Vowel Fronting in Ulumanda' (South Sulawesi)

[WORKING PAPER IN PROGRESS, PLEASE CITE BY URL AND DATE]
First posted: 15 Sep 2014
Last revised: 18 May 2018

Ulumanda' belongs to the Northern branch of the South Sulawesi languages and is spoken by approximately 30.000 speakers in the interior of West Sulawesi province, which latter largely coincides with the historical Mandar region.

The existence of Ulumanda' as a distinct language was long overlooked in the history of linguistic studies on Sulawesi languages. Van der Veen (1929) included Ulumanda' lects as dialects of Pitu Ulunna Salu, whereas Grimes & Grimes (1987) list several Ulumanda' lects as dialects of the Mamuju language. First explicit mention of the Ulumanda' language in a wider-distributed publication was made by Friberg and Laskowske (1989).

A highly distinctive feature of Ulumanda' is its eight-vowel system that includes the rounded front vowels ö [ø] and ü [y], which is very unusual among the languages of Sulawesi. Most commonly on Sulawesi, we find a simple five vowel system, as in all Celebic languages, Gorontalo-Mongondic, Makassaric, and most languages of the Northern branch of South Sulawesi. In Minahasan and Sangiric, and in Buginese, the five vowel system is expanded by the mid/high central vowel /ə/. Historically, the Ulumanda' eight-vowel system developed through regular sound change from a simple five-vowel system as found in related Mamuju or Toraja.

The data presented here is from the Kabira'an dialect of Ulumanda', which was collected from a single informant, Abdul Khaliq, who was 20 years old at the time when I elicited the data.

The synchronic vowel system of Ulumanda'

The structure of the Ulumanda' vowel system is largely identical to the one found in Finnish. It is composed of the three back/front pairs /a/:/ä/, /o/:/ö/, /u/:/ü/, and the two "neutral" vowels /e/ and /i/. The distiction between the back/front pairs is contrastive only in final closed root syllables: _[ŋ/ʔ]#.

appaʔ
uraŋ
linoʔ
soroŋ
kuttuʔ

('four')
('person')
('earthquake')
('bride price')
('knee')

vs. täppäʔ
vs. üräŋ
vs. mänöʔ
vs. möröŋ
vs. tändüʔ

('summit')
('shrimp')
('chicken')
('swim')
('horn')

The phonetic realization of the low pair /a/:/ä/ is [ɑ]:[æ]. Note that speakers of Ulumanda' and of neighboring lects which have the same contrast, use the back vowel [ɑ] also when speaking Indonesian as realization of the vowel /a/, which further contributes to the "foreign" impression that the Ulumanda' accent gives to outsiders.

In open final syllables, the back/front contrast is neutralized, and phonetically only the back variant occurs here. Phonemically, we can represent the vowels by distinct neutral archiphonemes. It is preferable though, to equate them with the back phonemes, matching the phonetic realization.

In non-final syllables, the pairs a:ä, o:ö, and u:ü are allophones of the archiphonemes /a/, /o/, and /u/. Here, fronting is a result of regular vowel harmony. By default, the back variant is found, while the fronted variant appears if the vowel of a contiguous syllable is fronted, i.e. ä, ö, or ü. This rule operates to both sides. To the left, it affects root vowels and prefixes, e.g. mäʔ‑bälüʔ /maʔ-balüʔ/ ('sell'), mälöäŋ /maloäŋ/ ('wide'), üräŋ /uräŋ/ ('shrimp'), whereas to the right, the suffix /-aŋ/ is affected by vowel harmony: bälüʔ‑äŋ /balüʔ‑aŋ/ ('sell'). The neutral vowels /i/ or /e/ block fronting, e.g. kaiäŋ ('big'), mamindöŋ ('run'), with the back variant a appearing in the initial syllable, inspite of /ä/ in the final syllable.

The unusual vowel system of Ulumanda' was recognized already by van der Veen, where data samples from several Ulumanda' lects are given. The Tubbi-Taramanu' dialect fully agrees with the Kabira'an dialect, as can be seen from examples such as lötöŋ ('black'), mänöʔ ('chicken'), bännäŋ ('thread'). The Besoangin dialect shows loss of *ŋ following fronted vowels, e.g. lötö, bännä.


The historical source of fronting in Ulumanda'

The eight-vowel system in Ulumanda' developed out of an earlier simple five-vowel system as a result of a fully regular and transparent sound change. In order to understand this sound change, we have to take a look at the inventory of final consonants in the low-level parent language of Ulumanda', viz. Proto Northern South Sulawesi.

PNSSul
Duri
Toraja
Ulumanda'
Mandar

*-n
-n
-n
-ŋ
-ŋ

*-ŋ
-ŋ
-ŋ
-ŋ
-ŋ

*-t
-ʔ
-ʔ
-ʔ
-ʔ

*-k
-k
-k
-ʔ
-ʔ

*-s
-h
-ʔ
-ʔ
-s

*-r
-h
-ʔ
-ʔ
-r

As we can see, Ulumanda' has gone furthest in reducing the original inventory from six to two.

The reduction of final consonant contrasts was compensated however by the phonemic split of the vowels /a/, /o/, /u/ that led to the present-day eight-vowel system. Originally, fronting was a non-phonemic sound change that affected back vowels preceding the velar consonants *k and *ŋ. In all cases where a Ulumanda' root has fronted vowel in the ultimate syllable, the cognate root in Toraja has a velar final consonant. With the loss of the contrast between *n and *ŋ, and between *k and *ʔ, the split between back and fronted vowels became phonemicized:

*-an
*-aŋ

*-aʔ
*-ak

> [-an] /-an/
> [-äŋ] /-aŋ/

> [-aʔ] /-aʔ/
> [-äk] /-ak/

> /-aŋ/
> /-äŋ/

> /-aʔ/
> /-äʔ/

*-on
*-oŋ

*-oʔ
*-ok

> [-on] /-on/
> [-öŋ] /-oŋ/

> [-oʔ] /-oʔ/
> [-ök] /-ok/

> /-oŋ/
> /-öŋ/

> /-oʔ/
> /-öʔ/

*-un
*-uŋ

*-uʔ
*-uk

> [-un] /-un/
> ???

> [-uʔ] /-uʔ/
> [-ük] /-uk/

> /-uŋ/
> ???

> /-uʔ/
> /-üʔ/

The reflex of *-uŋ is not documented in the data I have collected, the expected reflex in Ulumanda' would be /-üŋ/.

PNSSul
*anak
*kayok
*tanduk
*manaraŋ
*malotoŋ

Toraja
anak
kayok
tanduk
manaraŋ
malotoŋ

Ulumanda'
änäʔ
käjöʔ
tändüʔ
mänäräŋ
mälötöŋ

'child'
'scratch'
'horn'
'clever'
'black'

In the following discussion, I will refer to vowel fronting triggered by a following final velar consonant simply as "pre-velar fronting".

Pre-velar fronting also occurred in non-final position, with syllable closing ŋ as in päŋŋäŋ from *paŋŋan. Here, the penultimate vowel was fronted, with subsequent vowel harmony affecting the ultimate vowel as well. Geminate medial ŋ is the only instance of syllable-closing ŋ in Ulumanda', since in the cluster *-ŋk-, *ŋ underwent total assimilation to the following stop, e.g. takke < *taŋke.

Pre-velar fronting in related lects

Aralle-Tabulahan

Aralle-Tabulahan is the closest relative of Ulumanda' within Northern South Sulawesi. It agrees with Ulumanda' in most of the vocabulary and functors; likewise, it shares with Ulumanda' the strong Seko substratum, which sets Ulumanda'-Aralle-Tabulahan apart from the remaining Northern South Sulawesi languages.

Aralle-Tabulahan displays the same pattern of pre-velar fronting in root final syllables. However, the pre-velar reflexes of earlier *o and *u are not just fronted, but further also unrounded, which lead to merger with /e/ and /i/.

Toraja
tanduk
malotoŋ
anak

Ulumanda'
tändüʔ
mälötöŋ
änäʔ

Aralle-Tabulahan
tändiʔ
maloteŋ
änäʔ

'horn'
'black'
'child'

We therefore find a six-vowel system in Aralle-Tabulahan with /ä/ occurring next to five cardinal vowels.

In non-final syllables, Aralle-Tabulahan shows significant differences from Ulumanda'. First, vowel harmony only affects earlier *a, as can be seen in änäʔ and tändiʔ, whereas *o and *u remained unchanged, e.g. maloteŋ. Secondly, with the unrounding of fronted *o and *u, the fronting of *a in non-ultimate position became phonemic, as exemplified by the near-minimal pair tändiʔ / laŋiʔ.

PUS - Bambam dialect

Bambam is one of the main dialects of Pitu Ulunna Salu (PUS). PUS is closely related to Mamasa, but also shows traces of convergence with Aralle-Tabulahan. Like Ulumanda' and Aralle-Tabulahan, Bambam has reduced the number of final consonants to two. The unified final nasal is /-n/, which has the unusual surface realization [-m]. In Bambam, only *a and *o underwent pre-velar fronting, whereas *u remained unaffected. The fronted reflexes of *a and *o have merged in Bambam into /ä/, resulting in a six-vowel system as in Aralle-Tabulahan. Syllable closing *ŋ which triggered fronting was subsequently lost.

*-aŋ, *-oŋ
*-ak, *-ok

> /-ä/
> /-äʔ/

but:

*-uŋ
*-uk

> /-um/
> /-uʔ/

PNSSul
*anak
*tondok
*manaraŋ
*malotoŋ

Toraja
anak
tondok
manaraŋ
malotoŋ

Bambam
änäʔ
tondäʔ
mänähä
malotä

'child'
'village'
'clever'
'black'

As we can see from these examples, Bambam has the same rules for vowel harmony as in Aralle-Tabulahan, only affecting *a but not *o.

In Bambam, the phonemic distiction between /a/ and /ä/ is extended to non-final syllables as the result of the shift *-aŋC- to -äC-, which is quite common since *ŋ does not assimilate to a following obstruent.

PNSSul
*aŋkat
*taŋke
*jaŋkan
*paŋŋan

Toraja
aŋkaʔ
taŋke
daŋkan
paŋŋan

Bambam
äkäʔ
täke
däkäm
päŋäm

'lift'
'branch'
'span'
'chew betel'

Due to rightward vowel harmony, Bambam developed a secondary contrast between -äm and -am, which clearly emerged at a stage after the loss of the final nasal that triggered pre-velar fronting.

Unlike in Ulumanda' and Aralle-Tabulahan, Bambam has another source for fronting *a to ä. Here, fronting also occurred in roots of the type *(C)ai(C)/*(C)ae(C), mäpäiʔ < *mapait ('bitter'), mäsäe < *masae ('long'), täi < *tai ('feces').

The merger of *a and *o to ä in pre-velar position is found in most other PUS dialects. The Kiha dialect however has e for fronted *o (Smith 1993), which indicates that we have to reconstruct a distinct vowel *ö for Proto PUS, which merged with *ä in all PUS dialects except for Kiha.

Mambi

The Mambi lect is classified in the SIL surveys (and in the Ethnologue) as a dialect of Aralle-Tabulahan. Lexically and also phonemically, it is strongly influenced by Bambam, but the most basic functors generally agree with Aralle-Tabulahan.

With regards to pre-velar fronting, Mambi superficially follows the Bambam pattern: fronted *a and *o are represented by /ä/, and the final velar nasal is lost following /ä/. Ultimate *u is represented by a single phoneme /u/. There is however evidence for an earlier split of ultimate *u as in Ulumanda' and Aralle-Tabulahan. The Mambi reflex of Proto NSSul *tanduk is tänduʔ, with a fronted /ä/ in the penultimate syllable. The fronting of penultimate *a can only be explained as the result of vowel harmony to a following fronted vowel **ü at a pre-Mambi stage, with subsequent neutralization of the /u/-/ü/-contrast:

*tanduk > **tändüʔ > tänduʔ.

This re-merger of pre-Mambi **u and **ü resulted in the phoneme /u/, which is phonetically actually not a back, but a rounded high central vowel [ʉ]. Note that it is not the current fronted articulation of [ʉ] which triggered fronting of *a in tänduʔ ([tändʉʔ]), since ultimate [ʉ] reflecting earlier **u does not affect *a in the preceding syllable, e.g. *salu > salu ([salʉ] 'river').

Mamasa - Dama' dialect

Pre-velar fronting also appears in the Dama' dialect of Mamasa, which is spoken in the neighborhood of PUS (Smith 1993). Like in all other Mamasa lects, *k and *ŋ were retained as separate phonemes, e.g. illoŋ < *illoŋ, manuk < *manuk. However, ultimate /a/ becomes fronted to ä if followed by a final velar, which latter shows the same shifts as in PUS: /k/ is realized as a glottal stop, whereas /ŋ/ is dropped. Since final /k/ and /ŋ/ are unchanged following all other vowels but /a/, we can assume that pre-velar fronting and the concomitant changes of the final velar are just surface realizations of phonemic /-ak/ (= äʔ) and /-aŋ/ (= ä).


The emergence of pre-velar fronting as an areal feature

As I have indicated above, pre-velar fronting occurs in lects which belong to different branches of Northern South Sulawesi. Ulumanda' and Aralle-Tabulahan make up a branch of their own, which presumably also includes Pannei. I will refer to this branch as Ulumanda'-Aralle-Tabulahan (UAT). Bambam on the other hand belongs to the Greater Torajan branch, which comprises PUS (incuding Bambam), Mamasa, Toraja, Kalumpang, Rongkong and Luwu'. Pre-velar fronting is an original feature of UAT, and can well be reconstructed for the common Proto UAT stage. Due to contact, pre-velar fronting spread into the westernmost lects of Greater Torajan, viz. PUS and the Dama' dialect of Mamasa.

Next to correponding forms in PUS and UAT (e.g. Bambam malotä, Ulumanda' mälötöŋ; Bambam, Ulumanda' äkäʔ), pre-velar fronting also occurs in PUS words which have no cognate in UAT, e.g. illä < *illoŋ ('nose', cf. Ulumanda' puduŋ), or which display no pre-velar fronting due to a different phonetic environment, e.g. täke < *taŋke (Ulumanda' takke). This clearly shows that pre-velar fronting is neither inherited from a shared source, nor has it spread from UAT into PUS by lexical diffusion, but rather spread as a phonetic feature affecting all vocabulary where it applies.


Vowel fronting before velars in a wider perspective

In Austronesian languages, we can find several other examples of vowel changes in the ultimate syllable that were triggered by the place of articulation of the following final consonant. In Mandar for instance, a Northern South Sulawesi language spoken along the coast to the west and south of the Ulumanda' speech area, we find fronting of vowels preceding Proto NSS *t, e.g. akkeʔ < *aŋkat ('lift'), polleʔ < *pollot ('buttocks'), uttiʔ < *kuntut ('knee'). This is quite the diametrical opposite of pre-velar fronting in Ulumanda'.

Vowel change triggered specifically by following velars is quite rare in Austronesian languages. I have found three examples which are quite different from each other in their outcome. The affected languages are Nias, Rejang and Long Melah Kayan. Only in the latter, we observe a vowel shift that is similar to the one observed in Ulumanda'.

In Nias, ultimate *a shifted to o preceding earlier *k, *ŋ and *R (probably a voiced velar fricative in pre-Nias): ono < *anak ('child'), fino < *pinaŋ ('betel nut'), lofo < *lapaR ('hungry'). Here, the velar triggered a back articulation, which subsequently was accompanied by raising and rounding. At least the back shift is naturally motivated, and represents the opposite of the pre-velar fronting in Ulumanda'.

In Rejang, final velars actually did not trigger a vowel change, but rather blocked one. Ultimate *a shifted to *ə in Proto Rejang, unless followed by *q, *k and *ŋ. Most probably, it was the retracted position of the tongue in the articulation of the velar and post-velar consonants that blocked the raising to *ə found with other finals. Again, both motivation and outcome of this vowel (non-)change are of a different nature if compared to Ulumanda' pre-velar fronting.

The Long Melah dialect of Kayan shows fronting of long /ā/ before final *k and *ŋ. In Proto-Kayan, the qualiative contrast between *ə and *a in the ultimate syllable was shifted to a quantitative contrast between short *a and long *ā. In Long Melah, long /ā/ remains unchanged preceding labial and apical finals. Preceding final /k/ and /ŋ/ however, /ā/ is realized as [ɛə], e.g. /anāk/ [a'nɛək] ('child'), /tulāŋ/ [tu'lɛəŋ] ('bone'). Even though this is still a merely phonetic surface phenomenon, it is yet very similar in outcome to pre-velar fronting in Ulumanda'. The only difference is the centralizing schwa offglide. The latter is not restricted to long /ā/, but also occurs with the coda /-iŋ/, e.g. /iliŋ/ [i'liəŋ] ('ear'). No schwa offglide is found with /-ik/ and ultimate /u/ or /o/.


The phonetic mechanism of pre-velar fronting

As can be seen from the inventory of final consonants in Ulumanda' and its higher-order ancestor, viz. Proto NSSul, pre-velar fronting partially compensated the loss of consonant contrasts in final position, by transferring phonemic distinctions between final consonants to the preceding vowel. What remains unclear however, is the phonetic mechanism which lead to the fronting of back vowels preceding final velar consonants. Based on place of articulation, fronting is very unusual here, since regular velar stops are articulated quite much in the same area of the oral cavity which is characteristic for the formation of back vowels.

As a possible solution to this puzzle, we can speculate that final *k and *ŋ had a fronted - close to palatal - articulation at an earlier stage. In this case, fronting of back vowels at least to central position is quite natural, resulting at an intermediate stage [a], [ɵ], [ʉ]. Further fronting (and raising of [a]) then must have occcurred to maximize the contrast to the original back vowels. At least for I final /k/, I have observed a relatively fronted and tense articulation in Toraja and Mamasa. This assuption requires further research, especially a detailed survey of vowel fronting in other Ulumanda' and neighboring lects.



Friberg, T. and T.V. Laskowske. 1989. South Sulawesi languages. In: J.N. Sneddon (ed.), Studies in Sulawesi
linguistics part 1. Jakarta: Badan Penyelenggara Seri Nusa.

Grimes, Charles E. and Grimes, Barbara D. 1987. Languages of South Sulawesi. (Pacific Linguistics: Series D, 78.) Canberra: Dept. of Linguistics, Research School of Pacific Studies, Australian National University. (Materials in languages of Indonesia ; 38).

Smith, Beverly K. 1993. A phonological reconstruction of Proto-Pitu Ulunna Salu. University of Texas at Arlington. Ann Arbor: UMI.

Van der Veen, H. 1929. Nota betreffende de grenzen van de Sa'dansche taalgroep en het haar aanverwante taalgebied. Tijdschrift voor Indische Taal-, Land- en Volkenkunde 69:58-96.