Tone in Mabia Languages: Structure and Processes

Abstract

The Mabia languages belong to the Niger–Congo family and are spoken primarily across the savannah and Sahelian regions of West Africa, including northern Ghana, Ivory Coast, Togo, Benin, Nigeria, Burkina Faso, Niger, and Mali. Tone plays a crucial role in these languages, shaping both lexical meaning and grammatical structure. This study is a synthesis or an overview of previously described facts about the tonal phenomena in six Mabia languages, Dagaare, Dagbani, Gurenɛ, Kusaal, Likpakpaln, and Buli—highlighting their tonal inventories, structures, and distinctive tonal processes. Dagaare and Dagbani exhibit a two-tone system (high and low), with an additional down-stepped high tone. Kusaal, Likpakpaln, and Buli employ a three-tone system (high, mid, and low), while the tonal status of Gurenɛ remains contested: some scholars describe it as a two-tone language, whereas others provide evidence for a three-tone system. The mid tone, though relatively less productive, appears to represent a later innovation within the group. The mid tone performs both lexical and grammatical functions. Notable tonal phenomena across these languages include tonal polarity, low tone spreading, and restrictions on contour tones, with Buli exhibiting particularly productive low tone spread. In all six languages, tone plays both lexical and grammatical functions. This paper explores these features and offers basic theoretical explanations for their occurrence. Overall, tone is a defining characteristic of the Mabia languages, intricately shaping their phonological and morphological structures.

1. Introduction1

African languages have benefitted significantly from studies of tone and have contributed to phonological theories, refining our understanding of prosody (Leben, 1973; Hyman & Schuh, 1974; Goldsmith, 1976; Pulleyblank, 1986; Archangeli & Pulleyblank, 1994; Archangeli & Pulleyblank, 2022; Odden, 1995; Hyman, 2011, 2018; Kenstowicz, 2020 and many others). More ground remains to be covered in African tone systems because, as Odden (2020) observes, even basic questions such as tone inventory and featural properties remain controversial, even in tonal languages that have been studied.

The tone systems of Mabia languages require detailed investigation because, while the majority of them remain under-researched, those that have been studied show considerable variation in terms of inventory, tone classes, polarity, etc., which is worth exploring. Additionally, the search for tone universals over the past few decades (Hyman, 2007; Hyman & Schuh, 1974) makes investigations into the systems of languages such as Mabia a necessary endeavor. The tone system of each of the six languages considered in this paper has received some attention. As with studies on tone systems more generally, questions about the tone inventory and tone features of these and other Mabia languages sometimes raise controversy. Our goal is to describe and synthesize the structure and function of the tone systems of six Mabia languages: Dagaare, Dagbani, Gurenɛ, Kusaal, Likpakpaln, and Buli. With this, we hope to provide a resource that not only enhances our understanding of the tonal properties of these languages but also serves as a reference that encourages further exploration of other Mabia languages—both on a language-by-language basis and in a broader typological or comparative context.

The map below in Figure 1 shows the locations where the six Mabia languages under study are spoken in Ghana. Except for Buli, the other languages are spoken in neighboring West African countries—Burkina Faso, Ivory Coast, Togo, and Benin.

All six languages are relatively closely related within the Mabia family, sharing similarities in their vocabulary and grammar. Thus, within the broader Mabia family, five languages—Dagaare, Dagbani, Gurenɛ, and Buli—fall under the subgroup Mabia Central, while Likpakpaln is under the subgroup Mabia East (Bodomo, 2020). While they are sufficiently mutually intelligible, this does not extend to allowing for sustained conversations among speakers of these languages. Even within individual languages, there are community varieties that differ significantly in their tonal features. That is, the tone systems of these Mabia languages are fascinating due to the differences observed both within community varieties of individual languages and across the family as a whole. These differences have, in some instances, led to controversies in their treatment. For instance, Northern (Dagara/Lobr-Dagara) and Western Dagaare (Berefɔ [bɪ̀rɪ̀fɔ̀]) are reported to have three level tones (H, M, and L) (Bemile, 1983; Kuubezelle, 2013; Dundaa, 2013), while Central Dagaare (Anttila & Bodomo, 1996) and Southern (Waale [wáálɪ́]) dialects (Abdul Moomin, 2015) are described as having two contrastive tones (H and L), with rising and falling contour tones observed in the Central (Angsongna, 2021, 2023; Anttila & Bodomo, 1996). In contrast, although the Central dialect of Buli was previously argued to have two tones (H and L), Buli linguists over the past few decades have now recognized a third (mid) tone (Akanlig-Pare, 2005; Schwarz, 2005). Similarly, Dagbani was once described as toneless, but linguistic studies (Olawsky, 1999) have disproven this assumption in recent decades. Similarly, Kusaal (both the Agole and Toende dialects) and Likpakpaln are known to have three contrastive tones (H, M, and L). The tonal status of Gurenɛ remains contested, with some scholars (e.g., Atintono, 2004) claiming it has two tones, while others (e.g., Adongo, 2018) argue for a three-tone system. Tonal differences among the six languages explored here are also evident in how tones behave or interact within each language. For instance, while Buli does not exhibit automatic downstep (Connell, 2011), this phenomenon is attested in Dagaare, Dagbani, Gurenɛ, Kusaal, and Likpakpaln. The complexities and ongoing debates surrounding the tonal systems of the Mabia languages make them a particularly rich area for linguistic inquiry. This study focuses on six (6) of the approximately eighty (80) Mabia languages for two main reasons: (i) their geographical proximity coupled with notable linguistic differences, and (ii) the preliminary nature of the present work, which forms part of a larger project and therefore cannot yet represent the full range of Mabia languages. A more comprehensive study encompassing a larger number of languages remains a goal for future research. For further details on the genetic classification, linguistic characteristics, and geographical distribution of the Mabia languages, see Bodomo (2020).

We begin by defining the transcription conventions used consistently throughout this paper for all six languages. High tone (H) is indicated by an acute accent (á), mid tone (M) by a macron (ā), and low tone (L) by a grave accent (à). Contour tones are marked with a caron for rising tone (ǎ) and a circumflex for falling tone (â), while downstep is represented by a superscript down arrow (ꜜá). All data are presented in broad phonetic transcription. For clarity in tone marking, long vowels are consistently represented as VV throughout the paper.

The remaining sections are organized as follows. Section 2 presents an overview of tone systems in West African languages, covering what is known about their inventory (Section 2.1), tone-bearing units (Section 2.2), and contour features (Section 2.3). Section 3 deals with the general features of tone in the five languages, viz., Dagaare (Section 3.1), Dagbani (Section 3.2), Gurunɛ (Section 3.3), Kusaal (Section 3.4), Likpakpaln (Section 3.5), and Buli (Section 3.6), with a focus on the salient tonal structures, their functions, and draws on some basic theoretical assumptions to explain the existing tonal structures and processes. Section 3.7 presents a comparative summary of the six tonal systems, while Section 4 concludes the chapter.

2. Tone in West African Languages

2.1. Typology and Number of Levels

The typology of tone systems in African languages has been examined from various perspectives (Gussenhoven & Chen, 2021; Downing, 2005; Hyman & Schuh, 1974; Lojenga, 2018; Wedekind, 1985; Welmers, 1959, among others), and the primary criterion for classifying these systems is the number of underlying contrastive tone levels (Lojenga, 2018). Other criteria include (i) the presence or absence of downstep or downdrift, (ii) the presence or absence of surface contour tones, (iii) the presence or absence of depressor consonants, and (iv) the presence or absence of common tone rules, such as high tone spreading.

Tone languages are generally categorized into two types based on the shape of their pitch phonemes: register tone languages and contour tone languages (Katamba, 1989). In register tone languages, lexical tones maintain a steady pitch, typically high, mid, or low. In contrast, contour tone languages, like Mandarin and Cantonese, feature fluctuating pitch patterns. However, this distinction is not absolute, as many tonal systems exhibit features of both types. Register tone languages may display derived contours resulting from tonal interactions, such as tone sequences or phonological processes like tone sandhi. Conversely, contour tone languages often include level tones as part of their tonal inventory. For example, Dagaare can be described as primarily a register tone language, though it exhibits phonetically or phonologically derived contour patterns (see Section 3.1).

This overlap highlights the need for a more flexible understanding of tone typology. As Chen (2000) and Hyman (2007) argue, tone systems are better viewed as existing along a continuum rather than as discrete, mutually exclusive categories.

Within this continuum, West African languages are predominantly of the register type, although contour-like features may occasionally arise. Such features are relatively rare but have been noted in some Kru languages, including Grebo spoken in Liberia (Innes, 1960, 1981; Newman, 1986).

Within certain sub-areas in West Africa, languages typically exhibit two- or three-tone register systems. For instance, two-tone systems are widespread and found in languages such as Mandinka (Creissels, 2024), Fɔngbe (Lefebvre & Brousseau, 2011), Twi (Abakah, 2005), Igbo (Welmers, 1970; Bright-Ajoku & Okumo, 2020), Hausa (Newman, 1986), and Bambara (Courtenay, 1974; Green, 2010, 2013; Vydrin, 2016) among others. Languages with three-tone systems include Yoruba (Pulleyblank, 2003), Ewe (Deklu, 2021), Buli (Akanlig-Pare & Kenstowicz, 2002; Schwarz, 2003), Kusaal ( Musah, 2010, 2017), Etulo (Ezenwafor-Afuecheta & Inyani, 2021), and several others.

Tone systems with more than three levels have been reported in a number of languages. For example, Bini (Atlantic–Congo) has been variously described as having six, five, or four tone levels, depending on the idiolect (Wescott, 1962). Dan (Southern Mande) dialects exhibit either five or three tone levels (Bearth & Zemp, 1967). Akinlabi et al. (2022), for instance, propose five tonal categories for Dan, distinguishing four pitch levels and a contrast between modal low and creaky low tones. Similarly, Bwamu has been described as having three tone levels (Manessy, 1960) in some analyses, and four (Riccitelli, 1965) in others. Nupe is reported to have five tone levels (Kandybowicz, 2004), while Grebo, Egidi, and Basa have all been described as having four (Hobley, 1964).

Overall, two-tone systems are the most widespread cross-linguistically, including in West Africa (see Hyman, 2011; Hyman & Leben, 2021), although three-tone systems are also well represented. Languages with more than three level tones are comparatively rare and tend to occur in specific regions, such as along the Liberia–Ivory Coast area (Vydrin, 2008) and within the Kru language family.

2.2. Tone-Bearing Units

The tone-bearing unit (TBU) refers to the linguistic element on which tones are anchored (Hyman & Leben, 2017; Hyman, 2018). It is the entity that carries the tone. Identifying the TBU in a given language is not always straightforward. For instance, Yip (2002) suggests that tones may be associated with a segment, syllable, or mora as the TBU. Odden (1995, 2020) provides evidence from various languages showing that, in some, tones associate with the mora, while in others, they associate with the syllable, ultimately calling for further research to clarify the TBU in different languages.

In West African languages, the TBU is generally the syllable or the mora. Most of these languages, such as Dagaare (Angsongna, 2021, 2023; Anttila & Bodomo, 1996, 2022; Kennedy, 1966), Dagbani (Hyman, 1993; Hyman & Olawsky, 2004), Moore and Lama (Kenstowicz, 1994; Kenstowicz et al., 1988), Konni (Cahill, 2004), Mbelime (Melick, 2012), Nateni (Neukom, 1995), Lɛtɛ (Ansah, 2022), Tutrugbu (Gborsi, 2015), Bambara (Green, 2013), Yoruba (Kenstowicz, 2006), Degema (Rolle & Kari, 2022), Hausa (Newman, 1995; Newman & Jagger, 1989; Schuh, 1989), Nupe (Kandybowicz, 2004), Mbum (Meeussen, 1970), and many others are analyzed as having the syllable as the TBU. Other languages, including Kusaal ( Musah, 2017), Gurenɛ (Adongo, 2018), Kabiye (Roberts, 2022), and Kita-Maninka (Creissels & Grégoire, 1993), are claimed to have the mora as the TBU. However, these claims require further investigation, as the identification of TBUs in some languages remains inconclusive.

2.3. Contour Tones in West African Languages

In register tone systems, when there are more tones than tone-bearing units (TBUs), contour tones often arise. In many tonal languages, it is possible for a syllable to carry two or more tones. Typically, contour tones fall into two categories: rising (low–high) and falling (high–low). However, the capacity of syllables to support contour tones differs across languages. While most languages allow level tones on all syllable types, some restrict contour tones to specific syllables. For instance, in Hausa, contour tones occur only on heavy, bimoraic syllables (Odden, 2020). Conversely, languages like Grebo permit contour tones on both heavy and light syllables. Other languages, such as Kpele, Esako, Degema, Mbum, and Kambari, display limitations on rising contours (Meeussen, 1970), while in languages like Margi (Chadic), falling contours are exceedingly rare. For a comprehensive survey of contour tone restrictions across languages, see Gordon (2001). With this background in mind, now let us turn to the general tone features of the Mabia languages discussed.

3. Tone in Mabia Languages: General Features

This section provides an overview of the tone systems of the six languages under study, with a focus on their general characteristics. Specifically, it examines key aspects such as tone inventories, minimal pairs and triplets, distributional patterns, functional roles of tone, and other salient features. The discussion begins with the two languages that exhibit two-tone systems—Dagaare (Section 3.1) and Dagbani (Section 3.2)—followed by a language whose tone inventory is inconclusive—Gurunɛ with two or three tones (Section 3.3). This is followed by a discussion of the languages with three-tone systems, namely, Kusaal (Section 3.4), Likpakpaln (Section 3.5), and Buli (Section 3.6).

3.1. Tone in Dagaare

Given the existence of different varieties of Dagaare, tonal variations are expected, as each variety exhibits different tone inventories. Central Dagaare, the focus of this section, features two basic tonal levels: high (H) and low (L) (Kennedy, 1966; Anttila & Bodomo, 1996, 2022; Angsongna, 2023). This is exemplified with the following minimal pairs.

(1)	Minimal pairs
	H		L
	tú	‘dig’	tù	‘follow’
	kʊ́	‘kill’	kʊ̀	‘give’
	báá	‘dog’	bàà	‘grow’
	zɛ́lɪ́	‘poison’	zɛ̀lɪ̀	‘lift’

Additionally, downstep—a phenomenon in which the second of two high tones is lower than the first (Hyman, 1985; Selkirk & Tateishi, 1991)—is observed and represented as in bʊ́ꜜmá ‘things’. Though not a regular feature in Dagaare, two contour-like tones exist: rising and falling, exemplified by a few examples as in wɪ̀ɛ́ ‘farm’ and nɪ́ɛ̀ ‘person’, respectively.

As in many tone languages, tone in Dagaare serves both lexical and grammatical functions. The minimal pairs in (1) illustrate the lexical role of tone, while the sentences in (2) demonstrate its grammatical function. Specifically, sentence (2a) expresses non-future negation with a past-oriented modal, often conveying evaluative or counterfactual meaning, whereas sentence (2b) expresses future negation.

(2)	a.	ʊ̀	kʊ́ŋ	wá
		3sg	pst.irr.neg	come
		‘S/he should not have come’
	b.	ʊ̀	kʊ̀ŋ	wá
		3sg	fut.neg	come
		‘S/he will not come’

The distribution of tones in Dagaare differs between verbs and nouns, as will be discussed in subsequent sections. Before turning to that discussion, however, we first examine what constitutes the tone-bearing unit in Dagaare.

3.1.1. Tone-Bearing Unit in Dagaare

It is important to note that in Dagaare, the syllable functions as the tone-bearing unit (TBU), as observed by Kennedy (1966) and Anttila and Bodomo (1996). Further evidence supporting this claim is provided by Yip (2002), who argues that if a language has two different syllable weights—light and heavy, and both can bear the same number of tones—the syllable is the TBU.

Thus, Dagaare exhibits cases in which both light and heavy syllables can bear the same number of tones. Yip’s argument therefore accurately captures the Dagaare pattern, supporting the view that the tone-bearing unit (TBU) is the syllable rather than the mora.

Dagaare permits a maximum of two pitch levels per syllable. Phonologically, this corresponds to up to two tonal associations per syllable, realized as either HL or LH contours. This pattern holds across the syllable types—CV, CVV, and CVC—for HL contours, while LH contours are generally restricted to CVV and CVC syllables. Consequently, only HL contours are attested on CV syllables (as will be demonstrated in Section 3.1.4. below).

Consider the following examples from Angsongna (2023, pp. 203–205). In (3a), a monomoraic syllable bears two tones; in (3b), a bimoraic syllable bears a single tone; in (3c), a monomoraic syllable again bears two tones; and in (3d), a bimoraic syllable bears two tones.

(3)

For the LH contour to be possible we require two moras. This is probably where the mora may be relevant in Dagaare. See the following examples. If we follow the argument for the HL forms in (3) above, then the LH cases in (4) appear to show that the TBU is the mora.

(4)

The absence of CV monomoraic LH contours in Dagaare is probably related to the observation that the HL contour is the most common in tone languages (Cahill, 2007; Maddieson, 1978). Given the special case of discrepancy with rising LH contours, it can be assumed that the syllable being the TBU is not entirely straightforward in Dagaare. In this research, we will assume the syllable as the TBU and leave the debate open for further investigation in terms of whether it is the mora or the syllable that constitutes the TBU.

3.1.2. Distribution of Verbal Tone

The basic structure of the Dagaare verb consists of a verb root followed by an aspectual suffix. Verbs appear in two primary forms: the default form (often referred to as the perfective) and the imperfective (or progressive). The default form consists of either the bare verb root or the root followed by a final front high vowel. In contrast, the imperfective form consistently includes the root plus a suffix of the shape (-C)V.

With regard to tone, Angsongna (2021, 2023) reports that Dagaare verbs fall into three tonal classes, each exhibiting distinct surface patterns in the default and imperfective forms. The default form, which may be mono- or disyllabic, shows three tonal patterns: H, L, and HL. In the imperfective, which is always disyllabic, the three default patterns correspond to the following surface forms: a default H yields H-L; a default L surfaces as L-H; and a default HL becomes H-ꜜH (HꜜH), which is analyzed as HLH with a floating L tone (Kenstowicz, 1994; Pulleyblank, 1986).

Importantly, the default form does not exhibit an LH pattern, and neither HH nor LL patterns occur in the imperfective. However, a HH pattern may arise postlexically in the imperfective when certain preverbal particles precede the verb (see Angsongna, 2023; Anttila & Bodomo, 2007).

Table 1. Tone in verbs (default vs. imperfective).

Tone	Default	Gloss	Tone	Imperfective	Gloss
L	dì	‘eat’	L-H	dì-ré	‘eat-IPFV’
	tùò	‘carry’		tùò-ró	‘carry-IPFV’
	wù̀lì	‘show’		wùl-ó	‘show-IPFV’
H	lé	‘tie’	H-L	lé-rè	‘tie-IPFV’
	lɔ́ɔ́	‘drop/throw’		lɔ́ɔ́-rɔ̀	‘drop/throw-IPFV’
	dʊ́ɡɪ́	‘cook/boil’		dʊ́ɡ-rɔ̀	‘cook/boil-IPFV
HL	dɪ̂	‘take’	H-LH = H-!H	dɪ́-ꜜrɛ́	‘take-IPFV’
	dáà	‘push’		dáá-ꜜrá	‘push-IPFV’
	kɔ́rɪ̀	‘cough’		kɔ́r-ꜜɔ́	‘cough-IPFV’

, based on data from Angsongna (2023), summarizes the tonal correspondence between the default and imperfective verb forms in Dagaare.

Comparing the default and the imperfective, apart from the fact that the tone on the imperfective suffix is always opposite to the final root tone, there is the presence of vowel harmony between the root and the imperfective suffix in terms of tongue root, backness, and rounding feature. Since our paper is about tone, there is no further discussion of harmony. Following the tonal patterns observed in Dagaare verbs, the following generalization can be made: the imperfective suffix never has the same tone as the preceding root syllable. That is, if the root is L, the suffix must be H and if the root is H, the suffix must be L or ꜜH, where ꜜH is the result of a floating L in underlying HLH.

Building on the description and generalization of verbal tone, the claim is that the underlying tonal melodies of the verb root remain consistent with the surface tones of the default verb forms, regardless of the presence of a final vowel. This means that the underlying tones in the root do not change on the surface. The default forms may include a final vowel, which, if present, is always underlyingly toneless. The final or rightmost tone of the root then spreads onto the final vowel if it (the final vowel) is present. The mechanism for this tonal spreading is illustrated below.

(5)

Underlying and surface forms for the default verb

In the imperfective forms, an additional tone appears to the right of the root tone. The phonological system determines the form this additional tone takes. Specifically, the imperfective suffix exhibits tonal polarity, surfacing with a tone of opposite height to that of the root tone (Hoffman, 1963; Kenstowicz et al., 1988; Pulleyblank, 1983, 1986). This polarity is believed to result from the morphological process of suffixation, as the imperfective includes a dedicated aspectual suffix that is absent in the default form. The presence of this suffix consistently alters the tonal pattern of the verb. The proposal is that the imperfective suffix has an underlying H tone, which surfaces as either H, L, or ꜜH depending on Obligatory Contour Principle (OCP)-driven constraints (Leben, 1973). The tonal representation of the imperfective forms is as follows:

(6)

Underlying and surface forms of the imperfective

In (6a), the underlying H tone of the suffix is deleted, and an L is inserted due to a dissimilatory process driven by the OCP. In (6b), the suffix carries an associated H tone following a root with an L tone and this remains unchanged on the surface because tonal polarity is preferred. In (6c), the suffix H tone remains intact after a HL root, and the associated L is left floating on the surface, causing downstep. The imperfective forms consistently exhibit tonal polarity, with no H-H or L-L surface tone patterns in Dagaare. This absence of surface H-H or L-L tones is explained by dissimilation (see Suzuki, 1998), a process identical to the dissimilatory analysis of tonal polarity proposed by Kenstowicz et al. (1988).

3.1.3. Distribution of Nominal Tone

Most simple nouns in Dagaare belong to one of three primary tonal classes: H-H, H-L, and L-H. Interestingly, the L-L tonal pattern is systematically less productive for simple nouns (Anttila & Bodomo, 1996, 2000; Bodomo, 1997), although a small number of exceptions with an L-L pattern2 do exist. Note that by simple noun, we refer to a combination of a noun root and a number and class suffix. In other words, all simple nouns consist of a two-morpheme sequence: a root and a suffix (although in some cases, the suffix may be null). In terms of tonal contrast in nominal roots, as shown in

Table 2. Tone in nouns (singular vs. plural).

	Tone	Root	Singular	Gloss	Plural	Gloss
a.	H	bi-	bíé	‘child’	bíí-rí	‘children’
		bʊ-	bʊ́ɔ́	‘goat’	bʊ́ʊ́-rɪ́	‘goats’
		taŋ-	táŋ-á	‘mountain’	tán-nɪ́	‘mountains’
b.	LH	pì-	pìé 1	‘rock’	pìì-rí	‘rocks’
		ɲù-	ɲùó	‘navel’	ɲùù-rí	‘navels’
		bùɡ-	bùɡ-ó	‘silo’	bùɡ-rí	‘silos’
c.	HL	mí-	mí-rì	‘rope’	míè	‘ropes’
		nɔ́	nʊ́ɔ̀	‘hen’	nʊ́ʊ́-rɪ̀	‘hens’
		dʊ́ɡ-	dʊ́ɡ-ɪ̀	‘pot’	dʊ́ɡ-rɪ̀	‘pots’

¹ In Waale (Southern Dagaare), pìé–pììrí are reverse singular and plural, that is, pììrí (sg) and pìé (plural).

, three types can be identified: underlyingly toneless roots (a), high-toned roots (b), and low-toned roots (c). Unlike verbs, nouns generally maintain the same tonal melody in both singular and plural forms, with only a few exceptions. Table 2 below highlights these three predominant tonal classes observed in Dagaare nouns.

The tone patterns in Dagaare nouns exhibit cases where the singular/plural suffix either mirrors or contrasts with the tone of the root. In examples like those in (a) in the table, the H tone of the suffix contributes the H tone of the root, resulting in both sharing the same surface tone. However, in cases like (b) and (c), the tone of the root contrasts with that of the suffix, a phenomenon known as tonal polarity—a dissimilatory process in which a tone-bearing unit, typically an affix, adopts a tone value opposite to that of the adjacent root (Hoffman, 1963; Kenstowicz et al., 1988; Pulleyblank, 1983, 1986). Essentially, low (L) and high (H) roots lead to opposite tones on the suffix.

3.1.4. Contour Tones in Dagaare

Contours are tones that shift from one pitch to another over the course of a syllable. While Dagaare exhibits some contour tones, they are not a regular feature in the language. There are limited instances of contour-like tones—specifically falling (HL) and rising (LH) contours—attested in a small number of words. Both HL and LH contours typically occur on CVV syllables, as illustrated in the following examples:

• HL contours: nɪ́ɛ̀ ‘person.sg’, dáà ‘push’.
• LH contours: tùó ‘baobab.sg’, wɪ̀ɛ́ ‘farm.sg’.

In particular, LH contours are only found in nouns and are unattested in verbs. In addition to CVV syllables, a few CV syllables also exhibit HL contours, such as:

• nû ‘hand’, zû ‘head’, ɡbô ‘heart’, ʧî ‘millet’.

Nouns such as nû and zû are historically derived from sequences with long vowels, as núù and zúù, respectively. Through a vowel reduction process, specifically vowel deletion or contraction, one of the two vocalic morae is lost. The tonal specification of the deleted vowel, which bears a low tone, is not lost but instead reassociates (or docks) onto the remaining vowel. This tonal reassociation results in the realization of a high–low (HL) contour tone on the surface CV form.

LH contour tones do not occur on CV syllables as they are generally restricted to CVV and CVC syllables (see Section 3.1.1 above). Only one CV verb with a contour tone has been identified in Dagaare: dɪ̂ ‘take’, which bears a HL contour. No CV verb with an LH contour has been attested in the language.

Cross-linguistically (Hyman, 2011; Yip, 2002), contour tones typically occur on final syllables. In some dialects of Dagaare, however, there are a few exceptions where LH contours appear on initial syllables, as in ɡʊ̀ʊ́-rɪ̀ ‘thorn (SG)’ and tàám-à ‘shea fruit (PL)’. Given the highly restricted and infrequent occurrence of contour tones, which arise not as minimal tonal units but pattern as a sequence of level tones, Dagaare can be characterized primarily as a register tone language. Contour tones appear to be exceptional rather than systematic.

From the preceding description of nominal tone and contour patterns, we make the following generalizations:

• There is a restricted number of L-L nouns in Dagaare;
• There are no surface HLH patterns within simple words (nouns and verbs) in Dagaare;
• A contour tone in Dagaare has a maximum of two levels of pitch, HL and LH, that is, a maximum of two tones associated to a TBU;
• Both rising and falling contours are attested in nouns, but only falling contours are attested in verbs (=*LH in verbs);
• There is no evidence suggesting the presence of a floating H, but L tones can float between high tones, causing downstep.

Following the above discussion of the nouns, we adopt the view that in Dagaare, nominal roots may be toneless, H, or L, leading to three tonal types: (i) consistently L roots, e.g., [bùɡ-] ‘silo’; (ii) consistently H roots, e.g., [dʊ́ɡ-] ‘pot’; and (iii) alternating roots that are L in compounds and H with a number suffix, e.g., [pɔ̀ɡ]-bílé ‘small/young woman’ vs. [pɔ́ɡ]-ɔ́ ‘woman-sg’. Suffixes surface as H after low roots, L after high roots, and sometimes both root and suffix share the H tone. L and H roots reflect their underlying tones, while alternating roots are assumed to be toneless. Below is an illustration of each root type.

(7)

Underlying roots

Similar to the case of the imperfective suffixes, we assume that all nominal suffixes are underlyingly H-toned, not toneless as illustrated in (8) below.

(8)

Suffix

In summary, the default verb forms in Dagaare are derived from three underlying roots: H, L, and HL, all of which have toneless final vowels. In both the H and L cases, a single tone spreads across the root and the final vowel, which is preferred over assigning separate tones to each TBU. With the HL root, the L tone simply docks onto the final vowel from the root. For the imperfective forms, an additional underlying H tone appears, and based on OCP-driven constraints, the suffix surfaces as H, L, or ꜜH.

Similarly, simple nouns have three main underlying roots: toneless, H, and L. As with the imperfective forms, the nominal suffix is underlyingly H and remains H after an L root, spreads to a toneless root, or changes into L due to dissimilatory effects influenced by the OCP. Thus, the imperfective H-L and L-H patterns mirror the representations of nominal H and L roots.

Other aspects of Dagaare tone, such as downstep, OCP effects, and tone patterns in complex forms (e.g., associative constructions (see Anttila & Bodomo, 1996), nominalizations, agentive reduplication, noun–adjective compounds (see Angsongna, 2023), and ideophones (see Bodomo, 2006)) are noteworthy. However, at present, this paper does not delve into these aspects of Dagaare tone. Now, the attention shifts to Dagbani.

3.2. Tone in Dagbani

In terms of tone inventory, Dagbani is similar to Dagaare with two contrastive tones—high and low (Hudu, 2010; Hyman, 1993; Hyman & Olawsky, 2004)—as exemplified below.

(9)	Minimal pairs
	H		L
	dám	‘alcohol’	dàm	‘shake’
	kpáŋ	‘guinea fowl’	kpàŋ	‘axe’	(Issah, 2020, p. 16)

The examples above in (9) illustrate the lexical function of tone in Dagbani. With regard to its grammatical function, Olawsky (1999) acknowledges that grammatical tone is not absent in the language but does not provide empirical data to substantiate this claim, instead calling for further research. Consequently, there is currently no clear evidence to exemplify the grammatical role of tone in Dagbani, and this remains an area for future investigation.

In this section, we give a brief overview of Hyman and Olawsky’s (2004) account of verb tonology, which proposes four essential steps for analyzing verb tone. First, the underlying tonal representations of morphemes must be identified, revealing how verb roots exhibit tonal oppositions. Second, rules of morphological tone assignment are determined, whereby tense, aspect, or mood impose specific tonal melodies that may override root tones. Third, lexical phonological processes are examined, as these may further alter both underlying and morphologically assigned tones, often in highly morphologized ways. Finally, postlexical phonological processes are considered, accounting for additional tonal changes beyond the lexical level.

Hyman and Olawsky argue that Dagbani verb tone is grounded in underlying representations, with all subsequent tonal processes building on these patterns. Evidence from verb roots demonstrates systematic tonal differences (high vs. low) shown below.

(10)		H vs. L verbs
	a.	CV
		dú ‘go up’	lù ‘fall’
		ɲé ‘put down’	tò ‘pound’
	b.	CVN
		ném ‘grind’	wùm ‘hear’
		jém ‘yawn’	dèm ‘play’
		kpán ‘axe’
		dán	‘to carry with difficulty’
	c.	CVC(i)
		sábí ‘write’	làbì ‘return (intr)’
		lábí ‘throw’	tàɡì ‘exchange’
	d.	CVV(i)
		ɡbááí ‘catch’	ɡùùì ‘run’
		lóóí ‘cross’	jòòì ‘open’
	e.	CVCC(i)
		wálɡí ‘divide’	zàɡsì ‘refuse’
		sírɡí ‘go down’	zìlsì ‘wander’
	f.	CVC(i)C
		dáhím ‘taste’	kàrìm ‘read’
		jóhím ‘deceive’	làɡìm ‘meet’

The following observations are made:

• Within the morphological structure, a Dagbani verb stem can consist of a root only or of a root plus a suffix, e.g., /lab/ ‘return’ (intr.) vs. /lab-si/ ‘return’ (tr.), where /-si/ is a causative suffix.
• Within the syllable structure on the surface level, a verb stem can be monosyllabic (CV, CVN, CVV) or disyllabic (CV.Ci, CV.Vi, CVC.Ci, CV.Cim), noting that the only permissible word-final consonants in Dagbani are /m/ and /n/, as seen in (10b).
• Still on the syllable, the second syllable [i] is always predictable in Dagbani. That is, it occurs only after final oral consonants as in (8c–e) and long vowels as in (10d), and between an oral consonant and /m/ as in (10f).
• And finally, the root structures of the verbs as seen in the data above are thus /CV/, /CVN/, /CVC/, /CVV/, and /CVCC/. The different placement of the second syllable [i] as in (10e,f) is thus predictable; both can be argued to be underlyingly CVCC, e.g., /zaɡs/, /karm/, where /s/ and /m/ may be (frozen) suffixes.

It has further been reported (see Hyman & Olawsky, 2004) that certain verb paradigms were used to collect data on the verb tones in Dagbani. These paradigms are necessary as they assist the researchers to group and dichotomize the verb tones in the language for clearer understanding of the phenomenon in Dagbani. These verb paradigms are listed below.

(11)	Verb paradigms for which systematic tonal data was collected
	Tenses	Aspects	Clause type	Polarity
	Present (0)	Perfective (PFV)	Main Clause	Affirmative
	Today Past (P1)	Imperfective (IPFV)	Relative Clause	Negative
	General Past (P2)
	General Future (F)
	Incipient ‘about to’
hence: Main Clause Affirmative, (MCA), Main Clause Negative (MCN), Relative Clause Affirmative (RCA), and Relative Clause Negative (RCN).

From the above, five tenses are identified, which occur in both perfective and imperfective aspects, and are then premeditated in main and relative clause affirmatives and negatives in Dagbani. These paradigms are used to establish four different tone patterns in Dagbani verb tonology, as shown in

Table 3. Patterns of Dagbani verb tonology.

MCA		MCN		RCA		RCN
PFV	IPFV	PFV	IPFV	PFV	IPFV	PFV	IPFV
LH	LH	HL	LH	Lex	LH	HL	LH
zàgsìyá	kùrá	málì	sàɣsìrá		bàrá	dúɣì	sàbrá
‘refused	‘killing’	‘have/own’	‘shifting’		‘riding’	‘cooked’	‘writing’
LH	LH	HL	LH	Lex	LH	HL	LH
zàgsí	dìrá	chíhì	chìmdá		wàrtá	dúhì	ŋùbrá
‘refused’	‘eating’	‘mark’	‘frying’		‘splitting’	‘raised’	‘chewing’
LH	LH	HL	LH	Lex	LH	HL	LH
zàgsí	mìrá	chírígì	wùrìndá		yù:ná	gábì	bùɣsìrá
‘refused’	‘raining’	‘meet’	‘destroying’		‘watching’	‘mix’	‘measuring’
H	H	H	H	(H)	(H)	H	H
dí ‘ate’	ló ‘tie’	kú ‘kill’	yá	chím ‘fry’	zíŋ	bá ‘ride’	ӡí ‘carry’
			‘harvest’		‘miss’
HL	LH	HL	LH	HL	LH	(HL)	(LH)
zágsì	sìbrá	dónì	lòrá	díhyá ‘fed’	dàrá	gòyá	wàrá
‘refused’	‘writing’	‘lie’	‘tying’		‘buying’	‘travelled’	‘dancing’

below.

Basically, the verb tone patterns in Dagbani are as presented below:

• H: H tone pattern, merger of L and H lexical root tones of CV and CVC stems, as in ló ‘tie’ and ʧím ‘fry’, respectively;
• LH: LH tone pattern, merger of L and H lexical root tones in all stems, as in mè ‘build’ and sàbrá ‘writing’;
• HL: HL tone pattern, merger of L and H lexical root tones except on CV and CVC stems, as in móóì ‘fetch’ and bíì ‘heat’;
• Lex: L and H lexical root tones realized on each stem syllable.

Moreover, Hyman and Olawsky observe that the order of presentation, as shown in Table 3 above, tentatively suggests that there is a hierarchy of paradigmatic verb tone assignment in Dagbani (see (12) below).

(12)	Hierarchy of paradigmatic verb tone assignment
	Future		IPFV		{Neg, Incip}		MC PFV		RC PFV
	H	>>	LH	>>	HL	>>	LH	>>	Lex

This hierarchy may either be considered morphological (based on grammatical features that assign tones) or phonological (based on the tone patterns themselves). Hyman and Olawsky argue that the hierarchy cannot be purely phonological due to overlapping tone patterns (e.g., the imperfective LH overriding negative and incipient HL, while HL overrides the main clause perfective LH). Thus, the suggestion is that it should be seen as a combination of both morphological features and tone patterns, though the tones themselves are seen as superfluous to the hierarchy.

The hierarchy reflects a systematic tone assignment process influenced by both morphological and phonological features. However, the relationship between these features raises questions about the dominance of specific grammatical categories (like future tense) over broader intersecting features and why certain tonal patterns are shared across seemingly unrelated grammatical forms (e.g., negative and incipient). The authors propose a need for further clarification on whether the hierarchy should prioritize morphological structure or tonal realization.

3.2.1. Tone-Bearing Unit in Dagbani

Dagbani phonology treats the syllable as the primary tone-bearing unit (TBU), but specific data from verb tonology reveal moraic sensitivity in root structure (Hudu & Nindow, 2020). Key generalizations on the Dagbani TBU include the following:

• In Dagbani, monomoraic roots count as single TBUs: CV and CVC roots (e.g., dú ‘climb’, lù ‘fall’, ném ‘grind’, wùm ‘hear’ (L)) merge L and H lexical tones into a single surface H or L on one TBU in paradigms like future perfective, as tone spreads across the single mora regardless of underlying contrast.
• Bimoraic roots support multiple TBUs: CVV and CVN roots with long vowels (e.g., ɡbááí ‘catch’ (H), ɡùùì ‘run’ (L), píí ‘choose’, záɡsí ‘refuse’) realize distinct LH or HL patterns across two TBUs, as seen in future perfective, where L roots show doubly linked H (ò-n záɡsí kóꜜdú) vs. H roots with HL (ò-n tábsì kôdú).
• Finally, TBU count drives paradigmatic tone hierarchy: The number of TBUs in roots determines tone realization in hierarchies (Future IPFV > {Neg, Incip} > MC PFV > RC PFV), with monomoraic verbs uniformly H (ò-n tó kóꜜdú) but bimoraic ones differentiating via spreading rules like HTS/LTS, yielding downstep or falling contours only on multi-TBU structures.

Dagbani is also characterized by the presence of contour tones (Olawsky, 1999). However, as in Dagaare, these contour tones do not function as independent minimal units as they are derived from the combination of level high and low tones, as in bâ ‘father’. This suggests that Dagbani, similar to Dagaare, is primarily a register tone language.

3.2.2. Tone Assignment Rules in Dagbani

In this account, we follow the lexical phonology and morphology architecture with postlexical rules as proposed in Hyman and Olawsky (2004). These rules would mainly be High Tone Spreading (HTS) and Low Tone Spreading (LTS), as will be demonstrated here. There are two key postlexical tone rules—High Tone Spreading (HTS) and Low Tone Spreading (LTS)—as well as the implications for verb tone patterns in Dagbani. High Tone Spreading (HTS) causes an underlying high (H) tone to spread to the right, delinking the immediately following low (L) tone. In the sense of Hyman and Olawsky (2004), if two successive L tone-bearing units are available, HTS delinks the first L and creates a HL falling tone with the second unit. Consider the following examples:

(13)	Without HTS: ò zàɡsí yá ‘he has refused’. Here, the L tone subject pronoun and LH verb melody remain intact. This is in line with the proposal of Hyman (1993) and Olawsky (1999) that an underlying H tone will spread to the right, delinking an immediately following L tone.
(14)	With HTS: n zàgsì yá → ń záɡsî yá ‘I have refused’. Here, the H tone from the subject pronoun spreads, creating an H plus HL falling sequence.

On Low Tone Spreading (LTS), however, a low (L) tone spreads onto the following stressless H tone-bearing unit, delinking the H tone, but only applies when the sequence is not utterance-final. See examples below from their data.

(15)	Input: ò zàɡsí kòdú ‘he has refused a banana’.

✓

HTS applies first: ò zàɡsí kóꜜdú.

✓

Then, LTS spreads the L tone: ò zàɡsì kóꜜdú.

✓

This results in a downstepped H tone on kóꜜdú.

As argued by Hyman and Olawsky, the noun object kòdú ‘banana’ has a floating L tone that is between linked Hs, resulting in a downstepped H tone. Thus, their assumption is that the sequence H-ꜜH is simplified to H-L, yielding the form ò zàgsì kódù ‘he has refused a banana’.

3.2.3. Application to Future Perfective Forms

The future perfective constructions show that verbs are realized with an all-H tone pattern. Thus, for monosyllabic roots (CV or CVC), L and H tone verbs merge, and for disyllabic roots (CVV or CVN), L and H tone verbs do not merge with tonal differences, evident in the realization of the object kòdú ‘banana’. These applications are shown in examples (16a–b) for the monosyllabic roots and in (16c–d) for the disyllabic roots, respectively.

(16)	a.	L tone verb ò-n tó kóꜜdú (‘he will pound banana’)
	b.	H tone verb ò-n nyé kóꜜdú (‘he will put down banana’).
	c.	L tone verb ò-n záɡsí kóꜜdú (‘he will refuse banana’).
	d.	H tone verb ò-n tábsì kôdú (‘he will touch banana’).

Note that a downstep occurs before the second syllable of kòdú in all examples in (16a–c). This pattern is absent in (16d), where the initial syllable instead bears a HL contour tone. The downstep observed in (16a–c) may be conditioned by the sequence of preceding high tones, which triggers automatic downstep before the following LH-toned object. In contrast, the HL realization in (16d) may be attributable to the HL-toned verb tábsì ‘touch’, whose tonal contour spreads or conditions the HL pattern on the following syllable.

3.2.4. Complications in Dagbani Verb Tone Assignment

Several complications in tone assignment have been observed. For example, for (i) future tense marking, it is observed that in monomoraic verbs (e.g., ò-n tó kó!dú), the future tone must be a single H, but in bimoraic verbs (e.g., ò-n záɡsí kó!dú), the L tone root must have a doubly linked H. Furthermore, for (ii) differentiation in disyllabic verbs, it is noted that for L tone verbs (e.g., zàɡsì), HTS spreads the doubly linked H, producing a downstep on the object (e.g., kó!dú), but for H tone verbs (e.g., tábsì), a HL sequence is required, allowing HTS to apply twice and generate a HL falling tone on the object (e.g., kôdú). The data in (17) is a complete extract from Hyman and Olawsky (2004, p. 102).

(17)	MCA Future Perfective (“H tone pattern”): verb + /kòdú/ ‘banana’
	H Tone Verb Roots
	ô-n tó kóꜜdú ‘he will pound banana’	ô-n nyé kóꜜdú ‘he will put down banana’
	ô-n tág kóꜜdú ‘he will exchange banana’	ô-n láb kóꜜdú ‘he will throw banana’
	ò-n záɡsí kóꜜdú ‘he will refuse banana’	ò-n tásíb kóꜜdú ‘he will touch banana’
	ò-n píí kóꜜdú ‘he will choose banana’	ò-n gbáá kóꜜdú ‘he will catch banana’

Evidence from the above data provides an open problem, in which the questions are raised regarding how the future tense replacive H tone is realized differently across verb roots.

✓

Why does an L tone verb like záɡsí have a doubly linked H tone?

✓

Why does an H tone verb like tábsì instead show a H-L sequence?

Answering the questions above requires an understanding of the complex interactions of tone, which also reveals gaps in our knowledge of the mechanisms underlying Dagbani tonal grammar, particularly the interaction between postlexical rules such as HTS and LTS and lexical and morphological tone patterns. Ultimately, to deepen understanding of tone-morphology interactions in Dagbani and related languages, further research is required. Moreover, additional data, especially from dialectal variation and syntactic contexts, could further enrich research in this area of tonology. In the next section, we look at tone in Gurenɛ.

3.3. Tone in Gurenɛ

This section provides a summary of tone in Gurenɛ, spoken in the Upper East Region of Ghana and southern Burkina Faso. The discussion draws on two principal scholarly works: Atintono (2004) and Adongo (2018). Notably, while Atintono’s work centers on the morphosyntactic structure of the Gurenɛ verb, Adongo addresses broader aspects of Gurenɛ phonology, with special emphasis on tone.

Atintono (2004) examines the phonological systems that underpin verbal morphology in Gurenɛ, emphasizing tone as a crucial suprasegmental feature, while Adongo (2018) provides a comprehensive, multidimensional analysis of the language’s tonal system, situating it within the broader Mabia linguistic landscape and comparing its tonal patterns with those of related languages. Both studies acknowledge the centrality of tone in Gurenɛ but diverge in some of their claims/findings, as will be briefly shown in subsequent sections.

3.3.1. Tone Inventory

According to Atintono (2004), Gurenɛ possesses two principal contrastive tones: high and low. Tone is segmentally realized on the syllable nucleus, and there is grammatical tone that affects the surface realization of both nouns and verbs, contributing directly to grammatical meaning and function (cf. Adongo, 2018). At the lexical level, tone is distinctly contrastive for both nouns and verbs, as shown in the examples that follow.

(18)	Two-tone contrasts in Gurenɛ
	a.	Verbs
		kóné	‘knock (on head)’	kònè	‘miss something’
		bélégé	‘coax’	bèlègè	‘over apply oil (on body)’
		nyáŋɛ́	‘pluck leaves’	nyàŋɛ̀	‘overcome’
	b.	Nouns
		sánɛ́	‘melon’	sànɛ̀	‘debt’
		báŋá	‘ring’	bàŋà	‘lizard’
		(Atintono, 2004)

Verbs in Gurenɛ generally exhibit a predominance of low tone patterns over high tone patterns (see Adongo, 2018), a claim further supported by entries in Gurenɛ dictionaries. In addition, Atintono observes that tonal alternations occur in verbal constructions, noting that the tonal pattern of a verb may shift across contexts—particularly in affixation—yielding configurations such as low–high, high–low, or high. These alternations can signal differences in grammatical meaning, indicating that tone in Gurenɛ is both contrastive and grammatically distinctive.

This grammatical contrast operates through processes such as tonal polarity and spreading, whereby affixes may adopt tones that are contrastive with those of the root. For example, as illustrated in Dakubu (1996), the plural suffix -sí typically bears a high tone (or HL contour) in contrast to the root: bàà ‘dog’ → bàà-sí ‘dogs’, demonstrating an L-H polarity pattern. In effect, a low-toned root triggers a high-toned suffix, while a high-toned root may result in a low or downstepped high tone on the suffix. These polarity effects are comparable to those observed in Dagaare, where high-toned roots take low-toned suffixes and low-toned roots take high-toned suffixes.

A similar process is observed in the marking of progressive or imperfective aspects. The suffix -ti introduces a grammatical tone that either overrides or interacts with the default tone of the verb root. For instance, dí ‘buy’ becomes dí-tí ‘buying’, where tonal interaction yields a modified pattern via H tone spreading. Also, a default low-toned verb root may take a high-toned suffix, resulting in an L-H imperfective form.

Overall, grammatical tone in Gurenɛ emerges prominently in affixation, where plural and aspectual suffixes carry tonal patterns that either override the stem’s default tone or interact with it through tonal spreading and polarity.

Adongo (2018) provides compelling evidence for a three-level tonal system—high (H), mid (M), and low (L)—thus diverging from Atintono (2004) and earlier studies that propose a two-tone system (high and low). In addition, Adongo identifies contour tones (falling: HL; rising: LH), which she analyzes as surface realizations arising from phonological processes such as tone spreading, assimilation, and segment deletion at morpheme boundaries. This analysis therefore suggests that Gurenɛ does not possess underlying contour tones and, like Dagaare and Dagbani, is best characterized as a register tone language. The following data illustrates the three-level tonal system.

(19)	Three-tone contrasts in Gurenɛ (Adongo, 2018, p. 198)
	láá	‘bowl’	dāā	‘past’	làà	‘falling position’
	díá	‘food’	dìá	‘dance’	dīà	‘animal’

Note that in Gurenɛ, a mid-tone may occur utterance-initially, particularly when followed by a tone of a different height, resulting in ML contour realization such as dīà ‘animal’, which adds to the other attested contours in the language: HL lúà ‘well’ and LH, as in dìá ‘dance’. However, our data do not support a phonemic contrast between MH and high (H), and there is no MH contours phonetically attested.

Crucially, these contour tones do not function as independent minimal units, nor do they exhaust the tonal inventory of the language. Rather, they arise from the interaction of level tones. This pattern supports the classification of Gurenɛ as generally a register tone language, similar to Dagaare and Dagbani discussed above.

In terms of tonal sequences, these occur across monomorphemic, bimorphemic, and compound words, with simpler sequences appearing in short (one- or two-syllable) words and more complex tonal patterns found in longer words and compounds. It is important to note that Gurenɛ is analyzed as having the mora, rather than the syllable, as the tone-bearing unit (TBU). For evidence supporting this claim, see Adongo (2018). This contrasts with languages such as Dagaare and Dagbani, in which the syllable functions as the TBU. Consider the following tonal sequences across simple to compound words.

(20)	Tone sequences
	a.	One-syllable words	zóm (H) ‘flour’,	dā (M) ‘past’	zòm (L) ‘climb’
	b.	Two-syllable words	dʊ́kɔ́ (H-H) ‘pot’, zʊ̄ʊ̄rɛ̄ (M-M) ‘tail’		sìrā (L-M) ‘tail’
	c.	Three-syllable words	átíà (H-H-L) ‘tree’
	d.	Four-syllable words	pɔ̀ɣə̀dírí (L-L-H-H) ‘marriage’
	e.	Five-syllable words	átíābííté (H-H-M-H-H) ‘atia’s seed’

It is worthy to note additional possible tone melodies such as LL as in dààm ‘pito’, LH as in sànɛ́ ‘debt’, and HL as in nááfɔ̀ ‘cow’. A significant observation in Gurenɛ tonology is tonal stability and docking, in line with the broader tone literature, in compound words as in lúà + bílà → lûbílà ‘small well’. That is, compounding, often accompanied by vowel or syllable reduction in either constituent, tends to trigger tone shift, tone change, or tone spread, resulting in contour tones. When a segment is deleted, its tone may be preserved and realized at the morpheme boundary within the compound.

Tone in Gurenɛ is a fundamental carrier of lexical and grammatical information, distinguishing minimal pairs across word classes. For instance, the minimal pairs in (21) give an illustration of lexical function.

(21)	a.	báŋá	‘ring’	bàŋà	‘lizard’
	b.	kóné	‘knock (e.g., on head)’	kònè	‘to miss’

In verbs, the default tone pattern is typically low, though certain verbs form minimal pairs via tonal contrast. Moreover, tone may change contextually in specific grammatical constructions. For example, in negative imperative sentences:

(22)	a.	dá	kínɛ́		‘don’t go!’	(singular)
	b.	dá	kínɛ̀	yá	‘don’t go!’	(plural)

In these cases, a high tone is associated with a singular addressee, and a high–low sequence marks the plural. Functionally, tone interacts with morphosyntactic modifiers to mark tense, aspect, mood, negation, and imperative distinctions in verbal expressions.

3.3.2. Tonal Processes in Gurenɛ

Tone in Gurenɛ extends beyond the segmental, lexical, and grammatical levels, encompassing processes such as assimilation, downstep, deletion, and alternation (Adongo, 2018). Among these, tone assimilation (or tone spreading) is one of the most common phenomena in the language. It occurs when a tone assimilates to, or takes on, the tonal features of an adjacent segment. In effect, when uttered contiguously, high or low tones may spread onto neighboring TBUs. Tone spreading occurs mainly in compounding in Gurenɛ, particularly when a segment is deleted and its tone is preserved and spreads to adjacent segments. Consider the following examples in (23).

(23)	Tone assimilation
	a.	zóm + kòɁóm → zóŋkóɁóm ‘flour + water’ → ‘beverage’
	b.	sɔ̀gɔ́rɔ̀ + pììsɛ̀ + ká → sɔ̀ɣɔ̂píísə́ká ‘rubbish broom’

Segmental reduction at morpheme boundaries may also result in contour formation on a single TBU. Regarding downstep in Gurenɛ, when two high tone words occur in a sequence (e.g., in a phrase), there is a tendency of lowering the high tone of the first syllable of the second word—the word boundary, causing downstep as in (24). Tone spreading typically occurs in compounds but can also apply across affixal boundaries in multimorphemic words in Gurenɛ.

(24)	Downstep
	díá ꜜdʊ́kà ‘food cooking’
	dɔ́ɔ́rɔ́ ꜜkʊ́ká ‘wooden chair’

Overall, Gurenɛ exhibits a tonal system characterized by two to three level tones—low, (mid), and high—and their combinations, which results in contour tones (HL ‘falling’, LH ‘rising’). The verb typically carries a low default tone while noun tone varies based on its lexical, compound, or phrasal context. Tone contrast is prominent among nouns, with minimal pairs also found between nouns/verbs and nouns/adjectives. Tonal processes include assimilation, tone shift, and downstep.

3.4. Tone in Kusaal

Kusaal has long been classified as a tonal language with other sister Mabia languages including Dagbani, Dagaare, and Buli among others. Previous studies have established that there are three levels of tones in the language: high, mid, and low (Abubakari, 2018; Musah, 2010, 2018).

Table 4. Minimal pairs in Kusaal.

H		L
báŋ	‘bracelet/trap’	bàŋ	‘to know’
bʊ́k	‘to be weak’	bʊ̀k	‘to divine’
búɡ	‘to be drunk’	bùɡ	‘to carry on shoulder’
sáád	‘brooms’	sààd	‘pricing’
yɛ́l	‘trouble’	yɛ̀l	‘say’

below gives an illustration of minimal pairs.

It is important to add that although three levels of tones are established, minimal triplets are comparatively rare, particularly with nominal lexical items. Consider the following examples from Abubakari (2022) in

Table 5. Minimal triplets in Kusaal.

H		M		L
sáám	‘father’	sāām	‘to mash’	sààm	‘guests’
kʊ́	‘kill’	kʊ̄	‘cut/break into pieces’	kʊ̀	‘fut.neg’

.

Most often, verbs have a mid-tone, especially in the irrealis, but this may differ in other environments, including contexts that express actual, existing, or past events. The tone of a verb may change to high or low based on the influence of other adjoining lexical items. Nouns, pronouns, and other items, however, are often high, mid, or combinations of several tones. This is shown in the examples in (25) below.

(25)	ō	dì-t
	3sg	eat.hab
	‘S/he eats’
	ò	dì-tnɛ́
	3sg	eat-prog
	‘S/he is eating’
	ò	dì-yá
	3sg	eat.pfv
	‘S/he has eaten’

The various types of tones can be combined in a single lexical item. The possible permutations result in the following nine sequences: HH, HL, HM, LL, LH, LM, MM, MH, and ML. As illustrated in

Table 6. Possible tone classes/patterns in Kusaal.

HH	HM	HL	MH	MM	ML
váʊ́ɡ ‘leaf’	kɔ́dīɡ ‘slaughter’	áɡɔ̀l ‘high’	ātá ‘three’	ūdīɡ ‘sweep away’	ɡbànā ‘books’
kʊ́ɡʊ́s ‘chairs’	tʊ́ʊ́lʊ̄ɡ ‘heat’	bámà ‘those’		tʊ̄bīs ‘spit’	ɡʊ̄līs ‘write’
ɡbíɡím ‘lion’	líkīn ‘dark’	ánsìb ‘uncle’
LH	ML	LL
kʊ̀láʊ́ɡ ‘stream’	lūōbùg ‘sling’	kʊ̀rʊ̀ɡ ‘old’
màbíl ‘stepmother’		nìnsààl ‘human’

below, the data presented in Abubakari (2022) show that these combinations represent the surface realizations of tonal patterns (or classes) in Kusaal.

It is important to note that rising (LH) and falling (HL) contours are attested in Kusaal ( Musah, 2017) but are strictly limited to bimoraic syllables, since two distinct tones cannot be realized on a single mora. These contours arise from the combination of unlike level tones on adjacent moras—for example, LH in dàá ‘market’ and sʊ̀ḿ ‘good’, and HL in sáà ‘rain’ and múì ‘rice’. In light of this distributional constraint, and the fact that such contours are not underlying, they are best analyzed as derived surface realizations rather than independent phonemic units. Consequently, Kusaal is more appropriately characterized as a register tone language.

In addition to the three level tones, downstep may also occur in specific contexts. For example, in typical transitive constructions, the tone of the object is often influenced by the tone of the preceding aspectual suffix. Object nominals (often with high tones) that follow a falling tone are frequently downstepped, resulting in a mid-tone realization ( Musah, 2018, p. 46).

(26)	tī	dʊ̄ɡ-ʊ̄d-nɛ̀	ꜜsūm-ā
	1pl	cook-hab-foc	groundnut-pl
	‘We are cooking groundnut’
(27)	tĩ	wá’-ád-nɛ̀	ꜜɡē’ùŋ
	1pl	dance-hab-foc	type of dance
	‘We are dancing ge’uŋ’

3.4.1. Functions of Tone in Kusaal

Just as is the case in several other Mabia languages, tone in Kusaal has both lexical and grammatical functions (Abubakari, 2018, 2022; Musah, 2018). Most of the examples cited so far are instances where tone differentiates the meanings of segments that are composed of the same phonemes. Below in (28) are a few examples of minimal pairs with meaning differences realized based on tone (Abubakari, 2022).

(28)	Lexical functions
	a.	bá	‘father’	bà	‘to fix into’
	b.	ká	‘complementizer/conjunction’	kà	‘focus particle’

The grammatical distinction in meaning of a simple declarative negative construction and its interrogative counterpart is influenced by tone. One way of forming interrogative constructions in Kusaal is by lengthening the final vowel of the terminal word (Abubakari, 2018). While the declarative form ends with a high tone, the interrogative form ends with a low tone. Thus, the final H in the interrogative clause is a lowered L, which also spreads to the lengthened vowel: -à in (29b) and -ì in (29d). The entire clause ends with a low boundary tone, which is not optional but structurally required. This follows similar observations in previous studies (Spratt & Spratt, 1968; England & Ladusaw, 1985; Musah, 2018). The examples below are illustrations.

(29)	Grammatical functions
	a.	ò	pù	mɔ́r	bíígá
		3sg	neg	have	child
		‘S/he has no child’
	b.	ò	pù	mɔ́r	bíígà-à?
		3sg	neg	have	child
		‘Doesn’t s/he have a child?’
	c.	ò	pù	dá	dííbí
		3sg	neg	buy	food
		‘S/he did not buy food’
	d.	ò	pù	dá’	dííbì-ì?
		3sg	neg	buy	food
		‘Didn’t s/he buy food?’

The tonal differences observed here are best described as grammatical tone rather than intonation. In these interrogative contexts, the clause final boundary tone of the lengthened vowel replaces the declarative high tone, which results in a systematic tone lowering of the final morpheme. The observed low tone is grammatically conditioned and occurs in interrogative contexts where the final vowel of the terminal word is lengthened.

A recent development that can be analyzed as an instance of grammatical tone is the use of high tone kʊ́ as the future negative particle and low tone kʊ̀ as the perfective negative. Originally, the negative morpheme pù ‘not’ has been used as a negative marker until the addition of kù ‘not’. This development, though gaining popularity fast and often heard among speakers in and around Ninkongu, is used for differentiating the grammatical feature of the past negative event and a future negative event.

(30)	a.	bííɡ	lá	kʊ̀	dìì
		child	def	fut.neg	eat
		‘The child will not eat’
	b.	bííɡ	lá	kʊ́	dīī
		child	def	pfv.neg	eat
		‘The child did not eat’

It is important to add that there are speakers who find this development unusual, but the language is evolving, and this is becoming a commonly accepted grammatical feature.

Grammatical functions of tone can also be realized on the different forms of verbs in the language. Although verbs generally have mid tones, the realis and the irrealis forms often attract different tonal patterns. This is expressed in the following examples.

(31)	a.	Aduk	sá	dì	múì	lá	wúsá
		Aduk	pst	eat	rice	def	all
		‘Aduk will eat all the rice’
	b.	Aduk	ná	dī	múì	lá	wúsá
		Aduk	fut	eat	rice	def	all
(32)	a.	dáʊ́	lá	dà	búúg	lá
		man	def	buy	goat	def
		‘The man bought the goat’
	b.	dáʊ́	lá	ná	da¯	búúg	lá
		man	def	fut	buy	goat	def
		‘The man will buy the goat’

3.4.2. Tone-Bearing Units

The tone-bearing unit in Kusaal has remained one area of argumentation, as two opinions are formed by different researchers. While some studies (e.g., Musah, 2010, 2018; England & Ladusaw, 1985) argue for the mora as the tone-bearing unit, which serves as the basic phonological structure, others (e.g., Niggli, 2014, pp. 79–80) argue that the syllable is rather the tone-bearing unit. In support of their claim, England and Ladusaw (1985) explain that while a short vowel may be associated with only one, long vowels may bear two tones. They further add that monosyllabic lexical items support this assumption because there is no stem with tone shift on a short vowel, but there is a stem with tone shift on the long vowel.

Niggli (2014, p. 79), on the other hand, explains that Kusaal allows a maximum of two pitch levels per syllable: rising and falling. Phonologically, this corresponds to two tones associated with a syllable: HL or LH. If the mora were the TBU, then a vowel or a syllable-final nasal would each contribute to the mora count. This view would predict that the last syllable pattern, CVVN, would have three moras.

A final short or long vowel bears a single tone, with the latter having the possibility of bearing two tones.

(33)

The only attested syllabic nasal is the first-person possessive pronoun m, which also serves as a TBU with a low tone. By contrast, coda nasals are neither syllabic nor TBUs, as illustrated in the examples below.

(34)

sáán ‘stranger’

kèm ‘go!’

yùùm

‘year’

In summary, Kusaal, like other Mabia languages, exhibits a complex tonal system with three distinct levels: high, mid, and low. While minimal pairs and triplets demonstrate the role of tone in distinguishing lexical meaning, verbs generally default to a mid tone when produced in isolation. The tonal structure extends beyond lexical differentiation to grammatical functions, as seen in declarative and interrogative constructions, where tone influences sentence types and their meanings. These findings show the significance of tone as a fundamental feature of the linguistic structure of Kusaal. Our next focus is a Mabia language which is said to be understudied. We present Likpakpaln in the following section.

3.5. Tone in Likpakpaln

Research on the tonal system of Likpakpaln is far from advanced. Beyond a very preliminary work by Steele and Weed (1966), there is not much more accounting for the tone system of the language. As a result, improving upon our knowledge of the Likpakpaln tone structure is still very crucial at this moment. This section considers the salient tonal features of Likpakpaln. It focuses on the tone inventory, providing some minimal pairs/triplets confirming the existence of the identified tonemes. Aspects relating to tone distribution and functions are also considered. Where possible, Praat analyses are provided in support of the observed patterns. Parts of the analysis in this section are adapted from Bisilki (2022). A convention we adopt for tonal notations is to place a dot between tonemes that fall within separate syllables (e.g., H.L), whereas tonemes on adjacent vowels within the same syllable are left without any such marking between them (e.g., HL).

3.5.1. Tone Inventories

Three level tones (high, mid, and low) are identified in the tonal system of Likpakpaln. The syllable is the tone-bearing unit in Likpakpaln and can be a vowel or a syllabic nasal. Each member of a vowel sequence (VV) in a syllable receives a separate tone marking as notational convention. The forms in (35) provide an example. The minimal pairs/triplets in

Table 7. Minimal pairs/triplets based on the Likpakpaln three level tonemes. ¹

H	M	L
ńtúún ʿgourdʾ	ńtūūn ʿmessageʾ	ńtùùn ‘beans’
wáár ‘to cut food’	wāār ‘to be mad’	wààr ‘to open, to expand’
līwól ‘crowd’	līwōl ‘flute’	-
	gbʊ̄g ‘to bark’	gbʊ̀g ‘to shake’
táb ‘to be pasted to/to be affixed to’	-	tàb ‘to press’
lībííl ‘breast’	-	lībììl ‘cover/lid’
lībuul ‘pot’	-	lībùùl ‘pit’
bál ‘to reduce (price)’	-	bàl ‘to sling over’
líí ‘to allow/to drop’	-	lìì ‘to be in the lead’
káá ‘to not be afraid’		kàà ‘to sit, habitual’

¹ Note that in the minimal pairs/triplets in Table 7, the tonal differences establishing the pairs/triplets in the disyllabic words are the tones on the second syllables, since the tones of the first syllables are the same.

further exemplify the occurrence of the three level tones in Likpakpaln words.

(35)	ḿ.máál	‘birth’
	kàl	‘to sit’

As seen in Table 7, the data for this study as well as the general observations of one of the authors who is a native speaker linguist of Likpakpaln reveal that the involvement of the M tone phenomenon in lexical distinctions in Likpakpaln is far less elaborate compared to the role of the high and low tones. Note also that for the minimal pairs/triplets in Table 7, the lexical differences are based on the tones on the second syllables in the case of the disyllabic forms, since the tones on the first syllables of those words are the same.

Beyond the three tone levels, downstep has also been reported in Likpakpaln (Schwarz, 2007). This phenomenon is particularly evident in focus contexts, especially in responses to questions. In such cases, the high tone focus particle lá undergoes downstep when it is preceded by a focused object bearing a high tone. For example, in response to the question ‘Do you want his or my car?’, the answer may be ḿ bá w-àà-lɔ́ɔ́ ꜜlá ‘I want HIS car’. Furthermore, downstep also occurs when a high tone verb in focus precedes a noun class marker, as illustrated in ù ŋmán ꜜŋí-tùùm ‘She ATE beans’.

Unlike in the previous attempt to analyze Likpakpaln tone (Steele & Weed, 1966), we present pitch value estimates and sample pitch traces to give a more empirical perspective to the existence of three level tones in the language.

Table 8. Pitch values for the M tone and H tone ¹.

M Tone				H Tone
Hertz	lībɔ́l	līkál	līmúl	Hertz	lībɔ́l	līkál	līmʊ́l
Mean Hz of M tone	134.67	135.33	131.97	Mean Hz of H tone	154.26	158.75	141.48
Highest Hz of M tone	136.65	137.30	135.48	Highest Hz of H tone	156.25	161.85	144.12
Lowest Hz of M tone	132.69	133.36	128.47	Lowest Hertz of H tone	152.27	155.66	138.85

¹ The mean pitch values in Table 1 and Table 2 were calculated based on the highest and lowest Hz of the tonemes, as indicated in the respective tables.

shows pitch value estimates for the mid and high tones in Likpakpaln.

Table 9. Pitch values for the L tone.

Hertz	tàkɔ̀	bànààn	dàfìl
Mean Hz of L tone	109.07	72.94	96.37
Highest Hz of L tone	115.23	96.02	101.34
Lowest Hz of L tone	102.91	49.86	91.41

is also a representation of the low tone value estimates. Figure 2 is a pitch trace for two of the syllable samples from which the F0 (Hz) values for the mid and high tones in Table 7 were extracted. Figure 3, on the other hand, is a pitch trace of one of the syllable samples from which the low tone F0 (Hz) values shown in Table 8 were measured. The mid and high tone value estimates in Table 8 are based on the first and second syllables, respectively, of the words lībɔ́l /M.H/ ‘secret’, līkál /M.H/ ‘elephant grass’, and līmúl /M.H/ ‘vagina’. Similarly, the low tone value approximations in Table 9 are computed based on the L tones in only the onset syllables of tàkɔ̀ /L.L/ ‘wolf’, bànààn /L.LL ‘praise name’, and dàfíl /L.H/ ‘cheap’. All the words used in generating all the tonal pitch computations were forms uttered in isolation (i.e., in their citation forms). In this regard, a common set of three words were recorded from each of three speakers and the measurement of the pitches in each speaker’s set of words revealed the pitch variations, attesting to the three level tones. Nonetheless, for reasons of space, the sample incorporated in this section is drawn from one of the three speakers.

Thus, the F0 values in Table 8 and Table 9 buttress the existence of a three-way tonal contrast in Likpakpaln.

The question of whether or not Likpakpaln has a contour tone proper remains to be fully answered. However, we observe that non-identical adjacent vowel sequences consistently follow a HL order. This pattern is revealed in the examples involving the sequences /ui/ in ísúì /H.H.L/ ‘trees’ and /ua/ in līpúàl /M.H.L/ ‘liver’. This tonal contouring, at present, cannot also be attributed to any identifiable tonal process such as floating tone or tone deletion. We propose that this is a phonetic conditioning rather than a phonological occurrence as would have been the case of string-adjacent polar tones proper. If this situation were a typical string-adjacent polar tone pattern, there would be no basis to restrict the first of a non-identical vowel sequence to an H tone and the second to an L tone. At best, the polarization of tonal pitch on adjacent vowels would have been left open ended to equally recruit either a H tone or an L tone on either segment of a vowel sequence. The non-systematic and indeterminate status of contour tones in Likpakpaln thus supports its classification as a register tone language.

3.5.2. Functions of Tone in Likpakpaln

Tone performs both lexical and grammatical functions in Likpakpaln. There are common instances where words with identical segments are differentiated in meaning based (solely) on tonal pitch variation, as we see in Table 9 above. Grammatical tone often manifests in the Likpakpaln pronoun system (36)–(37) and then also in verbal aspect marking (38)–(39).

(36)	kɪ̄nà	lē	tɪ̄	bán
	this	foc	1pl	want
	‘This is what we want’
(37)	kɪ̄ná	lē	tɪ̄	bã́
	that	foc	1pl	want
	‘That is what we want’
(38)	bī-píí-b	chùùn	cháá	kī-nyá-ŋ
	cl.pl-woman-cl.pl	walk.hab	go	cl.sg-market-cl.sg
	‘Women walk to the market’
(39)	bī-píí-b	chúún	búèn	kī-nyá-ŋ
	cl.pl-woman-cl.pl	walk.pfv	go.pfv	cl.sg-market-cl.sg
	‘Women have walked to the market’

As can be seen from the above examples, the difference between the proximal and the distal forms in (36) and (37), respectively, is solely tone conditioned as the pronouns remain segmentally invariant. Both forms are M-toned on the first syllable. But whereas the proximal form is L-toned on the second and final syllable, the distal counterpart is H-toned on the corresponding second syllable. Similarly, the verb forms in (38) and (39) have different aspectual readings because of their different tonal inflections. Regarding the expression of aspect, note that individual Likpakpaln verbs select different means, such as tone, affixation, and suppletion, among others (see Bisilki (2022) for more on this subject).

As part of the grammatical functions of tone in Likpakpaln, it is used to distinguish àà, an imperfective cum adverbial particle from the áá negative marker. The àà imperfective cum adverbial particle is L-toned, whereas the áá negative marker is H-toned. The difference between (40) and (41) as positive and negative sentences, respectively, is a result of the tonal variation on the aa morpheme:

(40)	sī	àà	nyán	ḿ-máál	mún	pōnnɪ̄	ná
	2sg.sbj	ipfv	come.from	cl.sg-family-cl.sg	rel	inside	def
	‘The family that you come from’
(41)	sī	áá	nyán	ḿ-máá-l	múè	pōnnɪ̄
	2sg.sbj	neg	come.from	cl.sg-family-cl.sg	this	inside
	‘You’re not from this family’

Nonetheless, it must also be noted that the dynamic operation of tones in lexical and grammatical functions does not block homonymy in the semantic system of Likpakpaln. There are many instances where words are both segmentally and tonally the same yet have different denotations or unrelated meanings. Such forms are disambiguated based on context. The forms in (42) give some examples of homonymy in Likpakpaln:

(42)	Homonym	Unrelated meanings
	sū	‘to have sexual intercourse’ or ‘to steal’
	béé	‘these = prox’ or ‘to know’
	kíí	‘to respond, to accept’ or ‘to snap (of a rope)’
	kpáá	‘to nail’ or ‘to be deaf’

In conclusion, this section has provided brief highlights of the Likpakpaln tone system. It is established that a three-way level tone system operates in the language, with the tonemes being high, mid, and low. It is also shown that tone has lexical and a range of other grammatical functions in Likpakpaln, such as aspect configuration and pronoun distinctions.

3.6. Tone in Buli

Like Likpakpaln and Kusaal, Buli has a three-level tone system, comprising high, mid, and low tones. In nouns, all three tones show contrast, but this is not the case in verbs. Verbs bear mid tones when cited in isolation and exhibit no lexical tone contrast. The minimal pairs and minimal triplets in (43) and (44) demonstrate the lexical function of pitch distinction in nouns.

(43)	Minimal Pairs
	Low vs. High Tone
	bììk	‘language’	nààb	‘chief’
	bíík	‘child	nááb	‘cow’
	Mid vs. Low Tone		High-Low vs. Low-Low
	bāŋ	‘bangle’	búlùk	‘armpit’
	bàŋ	‘lizard’	bùlùk	‘hometown of Bulisa’
	Mid-Mid vs. High-High		Mid-Mid vs. Low-Low
	vīɔ̄k	‘leaf’	lālīk	‘bull’,
	vɪ́ɔ́k	‘wind’	làlɪ̀k	‘clay’
(44)	Minimal Triplets
	kɔ̀k	‘ghost’	síúk	‘path’
	kɔ̄k	‘feather’	sìùk	‘fish’ (special type)
	kɔ́k	‘mahogany’	sīūk	‘navel’

These minimal pairs and minimal sets demonstrate that tonal differences result in meaning differences in words. As seen in previous sections, this lexical function of pitch distinction is also present in other Mabia languages. However, pitch contrast is additionally used to indicate grammatical categories such as tense, aspect, and mood, as well as information structural functions. Grammatical functions are marked with both free morphemes and tone. For example, the morpheme à is employed to mark the imperfective aspect, and the verb in this context bears a default mid tone, as illustrated in (45). A similar strategy is used for marking the future tense, where the morpheme lì is used with a verb that also bears a mid tone, as shown in (46).

(45)	wà	à	dɛ̄	sāāb3
	cl.1	pfv	eat	saab
(46)	wà	lì	dɛ̄	sāāb
	cl.1	fut	eat	saab
	‘He will eat saab’

Tone also performs crucial grammatical functions (Akanlig-Pare & Kenstowicz, 2002; Schwarz, 2005; Kenstowicz, 2006; Sulemana, 2021; Adjong, 2025). For instance, as illustrated in (47), the verb is realized with a low tone to mark the perfective aspect.

(47)	bà	sùùrì	ní-sá
	cl.2	wash.pfv	hand-cl.6
	‘They washed (their) hands’
(48)	bà	sūūrī	ní-sá
	cl.2	wash	hand-cl.6
	‘They should wash (their) hands’

Grammatical tone function is also evident in both the imperative and subjunctive moods, which share identical tonal coding strategies (Adjong, 2025; Schwarz, 2005). Both mood categories are realized with a mid tone on verbs, as shown in (48). Imperative constructions are typically used as direct commands, where the addressee is usually not explicitly mentioned. As a result, these constructions can generally be identified and distinguished from subjunctive constructions by their lack of overt (usually second-person singular) subjects. In contrast, subjunctive constructions, which express desires, doubts, wishes, recommendations, suggestions, or hypothetical situations, usually include overt subjects.

Since imperative constructions are used as direct commands, their subjects (second-person singular pronouns) are typically not expressed overtly, but a second-person plural subject is always explicitly stated. Example (49) illustrates an imperative construction, whereas (50) demonstrates a subjunctive construction. However, as shown in (51), this distinction collapses in contexts where the subjects are explicitly present.

(49)	pɛ̄	wén
	swear	god
	‘Swear unto god’
(50)	bà	pɛ̄	wén
	cl.2	swear	god
	‘They should swear (unto) god’
(51)	nì	pɛ̄	wén
	2pl	swear	god
	‘Swear unto god’/‘You should swear (unto) god’

Tone Change

The pitch level of successive high (H) tones may decline gradually across an utterance but not according to any specific rules. This downdrift phenomenon differs from the downstep reported for Buli’s neighbors, such as Dagaare and Dagbani, in that the downstep is automatically triggered by intervening low tones, which causes the pitch levels of the high tones to decline systematically across the utterance. Example (52) demonstrates downdrift in Buli. Observe the gradual pitch decline of the high tones indicated by superscripts […³, ², ¹]. This results in each high tone being pronounced lower than the previous one. However, the occurrence of this downdrift phenomenon is unpredictable.

(52)	/…àtɪ̀ mí ɲà bííká tà mí-jéŋ/[…àtɪ̀ !mí ɲà !bííká tà !mí-jéŋ]4
		‘…and I saw that the child had a single rope’

Pitch variation creates contour tones in certain contexts, particularly in a low–high (LH) sequence where the syllable unit bearing the high tone is bimoraic. In such contexts, the pitch of the high tone initially declines but is suspended before the full realization of the bimoraic syllable. In Buli linguistics, these contour tones are surface-level outcomes of low tone spread affecting high tones (Adjong, 2025; Akanlig-Pare & Kenstowicz, 2002; Schwarz, 2003, 2009), indicating that they do not constitute part of the underlying tonal inventory, as is also the case in the other five languages discussed above. This phenomenon is most conspicuous in possessive constructions involving low tone pronouns and high tone nouns. The process, illustrated in the simplified autosegmental model in (53) and (54), is not cyclical; hence, as (54) shows, a high tone can cause a rising tone to be simplified to a low tone.

(53)
a.	b.
(54)
a.	b.

Low tone spread can also optionally occur at the syntactic level, where, as exemplified in (55) and (56), the low tone of a verb spreads to a nominal object.

(55)	/bà fòrì míí.ká/	[bà	fòrì	mìì-ká]
		cl.2	untie.pfv	rope-def.5
		‘They (people) untied the rope’
(56)	/sì nɛ̀ zúé/	[sì	nɛ̀	zù-é]
		cl.6	swallow.pfv	worm-cl.4
		‘They (chicken) swallowed worms’

The tonal processes described above are regular and systematic in normal speech, where no element of the sentence is focalized and the speaker’s emotions do not warrant sporadic variations in pitch. Focal and emotional expressions operate at special levels that are not always susceptible to tonal changes. Notably, in cases of emphasis, the high tone of a focal object is rarely affected by low tone spread. The constructions in (55) and (56) differ from those in (57) and (58), respectively, because the objects are given salience or prominence.

(57)	/bà fòrì míí.ká/	[bà	fòrì	míí-ká]
		cl.2	untie.pfv	rope-def.5
		They (people) untied THE ROPE (not the scarf)
(58)	/sì nɛ̀ zúé/	[sì	nɛ̀	zú-é]
		cl.6	swallow.pfv	worm-cl.4
		They (chicken) swallowed WORMS (not houseflies)

The first- and second-person plural emphatic forms, which make clusivity distinctions, differ segmentally and tonally from their unemphatic counterparts. Other pronouns differ only tonally. The high tone of an emphatic subject pronoun can cause the perfective low tone of a verb to be replaced with a high tone. Compare the tones of each pair of sentences in (59)–(61).

		Emphatic				Non-Emphatic
(59)	/bà fòrì míí.ká/	a.	[bá	fórí	míí-ká]	b.	[bà	fòrì	mìì-ká]
			cl.2	untie.pfv	rope-def.5		cl.2	untie.pfv	rope-def.5
			‘They untied the rope’				‘They untied the rope’
(60)	/nàmá fòrì míí.ká/	a	[nàmá	fórí	míí-ká]	b.	[nàmá	fòrì	mìì-ká]
			2pl.excl	untie	rope-def.5		2pl.excl	untie	rope-def.5
			‘You untied the rope’				‘You untied the rope’
(61)	/tàmá fòrì míí.ká/	a.	[tàmá	fórí	míí-ká]	b.	[tàmá	fòrì	mìì-ká]
			1pl.incl	untie	rope-def.5		1pl.incl	untie	rope-def.5
			‘We untied the rope’				‘We untied the rope’

While tone can indicate focus in a complex manner, the focus marker ká provides additional insights into this complexity. When ká introduces the salient element of a sentence, the optionality of low tone spread causes it to be realized with either a low tone (62) or a high tone (63).

(62)	wà	ɲù	kà	dáám
	cl.1	drink.pfv	foc	alcohol
	‘S/he drank alcohol’
(63)	wà	ɲù	ká	dáám
	cl.1	drink.pfv	foc	alcohol
	‘S/he drank alcohol’

Pitch variation at the syntactic level is undoubtedly versatile and complex, as regular tone spread affecting lexical and grammatical tones can be suspended or blocked due to focus-marking elements. Schwarz (2009) recognizes this and highlights the need to invoke postlexical phonology to explain the frequent irregularities between underlying tones and surface tones. Research on intonation remains a desideratum in Mabia linguistics, though it falls outside the scope of this article. The few studies that have addressed intonation (e.g., Cahill, 2016; Issah, 2015) demonstrate its crucial role in understanding the unique prosodic characteristics of Mabia languages.

3.7. Comparative Summary of the Six Tonal Systems

In summary, the six languages—Dagaare, Dagbani, Gurenɛ, Kusaal, Likpakpaln, and Buli—synthesized in the foregoing display both similarities and notable differences in their tone systems. The similarities largely reflect their common genetic origin within the Mabia Central and Mabia East subgroup, whereas the differences illustrate the extent to which tonal systems can diverge even among closely related languages.

With regard to shared features, a fundamental similarity across all six languages is the central role of tone in both their phonological and grammatical systems. In each language, tone functions lexically to distinguish word meanings, as evidenced by the presence of minimal pairs or triplets. With the exception of Dagbani—where the grammatical role of tone remains insufficiently documented—tone serves important grammatical functions in the languages, marking distinctions such as tense, aspect, mood, pronominal reference, and focus. Another salient shared characteristic is the presence of common tonal processes. Across the languages, phenomena such as tone spreading, downstep or downdrift, and constraints on the occurrence of contour tones are consistently attested.

The languages also exhibit typological tendencies characteristic of many West African tone systems. In particular, they employ register tone systems with level tones rather than complex contour systems. Although contour-like tones occur in some languages (e.g., Dagaare), they are typically restricted in distribution or arise as derived surface phenomena rather than forming part of the core tonal inventory. Across all six languages, contour tones are not phonemically contrastive; rather, they are generally derived through phonological processes and do not constitute underlying tonal categories.

Despite these shared characteristics, the languages differ significantly in their tonal inventories. Dagaare and Dagbani are generally described as two-tone systems consisting of high and low tones. In contrast, Kusaal, Likpakpaln, and Buli display three-tone systems involving high, mid, and low tones. The presence of a mid tone in these languages introduces an additional level of tonal contrast that expands the tonal space available for both lexical and grammatical distinctions. The tonal system of Gurenɛ occupies an intermediate position within this typology: its tonal inventory remains a matter of scholarly debate, with some analyses proposing a two-tone system (e.g., Atintono, 2004) and others suggesting a three-tone system (e.g., Adongo, 2018).

The languages also differ in the distribution and productivity of specific tonal processes. For example, tonal polarity—where the tone of a morpheme changes in response to its tonal environment—has been clearly documented in Dagaare and has been observed in Gurenɛ, but is less widely reported in the other languages considered here. Similarly, while downstep and downdrift are attested in Dagaare, Dagbani, Gurenɛ, Kusaal, and Likpakpaln, their presence in Buli is less clearly established. Tone spreading is productive in Dagaare, Dagbani, Gurenɛ, Kusaal, and Buli but is less evident in Likpakpaln. These differences suggest that although tone spreading is typologically common within the family, the mechanisms and domains of its application vary from language to language.

A further point of divergence concerns the tone-bearing unit (TBU). In Dagaare, Dagbani, and Likpakpaln, the syllable generally functions as the primary TBU. In contrast, Gurenɛ and Kusaal have been reported to allow the mora to function as the tone-bearing unit, while the status of the TBU in Buli remains unclear. Such structural differences influence how tonal processes—such as contour formation and tone spreading—operate within each language.

Table 10. Summary of the distribution of major tone phenomena across the six languages.

	Daagare	Dagbani	Gurenε	Kusaal	Likpakpaln	Buli
Number of contrastive tones	2	2	2/3	3	3	3
Tonal minimal pairs/triplets	√	√	√	√	√	√
Contour-like tones	√	√	√	√	(√)	(√)
Polarity/dissimilation	√	-	√	-	-	-
Tone-bearing unit	syllable	syllable	mora	mora	syllable	-
Downstep/downdrift	√	√	√	√	√	(√)
Tone spreading	√	√	√	√	-	√
Grammatical function of tone	√	-	√	√	√	√

below summarizes the distribution of the major tonal phenomena across the six languages, highlighting both areas of convergence and points of divergence.

With respect to grammatical tone,

Table 11. Summary of the grammatical function of tone across the six languages.

Language	Grammatical Functions
Dagaare	Grammatical distinction future negation past irrealis negation
Dagbani	Unclear
Gurenɛ	Grammatical distinction singular negative imperative plural negative imperative
Kusaal	Grammatical distinction negative declarative negative interrogative
Likpakpaln	(i) Grammatical distinction proximal pronoun distal pronoun (ii) Grammatical distinction present habitual aspect perfective aspect
Buli	Grammatical distinction perfective aspect non-progressive/non-perfect (deontic modality)

provides a comparative summary of the various grammatical features and functions expressed by tone in each language.

Overall, the similarities and differences identified in this study illustrate the dynamic nature of tonal systems within the Mabia language family. While the languages share a common typological foundation—including the central role of tone, similar tonal processes, and comparable grammatical functions—they also exhibit considerable structural variation in their tonal inventories, tonal processes, and phonological organization.

This combination of shared features and divergent developments provides valuable insights into both the historical relationships among Mabia languages and the broader typology of tone systems in West Africa. It highlights how closely related languages may retain core typological characteristics while simultaneously developing distinct tonal structures.

4. Conclusions

This study provides a comparative synthesis of the tonal systems of six Mabia languages—Dagaare, Dagbani, Gurenɛ, Kusaal, Likpakpaln, and Buli—focusing on their tonal inventories, processes, and functions. The findings demonstrate that although the tonal systems of these closely related languages diverge structurally, they nevertheless share several important features and processes. In particular, this paper shows that tonal assignment frequently arises from the interaction between morphological categories, such as tense and aspect, and phonological processes, including tone spreading and polarity effects.

Typologically, this paper adds to the comparative research on tone, especially in West African languages. Cross-linguistically, the various patterns (e.g., contour tones, downstep, tone spreading, etc.) discussed in the six languages confirm broader typological patterns across languages, while the discovery or emergence of mid tone in some Mabia languages such as Buli (Akanlig-Pare, 2005; Schwarz, 2005) highlights the evolving nature of tonal systems. Importantly, the presence of this mid tone raises the question of whether it should be treated as a phonologically primitive category or as a derivative outcome of high–low interactions.

One of the primary contributions of this paper is the comparative perspective it provides of tone systems across six Mabia languages. By examining their tonal inventories, processes, and grammatical functions, this study brings together insights from previously separate descriptions and identifies patterns of convergence and divergence within a single language family. This is particularly significant because many Mabia languages remain under-described or unevenly documented, and synthesizing the available analyses helps to establish a clearer comparative perspective.

More broadly, this study contributes to discussions of tone typology and tone universals. By examining the distribution of two- and three-tone systems, the productivity of tone spreading, and the restrictions on contour tones, this paper provides empirical evidence relevant to cross-linguistic generalizations about tonal structure. In this respect, this study offers an important comparative baseline for future research on Mabia tonal systems.

This paper also reinforces the typological observation that many African languages are register tone systems, with contour tones arising as derived surface phenomena rather than underlying units. This therefore has implications for the representation of tone in phonological theory and the distinction between underlying and surface tonal structures.

Despite these contributions, this study is not without limitations. The six languages examined are at different stages of documentation and scholarly investigation, and consequently the availability and depth of data are not uniform across them. Some languages have been more extensively studied than others, which may make certain tonal patterns or processes—such as tone polarity or downstep—appear more productive or better attested in some languages than in others. We acknowledge this imbalance as a methodological limitation, although the present study still offers a useful comparative synthesis that can guide future research.

A further limitation concerns the scope of the comparative sample. The analysis focuses on five languages within Central Mabia and one language within East Mabia subgroups, even though the broader Mabia family contains about eighty languages (Bodomo, 2020). The patterns identified here therefore cannot yet be assumed to represent the full range of tonal variation within the family.

Future research should extend the investigation to additional Mabia varieties and explore diachronic developments in tonal systems, which would help refine theoretical models and contribute to broader typological generalizations. In addition, further studies based on new field data could help verify earlier claims and address unresolved issues in existing descriptions. Ultimately, expanding and deepening the analysis of Mabia tonal systems will contribute to a more comprehensive understanding of the role of tone in African linguistic structures and in tonal typology more broadly.