Aspects of Ethiopian Komo (morpho)phonology

Manuel Alejandro Otero

University of Oregon

 

This paper presents the first comprehensive analysis of the phonological system of Komo (Koman of Western Ethiopia), including the tone system and morphophonological processes in the verb involving deictic directional morphology. The consonant inventory contains plain, ejective, implosive stops and an ejective affricate in three places of articulation. Komo has a seven-vowel inventory with Advanced Tongue Root (ATR) contrast in the high vowels and a typologically rare bi-directional ATR harmony system. Komo displays anticipatory [+ATR] harmony as well as progressive [–ATR] harmony, which, when taken together, call into question the notion of a single “dominant” ATR feature value in an ATR harmony system.

 

1      Introduction

 

Komo (also known as Koma, ISO 693-3 code xom) is a little-described endangered language spoken along the Ethiopia, Sudan and South Sudan border.[1] Determining the population of Komo speakers at present has been challenging for several reasons. Lewis et al.’s (2015) estimates (10,000 in Sudan and 5,000 in Ethiopia) date back to 1976 and the most recent Ethiopian census data from 2007, citing 15,000 ethnic ‘Komo’ in Ethiopia, groups together speakers of Komo and the related Koman language Gwama (FDCREPC 2008). In Ethiopia, speakers of Komo and Gwama identify ethnically as “Komo” yet recognize they speak two mutually unintelligible languages (Yehualashet 2008, Küspert 2015). Henceforth, “Komo”, will refer specifically to speakers of the Komo language, which can be found in two main areas in Ethiopia, the Mao-Komo special woreda (district) of the Benishangul-Gumuz Regional State and in Gambella, the capital of the Gambella Regional State and in a nearby village.[2]

 

In the Mao-Komo special woreda, which extends westward to the border of South Sudan, the majority of Komo speakers live in remote enclaves scattered along the border, communicable only by footpaths. Historically, communities residing in these borderlands, including the Komo, have lived in a continuum of subjection and migration (Meckelburg 2015:23), the result of being traded as slaves, a practice which dates back to the 19th century in this region (Corfield 1938:129). This was also witnessed in the late 1800’s by the Dutch traveller Juan María Schuver, who visited Komo villages in the Sudan (James et al. 1996). The slave trade led to cultural and social subjugation, which resulted in forced migration and fragmentation of the people groups of the Ethio-Sudan borderlands for decades. Slave trading of the Komo and other minority groups of this area continued well into the 1930s (Meckelburg 2015). More recently, due to the current federal policy towards ethnic diversity in Ethiopia, the Komo people, along with other minority populations inhabiting remote areas of western Ethiopia (e.g. Gwama and Opo), have undergone resettlement to villages nearer to larger towns. This process is ongoing and is increasing language shift in these communities to more dominant languages.

 

In Ethiopian villages, the Komo typically live among other ethnicities and intermarriage with these groups is common. In the Benishangul-Gumuz Regional State, Komo often intermarry with Gwama, Oromo and to a lesser extent, speakers of Dana.[3] In the Gambella State, Komo mostly live among the Anywa, Oromo, Nuer and Opo. Komo native speakers are by and large multilingual in Ethiopia, as they often learn the languages of the larger people groups they are in contact with. The contrary is almost never found. The current panorama in Ethiopia leads the present author to suggest that Ethiopian Komo is a highly endangered language, as many of its speakers are assimilating to neighboring dominant cultures and languages.

 

There has been ongoing controversy over the internal structure of the Koman family as well as its affiliation to the larger Nilo-Saharan phylum. This is evidenced as far back as Corfield (1938) and Tucker and Bryan (1966), who struggled with auto-determined and external nomenclature for the Koman groups they encountered, as well as the significant dialectal variation among clans speaking Koman languages.[4] Externally, Koman languages have traditionally been classified within the Nilo-Saharan phylum (Greenberg 1963). Bender (1991, 1994) places Koman coordinate with Gumuz, Kadu, and the East Sudanic branch within what he calls the ‘core’ group within Nilo-Saharan. By contrast, Ehret (2001) reconstructs Koman as one of two primary splits from Proto Nilo-Saharan.[5] Dimmendaal (2011), the most conservative, maintains Koman as an independent family until more extensive grammatical information becomes available. This paper does not directly address the external classification issue but the phonological analysis in the paper is an important step for any subsequent studies.

 

The Koman languages as a whole were largely undescribed until recently. The only comprehensive work on Komo is a short grammatical sketch by a missionary in the Sudan (Burns 1947). Burns’ pioneering work, both in terms of phonetics/phonology and morphosyntax, laid the groundwork for Bender (1983, 1994), who examined Komo within a Koman perspective employing prior data and his own field notes. Subsequently Kutsch-Lojenga & Otero (2011) provided a brief refinement of Komo phonology. While Negash (2015) provides a cursory phonological analysis, the work differs significantly from the current analysis, primarily in the vowels and tone. The major aims of this paper are to clarify analytical anomalies and provide a comprehensive overview of the major (morpho)phonological processes. The data for this paper come from roughly twenty months of fieldwork spread over various trips beginning in 2011 and continuing to the present. The current database contains over 3500 roots, 35 analyzed texts and countless verb paradigms.

 

This paper is organized as follows: the contrastive consonant inventory is discussed in §2. Contrastive vowels and ATR harmony are treated in §3. Syllable structure and ambiguous sequences are examined in §4 and contrastive tone melodies in monomorphemic roots and in polymorphemic inflected verbs are discussed in §5. Lastly, morphophonological processes in the verb are discussed in §6.

 

2      Consonants

 

The Komo contrastive consonant inventory contains 23 phonemic consonants (Table 1).[6] There is a three-way contrast for plosives in the bilabial, alveolar and velar places of articulation and a three-way contrast in manner of articulation, which includes plain, ejective and implosive stops. This inventory is largely consistent with previous analyses of Komo, with the exception of the velar nasal and the glottal stop, which are of marginal status.[7]

 

 

 

2.1  Marginally phonemic consonants: the velar nasal [ŋ] and the glottal stop [ʔ]

 

I consider the glottal stop /ʔ/ and the velar nasal /ŋ/ marginally phonemic consonants in the inventory due to their very limited distribution. Both the velar nasal and the glottal stop can occur phonetically in predictable environments: the glottal stop can occur before a word-initial vowel presumably to satisfy syllable onset requirements and the velar nasal can surface as a homorganic nasal before a velar stop. But there are instances where both the glottal stop and the velar nasal are phonemically contrastive segments discussed below. Burns (1947) only mentions the velar nasal as an allophone of the alveolar nasal assimilating to a following velar consonant.

 

The velar nasal occurs in two positions, word-internally and word-finally, where it may or may not alternate with the alveolar nasal (Table 2).[8] Word internally, the velar nasal occurs in two distinct environments: intervocalically, where it is marginally contrastive as very few near-minimal pairs exist; or as a homorganic nasal assimilating in place to a following velar stop across morpheme boundaries. In word-final position, the velar nasal [ŋ] can alternate with the alveolar nasal [n] in some lexemes but not all, an alternation that varies across different speakers.[9] It is difficult to know for certain whether or not the word-final velar nasal is the result of a prior homorganic [NC_velar] sequence from which the velar stop was lost over time. In total, there are only about twenty instances of an unconditioned velar nasal in my data.

 

 

Although also of marginal status, the glottal stop appears to have a larger presence in the phonological system than the velar nasal. Before discussing the contrastive instances of the glottal stop, the predictable phonetic realization of glottal stop is explained.

 

First, there is phonetic glottal stop epenthesis before word-initial vowels. This epenthetic glottal stop is non-contrastive and has a gradient phonetic realization across speakers (Table 3). Vowel initial words are not altogether common in Komo and one possible explanation of epenthesis of the glottal stop is to satisfy certain syllable onset requirements. Thus far there do not appear to be any co-occurrence restrictions regarding word-initial glottal stop epenthesis and vowel or tone.

 

 

There is a handful of monomorphemic verb roots that display what appears to be a contrastive word-final glottal stop. These CVʔ verb roots behave like CVC roots when they occur with inflectional morphology. To illustrate, near minimal pairs inflected with the deictic directional suffix /-í/ dd∅ are seen in Table 4.[10] In this paradigm, verb roots ending in a consonant take the dd∅ suffix /-í/ realized as [i] before bound pronominal suffixes that only consist of a consonant (e.g. 3sg.m /-r/). By contrast, roots that end in a vowel take the dd∅ suffix realized as zero.[11] In a near minimal pair, the verb /yàʔ/ ‘vomit’ takes the /-i/ form of the dd∅ suffix, whereas /yà/ ‘go.sg’ takes the zero form of the suffix.

 

 

In some roots, there is an alternation between the presence and absence of a glottal stop, as seen in the inflected verb /bɔʔ/ ‘hide’ (Table 4). Further, a very small portion of synchronically monomorphemic lexical roots exhibit a phonetic intervocalic glottal stop (Table 5). In natural speech, the intervocalic glottal stop can also be elided and is thus creating a lack of phonological contrast for the glottal stop in this position.

 

 

2.2  Aspiration and lenition: Plain stops and fricatives

 

As seen in Table 1, stops contrast for three points of articulation (bilabial, alveolar and velar), as well as for four laryngeal features (voiceless, voiced, ejective and implosive).[12] The allophonic realizations for these eleven plosives are discussed below.

First, the bilabial and alveolar implosives have a limited distribution, occurring only in word-initial position where they are realized as traditional implosives, with ingressive air following an oral release.[13] Second, plain voiceless stops are phonetically aspirated in word-initial and word-final position. Intervocalically, voiceless stops can be realized with aspiration or they can also lenite to voiceless fricatives. Third, voiced stops are realized with a delayed release before a pause. Aspiration is thus a key phonetic cue that distinguishes voicing in plain stops both word-initially and word-finally. Word-initial and word-final phonetic realizations in plain voiced and voiceless stops are seen in Table 6.

 

 

Intervocalically, it is common for the voiceless stops to lenite and weaken to fricatives (Table 7), creating non-contrastive free variation among the allophones of /p/ ([p, pʰ, ɸ]), /t/ ([t, tʰ]) and /k/ ([k, kʰ, x]). Plain voiced stops can also lenite intervocalically surfacing as the voiced fricatives and tap [β, ɾ, ɣ]. Intervocalic lenition of plain voiceless and voiced stops is non-categorical and gradient across speakers, more evident in casual or fast speech.

 

 

2.3  Lenition and neutralization: Implosives and ejectives

 

The implosives have a limited word-initial distribution and do not display any significant allophonic variation. As such, implosive consonants will not be discussed further in this section. The ejectives, on the other hand, are not restricted within the word and display both categorical and free phonetic variation. “Categorical variation” here refers to an ejective phoneme which has particular allophonic realizations in specific environments and within specific types of words. Free variation refers to allophones created via common phonetic processes, namely intervocalic lenition and word-final articulation without a glottal release.

 

Word-initially, the ejectives are always articulated with glottal release. Word-medially, the ejectives /pʼ, tʼ, kʼ, sʼ/ can either be released, or weaken to voiced allophones [b, d, g, z], respectively.[14] Word-finally, there is variation between released and unreleased ejective allophones. The behavior suggests an intermediate process of deglottalization in word-final position across place of articulation: the bilabial is the most affected, followed by the alveolar and lastly the velar ejective.

 

Fallon (1998:193) claims that crosslinguistically, deglottalization of ejectives is a natural type of lenition, often resulting in neutralization with stops. Deglottalization occurs most commonly in coda position and it is more common for ejectives to neutralize with voiceless stops. In the related Koman language Gwama, Goldberg (2015) describes ejective neutralization with voiced stops in medial position and complete neutralization with both voiced and voiceless stops in word-final position.

 

In Komo monomorphemic roots, intervocalic ejectives have an audible release, clearly contrasting with other plosives, as seen in the near minimal triplets in Table 8. Further, speakers have very little confusion identifying intervocalic ejectives and discriminating them from plain stops in monomorphemic roots.

 

 

In monomorphemic noun roots, word-final ejectives are also only released (Table 9). There is no allophonic alternation with an unreleased ejective pronunciation for the ejectives in noun roots word-finally.

 

 

Monomorphemic verb roots, by contrast, exhibit a different pattern word-finally (Table 10). Word-final sibilant ejectives are released [sʼ]. The alveolar and velar ejectives /tʼ, kʼ/ are in free variation with unreleased allophones [t̚, k̚] and the bilabial ejective /pʼ/ is unreleased [p̚].[15]

 

 

In polymorphemic inflected verbs, intervocalic ejectives exhibit considerable allophony. For instance, /tʼ, kʼ/ can be realized as [tʼ, kʼ] or as unaspirated [t, k]. By contrast, /sʼ/ largely retains its glottal release, being realized as [sʼ], though weakening to [z] also occurs.[16] The bilabial ejective /pʼ/ is very frequently, if not always, realized as [b] in natural fast speech.[17] The data in Table 11 show the verb roots seen in Table 10 inflected with the morpheme /-í/ dd∅, followed by the person marker /-r/ 3sg.m or /-n/ 3sg.n.

 

 

In sum, word-final ejectives in monomorphemic noun roots behave distinctly from word-final ejectives in monomorphemic verb roots. Noun roots preserve the laryngeal feature while in verb roots, ejectives are optionally unreleased word-finally and neutralize with voiced non-ejective allophones intervocalically. One observation is that Komo noun roots do not ever occur with suffixes or enclitics while verb roots are frequently inflected with vowel-initial suffixes. The following Table 12 contains a complete distribution of the allophones of the stops and sibilants in word-initial, intervocalic and word-final positions.

 

 

3      Vowels

 

3.1  Contrastive vowel inventory

 

Komo has seven contrastive vowels /i, ɪ, ɛ, a, ɔ, ʊ, u/ (Table 13). Members of this inventory are partially distinguished by the feature of Advanced Tongue Root (ATR), where phonemic ATR contrast occurs in the high vowels. Three [–high, +ATR] vowels [e, o, ə] occur on the surface as allophones of /ɛ, ɔ, a/, respectively, as a result of anticipatory [+ATR] harmony (see §3.2 for discussion). Evidence for the ‘basic’ status of [–ATR] is the fact that the vowels [e, o, ə] do not occur in monosyllabic monomorphemic roots and only occur in a specific environment which contains a [+high, +ATR] vowel /i, u/.[18] Burns (1947) proposes a ten-vowel inventory including seven oral vowels /i, e, ɛ, a, ɔ, o, u/ and three nasal vowels /ĩ, ɛ̃, ã/.[19] Nasal vowels, if they are indeed phonemic, are at best marginally phonemic and are always followed by a nasal consonant.

 

Table 13 Komo oral vowel inventory (allophones in brackets)

 

Evidence for a seven-vowel phonemic inventory is seen in the following (near) minimal pairs in Table 14 where first tier contains data contrasting the front vowels and /a/ and the second tier shows contrast in the back vowels and /a/.

 

 

A plot of the Komo vowel space reproduced here from Olejarczuk, Otero & Baese-Berk’s (2016) study is seen in Figure 1. The study measured a total of 6,815 total vowel tokens in a variety of consonantal environments extracted from 1,636 utterances produced by twelve adult male L1 Komo speakers. A total of 6,815 vowels (including those in the verbs and in the surrounding words) were measured at the midpoint for F1and F2.[20] Formants were measured manually using Praat software (Boersma 2002) and normalized using the Lobanov procedure to account for physiological differences across the speakers (Kendall & Thomas 2010). The ellipses in the vowel plot indicate regions within 1 standard deviation of the means. All plots collapse across prosodic and segmental environments. F1 was the most reliable correlate, where [+ATR] vowels had significantly lower F1 than their [–ATR] counterparts, which is consistent with other acoustic studies of languages with ATR harmony (e.g. Fulop et al. 1998, Guion et al. 2004).[21]

 

Figure 1 Komo acoustic vowel space (Olejarczuk, Otero & Baese-Berk 2016)

 

3.2        [ATR] Harmony

 

Advanced tongue root [ATR] harmony systems are often found in the Niger-Congo and Nilo-Saharan languages of sub-Saharan Africa (Casali 2008:497). Casali’s (2003) typological survey of 110 languages found a direct correlation between the phonemic vowel inventory and the type of ATR harmony system displayed. The specific criterion is whether or not the phonemic ATR contrast is in the high vowels. Languages like Komo, which have phonemic ATR contrast in the high vowels, overwhelmingly display [+ATR] ‘dominant-recessive’ harmony (Casali 2008:520). In this system, the feature [+ATR], considered ‘dominant’, spreads to ‘recessive’ [–ATR] vowels in a particular phonological domain (most often the word). Komo does in fact display this kind of [+ATR] dominance; but I show that the overall system also allows for [–ATR] spreading.

 

In fact, Komo has a highly productive and typologically unusual ATR harmony system that displays two distinct types of ATR assimilatory processes. The first process exemplifies [+ATR] dominant-recessive harmony and is anticipatory (leftward spreading), occurring in monomorphemic roots as well as in polymorphemic verbs. The second type is progressive (rightward spreading) [–ATR] harmony, where only high [–ATR] root vowels spread the [–ATR] feature to high [+ATR] suffix vowels. This paper summarizes Otero’s (2015) analysis and provides acoustic evidence by Olejarczuk, Otero and Baese-Berk (2016) to support the claims of an underlying seven vowel system and two distinct ATR harmony processes.[22]

 

Komo displays two categorical ATR harmony processes that are opposite both in terms of the ATR feature that spreads as well as the direction of the spreading. These processes are seen primarily in inflected verbs, as Komo lacks productive morphological processes on the whole outside of the verb. The verb roots in Table 15 contain [–high, –ATR] vowels /ɛ, a, ɔ/. Notice that the root vowels remain [–ATR] when preceding a [–ATR] vowel in a suffix, whether it’s [+high, –ATR] /ʊ/ as seen in the first set or [–high, –ATR] /a/ in the second. In these data, the entire domain remains harmonic for the [–ATR] feature, and [–high, –ATR] root vowels co-occur with [–ATR] suffix vowels.

 

Table 15 [-high, -ATR] Verb roots and inflected verbs

 

The first type of harmony spreads [+ATR] leftward from a [+high, +ATR] vowel /i, u/ to a [–high, –ATR ] vowel /ɛ, a, ɔ/, causing the latter to surface as [e, ə, o] [+ATR] allophones.[23] To illustrate [+ATR] harmony, Table 16 contains the same verb roots seen in Table 15, now inflected with suffixes containing a [+high, +ATR] vowel /i, u/. The result is anticipatory (leftward) [+ATR] dominant harmony, as the underlying [–high, –ATR] vowels /ɛ, a, ɔ/ in the roots become [+ATR] [e, ə, o] before [+high, +ATR] /i, u/. Notice also that the harmony is directionally constrained as the [+ATR] feature does not spread rightward to the 1pl.in suffix vowel /a/ at the right edge of the word.

 

Table 16 Anticipatory [+ATR] harmony in inflected verbs

 

Anticipatory [+ATR] harmoy in Komo not only occurs across morpheme boundaries but it is also evidenced in roots. Table 17 contains disyllabic monomorphemic noun and verb roots in which [+high, +ATR] vowels /i, u/ in the second syllable spread the [+ATR] feature value leftward to the [–high, –ATR] vowels of the first syllable.

 

Table 17 [+ATR] harmony in monomorphemic roots

 

The second ATR harmony process is what distinguishes Komo from Casali’s (2003, 2008) typology. This process only involves the high vowels and spreads [–ATR] rightwards from a [+high, –ATR] /ɪ, ʊ/ root vowel to a [+high, +ATR] suffix vowel /i, u/. In Table 18, the [+high, +ATR] vowels in the dd2 and dd∅ suffixes (-úk and –í respectively) assimilate to [–ATR] after [+high, –ATR] root vowels. Note that the harmony targets are the same suffix morphemes (containing the same suffix vowels) that triggered [+ATR] harmony in the environment seen in Table 18.

 

 

To illustrate that not only root vowels can trigger [–ATR] spreading, the data in Table 19 show the same verb roots seen in Table 15, now inflected with /-ʊ́/ dd1 followed by /-í/ 2sg. There are two phonetic outcomes in this case. The vowel /ʊ/ of the dd1 morpheme can either be realized as a bilabial glide [ʷ], presumably to avoid vowel hiatus, or /ʊ/ can be completely elided. What is relevant to the discussion of harmony is that in both cases, the [–ATR] feature of /ʊ/ is spread rightward to the following underlyingly [+ATR] vowel /i/ in the 2sg morpheme, causing it to surface as [–ATR] [ɪ]. Notice that in these data, the domain of harmony surfaces as entirely [–ATR].

 

 

Table 20 displays roots containing [+high, +ATR] vowels inflected with the same morphology as the verbs in Table 19, [–ATR] /-ʊ́/ dd1 followed by [+ATR] /-í/ 2sg. The data in Table 20 show how [–ATR] spreads rightwards from /ʊ/ to /i/ as seen in Table 19, but crucially the [­–ATR] feature does not spread leftwards to the potential [+high, +ATR] root vowels /i, u/. In this environment, the phonological word remains disharmonic for the ATR feature.

 

 

To summarize, both ATR harmony processes in Komo are directionally constrained: [+ATR] harmony spreads leftward while [–ATR] harmony spreads rightward. Furthermore, the domain of harmony is limited to the word and only spreads to the vowel of the adjacent target syllable as seen in (1) where [+ATR] harmony occurs in the inflected verb and does not spread to the potential vowel targets in the preceding or the following word.[24]

 

(1)	[hàr	ʃə̄bīr	kʼáw]
	hàr	ʃāpʼ-ī-r	kʼáw
	3sg.m	hit-dd∅-3sg.m	dog
	‘He hits/is hitting a/the dog.’

 

The data in Table 16 exemplify Casali’s (2003, 2008) typology for African languages with ATR contrast in the high vowels, displaying what can be characterized as anticipatory [+ATR] dominant harmony. He observes that one common characteristic of [+ATR] dominant harmony is the presence of underlyingly [+ATR] affixes (typically suffixes) that cause [–ATR] root vowels to assimilate to [+ATR]. In Komo, the [+ATR] vowels /i, u/ seen in the dd2 and dd∅ suffixes behave as prototypical triggers of [+ATR] harmony in one environment, yet non-prototypically become targets of [–ATR] spreading in another. While Casali (2008) finds that it is extremely common for languages with this vowel inventory to display [+ATR] dominance, he also notes that these languages may have a few cases of [–ATR] spreading. Casali refers to this as “superficial [–ATR] dominance", and argues that it is only found in limited circumstances within a [+ATR]-dominant system. Baković (2000) calls such exceptions “dominance reversals”, and claims that they are unproductive and restricted to a few idiosyncratic morphemes. The facts of Komo resist such "rule-and-exception" accounts; both ATR processes are extremely productive and appear to be phonologically conditioned. This apparent symmetry calls into question the typological validity of the notion of ATR “dominance”.

 

4      Syllable structure

A template of the Komo syllable is seen in Figure 2. I discuss the unambiguous syllable types (CV and CVC), complex codas (CVNC) and vowel-initial syllables in §4.1. Glides are treated in §4.2, by first describing phonetically conditioned labialization and palatalization then moving onto complex onsets (CGV, CGVC) in §4.3. Lastly, consonant clusters within polysyllabic words and across syllables types are examined in §4.4.

 

Figure 2 Komo syllable template

 

4.1  Unambiguous syllable types and complex codas

 

By far the most common and most unambiguous syllable type in monomorphemic words are CV and CVC (Table 21). Every consonant in the phonemic inventory can occur as an onset of a word-initial CV/CVC syllable.[25] Any consonant, with the exception of implosives /ɓ, ɗ/ and the glottal fricative /h/ can occur as codas of a word-final CVC syllable. Syllable nuclei in the CV and CVC shapes are always vocalic elements that function as tone bearing units.[26] Complex codas in monomorphemes are restricted to having a nasal consonant as the first of a two consonant NC coda. CVNC syllables are restricted to word-final position while CV and CVC syllables can occur both word-medially and word-finally.

 

 

Vowel initial syllables with V and VC shapes are also attested, though at a much lower frequency than CVC and CV syllable types.[27] In word-initial position a vowel-initial syllable may occur with a phonetic glottal stop [ˀV], as discussed in §2.1. This suggests that glottal stop epenthesis may be a strategy employed to derive a more common CV(C) syllable shape from a vowel-initial syllable to satisfy a phonotactic constraint of a consonant in syllable onset position word-initially.[28]

 

4.2  Glides: Phonetic labialization and palatalization

 

In Komo, glides can occur both phonetically and phonologically, and this has led to differing analyses in prior research. Burns (1947:3) asserts that all of the 23 consonants in the phonemic inventory can be labialized or palatalized word-initially but it is unclear whether he is positing phonologically-contrastive consonant phonemes, consonant allophones, or sequences of two separate consonants. Bender (1983:261), proposes non-phonemic labialized allophones [pʷ, tʷ, sʼʷ, ʃʷ, kʷ, gʷ, kʼʷ] restricted to word-initial position. In discussing complex onsets, Negash (2015) mixes predictable, phonetically-conditioned labialized and palatalized consonants with phonological consonant-glide (CG) onsets. I first discuss the predictable phonetic result of coarticulation from the following vowel onto a preceding consonant (i.e. “allophones” of the consonants) and then move on to complex onsets of the CG type. Lastly, I discuss ambiguous sequences.

 

There is non-contrastive labialization and palatalization occurring on word-initial consonants only in environments containing the mid vowels /ɛ, ɔ/. Phonetic labialization of a word-initial consonant occurs when followed by the mid back round vowel [ɔ, o], and similarly conditioned palatalization occurs with consonants preceding [ɛ, e]. Consonants which do not participate in these processes are nasals, liquids and semi-vowels. Examples of phonetic palatalization and labialization are in Table 22.

 

 

It is important to underscore that the above-mentioned processes are strictly the phonetic result of coarticulation. There is no evidence for phonemic contrast between “labialized” and non-labialized consonants before the vowel /ɔ/ ([ɔ, o]) (e.g. “kʷɔ” vs. “kɔ”) and likewise is there no contrast between “palatalized” and non-palatalized consonants before /ɛ/ [ɛ, e] (e.g. “sʼʸɛ” vs. “sʼɛ”). Prior descriptions of Komo have transcribed some of the sequences in Table 22 as if they were phonemic Cw/Cy sequences.[29] Thus the predictable phonetic labialization and palatalization do not have any bearing on the canonical CV and CVC syllable types.

 

4.3  Consonant-glide onsets

 

While the previous section showed how two particular vowels cause phonetic labialization and palatalization in word-initial consonants, I interpret these as a single consonantal phoneme in a simple syllable onset. The only complex onsets in Komo consist of an initial consonant followed by either a labiovelar approximant /w/ or a palatal glide /y/. Furthermore, the vowel /a/ is always the syllable nucleus of a Cy onset. Cw onsets occur primarily with /a/ as the nucleus, though there are few exceptions discussed below.[30] The near minimal pairs in Table 23 provide evidence for complex onsets, again restricted to word-initial syllables.

 

 

Complex onsets containing a labio-velar approximant (CwV) are headed by all attested phonemic stops and fricatives except for the bilabial implosive /ɓ/ and the voiceless bilabial ejective /pʼ/. The CyV sequence is much more limited and the attested C-glide sequences in word-initial position are seen in Table 24.

 

 

A peculiar phenomenon is the existence of a word-initial, phonetically palatalized glottal stop [ʔʸ]. I interpret this as a glottal stop followed by a palatal glide sequence occurring before the vowel /a/, which patterns consistently with the data in Table 24.[31] The palatalized glottal stop is found in three words, [ʔʸámʊ̄n] ‘seed’, [ʔʸāmàʔ] ‘sorghum’ and the verb [ʔʸákʰ] ‘go.dd2’, which appears to be a form of the motion verb ‘go.sg’ /yà/ that has been lexicalized with the Deictic Directional morpheme /-úk/ (dd2). A full paradigm for the defective verb [ʔʸák] is seen in Table 25. Notice that the palatalized glottal onset surfaces phonetically in the 1st person forms when the vowel of the second syllable is [a], but not in the rest of the paradigm when /a/ surfaces as [ə] as a result of anticipatory [+ATR] harmony.

 

 

Taking word-initial consonant-glide sequences into account, three additional syllable types can be derived (Table 26). These syllable types pattern with the unambiguous types Table 21 in that CG onset syllables allow a simple coda CGVC and a complex CGVNC coda. Further, CG onset syllables distribute like CV syllables in that CGV and CGVC can occur word-medially and CGVNC can only occur word-finally. There are no instances of a word final CGV syllable in which the glide element is a palatal glide /y/.

 

 

There is a handful of high-frequency words that don’t follow the general patterns described above for the distribution of syllables within a word. The exceptions are words that largely contain a high, front vowel in the nucleus, a contour tone and a complex Cw onset. Table 27 contains words with complex onsets followed by a vowel with a level tone (CGV) as compared to words with a vowel with a contour tone (CGV̌). Notice that the nuclei of the proposed CGV syllable types can contain a high, front vowel /i, ɪ/ as well as /a/.

 

 

While examining Semitic languages of Ethiopia, Goldenberg (2013:77-78) observes “[...] a general tendency of Ethiopian languages to decompose u into its components, viz. labiality or rounding and syllabicity (ʷ +V), and then to lose concomitant labiality when following a non-labializable consonant (C + V), but retain it when following a labializable (kʼ, x, g or k) consonant (Cʷ+V).” Comparative data from related Koman languages reveals the vocalic origin of certain ambiguous sequences. For instance, a CGV̌ sequence containing a high, front vowel quality in the nucleus in Komo, e.g.  ʃwǐ ‘beer’ has cognates containing /u/ as a vowel nucleus in related languages- sū in Uduk, ʃʊ́l in Highland Gwama and sùd̪ in Dana, which suggests a vocalic origin for some synchronic Komo labialized Cw onsets, specifically in this case when occurring with a contour tone and a high, front vowel nucleus. Further, Cw sequences in Komo occur frequently with “labializable” /k, g, kʼ, ʃ/ as the initial consonant and less so with /s, t, sʼ/. Further investigation and comparative evidence is needed.

 

4.4  Consonant clusters

 

The previous sections described consonant clusters in tauto-syllabic onsets and tauto-syllabic codas. Hetero-syllabic sequences of two distinct consonants (C₁C₂), are somewhat rare in isolated monomorphemic words.[32] Table 28 contains hetero-syllabic consonant sequences (C₁C₂) across syllables within monomorphemic words.[33] On the whole nasals and liquids form the initial consonant in a sequence followed by sibilants. Initial stops are rare and limited to the few examples in the table, some of which are place names that may not be of Komo origin. No medial or final clusters containing the following consonants in first position were found: /p, pʼ, ɓ, d, tʼ, ɗ, k, g, kʼ, z, h, w, y, ʔ/.

 

 

5      Tone

 

Komo displays three level tones as well as rising and falling contour tones. Overall, tone is quite stable in Komo: there is no evidence of downstep or downdrift in prosodic words or even across longer sentences and virtually every underlying tone and tone melody surfaces faithfully. The functional load of tone is moderate in the lexicon, more present in the verbal system where verbs can be subcategorized into classes by their tonal behavior. Further, many verb roots can also be tonally suppletive in number. In §5.1, I briefly discuss contrastive tone melodies in monomorphemic nouns and then move on to verbs in §5.2, where the discussion focuses on outlining the general patterns for verb tone classes and highlighting some anomalies in the system.

 

Burns (1947) claims there are four level tones in Sudanese Komo but I find no evidence, either phonetic or phonological, to support this in Ethiopian Komo. The current analysis proposes three level tones: (L)ow, (M)id and (H)igh. Phonemic contrast for the three level tones in monosyllabic words is seen in the near minimal monomorphemic triplets in Table 29.

 

 

There is ample phonological data to show contrast among the three level tones (cf. Appendix B). The acoustics also confirm these findings as the F0 values of the three level tone categories are significantly distinct from each other. The mean and standard deviation of the fundamental frequencies for each of the three level tones were measured from a total of 6,815 vowel tokens extracted from 1,636 utterances produced by twelve male L1 Komo speakers. The data were taken from Olejarczuk, Otero & Baese-Berk’s (2016) study following the same methodological protocols seen in the acoustic measurements of the vowels (see §3). F0 values were normalized by dividing each token’s F0 by the speaker’s mean F0 across all vowels. Results are seen in Figure 3.[34]

 

Figure 3 Fundamental frequency measures for the three level tones

 

Komo also displays (R)ising and (F)alling contour tones in monosyllabic monomorphemic roots as well as across morpheme boundaries in polymorphemic words, which can be monosyllabic or polysyllabic. In polymorphemic words, the contour tones can be analyzed as consisting of phonemic level tones. There does not appear to be more than one contrastive or rising tone in monomorphemic words. To be explicit, there are no minimal pairs of two falling tones (e.g. HL vs. HM) nor for rising tones (e.g. LM vs. LH).

 

 

5.1  Tone in nouns

 

All of the possible combinations of the three level tones, which comprise a total of nine tone patterns, are exhibited in disyllabic nouns (Table 31). There do not appear to be any restrictions with regard to consonants and tonal combinations as both voiced and voiceless consonants occur before all three of the tone levels. Out of the data in Table 31, the word-level tone combinations L.M, L.H, M.L, M.H, H.L, H.M can be extrapolated.

 

 

Monomorphemic words with more than two syllables are not common on the whole in Komo. The tonal patterns found in trisyllabic monomorphemic nouns are varied and some are rarer than others. Nonetheless, 26 of the 27 possible combinations are attested, with the exception of H.M.M (Table 32).

 

 

To generalize across the tone patterns on trisyllables, if a sequence of two identical tones followed by a distinct tone (e.g. L.L.H) can be analyzed as underlyingly L.H in autosegmental terms following the Obligatory Contour Principle (OCP), then a total of 12 contrastive patterns can be analyzed (L.M.L, L.M.H, L.H.L, L.H.M, M.L.M, M.L.H, M.H.L, M.H.M, H.L.M, H.L.H, H.M.L, H.M.H).

 

5.2  Tone in verbs: Tone classes and disyllabic roots

 

Tone has a heavier functional load in verbs, which is seen on the root level as well as in polymorphemic inflected verbs. Monosyllabic (C)V(C) verb roots are the predominant shape, though disyllabic CVCV(C) also occur.

 

Tone in Komo is rather stable overall, meaning that surface tone is almost always equivalent to the underlying tone, but there is an exception. In order to illustrate this, I examine monosyllabic verb roots. Monosyllabic verb roots fall into strict verb classes based on two criteria: the underlying tone of the verb root and how the root behaves tonally when inflected with dd and bound pronominal suffixes. The vast majority of monosyllabic verb roots occur with all three level tones (L, M, H).[35] Every finite verb must take one of three Deictic Directional (dd) suffixes, all of which contain a H tone vowel (Table 33). This suffix is followed by a bound pronominal suffix, most often realized as a consonant, which indexes an argument on the verb. The resulting shape of a verb with a monosyllabic root but also inflecting for a single argument is CVCV(C), containing two vocalic nuclei, each of which is a tone bearing unit.

 

 

Four tone classes emerge when examining monosyllabic roots across the three dd paradigms. This is specifically due to the behavior of a particular subset of L tone roots when occurring with dd∅. Table 34 schematizes the tone classes by examining the underlying tone of monosyllabic verb roots as compared to the disyllabic inflected forms across the three dd paradigms when inflecting a single argument. Class L₁, M and H roots (L, M and H tones respectively), retain their root tone when inflected across the three dd paradigms. Class L₂ roots also have a L tone when inflected with dd1 and dd2 but crucially when inflected with dd∅, the expected L.H pattern surfaces as M.M as seen in the shaded area in Table 34.

 

 

Table 35 provides Komo data to illustrate the tone analysis schematized in Table 34. The tone classes are indicated in the first column, followed by the root tone of the verb and a verb root from each class. The last four columns contain verbs inflected across the three deictic directional paradigms indexing a single argument (employing the 2pl suffix /-m/). Notice that the underlying L tone root in kɛ̀d ‘push’ (Class L₂) surfaces as M.M when inflected with dd∅ as indicated in the shaded portion.

 

 

The tone classes also hold when a verb indexes two arguments (Subject and Object), and ultimately creates a trisyllabic word. Surface tone behavior follows the patterns outlined in Table 34.[36] Class L₂ verbs are M.M in the dd∅ paradigm. Consider the following data that show the verb kàn ‘bury’ indexing two arguments on the verb: 3sg.m acting upon 1sg /-āk/, which is underlyingly M tone. Notice that the dd∅ form (shaded) surfaces as M.M(.M) while dd1 and dd2 are both L.H(.M).

 

 

The data show that not all L tone verb roots surface as L (as Class L₂ roots surface as M when occurring with dd∅), and not all dd∅ surface as H (as dd∅ surfaces as M with Class L₂ roots). Therefore, not only does the behavior of class L₂ roots occurring with dd∅ necessitate more complex underlying tones, but the underlying tones of the remaining classes must be reexamined as well as these roots do not change when inflected with dd∅. While an in-depth analysis of tone is beyond the scope of this paper, what is important for the current discussion of surface tone patterns is that there are four tone classes. These four tone classes hold and remain contrastive in other aspects of Komo verb tonology.

 

Disyllabic verbs also occur in the lexicon and like disyllabic nouns, all of the nine possible tone combinations are exhibited (Table 37). Disyllabic verbs inflect like monosyllabic verbs, taking directional and person-indexing suffixes.

 

 

5.3  Tonal suppletion in verbs: Number distinction

 

A large portion of verb roots, largely if not exclusively intransitive verbs, are tonally suppletive in number. The distinction is realized across the behavior of the four tone classes (even if not evident by the tone of the root alone), as not all L tone roots behave similarly when inflected. Thus, one particular tone class marks singular subjects while an opposing class marks plural subjects. Taking into account the four tone classes, nine of the ten possible combinations attested are seen in Table 38. The lack of an M (sg)/ L₁ (pl) combination may be due to insufficient data rather than a system-internal restriction. Interestingly there is no instance of an opposition within the L class of verbs (L₁ / L₂), which may suggest greater category affinity.

 

 

Disyllabic verb roots also display tonal suppletion in number as seen in Table 39. Number distinctions on the root level and/or tonally suppletive number on the root level in verbs is also attested in related Koman languages (e.g Opo, Smolders 2017).

 

 

6      Morphophonological processes

 

6.1  Metathesis, elision and voicing: /-úk/ dd2

 

This section examines the particular behavior of the Deictic Directional morpheme -úk (dd2) and its allophones [-úk /ʊ́k, -kú /kʊ́, -k, -ʷk].[37] Recall that a finite verb takes bound pronominal suffixes following the Deictic Directional suffix. The particular allomorph of -úk (dd2) is largely conditioned by the shape of the following bound pronominal. When there is no bound pronominal morphology on the verb, dd2 always surfaces as -úk. This is seen in an Auxiliary construction, which contains a fully inflected auxiliary verb followed by a verb root inflected with a Deictic Directional [aux-bp-(bp)-(bp) V-dd]. In this construction, seen in (2)-(4), the dd2 morpheme always surfaces as either [-uk] or [-ʊk], regardless if the verb root is consonant or vowel-final.

 

(2)	ʊ́-ná-g-ápʼ-ɪ̄	ʃɛ̀r-úk
	int-1sg-ben-3sg.f-3sg.nh	sweep-dd2
	‘I will have swept it for her.’

 

(3)	bāʃ-g-ɪ̀n	sɪ̀g-ʊ́k
	neg-1sg-3sg.nh	hear-dd2
	‘I haven’t heard it.’

 

(4)	bāʃ-g-ɪ̀n	ʃá-úk
	neg-1sg-3sg.nh	eat.sg-dd2
	‘I haven’t eaten it.’

 

Process rules for deriving allomorphs of /-úk/ are seen in Table 40. Metathesis of /-úk/ occurs before consonant-initial suffixes (i), and vowel-deletion or residual labialization occurs before a suffix containing /a/ that is either (H)igh or (M)id tone (ii).

 

Table 40 Process rules for -úk (dd2)

 

Table 41 contains a full paradigm of a consonant-final verb ìʃ ‘sleep’ occurring with the Deictic Directional morpheme /-úk/ (dd2) inflected with bound pronominal suffixes indexing the subject. When the bound pronominal suffix consists of a consonant, -úk surfaces as [-kú] as seen in 3sg.m/f, 3nh, 2pl and 3pl. Presumably, -úk metathesizes to avoid a sequence of three consonants in these cases. When the bound pronominal suffix following /-úk/ contains a vowel that is M or H tone, -úk surfaces as [-k] (1sg, 2sg, 1pl.ex). The [-k] allomorph also alternates with [-ʷk] before a mid or high tone /-a/ suffix in 1sg and 1pl.ex. The phonologically unpredictable behavior of /-úk/ is seen when the verb is marked with 1pl.in. In this case, /-úk/ always surfaces as [-úk]. Whether or not this has to do with the fact that the 1pl.in morpheme is low tone -(n)à or whether there is a diachronic explanation remains to be investigated, though there is no other evidence of tone or tone melodies constraining segmental structure in Komo.

 

Table 41 Paradigm of ɪʃ ‘sleep’ inflected with the dd2 suffix /-úk/

 

Further evidence of the allomorphic behavior of /-úk/ is seen in open syllable roots (Table 42). Notice that the bilabial offglide is present in 1sg in the verbs [ʃóʷká] ‘I sewed’ and [ɗīʷká] ‘I was angry’, but not [bʷɔ̀ká] ‘I hid’, where the initial consonant is phonetically labialized. When the verb is inflected with –(n)à 1pl.in, the vocalic element in /-úk/ can assimilate to the vowel of the root and a contour tone can emerge if the vowel of the root is not H tone. When inflected with -r 3sg.m, /-úk/ metathesizes to [-kú].

 

Table 42 Allomorphs of /-úk/ on open syllable verb roots

 

When the final consonant of a verb root is a voiced stop or voiced fricative, a following contiguous voiceless velar stop of the metathesized dd2 suffix [-kú] becomes voiced (Table 43). Notice that the bilabial nasal can also cause voicing assimilation but the alveolar nasal, lateral and trill do not.

 

Table 43  Voicing of /úk/ dd2 in verbs

 

Verb roots ending in ejectives behave similarly to those ending in voiced stops. Word-medially, ejectives are realized as voiced stops and cause the adjacent voiceless velar stop of the [-k(u)] allomorph to become voiced (Table 44). The sibilant ejective is realized as a voiceless ejective, and as such, does not participate in the voicing assimilation process.

 

Table 44 Ejective-final verb roots inflected with /-úk/ dd2

 

In §2.3 I discussed the distinct phonetic behavior of ejectives in monomorphemic noun and verb roots. Word-final ejectives are realized with glottal release in noun roots as opposed to verb roots. Intervocalically, the ejective phonemes are realized with a glottal release in monomorphemes. But in polymorphemic verbs with inflectional morphology, the ejective and voiced stop contrast can often be neutralized. There also appears to be a cline in terms of how the ejectives are realized. The sibilant ejective /sʼ/ is almost always realized as an ejective, followed by the alveolar and velar ejectives /tʼ, kʼ/ which alternate with voiced stops. The bilabial /pʼ/ has a wider range of allophones and has neutralized with /b/ word-medially in inflected verbs. This peculiar phonetic behavior of ejectives, specifically the bilabial ejectives, is also reported for related Koman languages.[38]

 

6.2  Consonant gemination in verbs: dd2 /-úk/

 

Consonant length is not a contrastive phonemic feature in Komo nor has it been reported in any other Koman language to date. However, heterosyllabic consonant gemination, a rare phenomenon in Komo, can occur within stems and across morpheme boundaries and is the result of morphophonological processes, namely metathesis, elision and partial reduplication which conspire such that the resulting surface forms contain a geminate consonant. The full word forms containing “geminate” consonants can ultimately contrast with non-geminate forms as minimal or near minimal pairs. Gemination in verbs can be seen in two environments: when a verb root ending in a velar stop is inflected with the /-úk/ dd2 suffix, and in verb stems that employ partial reduplication to code the pluractional.

 

The data in Table 41 show that certain morphophonological processes occur in verbs inflected with the suffix /-úk/ (dd2). When the inflected verb is formed by a root with a final velar consonant (C)VK, the result of these processes can create a sequence of contiguous velar stops, which in turn, is pronounced as a single geminate consonant. If other environmental factors, such as tone and vowel harmony align, length can be the only contrastive feature between inflected verbs (with roots that end in a velar stop). The data in Table 45 show pairs of verbs that contrast only in length. Notice that the vowel has elided in the forms inflected with the dd2 morpheme /-uk/. The result is a sequence of velar stops, which results in contrastive length when compared to the same verb inflected with dd∅.[39]

 

Table 45 Velar consonant gemination in verbs inflected with /-úk/ dd2 compared to dd∅

 

Similar contrast in length is seen in some verbs inflected for 3sg.m (Table 46). In these data, verb roots ending in a velar stop inflected with /-úk/ dd2 contrast in length with those inflected with /-ʊ́/ dd1. Note that the stem vowels do not surface identically for the ATR feature in ‘He sows’. In the pairs, ‘He crosses’ and ‘He waited’, the only phonological contrast is length. In these examples, the [−ATR] feature of the root vowel spreads to the [+ATR] /u/ vowel of the dd2 morpheme /-úk/, causing the vowel to surface as [ʊ́]. As such the sequences contain identical vowels and the pairs only contrast in length.

 

Table 46 Velar consonant gemination in verbs inflected with /-úk/ dd2 compared to /-ʊ́/ dd1

 

6.3  Consonant gemination in verbs: Pluractional

 

Pluractional derivations can be employed to express actions as happening to many subjects or to many objects, or it can describe events as occurring repeatedly or habitually (Newman 1990). As such, the pluractional is commonly considered a property that codes verbal number or rather, multiplicity of a verbal event. In Komo, the pluractional is expressed by morphological derivation which involves a partial reduplication of the verb root. Monomorphemic monosyllabic verb roots in Komo have a CV or CVC shape.[40] In the bare form (without any inflectional morphology), roots with a CVC shape reduplicate the final root consonant, after an epenthetic high [−ATR] vowel [ɪ̄]. Those roots with a CV shape reduplicate the first (i.e. only) consonant, followed by [ɪ̄] (Table 47).

 

Table 47 Pluractional derivation

The newly derived pluractional stem is able to inflect regularly with Deictic Directional morphology and bound pronominal suffixes. When the pluractional stem is inflected, deletion of the epenthetic vowel /ɪ̄/ occurs. As a result of the deletion, two identical consonants occur in sequence, pronounced as a geminate. In (5a) the derived pluractional verb is in the bare form following an auxiliary, as compared to (5b) where gemination occurs since the epenthetic vowel [ɪ̄] has deleted (but not before spreading the [−ATR] feature to the morpheme /-í/ dd∅).

 

(5)	a.	bāʃ-í-r	yìs-ɪ̄s	[yìsɪ̄s]	b.	yìs-ɪ̄s-í-r	[yìsːɪ́r]
		neg-ddØ-3sg.m	circle-redup.plu			circle-redup.plu-ddØ-3sg.m
		‘He doesn’t circle around (continually).’				‘He circles around (continually).’

 

Contrast between geminate and singlelton consonants involving the pluractional derivation is seen in Table 48 with the nominalizing enclitic =āgɪ́.

 

Table 48 Gemination in nominalized verbs

 

7      Conclusion

 

This paper is the first comprehensive examination of the core phonology of the Ethiopian variety of Komo, an endangered language of the Koman family. One notable feature of the Komo consonant system is the apparent deglottalization of ejectives, which results in voiced obstruents word-medially and unreleased obstruents word-finally. Surface contrast in consonant length, a feature that is not considered phonemically contrastive in the language, arises from morphophonological processes in the verb. Komo displays a phonemic seven-vowel inventory with ATR contrast in the high vowels, which is supported by acoustic measurements of over 6,000 vowels taken from twelve native speakers. One of the most unusual aspects of Komo phonology is the bi-directional ATR harmony system. The Komo system questions the notion of a single “dominant” ATR feature within an ATR harmony system, as the same vocalic elements that are triggers of anticipatory [+ATR] spreading in one domain can also be targets of progressive [–ATR] spreading in another.

 

References

 

Baković, Eric. 2000. Harmony, dominance and control. New Brunswick, NJ: Rutgers University dissertation.

Bender, M. Lionel. 1971. The languages of Ethiopia: A new lexicostatistic classification and some problems of diffusion. Anthropological Linguistics 13(5). 165-288.

Bender, M. Lionel. 1983. Proto-Koman phonology and lexicon. Afrika und Übersee 66(2). 259-297.

Bender, M. Lionel. 1991. Sub-classification of Nilo-Saharan. In M. Lionel Bender (ed.), Proceedings of the 4th Nilo-Saharan Linguistics Colloquium, 1-36. Hamburg: Helmut Buske Verlag.

Bender, M. Lionel. 1994. Comparative Komuz grammar. Afrika und Übersee 77(1). 31-54.

Boersma, Paul. 2002. Praat, a system for doing phonetics by computer. Glot international 5(9/10). 341-345. Online: http://www.fon.hum.uva.nl/praat/.

Burns, Samuel J. 1947. Notes toward a grammar of the Koma language. London: Sudan Interior Mission, MS.

Casali, Roderic F. 2003. [ATR] value asymmetries and underlying vowel inventory structure in Niger-Congo and Nilo-Saharan. Linguistic Typology 7(3). 307-382.

Casali, Roderic F. 2008. ATR harmony in African languages. Language and Linguistics Compass 2(3). 496-549.

Corfield, Frank D. 1938. The Koma. Sudan Notes and Records 21(1). 123-165.

Dimmendaal, Gerrit J. 2011. Historical linguistics and the comparative study of African languages. Amsterdam: John Benjamins.

Ehret, Christopher. 2001. A Historical-comparative reconstruction of Nilo-Saharan. Köln: Rüdiger Köppe Verlag.

Fallon, Paul D. 1998. The synchronic and diachronic phonology of ejectives. Columbus, Ohio: Ohio State University dissertation.

Federal Democratic Commission Republic of Ethiopia Population Census (FDCREPC). 2008. Summary and Statistical Report of the 2007 Population and Housing Census. Addis Ababa, Ethiopia.

Fulop, Sean, Ethelbert Kari, & Peter Ladefoged. 1998. An acoustic study of the tongue root contrast in Degema vowels. Phonetica 55. 80-98.

Goldberg, Justin. 2015. Ejective neutralization in Gwama. Paper presented at the 12th Nilo-Saharan Linguistics Colloquium, Nairobi, Kenya.

Goldenberg, Gideon. 2013. Semitic languages: Features, structures, relations, processes. Oxford: Oxford University Press.

Greenberg, Joseph H. 1963. The languages of Africa. Bloomington: Indiana University.

Grottanelli, Vinigi L. 1946. Materiali di lingua Coma. Rassegna di Studi Etiopici 5. 122-155.

Guion, Susan G., Mark W. Post, & Doris L. Payne. 2004. Phonetic correlates of tongue root vowel contrasts in Maa. Journal of Phonetics 32(4). 517-542.

Hellenthal, Anne-Christie, & Constance Kutsch-Lojenga. 2011. Phonology/Orthography statement for the Gwama language. Addis Ababa: MS.

James, Wendy, Gerd Baumann, & Douglas Hamilton Johnson (eds.). 1996. Juan Maria Schuver’s travels in North-East Africa, 1880-1883. London: The Hakluyt Society.

Kebebw, Lemi. 2010. A grammatical description of Opo. Addis Ababa: University of Addis Ababa MA thesis.

Kendall, Tyler, & Erik R Thomas. 2010. Vowels: Vowel manipulation, normalization, and plotting in R. R package, version 1.

Kievit, Dirk, & Erika Robertson. 2012. Notes on Gwama grammar. Studies in African Linguistics 41(1). 39-97

Killian, Don. 2015. Topics in Uduk phonology and morphosyntax. Helsinki: University of Helsinki dissertation.

Küspert, Klaus-Christian. 2015. The Mao and Komo languages in the Begi–Tongo area in Western Ethiopia: Classification, designations, distribution. Linguistic Discovery 13(1). 1-63.

Kutsch Lojenga, Constance & Manuel A. Otero. 2011. Phonology/Orthography statement for the Komo language. Addis Ababa: SIL Ethiopia, MS.

Lewis, Paul M., Gary F. Simons, & Charles D. Fennig. 2015. Ethnologue: Languages of the World, Eighteenth Edition. Dallas, Texas: SIL international.

Meckelburg, Alexander. 2015. Slavery, emancipation and memory: Exploratory notes on western Ethiopia. International Journal of African Historical Studies 48(2). 345-362.

Negash, Tesfaye. 2015. Documentation and grammatical description of Kwom. Addis Ababa: Addis Ababa University dissertation.

Newman, Paul. 1990. Nominal and verbal plurality in Chadic. Berlin: De Gruyter Mouton.

Olejarczuk, Paul, Manuel A. Otero, & Melissa Baese-Berk. 2016. Production and perception of the   [ATR] vowel system in Ethiopian Komo. Paper presented at the 171st Meeting of the Acoustical Society of America, Salt Lake City, UT.

Olejarczuk, Paul, Manuel A. Otero, &Melissa Baese-Berk. in prep. Acoustic correlates of unique bidirectional [ATR] system: The case of Ethiopian Komo

Otero, Manuel A. 2013. ATR Harmony in Komo [xom]. Paper presented at the 44th Annual Conference on African Linguistics, Georgetown, D.C.

Otero, Manuel A. 2015. [+ATR] dominant vowel harmony except when it’s not? Evidence from Ethiopian Komo. In Ruth Kramer, Elizabeth C. Zsiga, & One Tlale Boyer (eds.), Selected Proceedings of the 44th Annual Conference on African Linguistics, 212-220. Somerville, MA: Cascadilla Proceedings Project.

Otero, Manuel A. 2016. Reconstructing pronominal morphology in Koman languages. Paper presented at the 47th Annual Conference on African Linguistics, Berkeley, CA.

Payne, Doris L., & Manuel A. Otero. 2016. Deictic (Associated-)Motion and the tense-aspect domain: Evidence from Nilotic & Koman. Paper presented at the Syntax of the World’s Languages 7, Workshop on Associated Motion, Mexico City, Mexico.

Smolders, Joshua. 2017. A Wordlist of T'apo (Opuuo): Version 0.1. Addis Ababa: SIL Ethiopia.

Theis, Joachim. 1995. Nach der Razzia: Ethnographie und Geschichte der Koma. München: Trickster Verlag.

Tucker, Archibald Norman, & Margaret Arminel Bryan. 1966. Linguistic Analysis: The Non-Bantu Languages of North-Eastern Africa. London: Oxford University Press.

van Silfhout, Marijke. 2013. Opo: Towards a phonology. Leiden: Leiden University MA thesis.

Yehualashet, Teshome. 2008. A brief phonological description and nomenclature of Kwama/ Komo (A Nilo-Saharan language). Addis Ababa: University of Addis Ababa, MS.

 

Appendix A: Evidence for consonant contrast

 

p : b
initial
pá	‘shoe, sandal’	bà	‘be wide’
pàm	‘touch, feel’	bám	1sg.poss
pút	‘plaster’	būt	‘grasshopper (sp.)’
pít	‘vagina’	bìt	‘discard’
intervocalic
púpúʃ	‘tree (sp.)’	bùbūʃ	‘tree (sp.)’
ʃìpá	‘knife’	yībā	‘man’
final
ùp	‘bathe’	ùb	‘feed’
tʼàp	‘spit (v.)’	táb	‘kick’

 

p : pʼ
initial
pɛ̀l	‘untie’	pʼɛ́l	‘be red’
pàsʼ	‘soak’	pʼàʃ	‘tree (sp.’
intervocalic
púpúʃ	‘tree (sp.)’	nàpʼúʃ	‘soft, silky’
final
kɔ̀p	‘upper arm’	kɔ̀pʼ	‘eland antelope’
ɪ́p	‘drink.sg’	yɪ̄pʼ	‘spear, inject’

 

p : ɓ[41]
initial
pàsʼ	‘soak’	ɓàsʼ	‘be hot’
pú	‘decay, rot’	ɓú	‘be pregnant’

 

p : m
initial
pʊ́ʃ	‘sand’	mùʃ	‘draw water’
pá	‘shoe’	mà	‘do’
intervocalic
kùpú	‘cane’	kūmú	‘clitoris’
dúpɛ̀	‘skirt (trad.)’	dùmɛ̀	‘tree (sp.)’
final
páp	‘defecate’	pàm	‘touch
ʃɛ́p	‘tree (sp.)’	ʃɛ́m	‘stuff, pack in’

 

p : w
initial
pūd	‘wind’	wùd	‘fox (sp.)’
pàg	‘swim’	wàg	‘jeer, slander’
intervocalic
yīpāl	‘young man’	yīwáy	‘healer
final
káp	‘move fire’	kʼáw	‘dog’

 

pʼ : b
initial
pʼàʃ	‘tree (sp.)’	bàʃ	‘blood’
pʼīn	‘ashes’	bīní	2sg.poss
intervocalic
gʊ̀pʼɪ́	‘stew’	gùbí	‘house, hut’
final
bɪ́pʼ	3sg.f.poss	bɪ̀b	‘cow’

 

pʼ : ɓ
initial
pʼísʼ	‘rub hands together’	ɓísʼ	‘be strong.sg’
pʼàʃ	‘tree (sp.)’	ɓàsʼ	‘be hot’

 

b : ɓ
initial
bà	‘be wide’	ɓà	‘nape’
bɔ̀l	‘stretch’	ɓɔ́l	‘be good’

 

b : m
initial
bùl	‘drum’	mùl	‘hair’
intervocalic
kʊ́bá	‘pillow’	kʊ̄mà	‘path, road’
sʼābà	‘traditional healing’	sʼāmà	‘bird (sp.)’
final
dàb	‘be fragrant’	dàm	‘honey’
bàb	‘bury’	bàm	‘move together’

 

b : w
initial
būt	‘grasshopper (sp.)’	wǔt	‘ostrich’
final
dàb	‘be fragrant’	dàw	‘baboon’

 

m : n
initial
màʃ	‘marriage’	nàsʼ	‘grind (wet)’
intervocalic
ámà	‘there’	ánà	1sg.inclusive
final
tɔ̄m	‘spear’	tɔ̄n	‘pass, overtake’
súm	‘pound in’	sùn	‘point at, show’

     

m : w
initial
mādā	‘place’	wādá	‘fox (sp.)’
mālà	‘saliva’	wàlá	‘hide (of animal)’
final
dàm	‘honey’	dàw	‘baboon’

 

t : d
initial
tūt	‘return, reply’	dùt	‘elbow’
tɛ́n	‘slap, strike’	dɛ̀n	‘count, read’
intervocalic
wātá	‘be like, similar to’	wādá	‘fox (sp.)’
final
būt	‘grasshopper (sp.)’	bʊ̀d	‘unthatch’

 

t : tʼ
initial
tā	‘be, exist’	tʼā	‘mouth, language’
tín	‘fight, war (v)’	tʼìn	‘stretch’
intervocalic
pàtàgɪ́	‘to crawl’	pʼàtʼàgɪ́	‘ to be white, clean’
final
kàt	‘convince, lobby’	kʼátʼ	‘be easy’
dɔ̀t	‘ask’	dɔ̀tʼ	‘squat’

           

t : ɗ
initial
táb	‘kick’	ɗàb	‘adhere’
tɛ́	‘sun’	ɗɛ́	‘one’
tíʃ	‘sing and play’	ɗìʃ	‘be near’

 

t : n
initial
tɛ́g	‘not acknowledge.sg’	nɛ́k	‘Hammerkopf’
tā	‘be, exist’	ná	‘feel hopeless’
final
kát	‘cause trouble’	kàn	‘bury’

 

t : r
initial
tɪ́t	‘roughen (stone)’	rɪ̀tá	‘hyena’
intervocalic
kàtàkàtà	‘whisper’	kàrá	‘fish (sp.)’
final
bát	‘duck (sp.)’	bàr	‘lower back’
būt	‘grasshopper (sp.)’	bùr	‘ventilate, fan’

 

t : l
initial
tūs	‘push’	lùs	‘hide sth.’
tútút	‘crossroads’	lùtút	‘come free and fall’
intervocalic
lītī tɔ̄m	‘shooting star’	līlí	‘sink (v.)’
final
bìt	‘toss’	bɪ̀l	‘root’

 

t : sʼ
initial
ták	‘surpass’	sʼák	‘burn’
tásʼ	‘smith’	sʼásʼ	‘ant (sp.)’
intervocalic
dɪ̀tɪ̄n	dem.dist	dīsʼīn	‘pimple’
final
wǔt	‘ostrich’	wúsʼ	‘shut’
dɔ̀t	‘ask’	dɔ̀sʼ	‘suck’

 

tʼ : d
initial
tʼàʃ	‘salt’	dàʃʊ́	‘snake (sp.)’
tʼútʼ	‘freckle, spot’	dùd	‘elbow’
intervocalic
kátʼāgɪ́	‘be soft’	kádán	‘fish (sp.)’
final
tʼútʼ	‘freckle, spot’	dùd	‘elbow’
hátʼ	‘clear, flatten’	hàd	‘pull’

 

tʼ : ɗ
initial
tʼʊ́sʼ	‘choke’	dʊ̄sʼ	‘urinate’
tʼíʃ	‘finely grind’	ɗíʃ	‘be near.pl’

 

 

tʼ : sʼ
initial
tʼʊ́sʼ	‘choke’	sʼʊ́sʼ	‘scratch’
tʼádà	‘bake’	sʼábà	‘tree (sp.)’
final
dɔ̀tʼ	‘squat’	dɔ̀sʼ	‘suck’
hátʼ	‘clear, flatten’	hásʼ	‘ruminate, chew cud’

 

d : ɗ
initial
dàb	‘be fragrant’	ɗàb	‘adhere’
dɛ̌	‘thing’	ɗɛ́	‘one’
dɔ̀l	‘cry out’	ɗɔ̀	‘carry many things’

 

d : n
initial
dɪ̀	dem.prox.1	nɪ̀	dem.prox.2
dɔ̄kɔ̄	‘potato (sp.)’	nɔ̀gɔ́	‘worm (sp.)’
intervocalic
dàdāgɪ́	‘to complain’	dànāgɪ́	‘to growl’
final
dàd	‘refuse sth.’	dàn	‘menstrual cycle’
bʊ̀d	‘unthatch’	bʊ́n	3pl.poss

 

d : r
initial
dɪ̄dà	‘lance, spear’	rɪ̀tá	‘hyena’
dɔ̄kɔ̄	‘potato (sp.)’	rɔ̀kɔ́n	‘corner’
intervocalic
dɪ̄dà	‘lance, spear’	dìrà	‘fish (sp.)’
kādán	‘fish (sp.)’	kàrá	‘fish (sp.)’
final
ùd	‘hit’	ùr	‘be barren’
hàd	‘pull’	hàr	3sg.m

 

d : l
initial
dàm	‘honey’	làm	‘strain (e.g. food)’
dʊ̀	‘inhabit (temporarily)’	lù	‘sprout (v.)’
intervocalic
mādā	‘place’	mālà	‘saliva’
final
bʊ̀d	‘unthatch’	bùl	‘drum’
tɔ̄d	‘abort’	tɔ̄l	‘be fat.pl’

 

r : l
initial
rɔ́kɔ̀n	‘corner, angle’	lɔ̀gɔ́n	‘wall’
rɪ̀tá	‘hyena’	lɪ́tɪ̀	‘flute’
final
bár	‘heron’	bál	‘in order to’
tɔ̀r	‘slaughter, decapitate’	tɔ̀l	‘be fat.sg’

 

s : z
initial
sà	‘very, very much’	zà	‘dig’
sɛ̀s	‘delineate, mark on ground’	zɛ̀z	‘set in sights, measure’
intervocalic
bàsá	‘be good’	màzá	‘be tired’
final
sɛ̀s	‘delineate, mark on ground’	zɛ̀z	‘set in sights, measure’

 

s : sʼ
initial
sɪ̄	‘bowl, pot’	sʼɪ́	‘diarrhea’
sās	‘greet’	sʼàsʼ	‘chest’
sɛ̀s	‘delineate, mark on ground’	sʼɛ̀sʼ	‘termite’
final
kɪ̄s	‘tree (sp.)’	kɪ́sʼ	‘light fire’
hās	‘wash (sth.)’	hásʼ	‘ruminate, chew cud’

     

     

s : ʃ
initial
sɪ̄	‘bowl, pot’	ʃɪ̀	‘rain (v.)’
sá	‘tree’	ʃá	‘eat.sg (soft food)’
in CC sequence
ɪ́skɪ́r	‘squirrel (sp.)’	túʃkír	‘heart, internal organ’
final
yìs	‘circle, encircle’	yìʃ	‘penis’
tūs	‘force feed’	tùʃ	‘cotton, thread’

 

sʼ : z
initial
sʼɪ́	‘be black’	zɪ̀	‘be green’
sʼʊ́	‘river’	zù	‘pour’
intervocalic
bāsʼʊ́	‘snake (sp.)’	yàzʊ́	‘friend’
final
gwàsʼ	‘buffalo’	gwàz	‘men’
wásʼ	‘bite’	wʊ̀z	‘fish (sp.)’

 

sʼ : ʃ
initial
sʼɛ̀	‘shuck’	ʃɛ̀	‘tooth’
sʼǎk	‘burn’	ʃàk	‘catch fish (trad.)’
intervocalic
wàsʼāgɪ́	‘bite, pain’	wàʃɪ́kāgɪ́	‘boil’
final
dɔ̀sʼ	‘suck’	dɔ̀ʃ	‘stand.sg’
wàsʼ	‘fish’	wàʃ	‘wife’

 

k : g
initial
kíʃ	‘cocoyam’	gìʃ	‘hole’
kàm	‘brother’	gàm	‘find, meet’
kʊ́sʼ	‘be dry, infertile’	gʊ̀sʼ	‘swallow’
intervocalic
ɓākān	‘sister’	bágà	‘jerrycan’
kákā	‘thorn’	kágá	‘cackle’
final
ʃɔ̄k	‘move away, migrate’	ʃɔ̀g	‘leg, foot’
ták	‘overtake, surpass’	tàg	‘forehead’

 

k : kʼ
initial
kɪ̄	‘horn (of animal)’	kʼɪ́	‘remove shell’
kàm	‘brother’	kʼàm	‘feel’
final
sʼík	‘rat, mouse’	sʼɪ̄kʼ	‘be sour’
ʃúk	‘land, alight’	ʃʊ̄nkʼ	‘vein, tendon’

 

k : h
initial
kàm	‘brother’	hàm	‘yawn.pl’
kàʊ̄	‘bring’	hàʊ̄	‘come.sg’

 

g : kʼ
initial
gɪ̄	‘with’	kɪ̄	‘horn (of animal)’
gàm	‘find, meet’	kʼàm	‘feel’
final
sɪ̀g	‘listen, hear’	sʼɪ̄kʼ	‘be sour’
ʃʊ̀g	‘palm tree (sp.)’	ʃʊ̄nkʼ	‘vein, tendon’

 

Appendix B: Evidence for tone contrast: (near) minimal pairs

monosyllabic
L		H
ày	2sg	áy	‘remain, stay’
sʼɔ̀sʼ	‘be few, small.pl’	sʼɔ́sʼ	‘scratch’
twày	‘not sleep’	twáy	‘hunger’
bʊ̀m	‘sugarcane’	bʊ́m	3pl.poss
sà	‘very’	sá	‘tree’
L		M
sʼɔ̀sʼ	‘be few, small.pl’	sʼɔ̄sʼ	‘be few, small.sg’
bà	‘be wide’	bā	dem / poss
dìʃ	‘mushroom (sp.)’	dīʃ	‘three’
yàʔ	‘vomit.sg’	yāʔ	‘vomit.pl’
M		H
bāl	‘beehive’	bál	‘in order to’
wʊ̄l	‘together’	wʊ́l	‘miss mark’
F		M
ɓî	‘eye’	ɓɪ̄	‘mountian’
R		H
kɪ̌l	‘cattle egret’	kɪ́l	‘star’
R		M
ʃǐt	‘tree (sp.)’	ʃīt	‘wrinkle (on skin)’
R		L
bǎ	‘father’	bà	‘wake’
zǐ	‘excrement’	zì	‘be green.sg’
disyllabic
L.H		L.L
zàmá	‘walk gracefully’	zàmà	‘wound’
L.M		M.H
mɛ̀ʔī	‘food’	mɛ̄ʔí	‘goats’
M.L		M.H
mātà	‘perhaps’	mātá	‘hunt’
M.H		L.H
ʃɛ̄ʔí	‘kale, collard greens’	ʃɛ̀ʔí	‘teeth’
H.M		L.H
kákā	‘thorn’	kàká	‘grandmother’