the mora, not the syllable. […I]f the two different syllable weights can bear the same number of tones, then the syllable must be the TBU.

33Second, Kabiye tonal processes are sensitive to a unit smaller than the syllable, so a syllabic approach, if not less explanatory, quickly becomes cumbersome. The following sections investigate how these two observations play out in four different phonological domains: association of the LHL tone pattern (section 4.1), association of two-tone patterns (section 4.2), lexical L tone spreading (section 4.3), and vocalic elision (section 4.4).

4.1 Association of the LHL tone pattern

34The first phonological context that favours the mora as TBU is the initial tonal association of LHL words. Example (12)a is a canonical CV.CV.CV structure in which the identity of the TBU requires no choice between moras and syllables because the relationship between the two is isomorphic. In contrast, there is disparity between mora and syllable counts in the other examples: a sequence of one light and one heavy syllable (12b); and a sequence of one heavy and one light syllable (12c, 12d).

35In all these words, the three-tone LHL tone pattern associates to the three moras in a one-to-one fashion, and this straightforward correspondence is absent between tones and syllables. There is therefore no need to stipulate the direction of tonal association. Admittedly, a syllabic account in which tones associate left to right in a one-to-one fashion would fit the CV.CVV structures, simply forming a tonal contour at the right edge of the word (13).

36But such an analysis fails for examples with CVV.CV and CVV.V structures, because the same procedure generates unattested forms with contour tones on the final short syllable which are prohibited throughout the language (14). A syllabic approach would therefore need to stipulate a left-to-right tonal association for some words and a right-to-left tonal association for others.

37This context provides some initial evidence that Kabiye counts moras rather than syllables when associating tones to the segmental tier. Figure 2 compares formalisation of the moraic and the syllabic approach to LHL tone association. Syllables are light unless otherwise stipulated (here and in all the figures that follow). A syllabic approach has the double disadvantage that it requires three autosegmental representations and a tag to differentiate syllables of differing weights. A moraic approach, on the other hand, captures all phonotactic environments in a single formula.

Figure 2 — Syllabic and moraic formalisations demonstrating association of the LHL tone pattern

Figure 2 — Syllabic and moraic formalisations demonstrating association of the LHL tone pattern

4.2 Association of two-tone patterns

38Initial tonal association provides further evidence from another angle. As illustrated in example (4), the OCP is inoperative in the tone system of Kabiye nouns, allowing the exploitation of the full range of eight possible tone patterns on trimoraic forms, i.e. /HHH, LLL, HLL, HHL, LLH, LHH, LHL, HLH/. Among these, the four rising and falling patterns can all associate to a sequence of one light and one heavy syllable (15-18).

39A moraic account, because it is sensitive to a unit smaller than the syllable, permits the tonal contrasts between /HLL/ and /HHL/, and between /LLH/ and /LHH/ on trimoraic nouns. A syllabic account, on the other hand, is certainly possible (19-22) but, as before, requires the association of contour tones to heavy syllables at the right edge of the word (20, 21). This, in turn, forces us to ask again why tonal contours would be permitted on heavy syllables but not on light ones.

40Figure 3 compares formalisation of the moraic and the syllabic approach to the initial tonal association of the HHL pattern. As before, the syllable approach requires two formulas to account for structures of differing weights and the addition of a tag to differentiate between syllables of differing weights, whereas the moraic approach captures all the necessary information in a single formula. The same would be true of formalisations of the three other two-tone patterns.

Figure 3 — Syllabic and moraic formalisations of association of the HHL tone pattern

Figure 3 — Syllabic and moraic formalisations of association of the HHL tone pattern

4.3 Lexical L tone spreading

  • 7 The only other context in which L spreading occurs is in the associative noun phrase where it is a (...)

41Tone spreading rules also point towards the mora being the TBU. Let us compare what happens to the /HL/ tone pattern when it associates to verb roots of different structures. The imperative provides the underlying tone pattern of the verb root (Roberts 2021: 12-13). In the perfective aspect, the L tone of a subject pronoun spreads onto the verb root, delinking H association lines, until it is blocked by a singly-linked H. In a Lexical Phonology framework (Kiparsky 1982; Mohanan 1982), this process is deemed to be lexical because it only applies in this morphosyntactic environment, not across word boundaries throughout the language.7

42The important point for the purposes of a moraic analysis is that this spreading rule operates in the same way irrespective of the segmental structure of the verb root, whether it be a unambiguous sequence of three light syllables (23a) or an ambiguous sequence of one heavy and one light syllable (23b; 23c). The L tone spreading mechanism is blind to syllable count, and referring to moras rather than syllables allows the process to be explained more elegantly. As the L tone spreads rightwards across the verb root looking for association lines to eliminate, it only considers moras.

43In a syllabic account, on the other hand, the L tone spreading process leads to a LH contour with both tones cohabiting a single heavy syllable on CVVCV and CVNCV structures (24).

44But if this were the case, we would expect a similar tonal contour to occur on the first (light) syllable of the CVCVCV structure, and this is not the case — contour tones are never permitted on short syllables (25).

45Figure 4 compares formalisation of the moraic and syllabic approaches to lexical L tone spreading. The ellipses indicate that the spreading process is iterative and continues until it is blocked by a singly-linked H tone at the right edge. The syllabic approach has the disadvantage that it requires two formulas: one for light and one for heavy syllables. The moraic approach, on the other hand, can capture all phonotactic environments in a single formula.

Figure 4 — Syllabic and moraic formalisations of L tone spreading

Figure 4 — Syllabic and moraic formalisations of L tone spreading

4.4 Vocalic elision

  • 8 The surface forms vary according to the rules of ATR vowel harmony. Singular noun class suffixes a (...)

46Investigating vocalic elision provides more evidence in favour of the mora as TBU. The underlying form of the noun class 1 suffix is /u ~ ʊ/8, and this is the form that most frequently surfaces, often generating canonical CV.CV syllable structures (26). But following a nasal consonant the suffix elides, leaving a heavy CVN syllable (27), and the H tone re-associates to the final nasal mora. However, even when the syllable structure changes, the two-tone pattern (LH) and the two moras remain stable.

47A syllabic account of the CVN items is not difficult. It simply requires that the suffix tone, bereft of its segment, reassociates leftwards onto the root, with the result that two tones cohabit a single TBU (28). But, as before, this begs the question why contour tones would be permitted on heavy syllables when they are never found on light syllables.

48Dialect comparison is consistent with the vocalic elision argument. The standard pronunciation of example (29) is CVN.CV, that is a sequence of one heavy syllable followed by a light syllable. But in the Yadè dialect the same word may be pronounced as CV.CV.CV (CLNK 1999: 392), that is a sequence of three short syllables. Again, although the syllable structure changes, the LHL tone pattern, the mora count and the relationship between these two remain the same.

49As before, a syllabic account of the standard dialect form requires the reassociation of the H tone to the preceding heavy syllable (30), which would be an unremarkable process in an African language, except that no such contour tones ever occur on light syllables in Kabiye.

50Figure 5 compares formalisation of the moraic and the syllabic approach to vocalic elision following a nasal consonant. The syllabic approach requires stipulating the segmental composition of the second syllable, followed by resyllabification. The moraic approach, on the other hand employs a single formula to account for all environments, with no reference to the segmental structure at all.

Figure 5 — Syllabic and moraic formalisations of vocalic elision

Figure 5 — Syllabic and moraic formalisations of vocalic elision

5. Putative evidence in favour of the syllable as TBU

51We now turn to putative evidence in favour of the syllable as TBU, and this will be found in two post-lexical processes, contour realignment (section 5.1) and HLH plateauing (section 5.2).

5.1 Post-lexical contour realignment

  • 9 If the segmental structure does contain an intervening consonant, contour realignment does not occ (...)

52According to the post-lexical process of contour realignment (which we have already seen an example of in (6c)), before a L tone, the L of a HL sequence delinks, merging with the L tone to its right, and the H tone expands its domain to occupy its place. This process only occurs on condition that the structure does not contain an intervening consonant, i.e. only in the case of a heavy syllable with a rime composed of VV (31a) or VN (31b).9 It occurs wherever the condition is met, within words and across word boundaries, proving its status as a post-lexical process. Since the result of this process is that the contour aligns with the syllable boundary, a syllabic account has some merit especially when the process occurs within a word. It simply requires the delinking of the second constituent of a HL contour in the stipulated phonotactic environments.

53But a syllabic account of the process when it occurs across word boundaries either requires resyllabification (32, 33) or splitting into smaller syllabic units which are, in any case, the same weight as moras (i.e. CV.V, CV.N).

54All these examples demonstrate that the second element of a heavy syllable can operate somewhat independently of the first. A syllabic account of the contour realignment process is forced to stipulate that it only travels the distance of half a syllable, which is tantamount to stating that it travels the distance of a mora. Since the process is sensitive to a unit smaller than the syllable, a moraic account is preferable, whether within words (34) or across word boundaries (35, 36). In contrast to the syllabic approach, it can be described as a H tone spreading process in which the L of a HL sequence delinks before another L (with which it merges), and the H tone spreads one mora to the right.

55Figure 6 compares formalisation of the moraic and the syllabic approach to post-lexical contour realignment respectively. Both formalisations require the addition of segmental information on the timing tier, stipulating either that the syllable must be heavy or that the second mora must be non-branching, i.e. that it does not contain an onset. The syllabic approach has the advantage that it only refers to L-delinking, whereas the moraic approach requires a two-stage process of L-delinking and H-spreading. However, describing the process in terms of spreading adds unity to the overall analysis since this concept has already been used to articulate the L-spreading process in section 4.3.

Figure 6 — Syllabic and moraic formalisations of post-lexical contour realignment

Figure 6 — Syllabic and moraic formalisations of post-lexical contour realignment

5.2 Post-lexical HLH plateauing

56The second process that provides putative evidence of the syllable as TBU concerns the data already presented in (4h) and (8d). Roberts (2016; 2019; 2021) refers to it as “post-lexical HLH plateauing” and Lébikaza (1999:57f.) as “assimilation verticale”. The process again results in contour realignment, but this time it also creates downstep. The process stipulates that the underlying tone pattern /HLH/ is always realised as [HH], within words (37), across word boundaries (38) and even recursively, where the second H of a HLH sequence is also the first H of the next (39).

(37) a /hɔ́kʊ̀-ʊ́/ [hɔ́kʊ́ʊ́] rt-c3 ‘goitre’
b /ɲʊ́ʊ̀-dʊ́/ [ɲʊ́ʊ́dʊ́] rt-c1 ‘president’
(38) a /nɪ́mɪ́-yέ yɔ̀ɔ́/ [nɪ́mɪ́-yέ yɔ́ɔ́] rt-c7 loc ‘on the right’
b /ná yà-lá/ [ná yá-lá] rt[imp] rt-c7 ‘see the eggs!’
c /lé-sì fέyɪ́/ [lé-sí fέyɪ́] rt-c6 there.are.no ‘there are no funerals’
(39) /áfà kʊ́ɖʊ̀ḿ/ fá kʊ́ɖʊ́ḿ] rt[c1] num[c1] ‘one pig’

57Example (40) is unambiguous because the relationship between syllables and moras is isomorphic; it is included because it shows the process simply. The singly-linked L tone between the two H tones delinks, then the second H tone (with its downstepped register because of the floating L tone that precedes it) spreads one syllable/mora leftwards to fill the gap.

58But the two H tones associate in different ways depending on the segmental structure, so it will be instructive to compare a syllabic account with a moraic account of these ambiguous forms. According to a syllabic account, when /HLH/ maps to CV.CVV, the first constituent of the LH contour delinks, leaving only the second constituent, with its downstepped register due to the floating L tone (41).

59Conversely, when /HLH/ maps to CVV.CV, the second constituent of the HL contour delinks leaving only the first constituent, and the floating L tone causes downstep on the second H tone (42).

60In both cases, the downstep aligns with a syllable boundary, which is a potential point in favour of the syllable being the TBU.

61However, this leaves aside the important concern that contour tones such as those found in the underlying forms of examples (41) and (42) are never found on light syllables, and in any case a moraic account of the HLH plateauing process is also possible and arguably even preferable. According to this approach, when /HLH/ maps to CV.CVV, the singly-linked L tone between the two H tones delinks, then the second H tone (with its downstepped register) spreads one mora leftwards to fill the gap (43), which exactly matches the unambiguous CV.CV.CV representation in (40).

62Conversely, when /HLH/ maps to CVV.CV, the first H tone spreads one mora rightwards to fill the gap (40).

63In a moraic analysis, the spreading process — be it to the left or to the right — is sensitive to a unit smaller than the syllable, even though the phonetic outcome is that the downstep aligns with the syllable boundary. A moraic analysis has the advantage of emphasising the isomorphic correspondence between the three underlying tones and the three moras, a correspondence which is hidden in the surface forms. Admittedly, since a moraic account prohibits downstep occurring between two moras with no intervening consonant, one might legitimately ask why weightless consonant onsets should participate in the placement of downstep. The usual evidence for this comes from stress assignment patterns, where the presence of a consonant onset causes a syllable to attract stress, but this is not the case in Kabiye phonology because stress is not phonemic.

  • 10 Lébikaza (1999:191) provides no glosses for these two words. The first means ‘old’; the second mea (...)

64At this point, it will also be helpful to briefly return to the preceding literature. Lébikaza (1999: 191) endeavours to prove that the mora is the TBU by mapping the HLH plateauing process onto two words: /kɪ́bɪ̀nʊ́/ (an unambiguous series of three light syllables) and /kɪ́sὲḿ/ (an ambiguous light-heavy syllabic structure).10 His analysis is coherent, but it does not go the extra step and show what a syllabic analysis would look like and why it is to be rejected. This is important, because it is indeed possible to associate the three-tone pattern to the two syllables of /kɪ́sὲḿ/ as shown in (45).

65The only reasonable argument against the syllabic approach, which a careful reading of Lébikaza (1999) suggests he would almost certainly have agreed with, is the awkward inconsistency of allowing contour tones on heavy syllables and disallowing them on light ones.

66Figure 7 compares formalisation of the moraic and syllabic approaches to post-lexical HLH plateauing. This time, neither approach can avoid stipulating segmental information, but the moraic analysis is simpler and more coherent for three reasons. First, it requires only two formulas, whereas the syllabic analysis requires three. Second, it refers to a single process — H tone spreading — in both formulas, whereas this process is limited to only one of the three syllabic formulas. Admittedly, the direction of the spreading in the moraic account does depend on the phonotactic context. Third, the concept of spreading is in harmony with its use in the description of lexical L tone spreading (section 4.3) and in the moraic approach to post-lexical contour realignment (section 5.1), which helps to unify the overall analysis.

Figure 7 — Syllabic and moraic formalisations of post-lexical HLH plateauing

Figure 7 — Syllabic and moraic formalisations of post-lexical HLH plateauing

6. Discussion

67Overall, the evidence for the mora as TBU in Kabiye is compelling. On the one hand, four of the phonological domains investigated reveal an exact correspondence between moraic structures and tone patterns, while there is often a mismatch between syllable structures and tone patterns. In two post-lexical processes where the alignment of contour tones with the syllable boundary provides putative evidence in favour of the syllable as TBU, a moraic account is also possible and arguably even preferable. Crucially, throughout the analysis, if the syllable were the TBU we would expect heavy and light syllables to function similarly in terms of tonal association. Yet a syllabic analysis frequently leads to contour tones on heavy syllables whereas no such contours are ever attested on light syllables, unlike in some other Gur languages.

68It could be argued that the restriction against contour tones on short syllables does not constitute irrefutable evidence that the mora is the TBU since some languages in which the TBU is definitely the syllable also have a constraint against contours on short syllables. No universal principle insists that the same kinds of tone patterns must occur on all syllable types. A language can impose restrictions on co-occurrence of tone and syllable weight just as it can on other phenomena such as consonant voicing and place of articulation. So, of the two arguments presented in this paper, it is probably the independent behaviour of the second constituent of a heavy syllable with relation to the first that provides the strongest support for the mora as TBU.

69Evidence also emerges from the generation of formulas. Comparisons repeatedly show that the essential information can be captured with fewer moraic formulas than syllabic ones, and that the syllabic approach more often requires stipulating segmental information as a condition, as indicated by (C) in table 1.

Table 1 — Number of formulas required for each phonological domain

Syllabic analysis Moraic analysis
HLH association 3 (C) 1
2-tone association 2 (C) 1
L spreading 2 (C) 1
V elision 1 (C) 1
Contour realignment 1 (C) 1 (C)
HLH plateauing 3 (C) 2 (C)
Total: 12 7

70Some doubt remains about the status of final nasals, because it would seem preferable for the final N in CVN# and CVVN# structures to have the same phonotactic status: either as autonomous syllables or as codas of monosyllables. The former is not supported by evidence and seems counterintuitive; the latter would be in conflict with arguments against trimoraic syllables espoused in section 3.2. Moreover, parsing these structures as trimoraic syllables would have consequences for CVVV structures since vowels and tone-bearing nasals have been shown to behave similarly in several respects and identically in terms of tonal association. But postulating trimoraic syllables is controversial and is only possible if there as an undeniable reason to do so. As long as it has not been established that parsing these structures into multiple syllables is impossible, the upper binary limit adopted in this paper seems justified. An experimental study comparing and contrasting L1 speakers’ psycholinguistic perceptions of these structures would be an interesting avenue for future research but would not fundamentally change any of the arguments the nature of the tone-bearing unit presented in this paper.

71It is interesting to note that Hyman was partly led to develop his theory of phonological weight because earlier research (1982; 1983) had shown that the syllable was not a well-motivated functional unit in Gokana. Hyman has returned to this question more recently (2011) admitting that there is marginal evidence for syllable activation in Gokana, but has followed this up with a postscript (2015) that casts doubt on the more recent evidence, and still maintains that Gokana does not organise its tonal phonology with the syllable in mind. The Kabiye data concurs and makes a particularly interesting study for the sensitivity of tone to weight because it permits consecutive moraic slots of up to five vowels — an unusually elaborate quantity system by any measure.

72Admittedly, this paper does not reveal anything new about syllabic or moraic theory per se, and all the Kabiye tonal processes are described elsewhere. The moraic analysis is obvious and non-controversial, being accepted by all preceding authors who have stated their position on the subject. So what have we learned that is novel or noteworthy? The real interest in this paper is in laying out a moraic and a syllabic analysis side by side in order to evaluate their relative merits. This is an important stage in the research process because Kabiye is one of the few Gur languages in which it has been claimed that the mora is the TBU, and that claim has been made repeatedly with little to substantiate it. But it would be overhasty to consider Kabiye a Gur outlier as long as the tone systems of so many of the hundred or so languages in the family have not yet been analysed at all or have been given only cursory treatment. As more Gur researchers confront the question of the identity of the TBU in various languages, it should be borne in mind that, broadly, a syllabic approach is more suitable for languages in which the two constituents of a heavy syllable never differ from each other in their tonal association patterns. Conversely, moraic analyses are particularly appropriate for languages in which the tone of the second constituent in a heavy syllable operates with a certain autonomy with respect to that of the first.

73Anyone undertaking an autosegmental account of an African tone language needs to identify whether it is the syllable or the mora that is the TBU. But the literature is somewhat misleading in its portrayal of a clear-cut typological distinction involving a facile binary choice. Languages in which the mora must be the TBU or in which the syllable must be the TBU are probably untypical. Some languages have no branching nuclei or codas, so all syllables are monomoraic and the question does not arise, but researchers in such languages may still state that “tones associate to syllables”, giving the cursory impression that a choice has been made. In other languages it is possible to defend both positions equally, depending on one’s theoretical assumptions. In yet others, like Kabiye, arguments can be marshalled in favour of both, and it is for the researcher to present the two alternatives side by side as impartially as possible, then weigh the evidence.

Acknowledgments

74I wish to express my thanks to Hélène Ballarin, Rod Casali, Daveth Clark, Nick Clements, Christy Melick, Pamela Munroe, Jacques Nicole, David Odden, Cédric Patin, Steve Parker, Annie Rialland, Jann Russell, Keith Snider and five anonymous reviewers for their helpful comments during the preparation of this article. I am also grateful to my late research assistants Hénok Tchinguilou and Emmanuel Pidassa, without whose expertise and enthusiasm this research would never have been completed.

Abbreviations

C consonant
c noun class
cnd conditional
fut future
H high tone
imp imperative
inf infinitive
inj interjection
ipa imperfective past
ipf imperfective present
L low tone
floating low tone
loc locative
N nasal
neg negative
num numeral
OCP obligatory contour principle
op object pronoun
own ownership
pl plural
px adjectival prefix
rt root
sg singular
st stem
sp subject pronoun
TBU tone-bearing unit
V vowel
X vowel or nasal
Haut de page

Bibliographie

Adongo, E.A. 2018. Aspects of Gurenɛ phonology. Legon: University of Ghana (PhD dissertation).

Andersen, Torben. 1987. The phonemic system of Agar Dinka. Journal of African Languages and Linguistics 9. 1-27.

Anttila, Arto & Adams Bodomo. 1996. Stress and tone in Dagaare. Rutgers Optimality Archive 169.

Broselow, Ellen. 1995. Skeletal positions and moras. In John Goldsmith (ed.), The handbook of phonological theory, 175-205. Cambridge: Blackwell.

Broselow, Ellen, Su-I Chen & Marie Huffman. 1997. Syllable weight: Convergence of phonology and phonetics. Phonology 14. 47-82.

Cahill, Michael. 2004. Tonal polarity in Kɔnni nouns. Studies in African Linguistics 33(1). 1-34.

Chomsky, Noam & Morris Halle. 1968. The sound pattern of English. New York: Harper and Row.

Clements, George N. 2000. Prosodic systems and tone. In Bernd Heine & Derek Nurse (eds.), African languages: An introduction, 152-158. Cambridge: Cambridge University Press.

CLNK. 1999. Tɔm kpou kabɩyɛ-fransɩɩ (Dictionnaire kabiyè-français). Lomé: Ministère de l’éducation nationale et de la recherche, comité de Langue nationale kabiyè et SIL-Togo.

Crouch, Marjorie. 1985. A note on syllable and tone in Vagla verbs. Journal of West African Languages 15(2). 29-40.

Delord, J. 1976. Le kabiyè. Lomé: Institut national de la recherche scientifique.

Goedemans, Rob. 1998. Weightless segments: A phonetic and phonological study concerning the metrical irrelevance of syllable onsets. The Hague: Holland Academic Graphics.

Ham, William Hallett. 2001. Phonetic and phonological aspects of geminate timing. New York: Routledge.

Hayes, Bruce. 1989. Compensatory lengthening in moraic phonology. Linguistic Inquiry 20(2). 253-306.

Hulst, Harry van der & Nancy A. Ritter (eds.). 1999. The syllable: Views and facts (Studies in Generative Grammar 45). Berlin: Mouton de Gruyter.

Hyman, Larry M. 1982. The representation of length in Gokana. In Daniel P. Flickinger, Marlys Macken & Nancy Wiegand (eds.), Proceedings of the First West Coast Conference on Formal Linguistics, 198-206. Palo Alto, CA: Linguistics Department, Stanford University.

Hyman, Larry M. 1983. Are there syllables in Gokana? In Jonathan Kaye, Hilda Koopman, Dominique Sportiche & André Dugas (eds.), Current approaches to African linguistics, vol. 2, 171-179. Dordrecht: Foris.

Hyman, Larry M. 1985. A theory of phonological weight. Dordrecht: Foris.

Hyman, Larry M. 1993. Structure preservation and postlexical tonology in Dagbani. In Sharon Hargus & Ellen M. Kaisse (eds.), Studies in lexical phonology, 235-254. San Diego, CA: Academic Press.

Hyman, Larry M. 2011. Does Gokana really have no syllables? Or: What’s so great about being universal? Phonology 28. 55-85.

Hyman, Larry M. 2015. Does Gokana really have syllables? A postscript. Phonology 32. 303-306.

Jakobson, Roman, C. Gunnar M. Fant & Morris Halle. 1952. Preliminaries to speech analysis: The distinctive features and their correlates (Technical Report 13). Cambridge, MA: Acoustics laboratory, MIT.

Jakobson, Roman & Morris Halle. 1956. Fundamentals of language. The Hague: Mouton & Co.

Kassan, Balaïbaou Badameli. 1996. Système verbal et énonciation en kabiyè. Paris: Université de la Sorbonne Nouvelle (PhD dissertation).

Kassan, Balaïbaou Badameli. 2000. De l’influence du ton consécutif dans les formes de l’aoriste en kabiyè. Cahiers voltaïques/Gur Papers 5. 13-22.

Kenstowicz, Michael, Emmanuel Nikiema & Meterwa Ourso. 1988. Tonal polarity in two Gur languages. Studies in the Linguistic Sciences 18(1). 77-103.

Kiparsky, Paul. 1982. Lexical phonology and morphology. In I.S. Yang (ed.), Linguistics in the morning calm, 3-91. Seoul: Hanshin.

Labrune, Laurence. 2012. The phonology of Japanese. Oxford: Oxford University Press.

Lébikaza, K. K. 1999. Grammaire kabiyè: une analyse systématique - phonologie, tonologie et morphosyntaxe. Cologne: Rüdiger Köppe.

Lébikaza, Kézié Koyenzi. 1985. Phonologie, tonologie et morphosyntaxe du kabyè. Cologne: University of Cologne (PhD dissertation).

Melick, Christina. 2012a. Tone in the Mbelime verb system (SIL e-books 47). Dallas TX: SIL International (accessed 17 September 2022).

Melick, Christina. 2012b. The syllable contact law in Mbelime. GIALens, Electronic Notes Series 6(1). 1-14.

Mohanan, Karuvannur Puthanveettil. 1982. Lexical phonology. Cambridge, MA: Massachusetts Institute of Technology.

Musah, A. Agoswin. 2017. Syllable weight and tonal patterning in Kusaal: A moraic perspective. In Paul Newman (ed.), Syllable weight in African languages, 177-190. Amsterdam: John Benjamins.

Neukom, Lukas. 1995. Description grammaticale du naténi (Bénin) : système verbal, classification nominale, phrases complexes, textes (Arbeiten des Seminars für Allgemeine Sprachwissenschaft 14). Zurich: University of Zurich.

Neukom, Lukas. 2004. Esquisse grammaticale du mbèlimè, langue voltaïque du Bénin (Arbeiten des Seminars für Allgemeine Sprachwissenschaft 18). Zurich: University of Zurich.

Odden, David. 1995. Tone in African languages. In John Goldsmith (ed.), The handbook of phonological theory, 444-475. Oxford: Blackwell.

Odden, David. 2011. The representation of vowel length. In Marc Oostendorp, Colin J. Ewen, Elizabeth Hume & Keren Rice (eds.), The Blackwell companion to phonology, 465-490. Oxford: Wiley-Blackwell (accessed 1st October 2022).

Ourso, Méterwa Akayhou. 1989. Lama phonology and morphology. Urbana-Champaign, IL: University of Illinois at Urbana-Champaign (PhD dissertation) (accessed 1st October 2022).

Padayodi, Cécile M. 2010. A revised phonology of Kabiye segments and tones. Arlington, TX: University of Texas at Arlington (PhD dissertation).

Remijsen, B., Otto G. Ayoker & Signe Jørgensen. 2019. Ternary vowel length in Shilluk. Phonology 36(1). 91-125 (accessed 1st October 2022).

Remijsen, Bert, Otto G. Ayoker & Timothy Mills. 2011. Illustrations of the IPA: Shilluk. Journal of the International Phonetic Association 41(1). 111-125 (accessed 1st October 2022).

Remijsen, Bert & Leoma Gilley. 2008. Why are three-level vowel length systems rare? Insights from Dinka (Luanyjang dialect). Journal of Phonetics 36(2). 318-344 (accessed 1st October 2022).

Remijsen, Bert & D. Robert Ladd. 2008. The tone system of the Luanyjang dialect of Dinka. Journal of African Languages and Linguistics 29(2). 173-213 (accessed 1st October 2022).

Rialland, Annie. 2003. La syllabe ou la more en tonologie africaine, ou comment se fait l’interface entre segments et tons. In Jean-Pierre Angoujard & Sophie Wauqier-Gravelines (eds.), Phonologie : champs et perspectives, 105-129. Lyon: Editions ENS.

Roberts, David. 2003. Tone spreading in the Kabiye associative noun phrase. Cahiers voltaïques/Gur Papers 6. 95-100.

Roberts, David. 2013. La conjugaison des verbes en kabiyè, une langue du Togo : tableaux-types, règles d’emploi et index kabiyè-français français-kabiyè des verbes. Paris: L’Harmattan.

Roberts, David. 2016. Word tone and tonal overwriting in Kabiye nouns. Journal of African Languages and Linguistics 37(1). 131-160.

Roberts, David. 2019. The tone of Kabiye verbal extensions. Journal of West African Languages 46(2). 1-28.

Roberts, David. 2021. Polarity, low tone spread, and underspecification in the Kabiye verb phrase. Studies in African Linguistics 50(1). 8-37 (accessed 1st October 2022).

Schwarz, Anne. 2001. Low tone spreading in Buli. Cahiers voltaïques/Gur Papers 6. 121-130.

Vydrine, Valentin & Mongnan Alphonse Kességbeu. 2008. Dictionnaire dan – français (dan de l’Est) avec une esquisse de grammaire du dan de l’Est et un index français-dan. Saint Petersburg: Nestor-Istoria.

Watson, Janet. 2002. The phonology and morphology of Arabic. Oxford: Oxford University Press.

Yip, Moira. 2002. Tone. Cambridge: Cambridge University Press.

Haut de page

Notes

1 Toolbox is an integrated data management and analysis program (accessed 17 September 2022).

2 The only word in my database in which a long back unrounded vowel is not obviously at the morpheme boundary is [mɑ́ɑ́lέ] ‘type of initiation dance’.

3 The five long unrounded vowels are spelled <iɣ, ɩɣ, eɣ, ɛɣ, aɣ> respectively in the standard orthography. It is unfortunate that the <ɣ> symbol has found its way into so many linguistic papers (e.g. Lébikaza 1985; 1999; Kassan 1996; 2000; Roberts 2003) since it corresponds to no single phonetic value independently of the preceding vowel symbol.

4 In some languages with phonemic stress, it may be necessary to state “below the level of the foot”, but this is not the case in Kabiye.

5 The standard orthographies of (7d) and (7e) are <ɛhaɣaɣ> ‘was giving’ and <ɛhaɣaɣ-ɩ> ‘was giving him/her’, respectively.

6 A counter-argument is that when V-initial words are pronounced in isolation in Kabiye they are preceded by strong glottal stops, so phonetically these can be considered canonical CV syllable types (Steve Parker, p.c.). An anonymous reviewer has made a similar comment about word-initial glottal stops in Gokana.

7 The only other context in which L spreading occurs is in the associative noun phrase where it is also a lexical process (Roberts 2003).

8 The surface forms vary according to the rules of ATR vowel harmony. Singular noun class suffixes are toneless and receive their surface tones by means of the lexical root tone associating to the whole word (Roberts 2016: 141-149).

9 If the segmental structure does contain an intervening consonant, contour realignment does not occur, e.g. /há-tʊ̀ mɑ̀ɑ̀mɑ̀ɑ̀/ → [há-tʊ̀ mɑ̀ɑ̀mɑ̀ɑ̀] rt-c9 self ‘leaves themselves’; /ná kɪ̀yàk-ʊ̀/ → [ná kɪ̀yà-kʊ̀] rt[imp] rt-c3 ‘see the market!’ and many isolated words, e.g. [kpáŋɑ̀-ɑ̀] rt-c5 ‘donkey’. The fact that contour realignment does not occur in the verb [k͡pɤ́ɤ̀-ù] rt-inf ‘to pardon’ reinforces the claim that long back unrounded vowels are underlyingly VCV (see section 2.1).

10 Lébikaza (1999:191) provides no glosses for these two words. The first means ‘old’; the second means ‘red’, both in class 1. In both cases, his transcriptions have been slightly modified here to conform to the IPA.

Haut de page

Table des illustrations

Titre Figure 1 — Formalisation of Kabiye syllable structure
URL http://journals.openedition.org/lla/docannexe/image/2154/img-1.png
Fichier image/png, 16k
URL http://journals.openedition.org/lla/docannexe/image/2154/img-2.png
Fichier image/png, 66k
URL http://journals.openedition.org/lla/docannexe/image/2154/img-3.png
Fichier image/png, 67k
URL http://journals.openedition.org/lla/docannexe/image/2154/img-4.png
Fichier image/png, 61k
URL http://journals.openedition.org/lla/docannexe/image/2154/img-5.png
Fichier image/png, 46k
URL http://journals.openedition.org/lla/docannexe/image/2154/img-6.png
Fichier image/png, 20k
URL http://journals.openedition.org/lla/docannexe/image/2154/img-7.png
Fichier image/png, 30k
Titre Figure 2 — Syllabic and moraic formalisations demonstrating association of the LHL tone pattern
URL http://journals.openedition.org/lla/docannexe/image/2154/img-8.png
Fichier image/png, 41k
URL http://journals.openedition.org/lla/docannexe/image/2154/img-9.png
Fichier image/png, 76k
URL http://journals.openedition.org/lla/docannexe/image/2154/img-10.png
Fichier image/png, 80k
Titre Figure 3 — Syllabic and moraic formalisations of association of the HHL tone pattern
URL http://journals.openedition.org/lla/docannexe/image/2154/img-11.png
Fichier image/png, 34k
URL http://journals.openedition.org/lla/docannexe/image/2154/img-12.png
Fichier image/png, 79k
URL http://journals.openedition.org/lla/docannexe/image/2154/img-13.png
Fichier image/png, 50k
URL http://journals.openedition.org/lla/docannexe/image/2154/img-14.png
Fichier image/png, 35k
Titre Figure 4 — Syllabic and moraic formalisations of L tone spreading
URL http://journals.openedition.org/lla/docannexe/image/2154/img-15.png
Fichier image/png, 42k
URL http://journals.openedition.org/lla/docannexe/image/2154/img-16.png
Fichier image/png, 48k
URL http://journals.openedition.org/lla/docannexe/image/2154/img-17.png
Fichier image/png, 30k
URL http://journals.openedition.org/lla/docannexe/image/2154/img-18.png
Fichier image/png, 44k
URL http://journals.openedition.org/lla/docannexe/image/2154/img-19.png
Fichier image/png, 29k
Titre Figure 5 — Syllabic and moraic formalisations of vocalic elision
URL http://journals.openedition.org/lla/docannexe/image/2154/img-20.png
Fichier image/png, 23k
URL http://journals.openedition.org/lla/docannexe/image/2154/img-21.png
Fichier image/png, 50k
URL http://journals.openedition.org/lla/docannexe/image/2154/img-22.png
Fichier image/png, 88k
URL http://journals.openedition.org/lla/docannexe/image/2154/img-23.png
Fichier image/png, 143k
Titre Figure 6 — Syllabic and moraic formalisations of post-lexical contour realignment
URL http://journals.openedition.org/lla/docannexe/image/2154/img-24.png
Fichier image/png, 29k
URL http://journals.openedition.org/lla/docannexe/image/2154/img-25.png
Fichier image/png, 52k
URL http://journals.openedition.org/lla/docannexe/image/2154/img-26.png
Fichier image/png, 34k
URL http://journals.openedition.org/lla/docannexe/image/2154/img-27.png
Fichier image/png, 36k
URL http://journals.openedition.org/lla/docannexe/image/2154/img-28.png
Fichier image/png, 40k
URL http://journals.openedition.org/lla/docannexe/image/2154/img-29.png
Fichier image/png, 41k
URL http://journals.openedition.org/lla/docannexe/image/2154/img-30.png
Fichier image/png, 35k
Titre Figure 7 — Syllabic and moraic formalisations of post-lexical HLH plateauing
URL http://journals.openedition.org/lla/docannexe/image/2154/img-31.png
Fichier image/png, 80k
Haut de page

Pour citer cet article

Référence électronique

David Roberts, « Moraic tone-bearing units in Kabiye »Linguistique et langues africaines [En ligne], 8(1) | 2022, mis en ligne le 30 juin 2022, consulté le 21 mai 2026. URL : http://journals.openedition.org/lla/2154 ; DOI : https://doi.org/10.4000/lla.2154

Haut de page

Auteur

David Roberts

Independent researcher

Articles du même auteur

Haut de page

Droits d’auteur

CC-BY-SA-4.0

Le texte seul est utilisable sous licence CC BY-SA 4.0. Les autres éléments (illustrations, fichiers annexes importés) sont susceptibles d’être soumis à des autorisations d’usage spécifiques.

Haut de page
Rechercher dans OpenEdition Search

Vous allez être redirigé vers OpenEdition Search