Consonants

In the consonants chart, a voicing contrast is shown for obstruents. For the majority of speakers, and particularly those influenced by contact with French, there is a voicing or VOT contrast and a two-way tonal contrast. Voicing (for stops and fricatives) cannot be combined with H tone, so this leaves three contrastive groups when voicing and tone are taken into account. Among village women, there is less contact with other languages, and the tone/voicing relationship is different. There is no voicing contrast, but there is a three-way tonal contrast marking the same three groups of contrast as the other speech varieties. There is more discussion on this point in the section on tone and orthography, and in Pearce (Reference Pearce2006b). The importance of this discussion in this section is that for some speech varieties, a voicing contrast is part of the grammar, whereas for others, it is not. The consonant chart here applies to the variety with the voicing contrast.

In speech varieties where there is a voicing contrast, the phonetic realization of the contrast is usually between short and long positive VOT. However, some speakers (including the two speakers in the recordings for this Illustration) vary between short positive VOT and true voicing, similarly to the way English speakers often vary in the production of the voiced-stop series. Intervocalically there may also be true voicing if the surrounding tones are L.

The affricates /dʒ/ and /tʃ/ are analyzed phonologically as palatal stops (and are in fact treated as such in the orthography), but they are realized as affricates. The glottal stop only contrasts with absence of glottal stop, and only word medially (or finally in words with final vowel deletion).

Words beginning with a vowel are normally pronounced with an initial glottal stop, but this is not contrastive.

The inclusion of the labiodental flap (//) is somewhat controversial because the sound almost certainly entered the language from the influence of other languages, and is not used equally by all speakers (either because the vocabulary is not known or because a substitute such as /w/ is used). Kera is one of only 80 languages in the world where this sound is attested. In Kera, 95% of the 60 words containing the flap are ideophones. However, minimal pairs exist that show how it contrasts with similar sounds, and for this reason it is included in the phoneme chart.Footnote ¹ Kera was one of the languages cited by Olson & Hajek (Reference Olson and Hajek2003, Reference Olson and Hajek2004) in the argument for the inclusion of the labiodental flap in the International Phonetic Alphabet. Labiodental flap minimal pairs are illustrated in the following list (bold indicates the sounds in question):

Word-medially, almost all consonants can be in coda position (with the exception of the labiodental flap), but word-finally, only sonorants can occur:

Footnote ² To avoid an obstruent in word-final position, an epenthetic vowel [i] is added:

Vowels

Footnote ³ The ellipsis lines indicate allophones of one phoneme. In each case, the tense (more closed) vowel can only appear in light syllables before a heavy syllable while the lax (more open) vowel appears elsewhere. The light syllable with the tense vowel is phonetically shortened. Pearce (Reference Pearce2007a) claims that this is to heighten the contrast between the non-head and the head in an iambic foot. This means that the vowel in these syllables has a duration of 30–50 ms. The high vowels have no allophones. Table 1 demonstrates the changes of quality in light and heavy syllables.

Table 1 Changes of quality in light and heavy syllables.

The nasalized vowels are not common. It is possible that they were imported from the neighbouring language, Tupuri, or that they replace a coda [ɲ] which may have existed once in Kera as it does in other Chadic languages; thus *Vɲ becomes Ṽ

Further details of the vowel allophones are discussed in Pearce (Reference Pearce2008), including the continuum of formant values depending on the phonetic duration of the vowel. As seen in Figure 1, shorter vowels (approx. 30 ms) produce clear undershoot. Longer vowels, (approx. 70 ms, but still phonologically short) arrive at the target (or standard) value for the formant. Vowels between these two values (approx. 50 ms) also have limited undershoot. All phonologically long vowels (approx. 100 ms or more) are produced with the formants of the target value. Thus, the variation in quality is actually driven by the phonetic duration rather than by phonological length.

Note: The orthography transcribes shorter /a/ vowels as schwa, when in a light syllable preceding a heavy syllable. The other two non-high vowels are always written as 〈e〉 and 〈o〉 respectively. From a phonological point of view, this is confusing, but the orthography is now well accepted by Kera speakers and is unlikely to change.

Figure 1 Kera non-high vowels: Duration (ms) and First Formant (Hz).

Kera exhibits the following three types of vowel harmony:

(i) Total harmony within the root. Word-final CV syllables, the historical affixes a-, -a, the epenthetic -i following obstruents, and loans are exempt. Apart from these exceptions, all nouns and verbs comply.

(ii) Height harmony between the root and suffix in both directions, triggered by the vowels /i ɨ u/. [+high] dominance is rare in the world's languages (Archangeli & Pulleyblank Reference Archangeli, Pulleyblank and de Lacy2007), but the height harmony in Kera cannot be analysed as being the spread of [−high]. (See the non-shaded boxes in Table 2, where height harmony processes occur. Note that the only cases where there is no harmony are in the shaded boxes in the first column, where the input for both root and suffix is [−high]. Clearly, [−high] does not spread.)

Table 2 Inalienable nouns with −i, −u and −a possessive suffixes (Pearce Reference Pearce2007b).

Note: Inalienable nouns are nouns which carry possessive suffixes. They normally refer to body parts or family members. All other nouns stand alone with possession being marked by a separate pronoun.

(iii) Front ([–back]) and rounding ([+round]) harmony from the suffix to the central vowel of the root:

Rounding with height harmony:

• sar-u → suˊruˊu ‘by him’ (see transcribed text – 〈suruŋ〉 – definite article added)

Syllable structure

Vowel length is contrastive and the structure of the syllable is as follows: (C)V, (C)Vː, (C)VC.Footnote ⁴

Kera is weight-sensitive and in lexical words, strings of two light syllables are avoided in the surface form. Where two light syllables come together in the underlying form, either in the lexicon or as a result of morphophonemics, the vowel of the second syllable will either be lengthened (in phrase-final position) or deleted (mid-phrase, when not a suffix) so that the surface form has no examples of a CVCV string. This is true word-internally and across words with the exception of short function words. The transcribed passage includes the following examples:

The motivation for these processes appears to be a metrical parse at phrase level where light–heavy or heavy syllable sequences are formed by this process. As in some other Chadic languages, CV syllables are permitted to remain at the end of words without being lengthened.Footnote ⁵ Pearce (Reference Pearce, Remijsen and van Heuven2006a, Reference Pearce2007a, Reference Pearceb) claims that these processes demonstrate foot structure and iambicity, and that there is a strong relationship between the metrical structure and tone and vowel harmony.

Tone

In nouns, there are seven possible melodies. In verbs, there are only three possible melodies. The verb root has a restricted CV structure which can normally only carry one tone. There is an apparent relationship between voicing and tone, which is discussed below. The tonal melodies on nouns with mainly sonorants are as follows:

The tonal melodies on nouns with obstruents are as follows:

The relationship between voicing and tone applies equally to verbs, which have the surface melodies H, L, and LH in the citation form.

Most of the tone is lexical, but there is also some grammatical tone, most notably to distinguish between perfective and imperfective, though there are also segmental changes between these two forms. Some pronouns are also distinguished by tone.

There appears to be a maximum of two tones per mono-morphemic word, but this restriction may simply be a reflection of word length together with the fact that in words with three syllables or more, only heavy syllables can carry tone (Pearce Reference Pearce, Remijsen and van Heuven2006a). Syllables which are not specified for tone gain their pitch from interpolation.

As with many Chadic languages, there is a likely history of tonogenesis where voiced obstruents (known as depressor consonants) have lowered the pitch of the following syllable eventually developing into two tones.Footnote ⁶ At some point a further tone developed for Kera and at the same time the original voicing contrast was largely lost, as has happened in Musey (Shryock Reference Shryock1995). This is partly conjecture, but the process (though not an agreed time scale) has been affirmed for other Chadic languages.

The above examples show the tone and voicing relationship for male speakers who now live in an urban setting. In the Kera spoken by village women (and village men who are less influenced by French), there are three contrastive tones and no voicing contrast. In the speech varieties where French contact appears to play a role (in urban settings and among young males in the village), a voicing contrast has been reintroduced and the two lowest tones are pronounced at a similar or identical pitch. The distinction between words in these two groups is preserved with the voicing contrast.Footnote ⁷ For men who have moved to town from a village location, the L and M contrast is marked by both a difference in pitch and a difference in VOT. More details, including the sociolinguistic implications of these speech varieties and the results of production and perception tests, are given in Pearce (Reference Pearce2007b, Reference Pearce2009). A discussion of the relationship between VOT and voicing in Kera is also given in Pearce (Reference Pearce2007b). All of the words given in this Illustration as examples of consonants and vowels fit the description above, given that they involve voiced obstruents in syllables with L tone and voiceless obstruents in syllables with M or H tone. A more detailed investigation would show that the VOT of obstruents in M tone syllables is generally shorter than those in H tone syllables. In Figure 2, the voiced obstruent is clearly phonetically voiced and the F0 is low, whereas the voiceless obstruent is aspirated and the F0 is high. The spectrograms in Figure 2 are typical for this speech variety (although for many speakers the voicing is not true voicing, but rather low positive VOT).

Figure 2 Links between VOT and pitch.

The data in Figure 3 are for illustration only and includes the pronunciation of three words by two distinct speakers. Due to the recording conditions, there appears to be a voicing bar where I claim there should not be one, but the audio track confirms that these claims are in fact reasonable. In any case, the conclusions of this Illustration do not depend on these particular examples. Pearce (Reference Pearce2007b) gives details on the VOT measurements and significance. There is no significant difference between the VOT results in village women, and all three tones can co-occur with an onset VOT ranging from about 0 ms to about 50 ms. Town men do sometimes voice the onsets for L tone words, particularly if over-articulating for recordings, but the usual pattern in natural speech is that the VOT has a low positive value.

Note: Town man's Low voiceless value (137) is a little higher than expected, probably because the preceding tone was high, but the tone on the second syllable is as expected.

Figure 3 Comparison of two speech varieties.

Orthography issues

The orthography has been established by the Kera language committee in conjunction with SIL and with the approval of the Chadian Ministry of Education. It does not reflect the phonology perfectly, but is unlikely to change as a growing number of books have been produced with the present orthography and it appears to cause no problems to readers. The discrepancies with the phonology are as follows:

• • [ә] and /ɨ/ are both written as 〈ә〉.

• • The phonemes /ɔ/ and /ɛ/ are always written as 〈o〉 and 〈e〉, regardless of which allophone is used whereas the two allophones of the phoneme /a/ are written separately as 〈a〉 and 〈ә〉.

• • Technically, for the speech varieties where there is little contact with French, the voiced and voiceless obstruents are not contrastive and so should be written with one symbol, but the Kera orthography includes both voiced and voiceless obstruents. In practice, if the following tone is high or mid, the onset obstruent will be written as voiceless, whereas if the tone is low, the onset will be written as voiced. This means that, in general, tone does not need to be marked in an overt manner, because the consonants are already giving tonal information. For speech varieties with more influence from French, the voicing contrast in the orthography reflects a phonological contrast.

In certain words there is free variation between two tonal melodies; for example, puukul ‘type of vegetable’, cited above with the HM melody, can also have an HL melody, with no change in meaning, but in this case speakers insist on the spelling 〈puugul〉 (vs. 〈puukul〉 for the HM melody version). In the case of minimal pairs which are only distinguishable by tone (H vs. M or where there are no obstruents to mark L), the word with high tone is marked with an acute accent. There are around twenty words in the Kera lexicon which need to be marked in this way and the reader generally learns the word's spelling with the accent without necessarily realizing the meaning of the accent.

Other conventions in the Kera orthography include the affricates written as 〈c〉 and 〈j〉, which is common in Chadic languages in general. The approximant /j/ is written as 〈y〉. Nasalized vowels are written as 〈a˜〉, 〈ɔ˜〉, etc. Long vowels are written with a double letter, as 〈aa〉, 〈ii〉, etc. The medial glottal stop is written as an apostrophe, but the glottal stop before vowel-initial words is not reflected in the orthography. The best way of writing the labiodental flap is still under discussion, with suggestions including 〈vb〉, 〈wh〉 and the IPA symbol 〈〉.

‘The North Wind and the Sun’