Rangi Vowel System: Five or Seven?

Rangi Vowel System: Five or Seven? Oliver Stegen, SIL Tanzania ([email protected]) Paper presented at the 30th Colloquium of African Languages and Linguistics at Leiden University, August 28-30, 2000 1. Motivation The investigation described in this paper is part of the Rangi Bible translation project under the auspices of the Diocese of Central Tanganyika, Dodoma, Tanzania1. Rangi is a Bantu language, F.33 according to Guthrie’s classification. It is spoken as a first language by more than 300,000 people in the Kondoa District of Northern Central Tanzania. Prior to the start of the project, no serious efforts had been undertaken to establish an orthography for the Rangi language. When investigating the phoneme inventory of Rangi in order to come up with a workable writing system, one area of ambiguity was the vowel system. Seidel (1898) and Akhavan (1990) write five phonemic vowels while Dempwolff (1916), Tucker & Bryan (1957), and Kesby (1981) propose seven2. Moreover, among the seven vowel system advocates, Dempwolff writes [i, , , a, , , u]3, and Kesby [i, e, , a, , o, u]. Hyman (1999) on his map showing 7 vs. 5 vowel systems in Bantu languages, leaves Rangi unspecified. Of those linguists engaged in current research on Rangi, Yukawa (1989) analyzes it with seven phonemic vowels while Dunham (p.c.) advocates five. Finally, among the thirty or so Rangi mother tongue speakers who met in 1998 to establish a Rangi orthography, some wanted to write five vowels (as in Swahili, Tanzania’s national language) while others favoured seven vowels4. The consensus of the 1998 orthography meeting was to risk over-differentiation by writing seven vowels prelimirarily (using <i> and <u> for the degree 2 vowels) until further investigation would point in one of these two directions more decidedly. The participatory research approach was then believed to be a suitable methodology for such investigation. The main questions to be answered were: • Does Rangi have five or seven vowel phonemes? • In the case of a seven vowel system, is it of the / or of the e/o type? • What kind of feature spreading, if any, occurs? The corresponding questions for the Rangi orthography would be: • Should Rangi be written with five or seven vowel letters? • If seven vowels emerge, are the degree 2 vowels to be written based on i/u or e/o? • In cases of feature spreading, should the base form or the surface form be written? 2. Methodology The participatory research approach to phonological analysis has been developed and described in detail by Kutsch Lojenga (1996). Its main aim is “to raise the awareness of the native speakers for the sound and tone system of their language” (p.13). It is believed that native speaker input and insights “will not only speed up the analysis but also contribute to its quality” (p.15). The main steps of this approach are: • let native speakers transcribe the words of their mother tongue, • separate data according to word class as well as root and syllable structure, • let native speakers eliminate orthographic underspecification by checking the pronounciation of all sounds in the same position in the root structure, which are written with the same letter. 1 I am indebted especially to Saada Saidi and Samweli Ramadhani for their help with the Rangi language, and to Constance Kutsch Lojenga and Roderic Casali for comments on this paper. My involvement in the project was made possible under work permit C 0042314. 2 It should be noted that Akhavan’s data was collected by Paul Berger in 1935, and Kesby collected his data between 1963 and 1966. Also, Dempwolff and Akhavan/Berger both transcribe nine phonetic vowels, yet arrive at different conclusions as to their phonemic status. 3 Square brackets [] will be used for both phonetic transcription and phonological features, while pointed brackets < > will be used for orthographic symbols. 4 The sociolinguistic reasons for these two differing viewpoints cannot be discussed here. However, it could be ascertained that dialectal differences did not play a role in it. Rangi Vowel System - 1 - For this particular investigation on Rangi, the first step was omitted as a 2500 entry computerized dictionary already existed. Its entries had been spelled according to the Swahili alphabet (i.e. five vowels only) and subsequently checked with two Rangi speakers. It was decided to limit the scope of this stage of checking Rangi vowel quality to bisyllabic noun and monosyllabic verb stems regardless of vowel length. The main corpus were 707 -CV(V)CV noun stems and 213 -CV(V)C- verb stems. These were written on individual cards in order to enable several possibilities of sorting. Swahili glosses were included on the cards so as to avoid confusion of minimal pairs. Vowel quality for all positions in nouns (including V1 position in locatives) and for the stem position in verbs (including perfective forms) was then checked by one Rangi speaker, with double checking of V1 and V2 positions in nouns by another Rangi speaker5. For each vowel letter, the Rangi speakers were asked to state only whether its pronounciation was the same or different as compared to other items of the same group. Smaller parts of the corpus were checked by other speakers6. All of their responses were charted accordingly. 3. Results and Discussion First, it should be noted that the agreement between different speakers with regard to which words belong into which vowel quality group was always above 90%7.That means that for the vast majority of the words, Rangi speakers from different dialects agreed on the vowel quality of all vowels. It had been assumed that the stem vowels of verbs in the imperative would be least subject to possible harmonizing processes. Both high and mid vowels fell into two different groups, however, giving a total of nine phonetic vowels8. The more closed variant of the mid vowels occured only in causative verb forms9 which had not been removed from the supposedly monomorphemic sample. This included frozen forms like [ eia] ‘sleep’ and [k or a] ‘light (a fire)’ which do no longer exist in a simple, i.e. non-causative form. The distribution can be seen in Table 1, with examples given in (1). Table 1: -CV(V)Ca verb stems written number of distribution vowel groups i 2 [i, ] contrastive in analogous environment e 2 [e, ] more closed variant only before causative suffix -i- a 1 [a] o 2 [o, ] more closed variant only before causative suffix -i- u 2 [u, ] contrastive in analogous environment Examples (1): written more open variant gloss more closed variant gloss a. i f t a condole fi sa stir b. e r ra stagger re r a make stagger c. a l a la fall ill 5 These are Saada Saidi of Mnenya village, born in 1968, and Samweli Ramadhani of Kalamba village, born in 1971/72, both of whom have never lived outside of Kondoa District. They represent the Kolo and Haubi dialects of Rangi respectively. The other two dialects, Kondoa and Mondo, are represented through Mama Bahati and Andrew Lujuo respectively, who checked large parts of the computerized dictionary. It should be noted that vowel length and tone were checked too, yet are not a part of this investigation. 6 These were Rajabu Isangu of Mnenya village (Kolo dialect) for -CiCi, -CiCu, -CeCu, and -CuCu nouns, Pascal Ntomola of Kiteo village (Haubi dialect) for -CiCi nouns, Richard Kirimwi of Sakami village (dialect unknown) for -CiC- verbs, and Dr.Ikaji of Kondoa town (Kondoa dialect) for noun class prefixes of classes 1, 4, 7 and 11. 7 The exact figures of agreement on the word level are between Saada Saidi and Samweli Ramadhani 90,3% (393 of 435), both of these with Rajabu Isangu 91,3% (115 of 126), all three of these with Pascal Ntomola 95% (19 of 20), between Saada Saidi and Richard Kirimwi 97,7% (42 of 43), and between Saada Saidi and Dr.Ikaji 94,3% (133 of 141). The figures for Rajabu Isangu and Pascal Ntomola show that the disagreement between Saada Saidi and Samweli Ramadhani lie outside of the sample common to all three or four respectively. A number of instances of disagreement could possibly be explained by gradient assimilation. 8 Not all Rangi speakers were able to distinguish / versus e/o, probably due to their allophonic relation. 9 The causative suffix -i- is usually desyllabified, leading to palatalization of the preceding consonant. Rangi Vowel System - 2 -  d. o w na snore hon a miss (a target) e. u t la rest t u ba follow Stem vowels in nouns in V1 position showed a similar distribution as charted in table 2, with examples in (2). This gave rise to the assumption that high vowels spread either their [+high] or [+ATR] feature leftward. This observation will be discussed later. Table 2: V1 position in -CV(V)CV noun stems written number of distribution vowel groups i 2 [i, ] contrastive in analogous environment e 2 [e, ] more closed variant only before more closed variants of -i and -u a 1 [a] o 2 [o, ] more closed variant only before more closed variants of -i and -u u 2 [u, ] contrastive in analogous environment Examples (2)10: written more open tone gloss more tone gloss variant closed variant a. i ir ra llh cow path iriva lll lake b. e m r sa lll club mure d"i lll rope c. a m a sa lh tree sp. m a si rh Alagwa man d. o m k d" lfl urine #k o d"i hh bird sp. e. u s ha lh calabash sp. musuk a llh broken hoe It was the V2 position of noun stems which proved to be the neutral position, i.e. not being influenced by harmonizing processes. Its distribution can be found in table 3, and contrastive examples are given in (3)11. Table 3: V2 position in -CV(V)CV noun stems written number of distribution vowel groups i 2 [i, ] contrastive in analogous environment e 1[ ] a 1 [a] o 1[ ] u 2 [u, ] contrastive in analogous environment Examples (3)12: a. p a ri lll inheritance it a #p i lfl branch b. war ll porridge va #b lhh chaff c. isa r lfl word m a #% llh bird sp. d. k t ara llh batchelor’s hut m sa #p a lfl ewe e. l d&"al lll rafter m la #b lll trough f. #d&"at ? sheath m wa #% lhh tree species g. l varu lll rib 10 Tone, which is given in its phrase-medial surface form, is abbreviated as l(ow), h(igh), f(alling from high to low), and r(ising from low to high). 11 For a full paradigm, see appendix B. 12 ‘?’ in the tone column indicates either disagreement or that the tone was not elicited for this item. Rangi Vowel System - 3 -  l t a #bu lll pumpkin plant Additionally, a co-occurence chart of vowels in V1 versus V2 position as in table 4 consolidated a 7 vowel claim over against a five vowel system. Table 4: Co-occurence of V1 and V2 in -CV(V)CV nouns V2 i /e a /o u V1 i X X X X X X (X)? X X X X /e X X X X X X a X X X X X X X /o X X X X X (X)? X X X X X (X)? u X X X X X Uncertain instances of co-occurences of degree 1 and degree 2 vowels, which pointed towards gradient [ATR] spreading, are given in (4). Examples 413: a. m r 'mi lll ‘farmer’ b. m 'k i ll ‘smoke’ c. v 'du hh ‘dikdik’ [ , ] and [e, o] proved to be more easily distinguished by mother tongue speakers than by non- native researchers14. Nevertheless, the fact that they could be distinguished, and that [e, o] can be interpreted as allophones of [ , ] in following [i, u] environment, points towards a seven vowel system [i, , , a, , , u]15. It can now be supposed that [i, u] spread the feature [+ATR], rather than [+high]. Rightward [+ATR] spreading is ruled out by nouns like [lusi #d ] ‘grass species’ or [(iud ] ‘bird species’ as well as by verbal extensions like [t&)u #% ra] ‘tie to s.th.’ and [t&)u #% la] ‘untie’ which had been elicited outside of this investigation. Remains the question whether the assimilation of degree 2 vowels in a [+ATR] spreading domain is total or partial. Or phrased as its potential orthographic implication: Should <i, u> before <i, u> be written <i, u> or not? The two environments investigated were the locative suffix -íi for nouns and the perfective suffix -ire for verbs. For both, as shown in tables 5 and 6, the results were not conclusive. Rangi speakers declared stem vowels of different base nouns to be auditorily the same, yet insisted on their being different. That examples as in (5) could be distinguished might, of course, stem from native speaker knowledge about the respective base forms. Table 5: -CV(V)Cíi locatives (nouns) written number of groups distribution vowel i 1 [i] or 2 [i, i]? more closed variant than [ ] but still distinguishable? e 1 [e] only more closed variant a 1 [a] o 1 [o] only more closed variant u 1 [u] or 2 [u, u]? more closed variant than [ ] but still distinguishable? 13 All examples were infered from disagreement between Rangi speakers on their vowel quality, and for example a) from the fact that it is derived from [r ma] ‘to hoe’; cf. also the discussion of intermediate vowel quality for noun class prefixes. 14 Hyman (1999) gives a few instances of possible “confusion over the phonetics and phonology of [...] Bantu vowel systems”, which non-native speakers might be more prone to than those with mother tongue intuition. 15 The co-occurence of the seven vowel [i, , , a, , , u] system with [+ATR] spreading is in agreement with the observation of “ATR predictability” by Casali (1998). Rangi Vowel System - 4 - Table 6: -CV(V)Cire perfectives (verbs) written number of groups distribution vowel i 1 [i] or 2 [i, i]? more closed variant than [ ] but still distinguishable? e 1 [e] only more closed variant a 1 [a] o 1 [o] only more closed variant u 1 [u] or 2 [u, u]? more closed variant than [ ] but still distinguishable? Examples 516: a. m k s ? groin muk i s i ? in the groin b. ri k fl kitchen ri k i hf in the kitchen c. idu k a lfl shop idu k i lhf in the shop d. ib va lrh stalk ibu vi lrf in the stalk e. s #d&"a rl slaughter asi #d&"ir hhll he has slaughtered f. si *ba rl dig asi *bir hhll he has dug g. k la lh grow ak urir hhll he has grown h. k uva lh strangle ak uvir hhll he has strangled However, for some of the noun class prefixes, clearly three different vowel qualities were identified. [+ATR] spreading can happen gradiently as mentioned by Casali (2000), “both in the sense that the assimilation process frequently gives rise to vowels which are somewhat intermediate in quality between that of an underlying [-ATR] vowel and its [+ATR] counterpart, and in the sense that vowels which are further to the left of the underlying [+ATR] syllable (often referred to as the ‘trigger’) from which spreading occurs are likely to assimilate less fully”. The former sense could apply to the above examples of locatives and perfectives, while the latter sense might explain the results given in table 7 for noun class prefixes. Table 7: noun class prefixes noun written number of distribution class form groups 1 mu- 2 more closed variant only before more closed variants [mu, m of -i- and -u- ] 3 mu- 3 most closed variant only before more closed variants [mu, mu of -i- and -u-; “in-between” variant mostly before more ,m ] closed variants of -e- and -o- 4 mi- 3 most closed variant only before more closed variants [mi, mi+, of -i- and -u-; “in-between” variant mostly before more m] closed variants of -e- and -o- 5 i- 1 [i] 7 ki- 3 most closed variant only before more closed variants [k i, k i+, of -i- and -u-; “in-between” variant mostly before more k ] closed variants of -e- and -o- 8 vi- 1 [vi] 11 lu- 2 more closed variant only before more closed variants [lu, l ] of -i- and -u- 14 u- 2 [u, ] more closed variant only before more closed variants of -i- and -u- First, it can be noted that classes 5 and 8 show no variation as their underlying form contains the vowel [i], bearing the dominant feature [+ATR]. Then, the identification of only two vowel qualities for noun classes 1, 11 and 14 is due to the fact that these prefixes were checked first, yet 16 Of rounded vowels in V2 position, the feature [+round] is retained in the locative form as labialization of the preceding consonant. Rangi Vowel System - 5 - the speakers only later assigned an intermediate vowel quality. Disagreement between Saada Saidi and Dr.Ikaji on items like [mut ere,k i] ‘cook’, [muv - ri] ‘parent’, [k i+re,ri] ‘tail feather’, and [luwo,vij ] ‘tree species’, which one groups with [ , ] and the other with [i, u] and vice versa, could be due to the ambiguous status of the prefix vowel. The gradient nature of [+ATR] spread makes clear that the underlying forms of prefixes for noun classes 1, 3, 4, 7, 11, and 14 bear degree 2 vowels [ , ]17. Finally, some remarks by Samweli Ramadhani have to be mentioned. He reported that some Rangi speakers would write all instances of degree 2 vowel [ ] as <e>. Yet others would separate the group of degree 2 vowel [ ] into those to be written <i> versus those to be written <e>. The first observation could indicate a potential shift from a 7 vowel [i, e, , a, , o, u] system to a [i, , , a, , , u] one, i.e. [ , ] [e, o], or vice versa, as hypothesized by Hyman (1999, ex.16 a and c). The second observation could indicate a potential shift from a 7 vowel to a 9 vowel system18. Both would have severe implications for a Rangi orthography, especially as none of them involve a simplification towards a five vowel system. 4. Conclusion The investigation has consolidated the following facts: • Rangi has a 7 vowel system of the [i, , , a, , , u] type. • [+ATR] is spread leftward, resulting in the +ATR variants [i, u] of [ , ], and the [e, o] allophones of [ , ]. • [+ATR] spreading is gradient, resulting in intermediate vowel qualities between [i, u] and [ , ], respectively. For Rangi orthography, it is suggested to confirm the seven vowel letters <i, i, e, a, o, u, u>. Unless wide-spread evidence can be found for a vowel shift [ , ] to [e, o], basing the letters for the degree 2 vowels on the graphemes <e, o> could result in confusion with the [+ATR] allophones of [ , ]. It is further suggested to write the underlying form of degree 2 vowels as <i, u> regardless of surface realizations which result from [+ATR] spread. That way, the issue of gradient assimilation to [+ATR] is avoided. In light of potential vowel shifts, further investigation is strongly recommended. Such investigation should include instrumental studies as well as expand the sampled data to systematically include instances of vowel coalescence19 and all possible combinations of verbal extensions20. It should be borne in mind that the influence of language shift on orthography is not a one-way street. 17 For examples, see appendix C. 18 A few Bantu languages seem to have undergone such a development; for an overview, see Casali (2000). 19 Casali (2000) explains how vowel coalescence can shed light on vowel systems and [ATR] harmony. 20 Hyman (1999) uses verbal extensions to exhibit height harmony symmetries and assymetries, and explains its implications for the development of Bantu vowel systems from ProtoBantu which had seven vowels. Rangi Vowel System - 6 - Appendix A Verbs for short and long vowels of all qualities21 stem vowel gloss imperative perfective causative [i] descend k i ma ak i mir k ima [i] go out (of fire) rima arimir rim a [] [i] be long l ha ali hir li h a [] [i] hoe r ma arimir [ ] [e] go j #da aje #dir je #d a [ ] [e] be full m ma amemir mem a [a] refuse k a na ak a nir k ana [a] know ma a ama ir [ ] [o] boil (intr.) t t a at o t ir t ot a [ ] [o] cool down h la ahorir hor a [ ] [u] talk l sa alu sir [ ] [u] come out f ma afumir fum a [u] wake up (intr.) vu k a avu k ir vu )a [u] hit t&)ula at&)urir t&)ur a Appendix B Noun stems stem vowels gloss Rangi word tone [i, i] rope ludihi lhh [, ] length l h lhh [i, ] cooking stick lufir llh [i, a] lazy person muvira lll [ , a] path #d&" ra ll [i, ] tree sp. muk ir lll [, ] payment ir h lll [i, u] tree sp. muviru lhh [i, ] day sik ll [, ] calf #d&" k hh [e, i] sultan m 't emi lll [ , ] lizard sp. m h lhh [ , a] gazelle sp. #d&" ra hh [ , ] wind *p h hh [e, u] eagle *bewu ll [ , ] world w r ll [a, i] butter churn ba hi lh [a, ] zebra #d&"aj lh [a, ] friend mar lh [a, a] hunger #d&"ala ll [a, ] rafter l d&"al lll [a, u] python sat u hh [a, ] sheath #d&"at ? [o, i] shame soni ll [ , ] millet sp. m h n lll [ , a] feather l d&" ja llh [ , ] clan l k l lhh 21 Tone on imperatives is low-high rise followed by low tone for CVVCa stems, and low tone phrase-finally or low tone followed by high tone elsewhere for CVCa stems. Causative forms are also given in the imperative. Tone on perfectives (all 3sg) is two high tones followed by two low tones. Blank causative fields indicate that no causative in that category could be found. Rangi Vowel System - 7 -  [o, u] rubbish k osu hh [u, i] dog k uri hh [ , i] river l ' d&"i ll [ , ] dust ir r llh [u, ] bird sp. iwud ? [ , ] guinea fowl *b w ll [u, a] ritual medicine masula ? [ , a] bark, shell is la llh [u, ] dialogue *bu w hh [ , ] fly sp. *b /% hh [u, u] strength /%uru ll [ , u] dikdik v 'du hh [ , ] testicle #d l ll Appendix C Noun prefixes noun prefix gloss Rangi word tone class 1 m - poor person m k va lhh dependant m r r a llh old woman m dala llh fool m h h llh father-in-law m k l lhh mu- child musi/%a lll younger mudu di lhh sibling guest mujeni lll bridegroom mulo li lfl blind person muhok u lll 3 m - work m rm lll snake sp. m lal llh mouth m l m lll God m l /% lll mu- slope mu%irit ? whistle muluri lhh mu- rope mure d&"i lll building stick muli+/%o t i ? 4 m- hips mk s ? bread mk at lfl hands mk n lhh trees sp. mk j lhh mi- trees sp. mik ir lll chaff mik u /k u lhh mi+- trees sp. mi+be ruma lllh trees sp. mi+sek a ? 5 i- log ik i ni lhh sugarcane tip it #d ? tree leaf ibe ri llh dragon fly in h lhh rat sp. ifud a lhh bee bread il va lll Rangi Vowel System - 8 -  7 k- promise k lah ? tripod k t /% r lhll wound k l #da lll onion k t /% r llfl ki- bead k it&)i /% lll stepping gate k it ut u l lhfl ki+- counsel k ilo /%i lll snake sp. k i+t e /k u ? 8 vi- bald heads vip ara lhh chests vip *b lhh lamentations vir r lll elbows vik k l lhll insects vik o k i llh epidemics vik /% llh pipes vip u #d ? 11 l - cheek l saja llh laughter l s k lll bird sp. l v h lhh leather strip l k va lhh bean sp. l k sa lll lu- story lusim lhh shard luwi j lll tree sp. luwo#du ? knife luf lh calabash sp. luk uru lhh 14 - poverty k va lhh skin rash h r lll fierceness k ari lhh leprosy k ma lfl greatness k l lhh u- fibre uk iri lhh truth uk e ri lfl shortness uk ufi lhh References Akhavan, F. 1990 “Untersuchungen zur Grammatik des Irangi anhand des Materials aus dem Nachlass Dr.Paul Berger”, MA thesis, University of Hamburg Casali, R. 1998 “Predicting ATR Activity”, Chicago Linguistic Society 34/1, 55-68 Casali, R. 2000 “An Introduction to ATR Vowel Harmony in African Languages”, manuscript Dempwolff, O. 1916 “Beiträge zur Kenntnis der Sprachen in Deutsch-Ostafrika - Irangi”, Zeitschrift für Kolonial-Sprachen 6, 102-23, Berlin Hyman, L. 1999 “The Historical Interpretation of Vowel Harmony in Bantu” in: Hombert & Hyman (eds.) Recent Advances in Bantu Historical Linguistics, Stanford: C.S.L.I. Kesby, J. 1981 “The Rangi of Tanzania: An introduction to their culture”, New Haven: HRAF Kutsch Lojenga, C. 1996 “Participatory research in linguistics”, Notes on Linguistics 73, 13-27, Dallas: SIL Seidel, A. 1898 “Grammatik der Sprache von Irangi” in: Werther, W. (ed.) Die mittleren Hochländer des nördlichen Deutsch-Ost-Afrika, 387-434, Berlin Tucker, A. & M. Bryan 1957 “Linguistic Survey of the Northern Bantu Borderland”, London: OUP Yukawa, Y. 1989 “The verb in Bantu languages (3) - Langi” (in Japanese), ILCAA Newsletter 65, 15-23 Rangi Vowel System - 9 -