DigitalResources SIL eBook 62
®
Phonological and Phonetic Aspects of Enggano Vowels
Brendon E. Yoder Phonological and Phonetic Aspects of Enggano Vowels
Brendon E. Yoder
SIL International® 2014
SIL e-Books 62
2014 SIL International®
ISBN: 978-1-55671-372-9 ISSN: 1934-2470
Fair-Use Policy: Books published in the SIL e-Books (SILEB) series are intended for scholarly research and educational use. You may make copies of these publications for research or instructional purposes free of charge (within fair-use guidelines) and without further permission. Republication or commercial use of SILEB or the documents contained therein is expressly prohibited without the written consent of the copyright holder(s).
Editor-in-Chief Mike Cahill
Managing Editor Bonnie Brown
Compositor Margaret González
i PHONOLOGICAL AND PHONETIC ASPECTS OF ENGGANO VOWELS
by
Brendon E. Yoder Bachelor of Arts, Moody Bible Institute, 2006
A Thesis
Submitted to the Graduate Faculty
of the
University of North Dakota
in partial fulfillment of the requirements
for the degree of
Master of Arts
Grand Forks, North Dakota August 2011
Copyright 2011 Brendon Yoder
ii
This thesis, submitted by Brendon E. Yoder in partial fulfillment of the requirements for the Degree of Master of Arts from the University of North Dakota, has been read by the Faculty Advisory Committee under whom the work has been done and is hereby approved.
Chairperson
This thesis meets the standards for appearance, conforms to the style and format requirements of the Graduate School of the University of North Dakota, and is hereby approved.
Dean of the Graduate School
Date
iii
PERMISSION
Title Phonological and Phonetic Aspects of Enggano Vowels
Department Linguistics
Degree Master of Arts
In presenting this thesis in partial fulfillment of the requirements for a graduate degree from the University of North Dakota, I agree that the library of this University shall make it freely available for inspection. I further agree that permission for extensive copying for scholarly purposes may be granted by the professor who supervised my thesis work or, in his absence, by the chairperson of the department or the dean of the Graduate School. It is understood that any copying or publication or other use of this thesis or part thereof for financial gain shall not be allowed without my written permission. It is also understood that due recognition shall be given to me and to the University of North Dakota in any scholarly use which may be made of any material in my thesis.
Signature ______
Date ______
iv TABLE OF CONTENTS
LIST OF FIGURES ...... vii
LIST OF TABLES...... ix
ACKNOWLEDGEMENTS ...... xi
ABSTRACT ...... xii
CHAPTER 1. INTRODUCTION ...... 1
1.1 . Historical, geographical and sociopolitical context ...... 2
1.2. Previous research on Enggano...... 6
2. RESEARCH DESIGN AND METHODOLOGY ...... 12
3. OVERVIEW OF ENGGANO PHONOLOGY...... 21
3.1 . Consonants...... 21
3.2. Vowels ...... 31
3.3. Syllable structure ...... 36
3.4. Stress pattern ...... 37
4. SYLLABIFICATION OF VOCOID SEQUENCES...... 41
4.1. Attested sequences...... 41
4.2. Syllabification...... 44
4.3. Acoustic correlates of syllabification patterns ...... 49
4.4. Phonological analysis in Optimality Theory...... 64
5. VOWEL QUALITY...... 70
5.1. Methodology...... 70
5.2. Results ...... 72
v 5.3. Discussion ...... 74
6. VOWEL QUALITY TRANSITION IN VOCOID SEQUENCES ...... 82
6.1. Methodology...... 82
6.2. Results ...... 85
6.3. Summary...... 94
7. CONCLUSION...... 95
APPENDICES ...... 98
REFERENCES...... 145
vi LIST OF FIGURES
Figure Page 1. Enggano island in Indonesia...... 1
2. Ethnologue language map of Sumatra ...... 5
3. Villages on Enggano island...... 6
4. Vowel quality of stressed and unstressed vowels...... 39
5. Intensity contour of [ɘi ̯] in [pɘi̯s] ‘machete’ ...... 50
6. Intensity contour of [oi ̯] in [boh.oi̯] ‘rope’ ...... 51
7. Intensity contour of [iɘ] in [pi.ɘh] ‘massage’...... 52
8. Intensity contour of [iɨ] in [hi.ɨr] ‘woman’ ...... 53
9. Intensity contour of [jõ] in [jĩhjõn] ‘type of tool’ ...... 54
10. Intensity contour of [o̯e] in [boh.o̯e] ‘wild’ ...... 55
11. Duration of vocoid sequences in final open syllables...... 59
12. Duration of vocoid sequences in final closed syllables...... 60
13. Duration of vocoid sequences in final open syllables after glottal consonant...... 62
14. Duration of vocoid sequences in final closed syllables after glottal consonant...... 63
15. Plotted oral vowels...... 72
16. Plotted nasal vowels...... 73
17. Average value of oral and nasal vowels...... 74
18. Comparison of F1 ranges of oral and nasal vowels...... 76
19. Comparison of F2 ranges of oral and nasal vowels...... 77
20. Plotted nasal vowels as spoken by Adam Kurniawan...... 78
vii 21. Plotted nasal vowels as spoken by Josia...... 79
22. Plotted nasal vowels as spoken by Milson Kaitora...... 80
23. Plotted nasal vowels as spoken by Manogar RH ...... 80
24. Targets [u] and [a] in [pu.ak] ‘go’...... 84
25. Targets [a] and [ɨ] in [jarkɨ.aʔ.aɨ̯] ‘fence’...... 84
26. Disyllabic sequence [i.a] in [ki.ak] ‘blood’ ...... 85
27. Disyllabic sequence [ɨ.a] in [pɨ.ah] ‘face’...... 86
28. Disyllabic sequence [u.a] in [pu.ak] ‘go’ ...... 87
29. Disyllabic sequence [i.u] in [hi.ur] ‘dust’ ...... 88
30. Disyllabic sequence [u.i] in [kahə̆ʔu.is] ‘sea crab’...... 88
31. Diphthong [ai ̯] in [karaʔ.ai̯] ‘type of tree’ ...... 89
32. Diphthong [aɨ ̯] in [jarkɨ.aʔ.aɨ̯] ‘small fence’ ...... 90
33. Diphthong [au̯] in [baʔ.au̯] ‘guava’ ...... 91
34. Tautosyllabic diphthong [aɪ ̯] in [paɪ̯ʔ] ‘throw’ ...... 92
35. Targets [a] and [ɪ ̯] in [paɪ̯ʔ] ‘throw’...... 92
36. Diphthong [aʊ̯] in [bah.aʊ̯ʔ] ‘heart’ ...... 93
viii LIST OF TABLES
Table Page 1. Consonant inventory from Nothofer (1992) ...... 8
2. Difference between consonant inventories ...... 8
3. Vowel inventory from Nothofer (1992)...... 9
4. Difference between vowel inventories...... 9
5. Enggano consonant phonemes...... 21
6. Consonant co occurrences in underlying forms ...... 26
7. Contrast for /x/...... 29
8. Enggano vowel phonemes ...... 32
9. Syllable types in Enggano...... 36
10. Correlation of duration, intensity and pitch with word stress...... 38
11. Oral vocoid sequences...... 42
12. Nasal vocoid sequences ...... 43
13. Syllabification patterns in vocoid sequences...... 48
14. Final vocoid sequences for duration measurement ...... 57
15. /ʔa piah/ [ʔa.pi.ah] ‘graze’...... 65
16. /ka karai/ [ka.ka.rai̯] ‘chase’ ...... 65
17. /ki deok/ *[ki.deo̯k] ‘pinched’, incorrect prediction ...... 66
18. /ki deok/ [ki.de.ok] ‘pinched’, correct prediction...... 66
19. /karaʔai/ [ka.raʔ.ai̯] ‘type of tree’...... 66
20. /kõʔĩã/ [kõʔ.jã] ‘tree sp.’...... 67
21. /koʔoeʔ/ [koʔ.o̯eʔ] ‘devil’...... 67
ix 22. /ki deok/ *[kid.e̯ok] ‘pinched’, incorrect prediction ...... 67
23. /ki deok/ [ki.de.ok] ‘pinched’, correct prediction...... 68
24. Summary of OT constraint ranking...... 69
x ACKNOWLEDGEMENTS
I would like to express my appreciation to the people of Meok village on Enggano.
They graciously hosted me, showed me around their village and their island, and patiently answered my endless stream of questions about their language. Special thanks to my hosts Jhon Rafles K.N. and his wife. Thanks also to the other Meok residents who participated in the research: Adam Kurniawan, Josia, Milson Kaitora, Marhelam,
Manogar R.H., and two anonymous participants.
Thanks to SIL Indonesia for covering transportation and housing costs, as well as for providing logistical support throughout the research.
I am especially grateful for my thesis advisor Dr. John Clifton, who guided me through the entire research project. My understanding of phonology has grown tremendously as John has explained and applied concepts, taken my ideas apart and put them back together again. I also owe a big thanks to my other committee members Dr.
Steve Marlett and Dr. David Weber, who provided invaluable insight as research progressed.
My wife Naomi has been a great support to me, enduring my long absence during research trips during our engagement, and long mental absence at home in front of the computer screen. Thank you, and I love you.
Finally, I am grateful to God who gives wisdom, strength and life.
xi ABSTRACT
The Enggano language has received little attention in Austronesian linguistics. It is an isolate located geographically in a large area of related languages. A dedicated description of Enggano phonology has never been undertaken before. This thesis describes aspects of Enggano phonology and phonetics, primarily at the word level. It focuses mainly on vowels and vocoid sequences.
As a starting point for analysis of Enggano phonemes, a list of phonemes was compiled from previous research where phonemes are mentioned. For each reported phoneme, words were elicited with the phoneme in initial, prestress and final (stressed) position. An example word containing each vowel and consonant was recorded in frames for acoustic analysis. Since sequences of two or more adjacent vocoids are common, words containing examples of all vocoid sequences were recorded in frames as well.
The Enggano phoneme inventory consists of twelve consonants and fourteen vowels in a seven vowel oral system and an analogous seven vowel nasal system. There are seven possible syllable types. Word stress is consistently final in both monomorphemic and polymorphemic words. Acoustic measurements show that word stress is indicated by intensity in closed syllables, and possibly by length and pitch in both open and closed syllables.
There are a few allophonic processes in Enggano. An intrusive vowel (Hall 2006) is inserted in consonant sequences beginning with /ʔ/. A tentative analysis of the fricative
/x/ is that it is realized as [x], [ç], or [s] depending on the context. The glottal stop is
xii optionally palatalized after a high front vowel, and vowels are nasalized in words with a nasal consonant.
Vocoid sequences syllabify based on the preceding environment and the relative height of the two vocoids. Two vocoid sequences after medial non glottals are disyllabic except sequences beginning with a lower vocoid and ending with a higher vocoid (low high and mid high). These are realized as diphthongs. Vocoid sequences after medial glottal consonants [ʔ] and [h] are realized differently. Glottal consonants syllabify in the coda of the previous syllable. Syllable initial vocoids in sequences that are not low high are realized in the onset of the syllable, as in /kõʔĩã/ [kõʔ.jã] ‘tree sp.’. This process does not affect low high sequences, or sequences where the glottal consonant is word initial. Acoustic measurements show that the three syllabification patterns of vocoid sequences (vowel vowel, glide vowel, vowel glide) can be distinguished by both intensity and overall duration of the sequences.
Nasal vowels have a much greater range in the vowel space than oral vowels, and consequently there is much more overlap between adjacent vowels. This range can be attributed to variation between speakers in articulation of nasal vowels. Vocoids in sequences are very similar in place of articulation to their interconsonantal counterparts. Vocoids in disyllabic sequences are generally in more extreme areas of the vowel space, while vocoids in diphthongs are generally located in more central areas of the vowel space than plain vowels.
xiii CHAPTER 1
INTRODUCTION
The Enggano [eno] 1 language is an isolate spoken by about 1,500 people in
Indonesia (Lewis 2009). It is spoken only on the island of Enggano, which is about 80 miles (125 km) off the southwestern coast of Sumatra, Indonesia, in the Indian Ocean.
The island of Enggano is the southernmost of the Barrier Islands that extend along the western coast of Sumatra. Figure 1 shows a map of Indonesia, with an arrow marking
Enggano.
Figure 1. Enggano island in Indonesia 2
1 ISO 639 3 language identification codes are given in square brackets following language names. 2 Accessed on the Internet at https://www.cia.gov/library/publications/the world factbook/maps/maptemplate_id.html . I added the marker for Enggano island.
1 Enggano is linguistically and geographically isolated from surrounding language groups.
Because of this and the small number of people who speak it, the language has received little attention in the literature. This thesis, a description of Enggano word level phonology and phonetics, is the first research project dedicated to analyzing the phonology of Enggano.
The thesis is laid out as follows. The first three chapters are introductory in nature, giving background (Chapter 1) and methodology (Chapter 2), followed by an overview of phonemes, allophonic processes, stress patterns, and syllable structure in Chapter 3.
Chapter 4 is dedicated to a discussion of vocoid sequences, focusing on syllabification processes. Chapter 5 presents an acoustic analysis of vowel quality of single vowels, while chapter 6 lays out an acoustic analysis of vowel quality transition in vocoid sequences. Finally, conclusions are given in chapter 7.
1.1 Historical, geographical and sociopolitical context
Little is known about the island of Enggano or its people before the late nineteenth century. The name Enggano is popularly thought to come from the Portuguese engano
‘deception, error’. A story, told by local Enggano people as well as outsiders, says that the Portuguese explorers of the sixteenth century came upon Enggano from the West and mistakenly thought that they had found Sumatra. In their disappointment at finding that they had not arrived on Sumatra, they named the island Enggano. Any former language internal name for the Enggano people and language has been lost, as the exonym Enggano is the only name that is known.
The Museum Volkenkunde in Leiden, The Netherlands, houses a collection of artifacts and drawings from early explorers’ visits to Enggano (ter Keurs n.d.). Drawings from Modigliani (1894) show Enggano people living a Stone Age type existence with no evidence of influence from the Malay and Javanese civilizations a few hundred miles
2 away. They lived in beehive shaped houses similar to traditional houses of the
Andaman Islands of India 1,000 miles to the northwest in the Indian Ocean. The old
Malay name for the island of Enggano, Pulau Telanjang or Naked Island, reflects the fact that the Enggano had a lifestyle very different from that of the larger people groups of Sumatra and Java.
With the onset of contact with outsiders, the Enggano society went through major upheaval. At some point between 1866 and 1884 the population of the island was decimated, going from over 6,000 people in 1866 to 900 people in 1884 (Jaspan 1964,
110). Some authors believe that this was caused by cholera, venereal disease, and induced abortion (Helfrich 1888).
Whatever the cause of the decimation of Enggano’s population, the effects were devastating for the language. In his dictionary Kähler notes that when he conducted his
1937 field research on the island, only about 200 people still spoke Enggano as their mother tongue, “which, however, in the case of the younger generation . . . was already strongly influenced by Malay” (1987, 81). Concerning Enggano culture, Keuning (1955) said that the traditional culture had almost completely disappeared by the middle of the twentieth century. The Enggano culture and language were put in further jeopardy in the early 1960s when the Indonesian government began relocating convicts from Java and Sumatra to Enggano. By 1963 there were 2,600 offenders from other parts of
Indonesia on the island in addition to the 400 remaining Enggano people (Jaspan 1964,
110).
Contrary to predictions during the early 1900s, the Enggano population has actually grown in the past fifty years, and the language is still in vigorous daily use.
Estimates of the number of Enggano speakers today range from 700 (Wurm 2000) to
1,500 (Lewis 2009). The total population of the island is around 2,000, with about half
3 of the population being native and the other half being people of outside descent, mostly from Java and Sumatra.
Enggano has recently been heavily influenced by the culture, politics and languages of Java and Sumatra. The languages most influential on Enggano are the Malay and
Indonesian language varieties spoken on Java and Sumatra. The nearest language is
Bengkulu Malay, spoken in and around the city of Bengkulu (classified in the
Ethnologue as a dialect of Central Malay [pse]). The only ferries connecting Enggano to
Sumatra go to Bengkulu, so there is much direct contact between that city and the island. Minangkabau [min] is a regional trade language along the west coast of
Sumatra, and as such it also has influence on Enggano. The third prominent language influencing Enggano is Standard Indonesian [ind], the official language of Indonesia and the language of school, government, and all official business on Enggano. Figure 2 shows the Ethnologue language map of Sumatra. Enggano island is number 13 at the far bottom, Central Malay is spoken in the area designated by number 10 in the southwest, and Minangkabau is spoken in the area designated by number 28 along the western coast.
4 Figure 2. Ethnologue language map of Sumatra
Bengkulu Malay, Minangkabau, and Indonesian are all closely related to each other.
It is often difficult to tell which one has caused a particular effect in Enggano.
Speakers of Bengkulu Malay, Minangkabau and Indonesian now live on Enggano island, intermingled with the local Enggano population. Speakers of other languages, including Batak and Javanese, live there as well. Since few of the immigrants speak
Enggano, most interethnic communication takes place in Indonesian or Bengkulu Malay.
Enggano speakers only use their language among themselves, although many immigrants claim that they can understand some of the language.
There are six villages on Enggano island, all of them strung along the island’s only road on the northeast coast. Figure 3 shows the villages on the island. The lines mark the territorial boundaries of the villages, while the dots represent the residential area for each village.
5 Figure 3. Villages on Enggano island 3
The three central villages of Malakoni, Apoho and Meok have greater percentages of native Enggano residents, while the outlying villages of Banjarsari, Kaana and
Kahayapu have larger immigrant populations. Ferries to Bengkulu depart from
Kahayapu and Malakoni.
1.2 Previous research on Enggano
The first substantial documentation of life on Enggano began when Enggano’s contact with the outside world increased in the last half of the nineteenth century.
Dutch administrators and explorers made numerous trips to the island and published several general reports on the people, culture and language. These include Boewang
(1854), van der Straaten and Severijn (1855), von Rosenberg (1855), Walland (1864), van der Hoeven (1870), and Helfrich (1888). These reports culminated in an often cited
3 This map was created by SIL Indonesia and is used by permission. I added the points to represent the approximate place of each village’s residential area.
6 ethnographic description of the Enggano people, “The Island of Women”, by Italian researcher E. Modigliani (1894). Ethnograpic work continued with more recent work by
Tichelman (1942), Keuning (1955), Amran (1979), and ter Keurs (2006).
The first linguistic records from Enggano are the brief wordlists found in van der
Straaten and Severijn (1855) and von Rosenberg (1855). Both of these wordlists were recorded by non linguists unfamiliar with the Enggano language, so their accuracy is doubtful. Helfrich and Pieters (1891) published an extensive wordlist, with extensive corrections and additions published a few years later (Helfrich 1916). The complete wordlist contains over 1,000 items, with glosses in Malay and Dutch. Also available is the Holle List for Enggano (Stokhof 1987), which was collected in 1895 and also contains over 1,000 items. These two extended wordlists represent the first substantial records of the Enggano language.
The first and only major linguistic analysis of Enggano was conducted by German linguist Hans Kähler in 1937. The results of this research include a published grammar
(1940), dictionary (1987), and texts (1955; 1961; 1973; 1975). Kähler’s grammar and dictionary are the sources for most typological and comparative work on Enggano.
While the grammar and dictionary are both extensive, the only mention of phonology in either one is a list of consonant and vowel phonemes and a short paragraph describing some basic features. The list of phonemes in the dictionary does not agree with the one in the grammar, and some of the dictionary’s transcriptions are inconsistent. The font used in the publication causes poor legibility for some words. Prentice (1989), among others, questions the accuracy of the transcriptions.
The only literature that discusses Enggano phonology is Kähler’s grammar and dictionary, along with Nothofer’s more recent paper on loanwords in Enggano, which has a list of phonemes (Nothofer 1992). Table 1 shows Nothofer’s analysis of the consonant phonemes.
7 Table 1. Consonant inventory from Nothofer (1992)
Bilabial Alveolar Palatal Velar Glottal Stop p b t d k ʔ Nasal m n Trill r Fricative h Approximant w j
There are extensive differences between this analysis and Kähler’s. Kähler adds /ɲ/ and /x/ as phonemes but does not include /w/. He claims that /t/ and /r/ are marginal, occurring only in certain dialects, adding /tʃ/, 4 /dʒ/, /f/, and /l/ as marginal phonemes as well. These differences are summarized in Table 2.
Table 2. Difference between consonant inventories
Kähler dictionary Kähler grammar Nothofer 1992 p √ √ √ b √ √ √ t marginal marginal √ d √ √ √ tʃ marginal √ dʒ marginal marginal k √ √ √ ʔ √ √ √ m √ √ √ n √ √ √ ɲ √ √ r marginal marginal √ f marginal marginal x √ √ h √ √ √ w √ j √ √ √ l marginal marginal
Table 3 shows Nothofer’s analysis of vowel phonemes.
4 Kähler’s dictionary indicates /tʃ/ is marginal, while his grammar indicates it is not.
8 Table 3. Vowel inventory from Nothofer (1992)
Front Central Back High i ĩ ɨ ̃ u ũ Mid e ẽ ə ə̃ o õ Low a ã
As with the consonant phonemes, there are many differences between Nothofer’s and Kähler’s analyses of vowel phonemes. Kähler’s dictionary does not include the high central vowels / ̃/ and / ̃/. His grammar, on the other hand, includes a set of low mid vowels /ɛ/, /ɛ̃/, /ɔ/, and /ɔ̃/ but does not include /ẽ/, /ə/, /ə̃/, or /õ/.
Table 4 shows the difference between the vowel inventories in the same sources.
Table 4. Difference between vowel inventories
Kähler dictionary Kähler grammar Nothofer 1992 i √ √ √ ĩ √ √ √ ɨ √ √ ̃ √ √ u √ √ √ ũ √ √ √ e √ √ √ ẽ √ √ ə √ √ ə̃ √ √ o √ √ √ õ √ √ ɛ √ ɛ̃ √ ɔ √ ɔ̃ √ a √ √ √ ã √ √ √
The Kähler grammar also lists seventeen diphthongs as part of the phoneme inventory. All the diphthongs go from low to high, beginning with /a/, /e/ or /o/ and ending with /e/, /o/, /i/ or /u/. Diphthongs are not mentioned in any of the other sources.
9 Besides Nothofer (1992), recent discussion of Enggano has mostly been limited to brief mention of the language in historical comparative and typological work on
Austronesian languages. These begin with Dyen’s lexicostatistical analysis of
Austronesian wordlists (1965). Dyen’s analysis puts Enggano’s highest percentage of shared vocabulary at 11% with Murut, a language of Borneo. While the lexicostatistical method is not generally considered valid for determining language relationships,
Enggano’s extraordinarily low percentage of shared vocabulary still shows that it is not closely related to any of the surrounding languages. The comparative work with the most thorough discussion of Enggano is Nothofer’s article on the Sumatran Barrier
Islands languages (1986). Nothofer provides quantitative evidence for subgrouping
Sichule (Sikule), Nias, Mentawai, and Simalur (Simeulue). Enggano, however, is included only as a possible member of the subgroup because it is so deviant from the other languages in the Barrier Islands.
Other authors take widely differing approaches to Enggano’s classification. Mahdi
(1986) groups all Austronesian languages in Western and Eastern subgroups, with most of the Barrier Islands languages in the Western subgroup and Enggano in the Eastern subgroup. Capell (1982) takes a cautious approach, classifying Enggano as a non
Austronesian isolate. He concludes, “Enggano does not in fact belong to the [Proto
Austronesian] group” (p. 4). The Ethnologue summarizes the lack of consensus on
Enggano’s classification: “Not closely related to other languages. Not conclusively established as an Austronesian language, rather than an isolate with Austronesian loans.”
There are three other studies discussing other aspects of the Enggano language.
Nothofer (1992) lists sound changes that occurred in the adaptation of loanwords from
Malay/Indonesian and Minangkabau into Enggano. Schmidt (1988) discusses the sound changes that occurred in Enggano over the past fifty years. Enggano is undergoing rapid
10 change under influence from Malay and Indonesian in its grammar, lexicon, phonology and domains of use. A very recent study of bilingualism patterns among the Enggano people by Simanjuntak (2009) shows that Enggano is still a vital language used in conjunction with Standard Indonesian for daily life. Simanjuntak states that attitudes toward both Enggano and Indonesian are positive.
11 CHAPTER 2
RESEARCH DESIGN AND METHODOLOGY
The goal of this research is a description of Enggano phonology and phonetics that will be of typological interest and will be useful for language development, with a focus on vocoids and vocoid sequences. To accomplish this, I take what is already known about Enggano phonology and expand on that knowledge with original research. In this chapter I summarize what has been previously published about Enggano phonology. I then sketch out my objectives for further research and the methodology used to accomplish the objectives.
As indicated in Chapter 1, the three publications that have information on Enggano phonology are Kähler’s grammar and dictionary and Nothofer’s 1992 paper on loanwords. Following is a summary of the analyses presented there.
(a) Phoneme inventory: twelve to seventeen consonants and six to eight oral
vowels with nasal counterparts. Kähler (1940) lists seventeen diphthongs in
the phoneme inventory while other sources do not list these.
(b) Syllable structure: CV and V. Kähler’s dictionary does not list any syllable
codas. There are many vowel sequences.
(c) Stress patterns: The discussion in the 1895 Holle list says, “word stress mostly
is on the penultimate syllable; if not it falls on the final syllable” (Stokhof
1987, 189). Kähler’s dictionary also states that stress is penultimate.
(d) Morphophonology: There are many prefixes and suffixes in the language. As
affixation is difficult to decipher in Kähler’s dictionary because of
12 typographical issues, it is difficult to tell whether there is any phonologically
conditioned alternation in derived forms.
(e) Dialects: The Kähler grammar lists a few phonological differences between
two dialects, and the older Dutch sources also list dialectal differences. There
were apparently different speech patterns on several small inhabited outlying
islands, especially Pulau Dua. Today the outlying islands are no longer
inhabited. In 2007 Enggano speakers informed a sociolinguistic survey team
that there was no dialect variation at all in the Enggano language (Aprilani
2007). The same was reported to me during my field research in 2010. Only
the northeastern coast of the island is now inhabited.
Based on the goal of a descriptive phonology that is of typological interest, and based on what is already known from the literature, the following objectives guide the research. The focus here is on vowels and vocoid sequences.
(a) Establish the phoneme inventory, specifically by resolving areas of
disagreement between the various published phoneme inventories.
(b) List and discuss basic phonological patterns including syllable structure, stress
patterns, allophonic processes and positional neutralization of contrast.
(c) List and discuss syllabification patterns in vocoid sequences.
(d) Provide a phonological analysis of syllabification patterns of vocoid
sequences.
(e) Show the acoustic correlates (duration, intensity, formants) of all
syllabification patterns of vocoid sequences.
(f) Show how the vowel system maps out acoustically in the vowel space.
Because of external factors, I was able to spend only two weeks doing intensive data collection in Meok. I was able to collect a large amount of quality data during that time, thanks to the guidance of my advisor and the generosity of my Enggano hosts. However,
13 two weeks is a very short time span for conducting field work, and there are several areas of Enggano phonology that remain unclear because of a lack of time in field research. If more extended visits to Enggano are undertaken in the future, researchers will be able to build on the initial research presented here. I discuss in prose below the methodologies used to accomplish the six objectives presented above.
The phoneme inventories in Kähler (1940; 1987) and Nothofer (1992) give a total of fourteen vowels and eighteen consonants, including all of Kähler’s marginal consonants. For each of the vowels and consonants, native speakers were asked to think of words that contained that segment in various positions. Initially, the goal was to obtain ten examples of each consonant in the onset of syllables in three positions: initial, medial pre stress, and stressed; as well as word finally. This is shown in (1). (1) Four word positions for consonant elicitation (Example consonant: [p]) Initial /purik/ [purik] ‘grow’ Medial pre stress /kĩ pãʔĩõp/ [kĩ pãʔjõp] ‘eight’ Stressed /hã pɘ̃ʔ/5 [hã pɘ̃ʔ] ‘breathe’ Final /kõ p/ [kõ p] ‘grave’
Each vowel was to be elicited in initial, stressed, and post stress environments.
These parameters were set up with the assumption that, as per Kähler, stress was penultimate and there were no syllable codas. When it became evident that stress was final, elicitation for the post stress position was discontinued. Since word initial syllable onsets are mandatory, elicitation for vowel initial words was discontinued as well. The word positions for vowel elicitation are shown in (2). (2) Three word positions for vowel elicitation (Example vowel: [a]) Pre stress / tahaʔ/ [t ahaʔ] ‘write’ Final closed syllable /por aʔ/ [por aʔ] ‘split’
5 Formant measurements of the mid central vowel show that it is acoustically higher than the other mid vowels /e/ and /o/ (see section 5.2). Perceptually, the quality of this vowel sounds different from that of intrusive schwa (see section 3.3). So the high mid vowel symbol [ɘ] is used for the mid central phoneme, while the schwa symbol with the breve [ə̆] is used for intrusive schwa.
14 Final open syllable / kitah a / [kitah a] ‘centipede’
It also became evident during elicitation that there were many syllable codas in the language, including word internal codas. This created the possibility of medial consonant sequences across syllable boundaries. Medial consonant sequences are common. Elicitation focused on consonant sequences was therefore not necessary since there were many examples of the sequences in words elicited for other purposes.
Examples of each vocoid sequence 6 were elicited as well. The goal was to obtain at least five examples of all possible combinations, excluding homorganic sequences. The surrounding environment was not specified as it was for consonants and vowels.
Speakers could only think of one or two examples of some of the rarer sequences, especially the nasal and nasalized sequences.
A total of six men and two women, all from the village of Meok, participated in the research. All had spent most of their lives in Meok. All were initially thought to be first language Enggano speakers, although I learned later that one of the women had learned
Indonesian as her first language and had learned Enggano while in elementary school.
Data contributed by this speaker was not removed from the database since everyone agreed that she spoke Enggano no differently from anyone else, and had been fluent for a very long time.
The speakers worked in small groups. Together they filled out paper forms that listed a segment, a specific environment, and ten blank example spaces for examples.
The speakers worked in small groups to fill out the forms. One group member was informally assigned as the transcriber. Standard Indonesian orthography was used for
6 In this thesis, the term vocoid sequence is used to denote a sequence of underlying vowels. Vocoid sequences have surface realization as either two vowels or a vowel and an approximant.
15 transcription, with the addition of the apostrophe <'> for glottal stop, schwa <ə> for both central vowels, and tilde (e.g. <ã>) for vowel nasalization.
As lists of words were completed, all data was checked and reviewed with one of the speakers, and then entered into a FieldWorks database. 7 When all of the lists were complete and entered into the database, one of the speakers made an audio recording of the whole database. All recordings in the study were done on a Zoom H2 portable digital recorder with an Audio Technica Pro 70 lapel microphone. Recordings were made at 24 bit depth, 48 kHz sampling rate. The recordings were done in stereo mode, with right gain turned to zero. Using Audacity (Audacity Team 2010), the right channel was extracted and deleted, leaving a mono file. For the recording, I gave a prompt in
Indonesian and the speaker repeated the corresponding Enggano word twice. The only problem with this method was that it yielded words with “list intonation” – the first repeat of each word has a rising pitch and greater overall intensity, and the second has a falling pitch with lower overall intensity. All words in the FieldWorks database were later checked with the audio recordings to ensure phonetic accuracy.
The database resulting from this methodology contains 841 entries. 144 of these are complex words, loanwords, or place names. These were removed for the basic phonological analysis. The database of basic words for phonological analysis thus contained 697 items.
The audio files resulting from the research will be archived online in SIL
International’s REAP archive. As of this writing, the REAP archive is still very new and has not yet been made publicly available on the Internet. Plans are being developed to make this data publicly available, however.
7 SIL International’s FLEx (FieldWorks Language Explorer) was used for the database of lexical items. SIL’s Phonology Assistant was used to analyze phonological patterns from the words in the corpus. Both programs are freely available on the Internet at http://fieldworks.sil.org/ and http://phonologyassistant.sil.org/ .
16 Most Enggano verbs and adjectives have affixation. These words were included in the basic database, while other complex and compound words were excluded. In the analysis, care was taken to ensure all phenomena are attested within morphemes and not only across morpheme boundaries. In verbs with affixation, it is not clear what form of the verb is represented with the different affixed forms. Determining this would require a morphological analysis, which was impossible due to time constraints. The
Kähler grammar would have likely been of great help in this area, but it was largely inaccessible to me since it is written in German. All glosses of verbs, therefore, only contain the basic sense of the words and do not show what form of the verb is represented. Procedures for dividing morphophonological phenomena from phonological phenomena are discussed in Appendix A, along with a brief overview of affixation.
For each vocoid, consonant, and vocoid sequence, one example word was recorded in the sentence frame [ʔu pe __ ʔan pe janik] ‘Say __ clearly’ for acoustic analysis.
Words recorded in frames with a vocoid in focus are shown in (3), while those with a consonant in focus are shown in (4). The segment in focus is in bold. (3) Vocoids recorded in frames 8 /hɘdik/ [hɘdik] ‘turn around’ /k ɨx/ [k ɨx] ‘turtle’ /dud uk/ [dud uk] ‘burn’ /ka rep/ [kar ep] ‘boil’ /k ɘx/ [k ɘx] ‘mountain’ /kaʔ tok/ [kaʔə̆tok] ‘red’ /k ak/ [k ak] ‘person’ /kãʔ kĩh/ [kãʔə̆kĩh] ‘dry’ /p ̃ʔ/ [p ̃ʔ] ‘fireplace’ /k ũk/ [k ũk] ‘back’ /kã pẽp/ [kãpẽp] ‘shallow’
8 As discussed in sections 3.2 and 5.2 , the mid central vowel (represented as [ə] in other sources) is better represented as [ɘ] as it is acoustically higher than the other mid vowels. This convention is also useful to distinguish it from the schwa vowel inserted in consonant clusters, which I represent with the symbol [ə̆]. [x], [ç], and [s] were found to be allophones of /x/, as discussed in section 3.1.3.
17 /kãʔ kɘ̃h/ [kãʔə̆kɘ̃h] ‘black’ /k õp/ [k õp] ‘grave’ /k ãp/ [k ãp] ‘traditional leader’ /ʔa io iaʔ/ [ʔajo jaʔ] ‘pay out (a rope)’ /ʔauaʔ/ [ʔawaʔ] ‘mangrove’
(4) Consonants recorded in frames /ʔabeh/ [ʔabeh] ‘bamboo’ /ba tar/ [ba tar] ‘pillow’ /ka der/ [ka der] ‘soft’ /kaba keʔ/ [kaba keʔ] (a term of address) /ba ʔau/ [ba ʔ.au̯] ‘guava’ /ʔamak/ [ʔãmãk] ‘God’ /ka nam/ [kã nãm] ‘salt’ /ka rah/ [ka rah] ‘body’ /pɘix/ [pɘis ̯ ] ‘machete’ /karai x/ [kara ç] ‘cup’ /kaʔ dai x/ [kaʔə̆da ç] ‘white’ /kɘx/ [kɘx] ‘mountain’ /ta haʔ/ [ta haʔ] ‘write’ /ka lel/ [ka lel] ‘soft’
Words recorded in frames with a vocoid sequence in focus are shown in (5). Here and throughout the thesis, syllable breaks between two vowels are marked with a period. Glottal consonants syllabify with the preceding vowel (Chapter 4); these syllable breaks are also marked. 9 Syllable breaks not involving a vocoid sequence are unmarked.
Most of the sequences are sequences of two vowels, but several have a vowel and an approximant. Because of time limitations, I was not able to record an exhaustive set of vowel vowel and vowel approximant sequences. (5) Vocoid sequences recorded in frames /h iu r/ [h i.u r] ‘dust’ /ie / [je ] ‘vomit’ /h iɘr/ [h i.ɘr] ‘earthworm’ /k ia k/ [k i.a k] ‘blood’ /dud ia r/ [dud i.a r] ‘durian fruit’ /ki dadaʔɨɘh/ [kidadaʔ.ɨ̯ɘh] ‘care for deceased’s spouse’ /kaʔhɨɘ/ [kaʔə̆hɨ.ɘ] ‘female leader’
9 This analysis of glottal consonant syllabification is necessary for the analysis of vocoid sequence syllabification presented in Chapter 4. I assume that there is no articulatory or perceptual difference between glottal consonant as coda (e.g. [bah.aʊ̯ʔ] and onset (e.g. [ba.haʊ̯ʔ]). These syllabification marks are part of the analysis rather than a reflection of actual articulation.
18 /ka kɨɘ/ [kak ɨ.ɘ] ‘short’ /p ɨah/ [p ɨ.a h] ‘face’ /kahʔui s/ [kahə̆ʔwi s] ‘sea crab’ /ba ʔue h/ [baʔwe h] ‘sleep’ /ʔi ʔuo ʔ/ [ʔiʔwo ʔ] ‘there’ /p ua k/ [p u.a k] ‘go’ /kik eo r/ [kik e.o r] ‘lost’ /ki deo k/ [kid e.o k] ‘pinched’ /pa hit ɘi/ [pahit ɘi̯] ‘mock’ /k oi / [k oi̯] ‘wild pig’ /boh oe / [boh .o̯e] ‘wild’ /k oa r/ [k o.a r] ‘type of kitchen utensil’ /p ai do/ [p ai̯do] ‘cry’ /bah au ʔ/ [bah.aʊ̯ʔ] ‘heart’ /b ei a/ [b ej a] ‘table’ /t au ud/ [t aw ud] ‘year’
Words recorded in frames with a nasal or nasalized sequence in focus are shown in
(6). (6) Nasal and nasalized sequences recorded in frames /h ĩũ k/ [h ĩ.ũk] ‘louse’ /ki hĩẽ k/ [kĩhjẽ k] ‘sit’ /kĩʔ ̃ɘ̃k ̃ʔ ̃ɘ̃/ [kĩʔjɘ̃k ̃ʔjɘ̃] ‘sea creature sp.’ /h io n/ [h ĩ.õn] ‘scratch’ /ki hia m/ [kĩhjã m] ‘itchy’ /ĩũʔp ̃ãʔ/ [jũʔə̆p ̃.ãʔ] ‘fence’ /kũʔũĩ / [kũʔwĩ ] ‘great grandparent’ /kũhũẽ / [kũhwẽ ] ‘usefulness’ /m uo / [m ũ.õ] ‘flower’ /k ua n/ [k ũ.ãn] ‘when’ /h ẽõ k/ [h ẽ.õk] ‘cockroach’ /ʔẽã p/ [ʔẽ.ãp] ‘thick’ /ki mɘmɘi/ [kĩmɘ̃mɘ̃ ̯̃] ‘overcast’ /pinoh oi / [pĩnõh.õ ̯̃] ‘naïve’ /k oa n/ [k õ.ãn] ‘Lord’ /ʔamah ai / [ʔãmãh.ã ̯̃] ‘bed’ /p ai nɘn/ [p ã ̯̃nɘ̃n] ‘feelings ‘ /kĩpãʔãũ p/ [kĩpãʔ.ãʊ̯̃p] ‘ten’ /kãʔãũ / [kãʔ.ãũ̯] ‘fat (adj.)’
Six speakers were given the list of words written in five different orders on separate sheets of paper. Each speaker went through the lists one at a time, repeating each word in the frame [ʔu pe ___ ʔan pe janik] ‘Say ___ clearly’. For analysis I selected two recordings each from four men, totaling eight tokens of each word. I chose the second and third recordings from each speaker unless there was a problem with one of these
19 such as stuttering, stumbling or scrambling the words in the frame, a pause part way through the frame phrase, or background noise. In these cases I selected the fourth recording of the word unless it also was problematic, in which case I chose the first or fifth one. I also recorded all the words from the entire database in list format with one speaker. I spoke the Indonesian prompt once, and the speaker repeated the Enggano word twice. When the words recorded in frames were not sufficient for acoustic analysis, I used recordings of words from this comprehensive list.
20 CHAPTER 3
OVERVIEW OF ENGGANO PHONOLOGY
This chapter presents a basic overview of Enggano phonology as a background for the discussion of vowels and vocoid sequences in the remainder of the thesis. The chapter discusses consonants and vowels, along with their distribution and related allophonic processes. Syllable structure and stress patterns are also discussed.
3.1 Consonants
My analysis shows that Enggano has twelve consonant phonemes: six stops, two nasals, a trill, two fricatives, and a lateral approximant. Table 5 shows the inventory of consonant phonemes in Enggano. Evidence of contrast for the consonants is found in in
Appendix F, which lists examples of each consonant in initial, prestress medial, stressed, and final positions.
Table 5. Enggano consonant phonemes
Bilabial Alveolar Palatal Velar Glottal Stop p b t d k ʔ Nasal m n Rhotic r Fricative x10 h Lateral l11 approximant Approximants: [j], [ɨ ̯], [w], [e̯], [o̯]
10 I analyze surface [s], [ç], and [x] as allophones of /x/, although this analysis is tentative and awaits further study. See section 3.1.4 for a discussion. 11 Occurs in only a few words.
21 The glottal and velar stops are by far the most frequently occurring consonants in
Enggano.
Besides the consonants, many vocoids can occur in syllable onset position. These include three high approximants [j], [ɨ ̯] and [w], and two mid approximants [e̯] and
[o̯]. These are shown in (7). (7) Underlying Surface Environment /i/ [j] onset /u/ [w] onset /ɨ/ [ɨ ̯] onset (not word initial) /e/ [e̯] onset (not word initial) /o/ [o̯] onset (not word initial)
All five approximants occur in syllable onsets that are not word initial, while only [j] and [w] can occur word initially. 12
3.1.1 Restrictions on distribution
Voiced oral consonants (/b/, /d/, /r/, and /l/) do not co occur in the same word with nasal consonants and vowels; that is, forms like */bãbã/ [bãbã] and */naba/
[nãbã] do not exist. I do not have an explanation for this. The single exception to this is the word /parna/ [pãrnã] ‘paper’, where the sequence /rn/ occurs. This is probably a loanword; see Sanskrit /patra/ ‘paper’. A sample of words containing voiced oral consonants is shown in (8), while a sample of words with nasal consonants and vowels is shown in (9). (8) Words with voiced oral consonants /tabɨdaʔ/ [tabɨdaʔ] ‘jackfruit’ /ba/ [ba] ‘come’ /duduk/ [duduk] ‘burn (v.)’ /dix/ [diç] ‘earthquake’ /dɘr/ [dɘr] ‘river current’ /parur/ [parur] ‘festivity’
12 There does not seem to be a principled reason why mid approximants could not occur in word initial position since they do occur medially. This distribution could be evidence for a constraint against nonsyllabic mid vocoids in word initial position, separate from constraints against high vocoids in initial position and against high and mid vocoids in medial syllable onsets.
22 /bero/ [bero] ‘river’ /pa lauaʔ/ [palawaʔ] ‘spit (v.)’
(9) Words with nasal phonemes /pan/ [pãn] ‘nose’ /kahten/ [kãhtẽn] ‘disgusting’ /kĩk/ [kĩk] ‘necklace’ /kome/ [kõmẽ] ‘bridge’ /ʔãpũʔ/ [ʔãpũʔ] ‘snake’ /hion/ [hĩ.õn] ‘scratch’ /ʔẽp/ [ʔẽp] ‘left’ /ka pẽʔ/ [kãpẽʔ] ‘slanted’
Besides the two categories shown above, there are also many words that have neither voiced oral consonants nor nasal phonemes.
The consonant /t/ occurs mostly in loanwords. Out of 71 total occurrences of /t/ in the data, it occurs in loanwords 47 times and in native vocabulary 24 times. Initial /t/ is found almost exclusively in loanwords. Thse words appear to be of Malay origin, where the source consonant is initial /s/, /t/ or /tʃ/. For example, Malay
/sirih/→/tiri/ [tiri] ‘betel leaf’, Malay /tʃəŋkeh/→/teke/ [teke] ‘cloves’, and Malay
/tahun/→/tauud/ [tawud] ‘year’. There are three apparently native words with initial /t/ in the data: /tahaʔ/ [tahaʔ] ‘stab’, /teʔ/ [teʔ] ‘there’ and /tabɨdaʔ/
[tabɨdaʔ] ‘jackfruit’.
The alveolar stops /t/ and /d/ rarely occur in final position. Final /t/ occurs only after /i/, with seven instances in the data. Final /d/ only occurs in one native word:
/dahɨd/ [dahɨd] ‘lean’, and in four loanwords. The other alveolar consonants /n/ and
/r/ occur more freely in final position. /n/ occurs finally 38 times, following all vowels, and final /r/ occurs 61 times after all oral vowels but /i/.
The rhotic consonant /r/ occurs initially in only one native word: /roro/ [roro]
‘pick up’. It occurs frequently in medial and final positions and is realized as either a trill or a tap (section 3.1.6).
23 In a few words, /d/ and /r/ appear to be in free variation, as in [kader] ~
[karer] ‘soft’ and the loanword [dudiad] ~ [dudiar] ‘durian fruit’. Enggano speakers report that older speakers tend to favor pronunciation with /d/, while younger speakers favor pronunciation with /r/. 13
The glottal stop occurs in all positions. It contrasts with other consonants initially and finally, but it is somewhat predictable medially. In 87% of the instances of intervocalic glottal stop in the data (113 out of 130), the vowels on either side of the glottal stop are homorganic. Geminate vowels are rare in Enggano (with only one attestation /kããh/ [kã.ãh] ‘afraid’ in the data), while heterorganic vocoid sequences are common. The glottal stop in these cases could then be analyzed as a surface insertion between geminate vowels. The remaining 13% of occurrences of intervocalic glottal stop do not fit this, however. These occurrences are between heterorganic vowels and cannot be explained as surface insertions. There is no principled way of distinguishing between heterorganic vowel sequences and sequences of heterorganic vowels separated by a glottal stop, so all glottal stops must be regarded as phonemic.
The glottal stop also occurs word medially in consonant sequences, usually as the first member of the sequence as in /doʔra/ [doʔə̆ra] ‘sand’. 14
The lateral approximant [l] is rare in Enggano. It occurs in only five words in the data. They are listed in (10). (10) /pa lauaʔ/ [palawaʔ] ‘split (v.)’ /ka lel/ [kalel] ‘soft’ /kaʔ kaluʔ/ [kaʔə̆kaluʔ] ‘rumpled’ /ka leʔleʔ/ [kaleʔə̆leʔ] ‘muddy’ /kaʔ blau/ [kaʔə̆blau̯] ‘blue’
13 Some younger speakers also use the form [dudian] with final [n]. The use of final [n] is probably because of influence from the Malay/Indonesian form of the word [durian] . 14 The inserted schwa in the surface form is predictable. See section 3.3 for a discussion.
24 The lateral approximant occurs intervocalically four times, finally one time, and medially following another consonant two times. The word /kaʔ blau/ ‘blue’ contains the only underlying three consonant sequence in the data. This is an assimilated loanword, probably from Dutch /blau/ ‘blue’ and with the Enggano adjectival prefix
/kaʔ /. The lateral approximant alternates with /d/ and /r/ in the word [kader] ~
[karer] ~ [kalel] ‘soft’.15
The fricative /x/ only occurs finally, in thirty three words in the data. According to my tentative analysis, it contrasts with other consonants in final position, as discussed in section 3.1.3.
3.1.2 Consonant sequences
Sequences of two consonants are common in Enggano, but the range of consonants that can occur in a sequence is restricted. Consonant sequences only occur word internally and are found most frequently at morpheme boundaries between a prefix and stem. There are no tautosyllabic consonant sequences in native words. The first consonant in a sequence is nearly always /ʔ/, /h/ or /r/. There are three consonant sequences in the data whose first member is a different consonant. Two are reduplicated forms: /kaʔ riprip/ [kaʔə̆riprip] ‘muddy’ with the sequence /pr/ and /ka bikbik/
[kabikbik] ‘muddy’ with the sequence /kb/. The third is the borrowed word /kaʔ blau/ [kaʔə̆blau] ‘green’ with the underlying three consonant sequence /ʔbl/. Table 6 shows usual sequences attested in phonemic representations.
15 It is not clear whether these forms alternate based on speaker, or whether the same speaker uses different forms.
25 Table 6. Consonant co occurrences in underlying forms
_p _b _t _d _k _ʔ _m _n _x _h _r _l ʔ √ √ √ √ √ √ √ √ √ √ h √ √ √ √ √ √ √ √ √ r √ √ √ √ √ √
Out of thirty six possible consonant co occurrences, twenty five are attested and eleven are not. The fact that there are no geminate sequences explains three gaps. /x/ does not occur in sequences since its restriction to word final position does not overlap with the word internal restriction of consonant sequences. This explains three more gaps. There are no attestations of /hl/ or /rl/, presumably because /l/ is so rare in the language.
These can be considered accidental gaps. The sequence /rm/ is not attested, and this is expected since voiced oral consonants do not co occur with nasal consonants in the same word. The sequence /rn/ is attested only in the borrowed word /parna/ [pãrnã]
‘paper’. The sequences /rh/ and /rʔ/ remain the only unexplained gaps in consonant co occurrences. It is possible that these are accidental gaps.
3.1.3 Vowel intrusion
In consonant sequences where the first consonant is a glottal stop, a short schwa vowel [ə̆] is inserted between the two consonants of the sequence. This occurs both morpheme internally and at morpheme boundaries. Morpheme internally, it occurs in words such as /doʔra/ [doʔə̆ra] ‘sand’ and /koʔma/ [kõʔə̆mã] ‘porch’. At morpheme boundaries it occurs in words like /kaʔ tok/ [kaʔə̆tok] ‘red’ and /kaʔ he/ [kaʔə̆he]
‘go octopus hunting’. Insertion only occurs in clusters involving a glottal stop and a true consonant. There is no insertion between a glottal stop and a following surface approximant; e.g. /kõʔĩã/ [kõʔjã] ‘melinjo nut’, or in consonant sequences that begin with /h/ or /r/, as in /kah mih/ [kãhmĩh] ‘squeeze’ and /ʔarkix/ [ʔarkiç] ‘rice’.
26 There is one occurrence of the sequence /hʔ/. Schwa insertion occurs in this sequence:
/kahʔuis/ [kahə̆ʔu.is] ‘sea crab’.
Hall (2006) distinguishes between epenthetic vowels, which are phonologically visible, and intrusive vowels, which are phonologically invisble. Epenthetic vowels can be of any vowel quality and can occur in various types of consonant clusters, while intrusive vowels are consistently schwa like and only occur in heterorganic consonant clusters.
Enggano inserted vowels have several of the properties of intrusive vowels. They are schwa like, and only occur in heterorganic consonant clusters. The inserted vowel
[ə̆] does not interact with stress patterns or syllable structure. Enggano secondary stress is consistently placed on alternating syllables right to left (see section 3.4), regardless of whether an inserted vowel is present or not. This is seen in the words /ʔanoʔon/
[ˌʔãnõʔˈõn] ‘heel’ and /koʔnene/ [ˌkõʔə̆nẽˈnẽ] ‘flirt’. The inserted vowel also does not interact with syllable structure. There are several instances of a sequence of glottal consonants followed by a sequence of vocoids; e.g. /kaʔhɨɘ/ [kaʔə̆.hɨ.ɘ] ‘female leader’ and /kahʔuix/ [kahə̆ʔu.is] ‘sea crab’. High vocoids following post vocalic glottal consonants are nonsyllabic, as in /kõʔĩã/ [kõʔ.jã] ‘tree sp.’ (see section 4.2). If the inserted vowel were phonologically visible, the high vocoid following the second glottal consonant should be nonsyllabic; e.g. *[kaʔ.ə̆h.ɨ̯ɘ] and *[kah.ə̆ʔ.wis] . But since the intrusive vowel is not phonologically visible, the glottal consonant before the high vocoid is post consonantal and the vocoid is consequently syllabic.
Representation of intrusive vowels in relation to syllable structure is a matter of some controversy (Hall 2006, 397). Because the Enggano intrusive vowels vowels are invisible to stress patterns, I follow Hall in not including such vowels in syllable structure. I do not mark syllable boundaries in phonetic forms for the intrusive vowel, assuming that the vowel does not belong to either syllable. For example, the word
27 /kõʔkõʔ/ [kõʔə̆koʔ] ‘sago palm’, while not marked for syllabicity, is assumed to have two CVC syllables.
3.1.4 Place assimilation of /x/
The three fricatives [x], [ç], and [s] only occur word finally. They appear to be in complementary distribution. I present below an initial analysis based on the available data. The analysis presented here is tentative since there are several factors related to these segments that remain difficult to explain.
I analyze the three fricatives as allophones of the phoneme /x/. This phoneme is realized as [x] after /ɨ/, /ɘ/, and /u/, as [ç] after /i/, /ai/ and /ãĩ/, and as [s] after vocoid sequences ending in /i/, including /ai/. The distributions of [ç] and [s] overlap after /ai/. The phoneme only occurs after these six vowels in its thirty three occurrences in the data. The three allophones of /x/ are shown in (11), (12), and (13). (11) [x] after non front vowels /ʔabɨx/ [ʔabɨx] ‘already’ /kɨx/ [kɨx] ‘turtle’ /kɘx/ [kɘx] ‘mountain’ /ʔudɘx/ [ʔudɘx] ‘shampoo’ /ka pu x/ [kapu x] ‘sick’
(12) [ç] after /i/, /ai/, /ãĩ/ /kaʔ pi x/ [kaʔə̆pi ç] ‘sudsy’ /di x/ [di ç] ‘earthquake’ /kai x/ [ka ç] ‘box’ /pãĩ x/ [pã ç] ‘pestle’
(13) [s] after vocoid sequences ending in /i/: /ii/, /ai/, /ɘi/, /ui/ /ii x/ [ji s] ‘word’ /kai x/ [kais ̯ ] ‘sound’ /pɘix/ [pɘis ̯ ] ‘machete’ /kahʔui x/ [kahə̆ʔu.is] ‘sea crab’ /juʔui x/ [juʔ.wi s] ‘oar’
I analyze the sequence [aç] as coming from underlying [aix]. In several words, [ais] alternates with [aç]; e.g. /kaix/ [kaç] ~ [kais] ‘box’ and /maix/ [mãç] ~ [mãĩs]
‘gold’. There are some words ending in [aç] that have no known [ais] alternant (e.g.
28 /pãĩx/ [pãç] ‘pestle’ and /taix/ [taç] ‘bag’), but there are no known words ending in
[ais] that do not have an [aç] alternant. Alternatively, it could be argued that [aç] is underlyingly /ax/, but it would be difficult to find a motivation for a process where /x/ becomes [ç] after /a/. The process /ix/→[ç] is phonetically grounded. The alternation in words like [kaç] ~ [kais] ‘box’ is straightforward if both are analyzed as underlyingly /kaix/ .
Although /x/ only occurs finally, it contrasts with other consonants in this position.
There is contrast between /x/ and /h/, stops, and the absence of /x/. This is shown in
Table 7.
Table 7. Contrast for /x/
Example Contrast /x/ vs /k/ /ka pux/ [kapux] ‘sick’ /puk/ [puk] ‘navel’ /x/ vs /k/ /ki didix/ [kididiç] ‘too big (adj.)’ /ki didik/ [kididik] ‘diligent’ /x/ vs /h/ /kɘx/ [kɘx] ‘mountain’ /kɘʔɘh/ [kɘʔɘh] ‘plant (n.)’ /x/ vs /t/ /iix/ [jis] ‘word’ /ʔit/ [ʔit] ‘banana’ /x/ vs ∅∅∅ /ʔabɨx/ [ʔabɨx] ‘already’ /kapɨ/ [kapɨ] ‘cold’ /x/ vs ∅∅∅ /kaix/ [kaç] ‘box’ /ʔeai/ [ʔe.ai] ‘fish’ /x/ vs ∅∅∅ /iix/ [jis] ‘word’ /ii/ [ji] ‘behavior’
The distribution of /x/ does not overlap with that of /t/. /x/ occurs as [s] only after a vocoid sequence ending in /i/, while final /t/ only occurs after interconsonantal /i/.
Because the distributions of [s] and [t] do not overlap, they could be analyzed as allophones of a single phoneme /t/. However, [s] is better regarded as an allophone of
/x/ since it is associated with /x/ in the alternation [aç] ~ [ais].
A problem with this analysis of the fricatives is their unusual distribution, particularly the distribution of [ç] and [s] after post consonantal and post vocalic /i/.
Another problem is that they only occur word finally. This is an odd distribution for a phoneme, and it remains unexplained why /x/ does not occur in word initial or medial position. The inflected possessive form of words ending in [aç] may give a clue
29 regarding an alternate analysis. The word /tax/ [taç] ‘bag’ (probably from Malay /tas/
[tas] ‘bag’) has forms with a possessive suffix including the sequence [ahi], such as
[tahiʔ] ‘my bag’ and [tahib] ‘your (sg.) bag’. While the exact shape of the possessive suffixes is not clear (see Appendix A), it may be significant that the final [ç] of the stem is [h] or [hi] in suffixed forms. More research is needed in this area.
3.1.5 Glottal consonant palatalization
In some words, glottal stops are optionally palatalized in surface form. The existence and duration of the palatal offglide varies, with some speakers completely omitting palatalization and others producing a very short palatal glide. Examples (14),
(15) and (16) are verbs with optional prefix /ki /, while (17) and (18) have the obligatory locative prefix /ʔi /. (14) /ʔu/ [ʔu] ‘say’ /ki ʔu/ [kiʔu] ~ [kiʔʲu] ‘say’
(15) /hẽk/ [hẽk] ‘sit’ /ki hẽk/ [kĩhẽk] ~ [kĩhʲẽk] ‘sit’
(16) /hau/ [hau̯] ‘bite’ /ki hau/ [kihau̯] ~ [kihʲau̯] ‘bite’
(17) /ʔi ʔẽʔ/ [ʔĩʔẽʔ] ~ [ʔĩʔʲẽʔ] ‘here’
(18) /ʔi ʔõʔ/ [ʔĩʔõʔ] ~ [ʔĩʔʲõʔ] ‘there’
Presumably, glottal consonant palatization occurs morpheme internally as well. In this environment it would be difficult to distinguish from a [iʔj] sequence, as in /kiʔia/
[kiʔja] ‘mosquito’, which could potentially be analyzed as /kiʔa/ [kiʔʲa] . However, there are no words with the morpheme internal [iʔj] sequence that have an alternate form without the glide. Since the words in the data with this sequence do not appear to have the alternate form without the palatal glide, I tentatively analyze them as having the underlying post glottal high vowel. Further research may reveal words where there
30 is optional palatization morpheme internally. There are words where palatization of the glottal consonant does not occur after /i/, as in /kiʔuɘi/ [kiʔwəi̯] (a term of address).
Labialization of the glottal stop occurs morpheme internally in one word after a high back vowel: /kũʔĩ/ [kũʔĩ] ~ [kũʔʷĩ] ‘great grandparent’.
3.1.6 [r] and [ɾ] allophones of rhotic consonant
The consonant /r/ is realized as either a trill [r] or a tap [ɾ]. In word initial and final positions it is realized as [r], as in /roro/ [roɾo] ‘pick up’ and /ʔadɘhɘr/
[ʔadɘhɘr] ‘grass’. In word medial position it is most often realized as a single tap [ɾ].
This is true in consonant sequences as in /iurpuʔ/ [juɾpuʔ] ‘knee’, as well as intervocalically, as in /kare/ [kaɾe] ‘traditional dance’. But there is some variability in word medial /r/, as some instances of the consonant have two taps rather than just one.
This variability is even present in different productions of the same words, as in
/ʔarop/ [ʔaɾop] ~ [ʔarop] ‘four’ and /roro/ [roɾo] ~ [roro] ‘pick up’. In summary, word initial and word final /r/ are realized with two or more taps, while word medial
/r/ is realized with one or two taps. Since there is no straightforward way to transcribe
‘two or more taps’ as opposed to ‘one or two taps’ in phonetic transcriptions, I have chosen to simply use the trill symbol [r] for phonetic representation of the rhotic consonant in all word positions.
3.2 Vowels
Enggano has seven oral vowels and seven corresponding nasal vowels. 16 Table 8 shows the inventory of vowel phonemes. Examples of each vowel in prestress, final closed syllable, and absolute final positions are found in Appendix F.
16 As with consonants, previous authors do not agree on the number of vowel phonemes. Kähler’s dictionary lists six oral and six corresponding nasal vowels, apparently not distinguishing
31 Table 8. Enggano vowel phonemes
Front Central Back High i ĩ ɨ ̃ u ũ Mid e ẽ ɘ ɘ̃ o õ Low a ã
Diphthongs: /ai/, /aɨ/, /au/, /ei/, /ɘi/, /oi/
By far the most common vowel in the data is the low vowel /a/. The next most common vowel /i/ occurs about half as frequently as /a/. The rarest of the vowels is the high central nasal vowel / ̃/. It occurs only three times, in the words /p ̃ʔ/ [p ̃ʔ] ‘fireplace’,
/ĩũʔp ̃ãʔ/ [jũʔə̆p ̃.ãʔ] ‘fence’, and /ka ʔ ̃ʔ/ [kãʔ. ̃ʔ] ‘strong’. The second rarest vowel
/ɘ̃/ occurs eighteen times.
There are numerous surface diphthongs in Enggano. The vowel in the syllable nucleus is always low or mid, followed by a mid or high offglide. There is no contrast between mid and high offglides. Mid offglides only occur in closed syllables, while high offglides only occur in open syllables. Diphthongs in open syllables are shown in (19), while diphthongs in closed syllables are shown in (20). (19) Diphthongs in open syllables /ai/ [ai ̯] /aɨ/ [aɨ ̯] /au/ [au̯] /ei/ [ei ̯] /əi/ [əi ̯] /oi/ [oi ̯]
(20) Diphthongs in closed syllables /ai/ [aɪ ̯] /au/ [aʊ̯] /oi/ [oɪ ̯]
between the two central vowels [ɨ] and [ɘ]. Nothofer’s (1992) list is identical to mine, with seven oral and seven nasal vowels.
32 3.2.1 Vowel nasality
Oral and nasal vowels contrast in words containing only oral consonants. In words with nasal consonants all the vowels in the word are nasalized, as in /ʔaken/ [ʔãkẽn]
‘shark’. Nasalization spreads across morpheme boundaries as well, as in /kaʔ kĩh/
[kãʔə̆k ̃h] ‘dry’. Vowels can be divided into three groups based on nasality: 1) underlyingly oral vowels that remain oral (non nasalized) in their surface form, 2) underlyingly nasal vowels that retain nasality in their surface form, and 3) vowels that are predictably nasalized from nasal consonants in the same word. The first two groups of vowels only occur in words without nasal consonants. They are not influenced by surrounding context and are unambiguously either oral or nasal. The third group of vowels occurs in words with nasal consonants. The oral/nasal contrast is neutralized in this environment. Vowels following the nasal consonant are heavily nasalized, while vowels preceding the nasal consonant are more lightly nasalized. Examples (21), (22), and (23) are of oral vowels, nasal vowels, and vowels whose nasalization is predictable from context, respectively. (21) Words with oral vowels /pap/ [pap] ‘cheek’ /ʔɨaʔ/ [ʔɨ.aʔ] ‘tie’ /kit/ [kit] ‘try’ /kudiʔ/ [kudiʔ] ‘belt’
(22) Words with nasal vowels /pĩh/ [pĩh] ‘squeeze’ /ka ʔ ̃ʔ/ [kãʔ. ̃ʔ] ‘strong’ /kãp/ [kãp] ‘tribal head’ /kũkũ/ [kũkũ] ‘follow’
(23) Words with predictable vowels /mɘk/ [mɘ̃k] ‘many’ /noʔoe/ [nõʔ.õ̯ẽ] ‘spilled’ /kin/ [kĩn] ‘tall’ /kumu/ [kũmũ] ‘well (n.)’
33 As these examples show, oral and nasal vowels occur in words with only oral consonants. Nasality spreads from nasal consonants to all vowels in the word.
I analyze vowels in words with nasal consonants as underlyingly oral, based on two pieces of evidence. First is the morphophonemic spreading of nasality. The nasality of stem segments spreads leftwards across morpheme boundaries to the prefix vowel, as in
/kaʔ kĩh/ [kãʔə̆k ̃h] ‘dry’ and /kaʔ man/ [kãʔə̆mãn] ‘fragrant’ . The vowel of this adjectival prefix is oral otherwise, as seen in /kaʔ pix/ [kaʔə̆piç] ‘sudsy’. Underlyingly oral vowels are nasalized through nasal spreading. As nasal spreading is already present in Enggano as a morphophonemic process, it may well apply as a phonological process morpheme internally as well. Secondly, vowels preceding nasal consonants sound less nasalized than those in words with no nasal consonants. This difference in surface form could point to an underlying difference in the nasality of the two types of vowels.
Vowels in words with no nasal consonants are [+nasal] in underlying form and retain this specification in surface form. Underlyingly oral vowels in words with nasal consonants lose their [ nasal] specification and take on the [+nasal] specification of the nasal consonant. Nasalization spread leftward is articulated less clearly than than right spreading nasalization and underlying nasality.
Underlyingly oral vowels could also be analyzed as underlyingly unspecified for nasality, depending on the theoretical framework. In this case, all oral and predictably nasalized vowels would have no specification for nasality in the underlying form. They would receive the default specification of [ nasal] except when the [+nasal] feature spreads from nasal consonants. Nasal vowels would be specified as [+nasal] in the underlying form. The question of whether the oral vowels are underlyingly unspecified or oral does not have direct relevance for the points pursued in this thesis. It will not be pursued further here.
34 3.2.2 Restrictions on distribution
All vowels occur in medial and final positions, both after consonants and after other vowels. Only /i/, /ĩ/ and /u/ occur initially, and always as surface approximants.
Sequences of two vocoids can be analyzed as disyllabic sequences of vowels (e.g.
/kiak/ [ki.ak] ‘blood’), diphthongs (e.g. /ka karai/ [ka.ka.rai̯] ‘chase’), or sequences of approximant and vowel (e.g. /kõʔĩã/ [kõʔ.jã] ‘tree sp.’). Chapter 4 discusses vocoid sequences from the perspective of phonology while chapter 6 presents an acoustic analysis of vocoid sequences. The remainder of this chapter is a discussion of restrictions on the distribution of individual vowels.
There is no contrast between high vowels /i/, /u/ and mid vowels /e/, /o/ when they are the second part of a diphthong in a word’s final syllable. These vowels in closed syllables are realized as [ɪ ̯] and [ʊ̯] as in /pãĩk/ [pãɪ̯̃k] ‘river shrimp’ and
/bahauʔ/ [bah.aʊ̯ʔ] ‘heart’. In open syllables they are realized as [i ̯] and [u̯] as in
/ka karai/ [kakarai̯] ‘chase’ and /baʔau/ [baʔ.au̯] ‘guava’.
Examples of high vowel glide lowering are shown in (24), while the absence of the process in plain vowels and diphthongs in open syllables is shown in (25). (24) High vowel glide lowering in closed syllable diphthongs /p ai ʔ/ [p aɪ̯ʔ] ‘throw’ /i ai ʔ/ [j aɪ̯ʔ] ‘sew’ /iaʔkaʔai ʔ/ [jaʔkaʔ.aɪ̯ʔ] ‘war’ /p ãĩ k/ [p ãɪ̯̃k] ‘shrimp’ /ʔau p/ [ʔaʊ̯p] ‘four’ /bah au ʔ/ [bah.aʊ̯ʔ] ‘heart’ /kĩpãʔãũ p/ [kĩpãʔ.ãʊ̯̃p] ‘ten’
(25) No lowering in plain vowels or open syllables /kop iʔ/ [kop iʔ] ‘suck’ /i iʔ/ [j iʔ] ‘buttocks’ /nah ai / [nãh.ã ̯̃] ‘allow’ /ʔeai / [ʔe. ai̯] ‘fish’ /ka puʔ/ [kap uʔ] ‘collapsed’ /baʔau / [baʔ.au̯] ‘guava’ /ki hau / [kihjau̯] ‘bite’
35 3.3 Syllable structure
There are four main syllable types in Enggano: CV, CVC, CVV, and CVVC. Three additional, vowel initial syllable types (V, VC, and VV) occur after vowels. All underlying vocoid initial words have an initial approximant in surface form. There are no surface vowel initial words. Table 9 shows all seven types, along with examples of each and an explanation of restrictions on syllable positions within words. The notation
LH refers to height specification of vocoids in sequences, where the first vocoid is lower than the second.
Table 9. Syllable types in Enggano
Syllable Underlying Surface Gloss Environments type CV /be/ [be] ‘dog’ All environments /kapa/ [kapa] ‘child’ CVC /kũk/ [kũk] ‘back’ All environments /porpiʔ/ [porpiʔ] ‘cigarette paper’ CVV /pau/ [pau̯] ‘horizon’ All environments /painɘn/ [pã ̯̃nɘ̃n] ‘feelings’ CVVC /paiʔ/ [paɪ ̯ʔ] ‘throw’ Only occurs in monosyllabic /pɘix/ [pɘi ̯s] ‘machete’ words; vocoid sequence LH V /mea/ [mẽ.ã] ‘cat’ Only occurs in two vocoid /ʔeaka/ [ʔe.aka] ‘crab sp.’ sequences (not LH) VC /kããh/ [kã.ãh] 17 ‘afraid’ Only occurs in two vocoid /kiɘhɘr/ [ki.ɘh.ɘr] ‘shore crab’ sequences (not LH) VV /buai/ [bu.ai ̯] ‘crocodile’ only occurs finally in three /ʔeai/ [ʔe.ai ̯] ‘fish’ vocoid sequences
Only CV, CVC and CVV syllables occur in all environments. CVVC syllables occur only finally, with the exception of /ko koiʔea/ [ko.koɪ̯ʔ.e̯a] ‘lazy’ (see chapter 4 for a discussion of syllabification of glottal consonants). The only environment where VV
17 This word contains the only homorganic vocoid sequence in the data. Phonetically the sequence is a very long vowel with a sharp increase in pitch midway through the vowel. Intensity dips briefly at the halfway point, then rises to a point higher than the previous steady state.
36 syllables occur is in three vocoid sequences of the shape High Low High, which always occur finally.
3.4 Stress pattern
Stress in Enggano falls on the final syllable of the word. 18 Several of the most common word shapes are shown in (26), with primary stress marked. (26) /bibi/ [bi.ˈbi] ‘residence’ /hapɨʔ/ [ha.ˈpɨʔ] ‘day’ /pahpɨ/ [pah.ˈpɨ] ‘see’ /porpiʔ/ [por.ˈpiʔ] ‘cigarette paper’ /kiak/ [ki.ˈak] ‘blood’ /ʔɨdiaʔ/ [ʔɨ.di.ˈaʔ] ‘sell’ /paido/ [pai ̯.ˈdo] ‘cry’ /ʔanɨkɘ/ [ʔã.n ̃.ˈkɘ̃] ‘caterpillar sp.’ /kabakeʔ/ [ka.ba.ˈkeʔ] ‘immediate family’ /karaʔai/ [ka.raʔ.ˈai ̯] ‘tree sp.’
Secondary stress falls on alternating syllables from right to left. Therefore, in three and four syllable words, secondary stress falls on the antepenultimate syllable, as in
/iaʔkeʔei/ [ˌjaʔ.keʔ.ˈei̯] ‘boil (n.)’ and /ʔakiʔakin/ [ʔã.ˌkĩʔ.ã.ˈkĩn] ‘six’.
Stress is also final in loanwords, both assimilated and otherwise unassimilated words. This is seen in the assimilated loanword /tepaʔ/ [teˈpaʔ] ‘kick’ (probably from
Malay /sepak/ [ˈsepak] ~ [sepaʔ] ‘kick’) and in the otherwise unassimilated loanword /tikus/ [tiˈkus] ‘rat’ (from Malay/Indonesian /tikus/ [ˈtikus] ‘rat’). In my tentative analysis, the [s] in [tikus] should be [x], but it remains unassimilated. Strress, however, has shifted from penultimate to final.
To measure the acoustic correlates of stress, thirteen disyllabic words of the shape
CVCV and CVCVC were selected with identical vowels in the first and second syllables.
18 In the introduction to his grammar, Kähler claims that Enggano stress is penultimate. He says, “In Enggano stress in principle occurs on the penultimate syllable. If a word is lengthened through suffixes, the stress is often laid on the vowel of the penultimate syllable, but sometimes may be placed on the original syllable” (1940, 84). It appears that the stress pattern has changed in the past eighty years since modern Enggano stress is clearly final.
37 Acoustic measurements of both the first and second vowels were taken, and then compared. The words were recorded in isolation. This is problematic for measuring vowel length since it is common for phrase final vowels to be lengthened. However, no words of suitable shape were recorded in frames. Measurements of both the first and second vowel of the selected words in terms of duration, intensity, pitch, and first and second formants are shown in Appendix C.
All seven oral vowels are represented in the closed syllable words, and six of the seven are represented in the open syllable words. There were no examples of two high central vowels /ɨ/ in a CVCV word. No underlying nasal vowels were included, although in some of the words (e.g. /mama/ [mãmã] ‘chew’) both vowels are nasalized in surface form.
Duration and pitch show a correlation with the stressed syllable. Intensity does not show an overall correlation, although it is significant in closed syllables. Table 10 shows the average of each group, as well as the results of a t test (paired, 2 tailed) performed on the values for each of the three groups.
Table 10. Correlation of duration, intensity and pitch with word stress
Duration (ms) Intensity (dB) Pitch (Hz) 1st 2nd 1st 2nd 1st 2nd Average 133 271 74 77 109 165 T test p = <0.01 p = 0.16 p = <0.01
With p values below 0.01, duration and pitch both appear to correlate strongly with stress. However, neither duration nor pitch can be conclusively associated with stress because of the type of recordings that were used. The measurements were taken from the first of two repeats of words in isolation. The phrase final position of the second syllable could be the reason for the longer vowels there. This list intonation (first repeat rising, second repeat falling) is likely the reason for at least some of the difference in
38 pitch between the first and second syllables. 19 Intensity does not show a significant difference in Table 10. However, further analysis shows that there is a difference between closed syllable and open syllable words. Intensity between the two vowels in almost all the open syllable words is the same, but is significantly higher in the final syllable when it is a closed syllable. A paired, two tailed t test on only closed syllable words yields a p value of 0.04, while the p value for open syllable words is 0.96.
The place of articulation of vowels does not correlate with stress. Figure 4 charts the vowel quality of stressed and unstressed vowels, showing that the first and second formants of stressed and unstressed vowels are very similar. In the figure, unstressed vowels are represented by lowercase letters while stressed vowels are represented by uppercase letters. The stressed high central vowel is represented by <$>, and the mid central vowel by <&>. There are two unstressed and two stressed tokens of each vowel except [ɨ], which has one token of each.
Figure 4. Vowel quality of stressed and unstressed vowels
19 I assume here that list intonation primarily affects the pitch contour and not duration or intensity. While auditorily this seems to be correct, it has not been proven.
39 The tokens of the two front vowels and three central vowels are very close together. The tokens of the two back vowels are spread farther apart, but both have instances of stressed and unstressed vowel very close to each other.
To summarize, intensity correlates with stress in closed syllables. It remains unknown whether vowel length and pitch correlate with stress, even though the data appears to show a correlation. Vowel quality does not correlate with stress.
The consistent final stress pattern is helpful in determining the status of vocoid sequences as either tautosyllabic diphthongs or disyllabic sequences. For example, in the form [ki.ˈak] ‘blood’ (< /kiak/ ), stress falls on the second vowel [a], while in
[pãɪ̯̃k] ‘shrimp’ (< /paik/ ) it falls primarily on the first vowel [a]. This is the main articulatory factor that forms the basis of the distinction between disyllabic and tautosyllabic sequences discussed in chapter Chapter 4. This criterion is most helpful in determining the status of final syllable vocoid sequences since primary stress falls on the final syllable.
40 CHAPTER 4
SYLLABIFICATION OF VOCOID SEQUENCES
The syllabicity of all vocoids in Enggano is predictable by the phonological environment in which they occur. This yields surface complementary distribution between vowels and glides. In this thesis, I use the term vocoid to refer to underlying forms. The terms vowel and approximant are used to refer to surface syllabic and nonsyllabic vocoids, respectively.
Sequences of two or more vocoids are very frequent in Enggano. This chapter discusses possible attested vocoid sequences (section 4.1), syllabification of vocoid sequences (section 4.2), and the acoustic correlates of syllabification (section 4.3). The chapter ends with a formal analysis of the facts in the framework of Optimality Theory
(section 4.4).
4.1 Attested sequences
Table 11 and Table 12 show all possible two vocoid sequences in Enggano. Table
11 shows the oral sequences, while Table 12 shows the nasal sequences. Oral and nasal vowels do not occur together in the same word. The two tables below reflect counts of underlying forms. Vowels that receive nasalization in surface form from a nasal consonant are included in Table 11. The tables also include syllable initial high vocoids that are glides in surface forms. Each cell gives one or two example words containing
41 the sequence, followed by a number indicating the total number of occurrences of the sequence in the data. 20
Table 11. Oral vocoid sequences
_i _ɨ _u _e _ɘ _o _a T i_ [ji ] [jɨɁ.ɨ.aɁ] [ju ] [je ] [jɘb] [jo ʔ.oʔ] [ja ʔ.a] 67 ‘behavior’ ‘rope’ ‘ocean’ ‘vomit’ ‘fire’ ‘spear’ ‘knife’ [h i.ɨr] [h i.u r] [kar i.ɘ] [p i.a k] ‘woman’ ‘dust’ 1 ‘work’ ‘tired’ 4 6 12 1 9 34 ɨ_ [kak ɨ.ɘ] [ʔɨ.a ʔ] 16 ‘short’ ‘tie’ 3 13 u_ [juʔwi s] [ʔu.eh] [da wo h] [wa ʔ] 17 ‘oar’ ‘sleep’ ‘thunder’ ‘uncle’ [p u.a k] ‘go’ 4 1 2 10 e_ [jaʔə̆keʔ.ei̯] [ʔe.o k] [ʔe.a ] 22 ‘type of ‘vein’ ‘bone’ boil’ 1 3 18 ɘ_ [p ɘi̯s] 3 ‘machete’ [kɘh. ɘi] ‘spilled’ 3 o_ [k oi̯] [boh.o̯e] [d o.a b] 13 ‘pig’ ‘wild’ ‘lightning’ [ʔaj oj aʔ] ‘pay out’ 7 5 1 a_ [k ai̯s] [jar.kɨ.aʔ.aɨ̯] [p au ] 30 ‘sound’ ‘small fence’ ‘horizon’ [ʔaw aʔ] 1 ‘go’ 18 11 T 37 7 23 7 4 14 76 168
20 The total number of each type of sequence may be skewed since the source data is made up of words elicited for phonological features. It is most likely skewed in favor of the less common vocoid sequences since I elicited for at least five examples of all possible vocoid sequences.
42 Table 12. Nasal vocoid sequences
_ĩ _ ̃ _ũ _ẽ _ɘ̃ _õ _ã T ĩ_ [jĩ ʔ.ĩ] [jũ ʔ] [jɘ̃h] [kã jõ ] [kĩʔjã p] ‘type of ‘stem’ ‘cold’ ‘straight’ ‘grasshopper’ tool’ [hĩ.ũk] [kĩʔ.jɘ̃] [pãʔjõ p] ‘louse’ ‘paralyzed’ ‘hug’ 1 5 4 2 4 16 ̃_ [jũʔə̆p ̃.ãʔ] ‘fence’ 1 1 ũ_ [ʔũ.ẽ] ‘cry’ [kũh. wẽ ] ‘usefulness’ 2 2 ẽ_ [ʔẽ.õ] ‘mud’ 4 4 ɘ̃_ 0 õ_ [pãkõʔ.õ̯ãʔ] ‘know’ 1 2 ã_ [k ãjõ] [kãʔ.ãũ̯] [k ã.ãh] ‘straight’ ‘fat’ ‘afraid’ [p ãɪ̯̃k] ‘shrimp’ 1 1 1 4 T 3 0 7 2 4 6 6 29
Out of 49 possible co occurrences, 23 oral and 15 nasal sequences are attested. The smaller number of nasal sequences is congruent with the distribution of oral and nasal vowels in general, where oral vowels occur more frequently than nasal vowels. Nine oral sequences do not have nasal counterparts: /iɨ/, /ie/, /ɨɘ/, /ui/, /uo/, /ei/, /ɘi/,
/oe/, and /aɨ/. The only nasal sequence with no attested oral counterpart is in the word
/kããh/ [kã.ãh] ‘afraid’. This is the only homorganic vocoid sequence in the data.
Vowels at the edges of the vowel space tend to co occur more frequently with other vowels at opposite edges of the vowel space. This is the expected distribution since a
43 vowel transition moving across a large area of the vowel space is easier to recognize than one in which there is little movement. However, there are some co occurrences of vowels rather close to each other, such as /iɨ/ in /hiɨr/ [hi.ɨr] ‘woman’, /ei/ in
/iaʔkeʔei̯/ [jaʔə̆keʔei] ‘boil (n.)’, and /ɨɘ/ in /kakɨɘʔ/ [kakɨ.ɘʔ] ‘short’. These sequences are less common than sequences stretching across the vowel space such as
/ia/, /iu/, and /au/. There are no low mid sequences, although there are both low high and mid high sequences.
Vocoid sequences tend strongly to occur in final stressed position. HL and LH sequences are also more common than same height sequences. There are no same height sequences in non final position.
Besides two vocoid sequences, there are also sequences of three and five vocoids in
Enggano such as /iapaʔioi/ [japaʔ.joi̯] ‘traditional wedding’ and /ʔaioiaʔ/ [ʔajojaʔ]
‘pay out (a rope)’. There are no four vocoid sequences in the data, although this is assumed to be an accidental gap. Syllabification of three or more vocoids is discussed at the end of the discussion of two vocoid sequence syllabification below.
4.2 Syllabification
Vocoid sequences are classified here based on relative height in the vowel space:
LH (lower vocoid followed by higher vocoid), HL (higher vocoid followed by lower vocoid), and SH (two vocoids of the same height). LH sequences include both low high and mid high sequences, while HL sequences include high low, high mid, and mid low sequences. SH sequences are high high, mid mid, and low low. Below I discuss syllabification of vocoid sequences in three environments: word initial, non initial, and following a glottal consonant.
44 Word initial high vocoids, which are always followed by another vocoid, are realized as the onset of the initial syllable. Example word initial vocoid sequences are shown in (27).21 (27) /ua ʔ/ [wa ʔ] ‘uncle’ /ia ʔkaʔaiʔ/ [ja ʔə̆kaʔ.aɪ ̯ʔ] ‘war’ /ie / [je ] ‘vomit’ /io hor/ [jo h.or] ‘song’ /iɘb/ [jɘb] ‘fire’ /ii ʔ/ [ji ʔ] ‘buttocks’ /iu m/ [jũ m] ‘fly (n.)’ /iɨʔɨaʔ/ [jɨʔ.ɨ.aʔ] ‘rope’
The only Enggano vocoids occurring word initially are /i/, /ĩ/ and /u/. The word
/uaʔ/ [waʔ] ‘uncle’ is the only monomorphemic word in the data with initial /u/.
Initial /u/ also occurs stem initially in prefixed forms like /ka uaua/ [ka wawa]
‘open’ and /pa uap/ [pawap] ‘yawn’. The absence of /ũ/ is either an accidental gap or simply unattested because initial high back vocoids are so rare.
In non initial sequences of the shape HL and SH, both vocoids are realized as separate syllable peaks. This is shown in (28) and (29).
(28) HL sequences (VHVL→V σ.V σ) /juk ua / [juk u.a ] ‘pole’ /h io n/ [h ĩ.õn] ‘scratch’ /k ia k/ [k i.a k] ‘blood’ /ka kɨɘʔ/ [kak ɨ.ɘʔ] ‘short’ /ʔea ka/ [ʔe.a ka] ‘close’
(29) SH sequences (ViVi→V σ.V σ) /h ĩũ / [h ĩ.ũ] ‘fruit’ /b ui / [h i.ɨr] ‘woman’ /ki keo r/ [kike.o r] ‘lost’ /k ãã h/ [kã.ãh] ‘afraid’
LH sequences are realized as a vowel followed by an offglide. Examples are shown in (30).
21 Syllabification as marked here is based on subjective perception. The acoustic study in section 4.3 quantifies syllabification patterns.
45
(30) LH seequences(VLVH→V σVAPPROX ) /k ai x/ [k ai̯s] ‘sound’ /p ai do/ [p ai̯.do] ‘cry’ /kakar ai / [ka.ka.r ai̯] ‘chase’ /ʔau p/ [ʔaʊ̯p] ‘four’ /p ai ʔ/ [p aɪ̯ʔ] ‘throw’
As shown in the above examples, the syllablfication patterns are the same in all non initial positions. LH sequences are tautosyllabic while HL and SH sequences are disyllabic.
Vocoid sequences have a different syllabification pattern after non initial glottal consonants /ʔ/ and /h/. In HL and SH sequences, the first vocoid is realized in the onset of the syllable while the second vocoid receives syllable prominence. This is shown in
(31) and (32).
(31) HL sequences after glottal consonant (VHVL→V APPROX Vσ) /ʔaraʔia h/ [ʔaraʔja h] ‘livestock’ /kuʔia ʔ/ [kuʔja ʔ] ‘push’ /kaʔ nih ia / [kaʔə̆n ̃hjã ] ‘dream (v.)’ /ki ʔ ̃ɘ̃/ [kĩʔjɘ̃] ‘paralyzed’
(32) SH sequences after glottal consonant (ViVi→V APPROX Vσ) /iuʔui s/ [juʔwi s] ‘oar’ /boh oe / [boh. o̯e] ‘wild’ /ʔa ʔiu r/ [ʔaʔju r] ‘progress (v)’ /keh iɨ/ [keh jɨ] ‘thirsty’
LH sequences are phonetically realized the same as LH sequences in other environments, that is, as a vowel followed by an offglide. Examples in (33) show LH sequences in this environment.
(33) LH sequences after glottal consonant (VLVH→V σVAPPROX ) /iaʔkaʔai ʔ/ [jaʔə̆kaʔ.aɪ̯ʔ] ‘war’ /kah ai ʔ/ [kah. aɪ̯ʔ] ‘one’ /nah ai / [nãh. ã ̯̃] ‘allow’ /kɘhɘi/ [kɘh. ɘi̯] ‘spilled’
After a non initial glottal consonant, LH sequences are realized as a vowel followed by an offglide while HL and SH sequences are realized as an onglide followed by a vowel. The contrast between the two syllabification patterns is seen in the difference
46 between the LH sequence in /bohoi/ [boh.oi̯] ‘rope’ (shown in Figure 6 on page 51) and the SH sequence in /bohoe/ [boh.o̯e] ‘wild’ (shown in Figure 10 on page 55). The two words sound very similar in terms of vowel quality. However, the sequence in
[boh.oi̯] is syllabified as a LH sequence, with word stress falling on the [o] and the final [i] being realized as an offglide. The sequence in [boh.o̯e] is syllabified as a SH sequence. The initial [o] is realized as an onglide, and word stress falls on the final vowel [e].
Vocoid sequences following initial glottal consonants follow the syllabification pattern of sequences following other consonants, as in /ʔiah/ [ʔi.ah] *[ʔjah] ‘what’ and /hĩũ/ [hĩ.ũ] *[hjũ] ‘fruit’.
When a consonant sequence ending in a glottal consonant comes before a high vocoid, the vocoid is syllabic rather than nonsyllabic as it is after intervocalic glottal consonants; e.g. /kaʔhɨɘ/ [kaʔə̆hɨ.ɘ] *[kaʔə̆h.ɨ̯ɘ] ‘female leader’ and /kahʔuix/
[kahə̆ʔu.is] *[kahə̆ʔ.wis] ‘crab sp.’. Since the inserted schwa is invisible to phonological rules (see section 3.3), a nonsyllabic high vocoid here would create a sequence of three consonants. Three consonant sequences are not attested in Enggano.
In summary, word initial sequences are always HL, where the H vocoid is in the syllable onset. LH sequences are always realized as a vowel followed by an offglide.
Non initial HL and SH sequences syllabify differently depending on the preceding consonant. If the segment preceding the sequence is any consonant but glottal /ʔ/ or
/h/, both vocoids are realized as separate syllable peaks. If the preceding consonant is
/ʔ/ or /h/, HL and SH sequences are realized as an onglide followed by a vowel. Table
13 summarizes these patterns.
47 Table 13. Syllabification patterns in vocoid sequences
LH HL SH
After non initial /ʔ/, /h/ VσVAPPROX VAPPROX Vσ VAPPROX Vσ
Elsewhere VσVAPPROX Vσ.V σ Vσ.V σ
Sequences of three or more vocoids follow similar syllabification patterns. There are two main types of sequences: those beginning and ending with a low vowel, (shown in (34)), and those beginning and ending with a high vowel (shown in (35)).
(34) LHL Sequences (…V LVHVL…) /nan aia ʔ/ [nãnãjã ʔ] ‘ruin’ /ʔaua h/ [ʔawa h] ‘go’ /k aio / [k ajo ] ‘straight’ /k eie r/ [k eje r] ‘water plant sp.’ /d auo h/ [d awo h] ‘thunder’ /d oio ʔ/ [d ojo ʔ] ‘pot’ /k a uaua / [k awawa ] ‘open (adj.)’ /ʔa ioia ʔ/ [ʔajoja ʔ] ‘pay out (a rope)’
(35) HLH Sequence (…V HVLVH…) /iapaʔioi / [japaʔjoi̯] ‘traditional wedding’ /kiʔuɘi/ [kiʔwɘi̯] (a term of address for young girl) /kaʔ dɨhɨai / [kaʔə̆dɨh. ɨ̯ai̯] ‘bitter’ /ʔeai / [ʔe.ai̯] ‘fish’
There is a single word with a LHH sequence, /tauud/ [tawud] ‘year’, which is a loanword.
All these sequences can be explained with phonological processes already discussed for two vocoid sequences. In LHL sequences, the middle H vowel is nonsyllabic as in LH sequences, and is realized as the onset of the following syllable. The five vocoid LHLHL sequences are simply longer sequences where the same process occurs. The LHH sequence is a LH sequence followed by a SH sequence. The generalization that SH sequences after non glottals are two syllables does not apply here. However, SH sequences in word initial position are tautosyllabic, as in /ii/ [ji] ‘behavior’. This generalization can be extended here to say that a high vocoid before V is nonsyllabic initially or after V.
48 Three of the HLH sequences are word final and are preceded by a glottal consonant. The first H vowel is nonsyllabic because of the preceding glottal consonant, and the last is nonsyllabic because it is the second part of a LH sequence. The sequence is thus realized as a L vowel with a H onglide and a H offglide. In the word /ʔeai/
[ʔe.ai̯] ‘fish’ the first vowel remains syllabic as there is no vowel before the preceding consonant, and it is the first vowel in the word.
4.3 Acoustic correlates of syllabification patterns
This section presents acoustic findings that reinforce the perceptual judgements in section 4.2. In section 4.3.1 I show the differences in intensity between different types of vocoid sequences, while in section 4.3.2 I discuss duration differences between the different types.
4.3.1 Intensity
In vowel glide sequences, the first vowel is articulated with more intensity, and the intensity drops off for the glide. In sequences with separate syllable peaks, intensity rises to a peak midway between the two vowels and tapers off again. In onglide vowel sequences the onglide is short, with intensity rising steeply through the onglide to a peak in the following vowel. Below I show two examples of each syllabification pattern, one example from each of the cells in Table 13. Figure 5 shows the LH sequence in
/pɘis/ [pɘi̯s] ‘machete’.
49 Figure 5. Intensity contour of [ɘi ̯] in [pɘi̯s] ‘machete’
The intensity of the first vowel in the LH sequence is much greater than that of the second. This is the typical shape of intensity contours in LH sequences. Figure 6 shows a
LH sequence following /h/ in the word /bohoi/ [boh.oi̯] ‘rope’.
50 Figure 6. Intensity contour of [oi ̯] in [boh.oi̯] ‘rope’
The intensity contour of [oi ̯] in Figure 6 looks very similar to that of [ɘi ̯] in Figure 5.
The first vowel has greater intensity than the second vowel, whose intensity drops off to the end. The spectrogram shows that the transition from one vocoid to the next takes place about halfway through the sequence.
In contrast to the vowel offglide sequences shown above, sequences that syllabify as separate syllable peaks have an intensity contour that rises more gradually and reaches a peak midway through the sequence. The HL sequence in /piɘh/ [pi.ɘh]
‘massage’ is shown below in Figure 7.
51 Figure 7. Intensity contour of [iɘ] in [pi.ɘh] ‘massage’
The HL sequence here has most intensity in the middle of the transition between vowels, with a gradual rise to it and gradual tapering off after it. An example of a SH sequence with the same syllabification is shown in Figure 8 below, in the word /hiɨr/
[hi.ɨr] ‘woman’.
52 Figure 8. Intensity contour of [iɨ] in [hi.ɨr] ‘woman’
As with the HL sequence in Figure 6, the SH sequence here has most intensity in the middle of the transition between vowels, with gradual rise to it and gradual tapering off after it.
Sequences of an onglide followed by a vowel are characterized primarily by the short duration of the onglide. Intensity rises through the onglide to reach a peak in the following vowel. The HL sequence in /iihion/ [jĩh.jõn] ‘type of tool’ in Figure 9 illustrates this.
53 Figure 9. Intensity contour of [jõ] in [jĩhjõn] ‘type of tool’
The SH sequence in /bohoe/ [boh.o̯e] ‘wild’ in Figure 10 below has a similar pattern.
54 Figure 10. Intensity contour of [o̯e] in [boh.o̯e] ‘wild’
In this sequence, the [o̯] is very short in duration, transitioning quickly into the [e]. The
[o̯] has lower intensity than the [e], whose greatest intensity is at the beginning of its articulation. There is a great difference in segment duration and pitch contour between the sequence [o̯e] shown in Figure 10 and the sequence [oi ̯] shown in Figure 6. The only other pair of words in the data that share this contrast, /kõʔõĩʔ/ [kõʔ.õ ̯̃ʔ] ‘gray’ and /koʔoeʔ/ [koʔ.o̯eʔ] ‘devil’, show spectrograms and intensity contours that are nearly identical to the ones above.
To summarize the findings, intensity contours are different in vowel offglide, vowel vowel, and onglide vowel sequences. Disyllabic vowel vowel sequences have greatest intensity in the middle of the vowel transition. Vowels with offglides have greatest intensity in the vowel, with intensity tapering off at the end. Vowels with
55 onglides have a short transition into the vowel, with intensity rising through the onglide to reach a peak in the following vowel.
4.3.2 Duration
Auditorily, sequences of vowels with separate syllable peaks sound longer in duration than vowels with onglides and offglides. To test my auditory perception, several sequences of each type were measured for duration. Since sequences occur most frequently in final position, only final sequences were chosen for analysis. Closed and open syllables were separated since this affects duration of vowels. Words with a medial glottal consonant were separated from those with none. This gave a four way distinction. Combined with the three way height distinction (LH, HL, SH), this gave twelve categories for final vocoid sequences. The words selected for each of the twelve categories are shown in Table 14.
56 Table 14. Final vocoid sequences for duration measurement
No preceding glottal /ʔ/ or /h/ Preceding /ʔ/ or /h/ Closed Open Closed Open LH [j aɪ̯ʔ] ‘sew’ [kakar ai̯] ‘chase’ [jaʔə̆kaʔ.aɪ̯ʔ] ‘war’ [karaʔ.ai̯] ‘tree sp.’ [p ãɪ̯̃k] ‘shrimp’ [kõʔə̆mã ̯̃] [kah.aɪ̯ʔ] ‘one’ [ʔãmãh.ã ̯̃] ‘bed’ [k ai̯s] ‘sound’ ‘immigrant’ [kabakah.aɪ̯ʔ] ‘nine’ [nãh.ã ̯̃] ‘allow’ [ʔaʊ̯p] ‘four’ [kãnã ̯̃] ‘all [bah.aʊ̯ʔ] ‘heart’ [baʔ.au̯] ‘guava’ [p ɘi̯s] ‘machete’ gone’ [pãmãh.ãʊ̯̃m] [haʔ.au̯] ‘ocean’ [jor oi̯] ‘husk’ ‘afternoon’ [mãʔ.ãũ̯] ‘local [ʔiht ɘi̯] ‘mock’ [kipaʔ.aʊ̯p] ‘ten’ people’ [kak ɘi̯] ‘nearly [kõʔ.õɪ̯̃ʔ] ‘gray hair’ [jaʔə̆keʔ.ei̯] ‘type of breaking’ boil’ [kĩmɘ̃mɘ̃ ̯̃] [pajɘʔ.oi̯] ‘overcast’ ‘nauseous’ [pĩnõh.õ ̯̃] ‘naïve’ [kɘh.ɘi̯] ‘spilled’ HL [ʔap i.a h] [kar o.a ] ‘pants’ [ʔaraʔja h] ‘livestock [kiʔja ] ‘mosquito’ ‘graze’ [juk u.a ] ‘pole’ of deceased’ [kõʔjã ] ‘goiter’ [ʔɨdi.a ʔ] ‘sell’ [kor e.a ] ‘gull’ [kuʔja ʔ] ‘push’ [kãʔə̆n ̃hjã ] ‘dream’ [p i.a k] ‘tired’ [kar o.a ] ‘pants’ [kĩʔjã p] ‘grasshopper’ [kokoɪ ̯ʔ.e̯a] ‘lazy’ [kar u.a ʔ] ‘go [kar i.ɘ] ‘work’ [kũʔwã h] ‘vehicle’ [kaʔeaʔ.e̯a] ‘skinny’ first’ [koh. e̯aʔ] ‘hut’ [pahkoʔ.o̯a] ‘love’ [p u.a k] ‘go’ [paʔ.o̯aʔ] ‘shout’ [doh. o̯a] ‘boat’ [p u.a h] ‘stir’ [karkoʔ.o̯aç] ‘night’ [kokoʔ.o̯a] ‘love’ [ʔõnẽ.ãʔ] ‘go [jamakaʔ.o̯aʔ] [kĩʔjɘ̃] ‘paralyzed’ together’ ‘thought’ [kĩʔjɘ̃k ̃ʔjɘ̃] ‘sea [kapar e.a k] [ʔĩʔjẽ ʔ] ‘here’ creature sp.’ ‘banana leaf’ [kiʔjo p] ‘face down’ [d o.a b] [ʔĩʔjõ ʔ] ‘there’ ‘lightning’ [jĩhjõ n] ‘type of tool’ [k o.a r] ‘type of [kipeh jɘr] ‘deaf’ utensil’ [k õ.ãn] ‘The Lord’ [p i.ɘh] ‘massage’ SH [kid e.o k] [b u.i ] ‘prison’ [ʔaʔju r] ‘progress’ [keh. jɨ] ‘thirsty’ ‘pinched’ [juʔwi s] ‘oar’ [boh. o̯e] ‘wild’ [kik e.o r] ‘lost’ [kahə̆ʔu.is] ‘crab sp.’ [joʔə̆ho.e] ‘bed’ [koʔ.o̯eʔ] ‘devil’ [nõʔ.õ̯ẽ] ‘spilled’
Most words used for analysis are disyllabic or monosyllabic, with a few three and four syllable words.
57 Final vowels in monosyllabic words tend to be longer than final vowels in disyllabic words. Fourteen vowels in monosyllabic words were compared with fourteen vowels in comparable disyllabic words. The vowels in monosyllabic words were an average of 37 ms longer than in disyllabic words. While word length does have an influence on final vowel length, it was not possible to control for this since some categories had only monosyllabic or only multisyllabic words. The possible ramifications of this are dealt with in the discussion of results below.
Final plain vowels (i.e. vowels not in sequences) were also measured in both open and closed syllables to compare with the vocoid sequences. These are shown in (36) and
(37). All duration measurements for both vocoid sequences and plain vowels are found in Appendix C. (36) Plain vowels in closed syllables /p ik/ [p ik] ‘open’ /p ẽʔ/ [p ẽʔ] ‘frog’ /k ak/ [k ak] ‘person’ /kɨʔ ɨk/ [kɨʔ ɨk] ‘narrow’ /k ɘx/ [k ɘx] ‘mountain’ /jũkũʔ/ [jũkũʔ] ‘wall’ /kõʔkõʔ/ [kõʔə̆kõʔ] ‘sago palm’
(37) Plain vowels in open syllables /ʔɨki/ [ʔɨki] ‘mango’ /k e/ [k e] ‘throwup’ /baʔa/ [baʔa] ‘die’ /p ɨ/ [p ɨ] ‘see’ /ka pɘ̃pɘ̃/ [kãpɘ̃pɘ̃] ‘foggy’ /kĩ kãʔũ/ [kĩkãʔũ] ‘nearly fainted’ /ʔẽnõ/ [ʔẽnõ] ‘dig’
The results are presented in the box plots below. Each box plot shows LH, HL, and
SH sequences in one of the four environments. The leftmost entry in each figure shows the duration of plain vowels in the same environment. The horizontal line across each box represents the median of the duration measurements, while the short dash in the center shows the mean. The bottom and top of each box represent the lower and upper quartiles, and the ends of the whiskers represent the minimum and maximum of all the
58 data. Diamonds represent outliers that are more than three times the inter quartile range.
Figure 11 shows duration of vocoid sequences in final open syllables.
Figure 11. Duration of vocoid sequences in final open syllables
The syllabification processes above predict that the LH sequences will be shorter, while the other sequences will be longer. LH sequences are indeed very close to the duration of plain vowels, while both HL and SH sequences are much longer.
Figure 12 shows duration of sequences in final closed syllables.
59 Figure 12. Duration of vocoid sequences in final closed syllables
The HL and SH sequences are, as predicted, substantially longer than plain vowels in the same environment. LH sequences fall somewhere between the two, but closer to the disyllabic sequences than expected since they are predicted to be tautosyllabic while the others are disyllabic. HL sequences range from 250 400 ms, while LH sequences range from 250 300 ms. The mean and median of the HL is quite a bit higher than for the LH, and the quartile below the median is considerably more spread out than the quartile above the median. The lower range of the HL sequences overlaps completely with the range of the LH sequences. One possible explanation for the extra length in LH sequences is that vowels in closed syllables are shorter than those in open syllables. A sequence of two vocoids is more difficult to pronounce in this more restricted environment, so they are phonetically longer than plain vowels although not as long as
HL and SH sequences in the same environment.
60 There are four HL sequences whose duration overlaps with that of the LH sequences. These are in the words [pu.ah] ‘stir’ (< /puah/ ) with duration of 250 ms,
[ʔapi.ah] ‘graze’ (< /ʔapiah/ ) with a duration of 255 ms, [ʔɨdi.aʔ] ‘sell’ (<
/ʔɨdiaʔ/) with a duration of 285 ms, and [ʔõnẽ.ãʔ] ‘go together’ (< /ʔoneaʔ/), with a duration of 280 ms. Of these four, the two sequences with the shortest duration are both preceded by a labial consonant. It is possible that there is a phonetic shortening process after labial consonnts. However, if this is the case, the shortening process is optional since [pu.ak] ‘go’ (< /puak/ ) has a midrange duration of 330 ms. The short duration of the two sequences above not preceded by labials might simply be due to quick speech. They are both preceded by a separate syllable. Plain vowels in final syllables preceded by another syllable are shorter than those not preceded by another syllable. By comparison, vocoid sequences in this environment are likely shorter as well.
If these four sequences with short duration are removed, there is no overlap between the duration of the remaining eight HL sequences and the LH sequences in closed syllables.
Vocoid sequences following a medial glottal consonant syllabify differently from other sequences. All sequences in this environment syllabify as a single syllable. LH sequences are realized as a vowel with an offglide as they are after non glottals. All other sequences are realized as an onglide and a following vowel. Therefore, all sequences are expected to be only slightly longer than plain vowels in this position, as shown in Figure 13 for open syllables.
61 Figure 13. Duration of vocoid sequences in final open syllables after glottal consonant
As predicted, all the vocoid sequences in this environment are slightly longer than corresponding simple vowels. The HL sequences here are much shorter than the HL sequences in the same position preceded by a non glottal consonant, while the other three types are quite similar (Figure 11). This evidence supports the claim that HL sequences are tautosyllabic after glottals. There are only a handful of examples of SH vocoid sequences, but they fall where expected. There are some very short HL sequences, even shorter than the corresponding plain vowels. The shortest are [kõʔ.jã]
‘tree sp.’ (< /kõʔĩã/) with a duration of 360 ms and [ko.koɪ̯ʔ.e̯a] ‘lazy’ with a duration of 385 ms. As with the HL sequences in Figure 12 above, this may be the result of a fast speech phenomenon where longer words are spoken more quickly.
There are similar results for sequences in closed syllables, shown in Figure 14 below.
62 Figure 14. Duration of vocoid sequences in final closed syllables after glottal consonant
All sequences are only slightly longer than corresponding plain vowels. This supports the claim that these sequences are tautosyllabic.
In summary, duration and intensity data support the claims about syllabification of vocoid sequences made in section 4.2. LH sequences are only slightly longer in duration than plain vowels in the same environment, which supports their analysis as tautosyllabic sequences. HL and SH sequences are much longer than corresponding plain vowels when preceded by a non glottal consonant, but are only slightly longer than corresponding plain vowels when preceded by a glottal consonant. This supports their analysis as disyllabic after non glottal consonants, and tautosyllabic after glottal consonants.
Intensity contours also support the claims about syllabification. Sequences claimed to be disyllabic have greatest intensity in the middle of the vowel transition. Sequences claimed to be vowels with offglides have greatest intensity in the vowel, with intensity
63 tapering off at the end. Sequences claimed to be vowels with onglides have a short transition into the vowel, with intensity rising through the onglide to reach a peak in the following vowel.
4.4 Phonological analysis in Optimality Theory
This section presents a phonological analysis of syllabification of vocoid sequences within Optimality Theory.
In Optimality Theory (Prince and Smolensky 1993), underlying forms are evaluated by a ranked set of constraints in order to achieve an optimal surface form. In order to find the necessary constraints, the phonological processes are stated below in OT friendly descriptive generalizations (McCarthy 2008), with an eye toward the ultimate constraints. (38) Syllables must have onsets (e.g. /ka karai/ [kakarai̯] ‘chase’, /iok/ [jok] ‘shore’), except those beginning with the second vowel in a non LH sequence (e.g. /kiak/ [ki.ak] ‘blood’). This requirement is enforced by requiring that vocoids in initial position are nonsyllabic.
(39) Vocoid sequences are always disyllabic (e.g. /ki deok/ [kide.ok] ‘pinched’) except when the sequence is LH (e.g. /ʔamahai/ [ʔãmãh.ã ̯̃] ‘bed’). In this case the second vocoid is nonsyllabic.
(40) The first vocoid in a sequence following a non initial glottal consonant /ʔ/ or /h/ must be in the syllable onset (e.g. /kõʔĩã/ [kõʔ.jã] ‘tree sp.’) except when the sequence is LH (e.g. /bohoi/ [boh.oi̯] ‘rope’). This requirement is enforced by a requirement that the first vocoid be nonsyllabic.
The first descriptive generalization suggests a constraint against initial vowels. The constraint Onset, a very commonly invoked constraint in OT, is repeated here from
Prince and Smolensky (1993).
(41) Onset: syllables must have onsets.
64 The exception clause in (38) suggests a constraint against certain kinds of diphthongs. All but LH sequences are disyllabic. Rosenthall (1994) proposes a constraint
SonFall, which disallows rising sonority in a diphthong.
(42) SonFall: sonority may not rise in a diphthong.
This constraint presupposes ranking of vocoids in a sonority hierarchy, where low vocoids are the most sonorous, followed by mid vocoids and then high vocoids (see
Parker (2002)). SonFall thus disallows a higher vocoid from preceding a lower vocoid in a diphthong.
Along with this, generalization (39) suggests a markedness constraint requiring vocoids to occupy syllable nuclei. This constraint is violated in LH sequences.
(43) V(mora): [ cons] segments must be moraic in the output.
Table 15 and Table 16 show the crucial ranking SonFall » Onset » V(mora) for HL and LH sequences. Each asterisk * marks one violation of a constraint. The exclamation point marks where a candidate loses against another more optimal candidate.
Table 15. /ʔa piah/ [ʔa.pi.ah] ‘graze’
/ʔa piah/ SonFall Onset V(mora) →ʔa.pi.ah * ʔa.pi ̯ah *! *
Table 16. /ka karai/ [ka.ka.rai̯] ‘chase’
/kakarai/ SonFall Onset V(mora) →ka.ka.rai ̯ * ka.ka.ra.i *!
Since SonFall disallows rising sonority in a diphthong, a SH diphthong would not violate the constraint. This predicts incorrect syllabification of SH sequences, as shown in Table 17.
65 Table 17. /ki deok/ *[ki.deo̯k] ‘pinched’, incorrect prediction
/ki deok/ SonFall Onset V(mora) × ki.deo̯k * ki.de.ok *!
To correctly predict Enggano SH sequences, SonFall has to be revised to require rising sonority rather than disallowing falling sonority.
(44) SonFall (revised): sonority must fall in a diphthong.
The revised version of SonFall requires the more sonorous vocoid to be first in a diphthong.With the revised version of SonFall, SH sequences are correctly predicted to syllabify as separate syllable peaks, grouping with the HL sequences rather than the LH sequences. This is shown in Table 18.
Table 18. /ki deok/ [ki.de.ok] ‘pinched’, correct prediction
/ki deok/ SonFall Onset V(mora) →ki.de.ok * ki.deo̯k *! *
The third descriptive generalization in (40) above suggests that glottal consonants syllabify as the coda of the previous syllable, causing the following vocoid in non LH sequences to be nonsyllabic to satisfy Onset. This constraint is formulated in (45).
(45) *[ σʔ,h: glottal consonants cannot be in syllable onsets.
This constraint crucially dominates V(mora), as seen in the LH sequence in Table 19.
Table 19. /karaʔai/ [ka.raʔ.ai̯] ‘type of tree’
/karaʔai/ *[ σʔ,h SonFall Onset V(mora) →ka.raʔ.ai ̯ * * ka.raʔ.a.i **! ka.ra.ʔa.i *! *
Table 20 shows a HL sequence and the correctly predicted output form. *[ σʔ,h,
SonFall and Onset all crucially dominate V(mora).
66 Table 20. /kõʔĩã/ [kõʔ.jã] ‘tree sp.’
/kõʔĩã/ *[ σʔ,h SonFall Onset V(mora) →kõʔ.jã * kõʔ.ĩ.ã *!* kõʔ.ĩã *! * kõ.ʔĩ.ã *! *
Table 21 shows a SH sequence, whose syllabification is the same as that of the HL sequence.
Table 21. /koʔoeʔ/ [koʔ.o̯eʔ] ‘devil’
/koʔoeʔ/ *[ σʔ,h SonFall Onset V(mora) →koʔ.o̯eʔ * koʔ.o.eʔ *!* koʔ.oe̯ʔ *! * ko.ʔo.eʔ *! *
Returning now to vocoid sequences following non glottal consonants, the current constraint hierarchy predicts that the first vocoid in a HL or SH sequence is nonsyllabic.
This incorrect prediction is shown in Table 22.
Table 22. /ki deok/ *[kid.e̯ok] ‘pinched’, incorrect prediction
/ki deok/ *[ σʔ,h SonFall Onset V(mora) × kid.e̯ok * ki.de.ok *!
The incorrectly predicted form has more codas than the correct form. The markedness constraint NoCoda (Prince and Smolensky 1993) appeals to the universal dispreference for syllable codas.
(46) NoCoda: syllable codas prohibited.
NoCoda is crucially ranked above Onset to predict the correct form, as shown in Table
23.
67 Table 23. /ki deok/ [ki.de.ok] ‘pinched’, correct prediction
/ki deok/ *[ σʔ,h SonFall NoCoda Onset V(mora) →ki.de.ok * * kid.e̯ok **! *
Glottal consonants syllabify as codas in spite of the constraint NoCoda, so *[ σʔ,h crucialy dominates NoCoda.
The complete constraint ranking is *[ σʔ,h, SonFall » NoCoda » Onset » V(mora).
This ranking correctly predicts all attested syllabification patterns of vocoid sequences.
Table 24 summarizes the constraint ranking as it applies to the various vocoid sequences.
68 Table 24. Summary of OT constraint ranking
*[ σʔ,h SonFall NoCoda Onset V(mora) /ʔa piah/ →ʔa.pi.ah * * ʔa.pi ̯ah *! * ʔap.jah **! * /ka karai/ →ka.ka.rai ̯ * ka.ka.ra.i *! ka.kar.ai ̯ *! * * /ki deok/ →ki.de.ok * * ki.deo̯k *! * kid.e̯ok **! * /karaʔai/ →ka.raʔ.ai ̯ * * * ka.raʔ.a.i * **! ka.ra.ʔa.i *! * ka.ra.ʔai ̯ *! * /kõʔĩã/ →kõʔ.jã * * kõʔ.ĩ.ã * *!* kõʔ.ĩã *! * * kõ.ʔĩ.ã *! * /koʔoeʔ/ →koʔ.o̯eʔ ** * koʔ.o.eʔ ** *!* koʔ.oe̯ʔ *! ** * ko.ʔo.eʔ *! * * /iɘb/ →jɘb * iɘb *! * i.ɘb *!*
Using constraints already proposed in the literature on OT with some modifications, this analysis correctly predicts syllabification of Enggano vocoid sequences.
69 CHAPTER 5
VOWEL QUALITY
This chapter presents an acoustic analysis of the positions of Enggano vowels, using measurements of the first and second formants. An analysis of vowel position is undertaken here for reasons of typological interest. Enggano’s seven vowel system is larger than the basic five vowel system found in many Austronesian languages. The placement of the vowels, particularly the high and mid central vowels, is therefore of interest. It is also of interest to see how the place of oral vowels compares with their nasal counterparts.
The results show that the oral vowels can be readily grouped into high, mid and low vowels, as well as front, central and back vowels. There is more inter speaker variation in formants of the nasal vowels. Section 5.1 discusses the methodology used in the study, followed by the results in section 5.2, and a discussion of the results in section 5.3.
5.1 Methodology
Examples of all seven oral vowels and six of the nasal vowels were recorded in the frame [ʔu pe ___ ʔãn pe janik] ‘Please say ___ again’. Two tokens of each word from four speakers were selected for acoustic analysis, yielding eight tokens of each vowel.
Each word selected has the target vowel in a closed stressed (final) syllable. Vowels between voiceless stops were chosen whenever possible for maximum differentiation between the vowel and the surrounding segments. If there was no such word in the
70 data, the word most closely matching these criteria was chosen. The words selected for analysis are listed in (47). (47) Vowels for acoustic analysis /hɘdik/ [hɘdik] ‘turn around’ /k ɨx/ [k ɨx] ‘turtle’ /dud uk/ [dud uk] ‘burn’ /kar ep/ [kar ep] ‘boil’ /k ɘx/ [k ɘx] ‘mountain’ /kaʔ tok/ [kaʔə̆tok] ‘red’ /k ak/ [k ak] ‘person’ /kãʔ kĩh/ [kãʔə̆kĩh] ‘dry’ /p ̃ʔ/ [p ̃ʔ] ‘fireplace’ /k ũk/ [k ũk] ‘back’ /kã pẽp/ [kãpẽp] ‘shallow’ /kãʔ kɘ̃h/ [kãʔə̆kɘ̃h] ‘black’ /k õp/ [k õp] ‘grave’ /k ãp/ [k ãp] ‘traditional leader’ /ʔa io iaʔ/ [ʔajo jaʔ] ‘pay out (a rope)’ /ʔauaʔ/ [ʔawaʔ] ‘mangrove’
I failed to record the vowel [ɘ̃] in a frame. In order to include this vowel in the study, I took four words from the main wordlist where two repeats of each word were recorded in isolation, and used the eight recordings for analysis. The selected words are shown in (48). All tokens for these words are from the same speaker. (48) Words with mid central vowel, said in isolation /p ɘ̃ʔ/ [p ɘ̃ʔ] ‘shoot’ /p ɘ̃ʔ/ [p ɘ̃ʔ] (a village name) /hãpɘ̃ʔ/ [hãpɘ̃ʔ] ‘breathe’ /ki pɘ̃ʔ/ [kĩpɘ̃ʔ] ‘weave’
I measured the first and second formants of each vowel. Using Praat, I first marked the beginning and ending boundaries of the vowel. I then took formant readings in the center of the vowel. All formant measurements were rounded to the nearest five hertz
(Hz), and are found in Appendix D.
71 5.2 Results
The oral vowels show consistency in their places of articulation in the vowel space, both between utterances of the same speaker and between speakers. The positions of the oral vowels are charted below in Figure 15.
Figure 15. Plotted oral vowels 22
All seven vowels are in the general areas expected, with some variations. In terms of height distinction, the high vowels [i], [ɨ] and [u] have a fairly uniform F1 of 250 400
Hz. The mid vowels [e] and [o] have an F1 of 450 600 Hz while the mid central vowel
[ɘ] is higher, with an F1 of 350 475 Hz. The mid and high central vowels [ɨ] and [ɘ] are very close together in height, but the height for each does not vary much. The lower limit of the F1 of [ɨ] is 350 Hz, while 350 Hz is the upper limit of the F1 of [ɘ]. 23 While the height range of [ɘ] overlaps considerably with the height ranges of [i] and [u], it does not overlap with that of [ɨ]. So height here is relative, being different for central vowels than for front and back vowels.
22 I used JPlotFormants (Billerey Mosier 2001) for all the vowel plot diagrams in the thesis. 23 The mid central vowel is acoustically a high mid central vowel, as reflected in its IPA representation [ɘ].
72 The low vowel [a] has an F1 value of 700 900 Hz. There is no height overlap between mid and low vowels. There is very little variation in the place of articulation of
[a].
In terms of backness, the back vowels [u] and [o] have a narrow range, with F2 readings between 700 1000 Hz. The central vowels [ɨ] and [ɘ] a have F2 readings of
1100 1700 Hz, while the front vowels [i] and [e] have F2 values of 1700 2600 Hz. The ranges of front and central vowels come very close to each other, with near overlap in the ranges of [e] and [ɘ]. 24
Compared to the relatively neat distribution of the oral vowels, the nasal vowels have a considerably larger spread. The vowel space of adjacent vowels often overlaps as seen in the position of the nasal vowels charted below in Figure 16.
Figure 16. Plotted nasal vowels
The low vowel [ã] is the only one with a clearly defined space distinct from all the other vowels. For all other vowels, both F1 and F2 ranges overlap in each adjacent pair.
24 While the ellipses encircling the ranges of [e] a nd [ɘ] overlap, the actual F2 values do not overlap.
73 F1 ranges overlap between [ĩ] and [ẽ]; [ ̃] and [ɘ̃]; and [ũ] and [õ]. F2 ranges overlap between [ĩ] and [ ̃]; [ẽ] and [ɘ̃]; [ ̃] and [ũ]; and [ɘ̃] and [õ].
5.3 Discussion
The placement of oral vowels is relatively straightforward. All the vowels fall neatly into three degrees of backness (front, central, back) and three heights (high, mid, low) except for the mid central vowel, which is phonetically high mid central. The nasal vowels, while occupying similar places to their oral counterparts, show substantially greater spread in the vowel space and much greater overlap in formant values between adjacent vowels. In order to explain this difference, I first compare the mean values of oral and nasal vowels to show that the central area in the vowel space is the same for both. I then show that individual speaker variance can account for the variation in the nasal vowels.
The mean F1 and F2 values of the oral and nasal counterparts for each vowel were compared. Figure 17 shows the results. Each symbol represents the mean formant values of that vowel’s eight tokens.
Figure 17. Average value of oral and nasal vowels
74
All seven of the nasal vowels have a higher mean F1 than their oral counterparts, and all but the low nasalized vowel have a lower F2 mean value. This suggests that the nasal vowels are articulated lower and farther back in the vowel space. The low vowel shows the least distance between mean values of the oral and nasal vowels. The mid front and back vowels have more distance between their oral and nasal counterparts, while the mid central vowel and all three high vowels have the greatest discrepancy between the nasal and oral counterparts. The greatest distance is between the F2 values of the high back vowel. The mean F2 of [u] is 929 Hz, while the mean F2 of [ũ] is 574
Hz.
While the means of oral and nasal vowels are systematically related, oral vowels show comparatively less variation in their formant values than do nasal vowels, where the ranges of adjacent vowels often overlap. Below I further analyze the difference between the ranges of oral and nasal vowels, making separate comparisons of F1 and F2 values. Figure 18 compares the spread of F1 values for oral and nasal vowels.
75 Figure 18. Comparison of F1 ranges of oral and nasal vowels
i ĩ
ɨ ̃ u
ũ e ẽ ɘ
ɘ̃ o
õ
a ã
0 100 200 300 400 500 600 700 800 900
F1 (Hz)
The range of F1 of the nasal vowels (represented by dark bars) is much greater than that of the oral vowels (represented by light bars). This is particularly true of [i] ~ [ĩ],
[e] ~ [ẽ], and [o] ~ [õ]. Since the first formant is inversely correlated with vowel height, the high nasal vowels are lower than their oral counterparts and thus potentially overlapping with the F1 space of the mid vowels. All of the nasal vowels except low [ã] and mid central [ɘ̃] have part of their F1 range between 400 and 500 Hz.
The second formant correlates with backness. We expect to see [i] and [e] group together as front vowels, [ɨ], [ɘ] and [a] as central vowels, and [u] and [o] as back vowels. Figure 19 compares the spread of the F2 values.
76 Figure 19. Comparison of F2 ranges of oral and nasal vowels
i
ĩ
e ẽ
ɨ