Upload
others
View
10
Download
0
Embed Size (px)
Citation preview
ABSTRACT
THE STRESS SYSTEM OF CHUKCHANSI YOKUTS
The stress system of Chukchansi, a variety of Yokuts, has never been
studied in any detail. In this thesis, I illustrate how primary stress works followed
by its acoustic correlates. I first illustrate that primary stress is attracted first and
foremost to long-vowel syllables. When no long vowel is present, stress is, by
default, on the penultimate syllable. Stress acoustically manifests itself most
strongly with greater intensity. Intensity is shown to be a strong correlate of stress
as it consistently makes inherently less-intense vowels more intense than
neighboring inherently more-intense vowels. Pitch too is shown to be a consistent
and reliable correlate of stress; every stressed syllable contains higher a pitch than
non-stressed syllables. Vowel lengthening and vowel quality differences are
shown to not be reliable acoustic correlates of stress in Chukchansi.
Daniel Correia Mello May 2012
THE STRESS SYSTEM OF CHUKCHANSI YOKUTS
by
Daniel Correia Mello
A thesis
submitted in partial
fulfillment of the requirements for the degree of
Master of Arts in Linguistics
in the College of Arts and Humanities
California State University, Fresno
May 2012
APPROVED
For the Department of Linguistics:
We, the undersigned, certify that the thesis of the following student meets the required standards of scholarship, format, and style of the university and the student's graduate degree program for the awarding of the master's degree. Daniel Correia Mello
Thesis Author
Sean Fulop (Chair) Linguistics
Brian Agbayani Linguistics
Xinchun Wang Linguistics
For the University Graduate Committee:
Dean, Division of Graduate Studies
AUTHORIZATION FOR REPRODUCTION
OF MASTER’S THESIS
X I grant permission for the reproduction of this thesis in part or in
its entirety without further authorization from me, on the
condition that the person or agency requesting reproduction
absorbs the cost and provides proper acknowledgment of
authorship.
Permission to reproduce this thesis in part or in its entirety must
be obtained from me.
Signature of thesis author:
ACKNOWLEDGMENTS
Many, many people have helped me out in some way since this thesis was
in its embryonic stage. I would especially like to thank my committee members for
their continuous help and support. Sean Fulop, Brian Agbayani and Jean Wang put
up with numerous questions of mine and patiently gave me much feedback. I
would also like to thank Chris Golston who effectively functioned as a fourth
member of my committee. Others who have helped me out at some stage of the
process include Niken Adisasmito-Smith, Chuck Radke, Peter Guekguezian, Matt
Gordon and Isaac Martin.
Many non-linguists have helped me out as well. Most notably, however, is
my family who from the beginning has always been extremely supportive of my
work. I owe a debt of gratitude to my wife Sarah for putting up with my constant
staying up late and half-drunk cups of coffee lying around the house. Your
patience and accommodations have not gone unnoticed.
Last but not least I would like to thank my Chukchansi language
consultants, Holly and Jane Wyatt, for graciously giving me all the time and help I
needed.
TABLE OF CONTENTS
Page
LIST OF TABLES ......................................................................................................... vii
LIST OF FIGURES....................................................................................................... viii
CHAPTER 1: INTRODUCTION................................................................................... 1
The Yokuts Language Family................................................................................. 1
Phonetics and Phonology......................................................................................... 3
Speakers ..................................................................................................................... 4
Data Collection ......................................................................................................... 5
Literature Review ..................................................................................................... 6
Structure of the Thesis ............................................................................................. 9
CHAPTER 2: LEXICAL STRESS IN CHUKCHANSI ........................................... 11
Stress at the Prosodic Level .................................................................................. 11
Chukchansi Syllable Rhymes................................................................................ 12
Why Not Quantity-Sensitivity? ............................................................................ 13
Native Speaker Intuition ........................................................................................ 13
Conclusion ............................................................................................................... 16
CHAPTER 3: PITCH AND VOWEL QUALITY ..................................................... 18
Introduction ............................................................................................................. 18
Pitch ....................................................................................................................... 18
Vowel Quality ......................................................................................................... 23
Conclusion ............................................................................................................... 27
CHAPTER 4: THE ABSENCE OF VOWEL LENGTHENING AS AN ACOUSTIC CORRELATE OF STRESS ....................................................... 28
Introduction ............................................................................................................. 28
Method ..................................................................................................................... 28
Results ...................................................................................................................... 31
Page
vi vi
Discussion................................................................................................................ 36
Conclusion ............................................................................................................... 39
CHAPTER 5: INTENSITY AS THE MOST RELIABLE ACOUSTIC CORRELATE OF STRESS .............................................................................. 41
Introduction ............................................................................................................. 41
Method ..................................................................................................................... 43
Results ...................................................................................................................... 45
Discussion................................................................................................................ 48
Conclusion ............................................................................................................... 49
Possible Avenues for Further Research ............................................................... 50
CHAPTER 6: CONCLUSION ..................................................................................... 51
Future Research ...................................................................................................... 52
REFERENCES ............................................................................................................... 53
APPENDIX ..................................................................................................................... 56
LIST OF TABLES
Page
Table 1 Chukchansi Consonant Inventory ................................................................... 3
Table 2 Chukchansi Vowel Inventory ........................................................................... 4
Table 3 Authentic Chukchansi Test Words ................................................................ 14
Table 4 Fictional Chukchansi Test Words ................................................................. 15
Table 5 Formant Averages ........................................................................................... 25
Table 6 Xata’an ............................................................................................................. 26
Table 7 Tesech ............................................................................................................... 27
Table 8 Syllable structure............................................................................................. 38
LIST OF FIGURES
Page
Figure 1. Chukchansi varieties ....................................................................................... 1
Figure 2. The distribution of Yokuts tribes in central California. ............................. 2
Figure 3. Pitch differences between stressed and unstressed syllables. .................. 20
Figure 4. Non-final pitch ............................................................................................... 20
Figure 5. Jukukut............................................................................................................ 21
Figure 6. Cheexa. ........................................................................................................... 21
Figure 7. Central Catalan vowel neutralization.......................................................... 23
Figure 8. Xata’an............................................................................................................ 25
Figure 9. Tesech. ............................................................................................................ 26
Figure 10. Hewetit. ........................................................................................................ 29
Figure 11. Cheexa. ......................................................................................................... 30
Figure 12. Short vowel averages. ................................................................................. 32
Figure 13. Long and short vowel averages. ................................................................ 33
Figure 14. Penultimate and final vowel duration measurements. ............................ 34
Figure 15. Penultimate and non-final vowels. ............................................................ 35
Figure 16. Napash. ......................................................................................................... 36
Figure 17. Sumk’unut. ................................................................................................... 38
Figure 18. Ugugu. .......................................................................................................... 38
Figure 19. Napash. ......................................................................................................... 39
Figure 20. Moída. ........................................................................................................... 42
Figure 21. Caço. ............................................................................................................. 42
Figure 22. Shuto. ............................................................................................................ 44
Figure 23. Short and Long vowel intensification....................................................... 46
Page
ix ix
Figure 24. Intra-word vowel intensification. .............................................................. 47
Figure 25. Tootono. ....................................................................................................... 48
CHAPTER 1: INTRODUCTION
The Yokuts Language Family
Yokuts is a family of mutually intelligible dialects spoken in and around the
Central Valley and Sierra Nevada in inland central California. Though there are
many dialects of Yokuts, the focus of this study is on the Chukchansi variety.
Chukchansi, by all known accounts, groups with the Yokuts sub-familyValley
Yokuts, and more specifically, Northern Valley Yokuts. A few diagrams illustrating
the genetic relationship between the Yokuts varieties exist, however the most
recent and perhaps most accurate is from Blevins (2004)1, which follows in Figure
1.
Yokuts
Poso Creek: Palewyami
General Yokuts:
Buena Vista: Tulamni, Hometwoli
Nim-Yokuts:
Tule-Kaweah: Wikchamni, Yawdanchi
Northern Yokuts:
Gashowu
Kings River: Chukaymina, Michahay, Ayticha,
Choynimni
Valley Yokuts:
Far Northern Valley: Yachikumne (Chulamni),
Lower San Joaquin, Lakisamni?, Tawalimni
Northern Valley: Noptinte, Merced?,
Chawchila, Chukchansi, Kechayi, Dumna
Southern Valley: Wechihit, Nutunutu, Tachi,
Chunut, Wo’lasi, Choynok, Koyeti, Yawelmani
Figure 1. Chukchansi varieties
1 Also see Martin (2011) and Guekguezian (2011) for more detailed analyses of the genetic
relationships of Yokuts varieties.
2 2
Figure 2 shows the geographical distribution of Yokuts in central
California. The main distinction in the map (highlighted in green) is between
Southern Valley Yokuts and Northern Valley Yokuts, with which Chukchansi
belongs.
Figure 2. The distribution of Yokuts tribes in central California.
2
2 http://bss.sfsu.edu/calstudies/nativewebpages/ca%20web%201.html
3 3
Though a systematic understanding of the genetic relationships between the
varieties of Yokuts is not needed for the following study, relevant genetic
relationship information will be discussed in the following sections when
important or needed for the study.
Phonetics and Phonology
Until very recently (2011), little scholarly work has been undertaken
focusing on the Chukchansi variety of Yokuts. The first and most comprehensive
scholarly analysis of Chukchansi is Collord’s Yokuts Grammar: Chukchansi
(1968). This thesis is a systematic overview of Chukchansi ranging from phonetics
to syntax. Though dated, it is still a thorough first-hand resource for Chukchansi.
In regards to phonetics, which is of principle interest for this study, a more
accurate account of Chukchansi (which had the benefit of utilizing modern
technology) is given in Martin (2011). The following phoneme inventories (Tables
1 and 2) are taken from that aforementioned study (albeit with slight
modifications).
Table 1
Chukchansi Consonant Inventory
Phoneme Bilabial Dental Alveo-Palatal Velar Glottal
voiceless unaspirated p t dʒ k ʔ
voiceless aspirated pʰ tʰ tʃʰ kʰ
ejective p’ t’ tʃ’ k’
voiceless fricative s x h
nasal m n
glottalized nasal m’ n’
approximant w l j
glottalized approximant w’ l’ j’
4 4
Table 2
Chukchansi Vowel Inventory
Short ɪ e ɜ o u
Long iː eː ɜː oː uː
The distinction between short and long vowels is important in regards to
stress and will be discussed in further detail in the following sections. The high-
front vowel has been traditionally transcribed as /i/, including in Martin (2011),
however it is more accurately characterized as a near-front unrounded vowel /ɪ/3.
The long version of this vowel, however, is undoubtedly a close-front unrounded
vowel /iː/. The use of /ɜ/ is adopted from Martin (2011) to replace the less accurate
low-vowel symbols /a/ or /ɑ/.
Speakers
Chukchansi is a moribund variety of Yokuts (similar to other Yokuts
varieties). At the time of this study (2010-2012) only two native (i.e., first-
language) speakers could be confirmed. Both speakers, who are sisters, were the
language consultants for this study. Though only two speakers are immediately
known, they both claim “a few more [Chukchansi speakers] exist.” This dearth of
language consultants limits the current study, though this often is the nature of
fieldwork with endangered languages.
Both sisters, named “Holly” [born 1941 in Coarsegold] and “Jane” [born
1943 in Madera] currently live in Fresno, California, about 35 miles south of
Coarsegold at the Picayune Rancheria of the Chukchansi Indians. The sisters
acquired Chukchansi as a first language while living with their grandmother
3 /ɪ/ should be regarded as a phoneme, not a variant of /i/. In some environments, especially
preceding a glottal stop, /ɪ/ appears to allophonize to /i/.
5 5
during childhood. They then acquired English when they entered elementary
school at the age of 5 or 6 (Martin, 2011). Similar to other Native American
communities, English is the dominant and de facto language of not only Holly and
Jane, but virtually all Chukchansi Indians. The sisters primarily speak English
with each other and with all others around them.
Data Collection
The entirety of the data used for this study is primary. I elicited the majority
of the recorded data on numerous occasions at California State University, Fresno
(CSUF) with the language consultants. A sizeable number of recordings used in
this study came from Isaac Martin. The majority of the data is composed of lists o f
words recorded in isolation, three times. Most word classes (nouns, verb
[paradigms], adjectives and adverbs) were recorded. A Chukchansi translation of
Aesop’s fable North Wind and the Sun was also used as data.
The majority of the words used for this study are in citation form, with the
exception of the recording of North Wind and the Sun. An attempt was made to
record a narrative (i.e., more “naturalistic” data), spoken between the sisters, but
this has proved incredibly difficult for various reasons (the personalities of the
consultants, language fluency, comfort levels, etc.).
All the recordings, except the ones from Martin (2011), which were
recorded at the Picayune Rancheria, were recorded at California State University,
Fresno. The elicitations mentioned above, save the North Wind and the Sun
recording, produced a corpus of 141 words. Once the words were recorded, they
were then phonemically transcribed in IPA (International Phonetic Alphabet)
format and analyzed. The recordings took place in a sound-proof booth at CSUF
between the fall of 2010 and fall 2011. As mentioned earlier, the words used for
6 6
elicitation were culled randomly from the Bilingual English-Chukchansi
dictionary (3rd
ed.). Subsequent elicitations, however, included words with certain
phonological forms that were lacking in earlier elicitations. For example, an
elicitation in the fall of 2011 focused primarily on words with the high-back vowel
/u/ in the final syllable.
Aside from North Wind and the Sun, each of the 141 words recorded were
recorded in isolation with three tokens being produced for each word. The reason
for this was the emphasis on lexical stress. The recording of sounds in carrier
sentences, it was felt, would add the extra and unwanted (possible) dimension of
higher-level stress (i.e., phonological phrase and utterance level stress).
Literature Review
Stress
Existing academic research on Chukchansi is slim; not much scholarly
research has been done on Chukchansi Yokuts, much less stress in Chukchansi.
The most systematic linguistic analysis of Chukchansi taken up thus far has been
by Collord (1968) who wrote a general grammar of the language. Within the
grammar, the only reference to stress is:
In a word uttered in isolation three intensities of stress can be detected.
Primary stress is on the penultimate syllable and, as a rule, is accompanied
by a higher pitch than preceding syllables. A secondary stress is found on
the closed syllables (non-penult) with pitch if following primary stress
elsewhere. Non-penultimate open syllables are weakly stressed with pitch
following primary stress and about the same pitch as surrounding syllables
elsewhere. It must be borne in mind that these features characterize the
7 7
isolated word of two or more syllables, and they do not necessarily hold
true in longer utterances. (Collord, 1968)
Stress has also been briefly investigated in other Yokuts languages. In
Yawelmani (Yowlumne) Stanley Newman (1944) similarly concluded that stress
is mainly on the penultimate syllable. The vast majority of scholarly research on
Yokuts has been on the Yawelmani dialect. Since Stanley Newman’s early
research on Yawelmani in the early- and mid-20th
century, virtually all research on
Yokuts has been about the language’s rich morphophonology. If stress is
mentioned in any of these studies, it is only to give background to the language.
However, no original research on stress is presented and it is “assumed” that stress
is on the penultimate syllable, following both Newman and Collord.
The last year has seen a proliferation of original data on the Chukchansi
dialect. Two theses, Guekguezian (2011) and Martin (2011) are rich in original
data, with the former about morphophonology and the latter a phonetic overview
of the Chukchansi vowel space. Both theses follow the literature in assuming
penultimate stress.
Correlates of Stress
Stress is defined as the use of extra respiratory energy during a syllable in
relevance to neighboring syllables (Ladefoged & Johnson, 2010). Acoustically
speaking, many correlates of stress occur, though it is agreed that no single
acoustic correlate exists (Ladefoged, Draper, & Whitteridge, 1958). According to
Fry (1955, 1958, 1964) the most common, however, include duration (length),
intensity (loudness), F0 (pitch) and vowel quality. Fry’s early studies of stress in
English pioneered the studies of acoustic correlates of stress and have not been
overruled since.
8 8
Individual languages make use of their own set of acoustic correlates. For
example, Dutch makes use of F0 movement, duration, intensity and vowel quality
(Sluijter & Van Heuven, 1996). Correlates can also be highly language-specific. In
Dutch, vowel quality appears to only be affected by stress in lexical items while in
German only tense vowels are lengthened when stressed. Dutch, English and other
Germanic languages tend to make far more use of stress than most languages of
the world. The high variability of many Germanic stress systems allows for these
languages to use stress as an important, lexically-significant suprasegmental
(Ladefoged & Johnson, 2010). In English for example, stress is used to distinguish
homographs of different word classes (cp. /ˈɹɛ.kɚd/ (n.) and /ɹə.ˈkɔrd/ (v.))4.
Though this does occur in Italic languages, albeit much less frequently, as in
Portuguese and Spanish, the stress systems are less variable.
Though a language may have different acoustic correlates of stress, not all
the correlates are considered “equal.” Some correlates contrast more than others.
For example, vowel duration and intensity are correlates in German and English,
but the intensity contrast between stressed and unstressed syllables is less than the
vowel duration contrast between stressed and unstressed syllables in those
languages (Fry, 1955; Jessen, Marasek, Schneider, & Classen, 1995). Therefore,
vowel duration is considered a more consistent and reliable correlate of stress than
intensity in both German and English.
It is also believed (controversially) that specific languages do not have
word stress. Most of these languages include tone languages indigenous to Africa
(Hyman, 2009, 2010), though certain North American indigenous languages like
Bella Coola also are said to not have lexical stress (Newman, 1947).
4 Note the vowel quality changes of the two words as a result of stress or stresslessness.
9 9
Intensity is another correlate of stress in many languages. When intensity is
documented as an acoustic correlate, for example as in Jessen et al. (1995), it is
seen as a general intensification over the entire stressed syllable regardless of
vowel quality (or the inherent vowel loudness). Therefore in most languages that
have stress, a stressed syllable will tend to show a slightly greater intensity than a
non-stressed syllable.
Though intensity is seen in many languages as a correlate of stress, this
correlation is not as consistent or robust in other languages that have stress. The
reason intensity generally follows behind other correlates (especially length and
F0 movement) as a reliable correlate of stress is because of the important variable
of vowel quality. As will be shown, not all vowels (which are the loudest sounds
in a language) are of the same intensity level. Therefore in some languages an
unstressed syllable with a low vowel, which is inherently louder than a high or
mid vowel, will sometimes have greater intensity than the stressed syllable (if it
contains a mid or high vowel). Further studies of stress and its correlates will
sporadically be discussed in the following chapters when appropriate.
Acoustic research on the correlates of stress in the Yokuts family of
languages, however, is nonexistent. The present thesis is an attempt to study an
area of Yokuts that is need of analysis.
Structure of the Thesis
In the following chapter, I give a general overview, of lexical stress in
Chukchansi illustrating how the system works in the language. The chapters
following the overview of stress will focus on the acoustic correlates of stress. In
chapter 3, I discuss the role of pitch as a correlate of stress as well as vowel quality
(i.e., vowel movement under stress). Chapter 4 investigates the ever-present
10 10
acoustic correlate of vowel length. In the final chapter focusing on the acoustic
correlates of stress, intensity is investigated as a possible correlate. The conclusion
in chapter 6 wraps up the thesis and illuminates some possible avenues of future
research.
CHAPTER 2: LEXICAL STRESS IN CHUKCHANSI
Stress at the Prosodic Level
In a typical Chukchansi prosodic word, which can vary from one to five
syllables,1 one primary stress is used while secondary and tertiary stress is
presumable. However, the focus of this chapter, and thesis in general, is on
primary stress. Though Chukchansi has a rich morphology, inflection has no effect
on stress (i.e., stress is not sensitive to the language’s morphology).
As mentioned in chapter 1, Chukchansi distinguishes between long and
short vowels. A large number of Chukchansi words that are at least bisyllabic
usually contain one long vowel.2 Of the words that have a long vowel, stress
automatically falls on the long vowel syllable. Examples (1-3) illustrate this:
1) gosneeno’hiy /gos.ˈne:.noʔ.hij/ “kitchen”
2) teesa'hi' /ˈte:.ʃɜʔ.hɪʔ/ “lizard”
3) boyiida' /bo.ˈji:.dɜʔ/ “chick”
Following the primary ranking of long-vowel stress, penultimate stress
follows. Many words in Chukchansi do not contain long vowels in which case
stress automatically falls on the penultimate syllable. Examples 4-5 below
illustrate this.
1 A look through the Bilingual English-Chukchansi Dictionary, v.3 (Sept. 2011) reveals that no
words known exceed five syllables. However, it may be theoretically possible for “as of yet elicited”
inflected verb forms to exceed five syllables.
2 The vast majority of recordings available show only one long vowel per word, if a long vowel is
present. However Collord (1968) claimed the word luucaalewse, presumably /luː.kɜː.lew.se/ “to wrestle,”
contains two long vowels. It has also been noted by other linguists working on Chukchansi that the
causative suffix, when affixed, sometimes results in two long vowels per word. However no recordings
exist to confirm this, though I do not discount this. The surfacing of two long vowels in a word would
result in stress on the right-most long vowel. Therefore luucaalewse would be transcribed as
/luː.ˈkɜː.lew.se/.
12 12
4) kalwansa /kʰɜl.ˈwɜn.sɜ/ “pumpkin”
5) lopish /ˈlo.pɪʃ/ “fish”
The fact that long vowels tend to already occur in penultimate position
leads to primary stress overwhelmingly occurring on the penultimate syllable. The
reason for antepenultimate stress is discussed below.
Chukchansi Syllable Rhymes
Guekguezian (2011) lists the surface forms of Chukchansi syllables as
CV(X) with “X” either being a long vowel in a CVV syllable or a coda in a CVC
syllable (though not *CVVC which violates *SUPERHEAVY). Onsetless
syllables and complex onsets and codas are not to be found. Though I agree that
codas add weight to the Chukchansi syllable (i.e. CVV and CVC are both heavy),
I argue that a long-vowel rhyme (VV) is instrumental in attracting stress. By
reexamining the Chukchansi syllable to make a long-vowel rhyme (VV) more
attractive for stress than a short vowel rhyme (V), we could explain the stress shift
from the penultimate (V) syllable to the antepenultimate (VV) syllable.
Although it appears as if diphthongs exist in Chukchansi, /aj/ /ɪw/ /aw/ /ew/
/ej/ /ij/, these digraphs should be analyzed as having a short-vowel rhyme as they
are more accurately characterized as a (short) vowel followed by a glide consonant
(making them fit a CVC syllable structure). These Chukchansi digraphs do not
appear to be restricted to any syllable and often tend to monopthongize in many
environments. The fact that they do not attract stress from the penultimate syllable,
unlike a long vowel, and that they fit a typical CVC syllable structure are reasons
to be considered oblivious to stress. Furthermore, many examples exist of words
that contain both CVV syllables and CVC syllables (-VC being a diagraph) as in
13 13
shopeeyanaw /ʃo.ˈpe:.je.nɑw/, sawaadanaw /sɜ.ˈwɜ:.dɜ.nɑw/ and yuk'shuusha'hiy
/juk’.ˈʃu:.ʃɜʔ.hij/. In these words, stress is on the long-vowel syllable.
The reanalysis of the Chukchansi syllable to consider CVV as the only
syllable that attracts stress from the penultimate syllable explains the stress shift
from the default, penultimate position to antepenultimate position. As the data in
the preceding section show, primary stress is confined to mainly the penultimate
syllable. When a long syllable rhyme presents itself outside the penultimate, the
main stress will shift to the antepenultimate, CVV syllable.
Why Not Quantity-Sensitivity?
Why should the above data not lead to the belief that Chukchansi stress is
quantity-sensitive? This would work in two conceivable ways. First, Chukchansi
can be considered quantity-sensitive if it abided by the Latin criterion (Gordon,
1999) of syllable weight in which CVV and CVC are both considered heavy, but
CVC syllables do not attract stress as has been seen. The other main criterion is
the Khalkha criterion in which CVV is considered heavy while other syllables (i.e.
CVC and CV) are considered light. This is the most conceivable option except for
the fact that codas in Chukchansi are moraic as codas shorten underlying long
vowels to avoid an undesirable *CVVC syllable. With this in mind, stress is
attracted to syllables with long vowels (CVV), not necessarily bimoraic syllables,
which are both CVV and CVC in Chukchansi.
Native Speaker Intuition
Native speaker intuition is highly valuable in studies of stress. During two
separate recording sessions, one of the two informants was asked about her
intuition of stress placement. After she was given a fundamental definition of
stress and how native speakers of a language “figure out” stress by way of a
14 14
certain technique (i.e. pencil tapping), she proceeded to confirm the above
findings. Out of a total of 21 common Chukchansi words tested (as opposed to
“test words” discussed below), she correctly identified the stress in all 21 words.
Another native speaker intuition task involved the repetition of authentic
Chukchansi words by a non-Chukchansi speaker (myself). The words were spoken
with stress on different syllables and the native speaker was asked to determine
which token sounded more “native” or “accurate.” The words in this task are listed
in Table 3.
Table 3
Authentic Chukchansi Test Words
Test Word Spoken Tokens
aabula /ˈɜː.bu.lɜ/ ✓ /ɜː.ˈbu.lɜ/ /ɜː.bu.ˈlɜ/
baabas /bɜː.ˈbɜs/
/ˈbɜː.bɜs/ ✓ nopop /ˈno.pop/ ✓
/no.ˈpop/ sawaadanaw /sɜ.wɜː.ˈdɜ.nɑw/
/sɜ.ˈwɜː.dɜ.nɑw/ ✓ /sɜ.wɜː.dɜ.ˈnɑw/
The checkmarks (✓) represent the token selected by the native speaker as
being the most “native” or “accurate” sounding. Each token selected by the native
speaker was the token with correct stress (i.e., stress on the PU syllable or APU
that contains a long vowel).
It is important to note that even though a non-Chukchansi speaker spoke
each word, the phonology of each word was accurate (i.e. Chukchansi phonemes
that conformed to Chukchansi syllable structures (CV, CVC, or CVV) see above).
15 15
Furthermore each token of each word was produced (and thus heard) twice. When
requested by the native speaker, the tokens were replayed.
However, the most concrete evidence comes from the “test words” utilized
during the second session. Twelve words were created that either conformed to
Chukchansi phonology or did not. The words that did not conform to Chukchansi
phonology, which are marked by an asterisk, included long vowels in syllables
outside the antepenultimate or penultimate syllables. The words are included in
Table 4.
Table 4
Fictional Chukchansi Test Words
Test Word Speaker Approximation
*aabadoxbu /ʔɜ.bɜ.ˈdox.bu/
*beebeshuto /bi.bes.ˈhu.to/
*shatoo /ˈʃɜː.tu/
*xooshutoshu /xoː.tu.so:.ʃu/3
kokoko /ko.ˈko.ko/
shashusha /ʃɜ.ˈʃu.ʃɜ/
sumsutu /sum.ˈsu.tu/
tatata /tɜ.ˈtɜ.tɜ/
shufaadi /su.ˈfɜː.di/
baababa /bɜ.ˈbɜ.bɜ/4
shafuda /sɜ.ˈfuː.dɜ/
shafida /ʃɜ.ˈfi.dɜ/
3 The speaker had great difficulty with this word. It was effectively produced as two separate
words, i.e. a large break was taken between the 2nd
and 3rd
syllables. This explains why the transcription
above illustrates two long vowels in a word. A more accurate way to transcribe what was spoken by the
speaker is: /ˈxoː.tu/ /ˈso:.ʃu/. In this formalism, which more accurately represents what was spoken,
Chukchansi phonology is unsurprisingly not violated.
4 Though Baababa conforms to standard Chukchansi phonology, the word was produced three
times as /bɜ.ˈbɜ.bɜ/. A likely reason for this may simply be a misreading of the test word.
16 16
A quick look at the speaker approximations shows that the phonology in the
test words, when in contrast to the speaker’s native phonology, was violated. The
long vowels in the words *aabadoxbu, *beebeshuto and *shatoo are all in
violation as they exist outside the PU or APU syllables. When approximated, the
long vowels were either shortened, as in /ʔɜ.bɜ.ˈdox.bu/ and /bi.bes.ˈhu.to/, or
metathesized as in /ˈʃɜː.tu/. These three approximations all conform to Chukchansi
phonology discussed in this chapter. These three examples, plus the example of
xooshutoshu (see footnote 3, chapter 2), clearly illustrate the aforementioned
violation of long vowels in any syllable outside of the PU or APU. Unsurprisingly,
but extremely important for the current study, stress was either on the CVV
syllables or the PU syllables for each approximation as the transcriptions show.
The non-asterisked words all conform to Chukchansi phonology. The
reason for their inclusion is simple; to test which syllable would be stressed when
the words were approximated. Again unsurprisingly the non-asterisked test words
provide solid evidence of the two-step hierarchy of stress mentioned in this
chapter. With a couple of minor (and unimportant) phonetic differences in
shufaadi and shafuda (/s/ for /ʃ/), stress is either on the CVV or the PU syllable.
Also of note is the vowel lengthening of the PU /u/ in /sɜ.ˈfuː.dɜ/. However
it is important to show that the final vowel is not lengthened. This example could
be indicative of the overwhelming tendency to stress a PU syllable (i.e. a stress
“overextension”).
Conclusion
In conclusion, stress in Chukchansi can be understood as occurring by
default on the penultimate syllable. When stress does not occur on the penultimate
syllable, it is on the antepenultimate syllable. What attracts stress off the
17 17
penultimate syllable and on to the antepenultimate is a long vowel rhyme (a CVV
syllable). Primary stress has not been found to occur on a non-long vowel syllable
when a long vowel is present. Therefore, we can simply formalize Chukchansi
stress in the following hierarchy:
1). Stress the penultimate syllable.
2). If a long vowel occurs in any syllable outside the penultimate,
stress the long vowel syllable.
The preceding claims are all supported by two native speaker intuition tests.
When the consultant heard Chukchansi words with stress on different syllables,
her identification of a stressed syllable matched my presupposed theory of stress
discussed above. When fictional words were created to test her stress placement,
again the speaker’s intuition judgments matched my presuppositions of stress.
Phonetic analysis, which follows in the next few chapters, illuminates the
correlates of stress, further providing evidence for the theory laid out in this
chapter.
CHAPTER 3: PITCH AND VOWEL QUALITY
Introduction
Acoustic correlates of stress vary across languages and are usually language
specific. Some correlates, like vowel duration, F0 movement and intensity, seem
to be fairly common in most of the world’s languages (see chapter 1). In the
following sections, pitch and vowel quality are investigated to show their
effectiveness in diagnosing stress acoustically (vowel duration and intensity will
be discussed in chapters 4 and 5 respectively).
Data Analysis
The words used for the following two sections as well as the remainder of
this thesis were recorded with a Digital Reference DRV100 microphone and
recorded onto a personal MacBook laptop1 via the speech analysis software Praat
(Boersma & Weenink, 2012). With Praat, six formants were tracked with a
sampling rate of 16000 hertz (HZ). The max formant value was set at 5500 HZ as
both informants where women. All acoustic analysis for this study was done with
Praat. The methods of analysis of the recorded data will be discussed in the
respective chapters.
Pitch
Background
Pitch (and F0, its acoustic correlate) is commonly defined in phonetics as
an “auditory [perceptual] property of a sound corresponding to a musical note that
enables a listener to place it on a scale going from low to high. Since an increase
1 Version 5.1. The recordings took place with a 2007 MacBook (model ID MacBook 2.1) but
where then analyzed on a 2011 iMac (model ID iMac 12.2).
19 19
in the flow of air out of the lungs will also cause an increase in pitch, stressed
sounds will usually have a higher pitch” (Ladefoged & Johnson, 2011). However,
the reliability and consistency of pitch as an acoustic correlate of stress varies
across languages, as do virtually all acoustic correlates of stress. In this section
pitch is investigated as a possible acoustic correlate of stress in Chukchansi.
Method
For this study, the pitch of all syllables of each of the 141 words available,
whether stressed or unstressed, was measured (in Hertz). To accurately measure
the pitch of a syllable, the syllable was first highlighted and then a pitch reading
was acquired at two points of the syllable; at 25% and 75% of the syllable. This
offers a more accurate measurement of pitch in a syllable. These measurements
were then averaged for each syllable.2
Results
Pitch appears to function well as a correlate of stress in a way similar to
intensity as will be seen. A stressed syllable will always contain a higher pitch
than neighboring unstressed syllables. In a very small number of words, the pitch
of a stressed syllable was the same as the pitch of a preceding unstressed syllable.
However this is not common. Figure 3 below shows the difference in pitch with a
stressed syllable and the pitch of unstressed syllables. It should be noted that
Figure 3 below illustrates the pitch of all unstressed syllables in a word, before
and after the stressed syllable. Figure 4 illustrates the pitch of stressed syllables
and non-final unstressed syllables:
2 It should be noted that the pitch of long vowels, which are always stressed, was taken into
account as well as the pitch of short vowels. No research has been found showing an inherent difference in
the pitch of short and long vowels.
20 20
Figure 3. Pitch differences between stressed and unstressed syllables.
Figure 4. Non-final pitch
0
50
100
150
200
250
300
Pitch
Stressed
Unstressed
210
215
220
225
230
235
240
245
Non-final Pitch
Stressed
Non-final
21 21
There is a significant 43.5 Hz difference in pitch between a stressed
syllable, with an average of 241.5 Hz, and unstressed syllables with an average of
198 Hz. An example of this difference, in a spectrogram, comes from the word
jukuukut in Figure 5.
Figure 5. Jukukut
The drop in pitch with the final, unstressed syllable of jukuukut
/dʒu.ˈkuː.kut/ is comparatively subtle (a drop of 26 Hz) and an average pitch of
211 Hz while the average pitch in the initial syllable is 232 Hz. The stressed
penultimate syllable has the highest pitch reading at 239 Hz. Another example
follows in Figure 6 with the word cheexa.
Figure 6. Cheexa.
The final, unstressed syllable of cheexa /ˈtʃeː.xɜ/ has an average pitch of
222 Hz. Also noted is a dramatic drop in pitch (a drop of 54 Hz) with the final,
22 22
unstressed syllable. The stressed penultimate syllable has a greater average pitch
of 238 Hz.
Often times any syllable that follows a stressed syllable has a lower,
declining pitch. With the majority of penultimate syllables being stressed, this
leaves only the final unstressed syllables producing a lower pitch, as noted above
with jukukut and cheexa. However, with long vowels in the antepenultimate
syllable, two unstressed syllables follow (the penultimate and final). In words with
antepenultimate stress, it is common, though not absolute, for each following
unstressed syllable to contain lower pitch. The average pitch for pre-stressed
syllables is 15 Hz. This lower pitch in antepenultimate pitch (in respect to the
following ultimate pitch) is expected. The greater drop in ultimate-syllable pitch
could identify the final syllable as the weakest-stressed syllable. It could also,
however, be the result of the common prosodic feature in which pitch naturally
drops at the end of prosodic words and utterances. To see if the pitch of stressed
syllables differs from unstressed syllables when not taking into account final
syllables, final syllables were factored out. The results below in table 7 again
include pitch for stressed syllables and pre-stressed syllables (i.e. non-final
syllables):
With the final syllables taken out of the measurements, the pitch of a
stressed syllable is still greater than non-final (and non-stressed) syllables.
Unsurprisingly, however, the difference between stressed syllables and non-final
syllables, or any syllable that comes before a stressed syllable, is slightly less at
18.5 Hz.
Pitch can therefore be seen as an acoustic correlate of stress in Chukchansi
in that every stressed syllable has higher pitch than non-stressed syllables. Even
23 23
with the variable of final syllable pitch decrease, stressed syllables still attain
higher pitch.
Vowel Quality
Background
Many languages employ vowel shifting (neutralization) as a correlate of
stress. In many dialects of English, it is common for unstressed vowels to
centralize to schwa while stressed vowels tend to marginalize. In Central Catalan,
for example, /e/ /a/ and /ɛ/ reduce to an unstressed central /ə/, while /o/
marginalizes to unstressed /u/ (Ortega-Llebaria & Prieto, 2010). Figure 7
illustrates this.
Figure 7. Central Catalan vowel neutralization.
Vowel reduction is also common in Portuguese. In European Portuguese
for example, /a/ /e, ɛ/ /o, ɔ/ are all raised to /ɐ/ /ɨ/ /u/ in unstressed syllables
(Barbosa & Albano, 2004). It is also common for these unstressed syllables to be
elided in fast or casual speech.
Though Catalan and Portuguese have similar vowel reducing processes,
Castilian Spanish (both a neighboring and related language to Catalan and
24 24
Portuguese) does not share these processes. Castilian Spanish as well as many
other dialects of Spanish maintain vowel quality across many if not most stress
contexts (Ortega-Llebaria & Prieto, 2010). In the following sections I show that
these vowel reduction processes are not systematic in Chukchansi, therefore
illustrating that vowel reduction is not a correlate of stress in Chukchansi.
Vowel Quality in Chukchansi
As illustrated in the introduction, Chukchansi distinguishes ten vowels, five
short vowels /ɪ/ /e/ /ɜ/ /o/ /u/ and their long counterparts /i:/ /e:/ /ɜ:/ /o:/ /u:/. The
five long vowels, of course, only occur in stressed syllables. Therefore the vowel
quality of long vowels is not discussed in this section, as there are no unstressed
long vowels with which to compare.
For this study, each word available to analyze, a total of 141 words, were
measured by formant value3. Since each recorded word has three tokens, all three
tokens were similarly measured and then averaged. The data from both consultants
were used. The averages, from each consultant for each word, were then plotted on
vowel charts and compared (i.e., stressed vowels to unstressed vowels). Each
vowel formant was measured at 25% and 75% of the formant. This method
assured that outliers, vowels that had radically different measurements compared
to the others for whatever reasons, did not appear in the overall averages for F1
and F2 formant values for each vowel. These values were then averaged out and
appear below in Table 5.
3 Only F1 and F2 were measure as these are the two formants which best and most accurately
describe a vowel’s quality.
25 25
Table 5
Formant Averages
Stress Formant /ɪ/ /e/ /ɜ/ /o/ /u/ Stressed F1 (Hz) 495
2038 647
1905 720
1368 664 1197
467 1049
F2 (Hz)
Unstressed F1 (Hz) 512
1992
630
1927
742
1317
602
1066
461
1144 F2 (Hz)
Results
The measurements in Table 5 include the first and second formant averages
for each short vowel in question. The averages are in hertz.
As the table above shows, there is little significant distinction between
stressed and unstressed vowels. Small correlations, however, are noted with the
two back vowels, /o/ and /u/. When stressed, it appears as if /o/ centralizes while
/u/ is marginalized a bit. Though it may appear as if stress is the result of this
movement, no systematic patterns were found to explain this movement.4
Spectrograms can illustrate the resiliency of Chukchansi vowels in both
stressed and unstressed positions. The following image of xata’an illustrates this
(Figure 8).
Figure 8. Xata’an.
4 Though it may appear as if stress is the systematic pattern, it is not considered because many
unstressed syllables have similar vowel quality leaving the vowel quality change to chance.
26 26
Since stress is not contrastive in Chukchansi, no minimal pairs appear to
exist. Therefore, words that contain complete vowel harmony are ideal to study. In
xata’an /xɜ.ˈtɜʔ.ɜn/ all three low central-vowels have similar vowel quality despite
stress differences. The first and second formant values are listed in Table 6.
Table 6
Xata’an
xata’an xɜ- -ˈtɜʔ- -ɜn
F1 (Hz) 710 692 744 F2 (Hz) 1277 1337 1330
As the formant values above show, there is little movement of the stressed
penultimate low-central vowel /ɜ/. The unstressed initial and final low central-
vowels also show very little movement in comparison to the stressed medial
vowel. Another example illustrates the front-central vowel, /e/ (Figure 9).
Figure 9. Tesech.
Tesech, transcribed as /ˈte.setʃ/ clearly is stressed on the initial penultimate
syllable. The formant values for each vowel are listed in Table 7.
Even though the F2 values listed above are uncharacteristically “back” for a
front-mid vowel, this is the case for both vowels. When both vowels are
compared, little significant movement (generally >100 Hz) is seen with either
27 27
Table 7
Tesech
tesech te- -ˈsetʃ
F1 (Hz) 673 598 F2 (Hz) 1754 1682
vowel. As noted earlier in footnote 4 with the mid-back vowel /o/, this difference
in formant values is most likely a result of chance; no systematic difference has
been found to show a complete vowel shift in specific environments (i.e. when
stressed or unstressed) for this vowel or any Chukchansi vowel. Furthermore, table
6 shows only the differences in the averages of each vowel and are not consistent,
therefore not much of an effect exists.
Conclusion
The preceding sections illustrate a correlate of stress [pitch] and a non-
correlate of stress [vowel quality]. Pitch is seen as a correlate of stressed syllables
in that every stressed syllable has a greater pitch than every non-stressed syllable.
Factoring out final vowels, stressed syllables still have a higher pitch than non-
final syllables.
Vowel quality, or more specifically, formant movement as a result of stress
or stresslessness, does not appear to be a correlate of stress. As mentioned above,
many languages mark stress with vowel reduction (commonly by way of
neutralization), however, vowels in Chukchansi appear to be fairly resilient when
under the pressure of stress. The small differences illustrated above in table 6
appear to be the result of chance as opposed to any systematic vowel shift.
CHAPTER 4: THE ABSENCE OF VOWEL LENGTHENING AS AN ACOUSTIC CORRELATE OF STRESS
Introduction
The fact that greater respiratory energy used in a stressed syllable tends to
lengthen the nucleus of that syllable makes vowel length a consistent correlate of
stress cross-linguistically. However not all languages lengthen stressed vowels.
Languages like Czech (cited in Crosswhite, 2001) and Polish (Dogil & Williams,
1999; Lukaszewicz & Rozborski)1 for example, do not correlate vowel
lengthening with stressed syllables with any consistency.
Hayes (1995) stated “languages with phonemic vowel length contrasts have
been shown to avoid using duration as a correlate for stress; see Bernstein 1979.
This makes sense, since using duration to mark stress in these languages would
obscure the phonemic vowel length contrasts.”2 Hayes went on to illustrate work
on Finnish stress by Carlson (1978). In this case, which Hayes called a “dramatic
example” of the preceding statement, unstressed vowels could sometimes be
lengthened with the syllable in question is uttered in a prosodic word that is
emphasized.
In the following subsections, I investigate whether or not vowel
lengthening as an acoustic correlate of stress in Chukchansi.
Method
As was mentioned in chapter 1, a total of 141 Chukchansi words was
recorded and analyzed via Praat. Each word was spoken three times by the
informants.
1 This has recently be challenged, however. See Newlin-Lukowicz (2012).
2 This would explain the Czech case mentioned above.
29 29
All three tokens were used in the counting though certain outliers were not
counted. The measurement of vowels is often times difficult and thus
measurement methods must be consistent. While looking at vowels acoustically,
both pitch and intensity were used to accurately determine vowel boundaries (i.e.
where the vowel begins and ends).
To accurately measure them, vowels in the immediate environment of a
semi-vowel or liquid (/l/ /w/ /j/) were not measured. Vowels in all other
environments, however, were measured. Pitch was used to help determine vowel
boundaries in a way that often worked with intensity, which is discussed in more
detail below. A rise in pitch often, though not always signals the beginning of a
vowel and a dip in pitch often signals, though again not always, the end of a
vowel.
In many spectrograms of words, no rise in pitch is seen, though a fall in
pitch is fairly common. A process of [unstressed syllable] elimination followed
from this observation. Figure 10 shows an example of pitch in the word hewetit.
Figure 10. Hewetit.
30 30
In Figure 10 above, hewetit, which is transcribed as /he.ˈwe.tɪt/, has stress
on the penultimate syllable. The falling pitch on the final syllable, along with
lower intensity, indicates an unstressed syllable.
Often times, more diagnostics were needed to further define the boundaries
of a vowel. As mentioned above, the measuring of intensity also proves a valuable
method. With this method, the intensity (measured in decibels (dBs)) of a vowel
was measured and then compared to the presumed boundaries of the vowel to
more accurately judge where a vowel begins and ends.
For example, if the middle of a vowel measured 80 dBs, the presumed
boundaries of the vowel would be measured to see how much lower the intensity
would be. Generally, high intensity (between 70-85 dBs) signifies a vowel; while
anything less than 70 dBs does not signify a vowel. It should be noted that this
method varied by a case-by-case basis; some vowels had higher or lower
intensities than other vowels.
As was mentioned earlier, a rise in intensity often occurs with a rise in pitch
and a dip in intensity often occurs with a dip in pitch. Figure 11 below, which is of
the word cheexa, illustrates this.
Figure 11. Cheexa.
31 31
The image above of the word cheexa /ˈʧe:.xɜ/, with stress on the first or
penultimate syllable, shows pitch correlating with lower intensity on the second,
unstressed syllable.
These peaks of intensity, along with the often times accompanying pitch,
naturally follow the vowels of the syllables. The intensity levels of the areas
outside the vowels, signified by the black arrows, measure from left to right, 54
dBs and 55 dBs while the areas that consist of greater intensity are the vowels
themselves signified by blue arrows, and measure 74 dBs and 71 dBs. As the
numbers show, the intensity of the first syllable (74 dBs) is greater than the
intensity of the second, unstressed syllable (71 dBs). It should be noted that the
intensity measurements above are the loudest (i.e. highest) measurements
obtained.
Results
The following information describes the average vowel lengths of all five
short vowels in Chukchansi. Vowels in both stressed and unstressed positions are
given. Long vowels, which are always stressed, were obviously not measured. The
number measurements are listed in seconds. A more detailed analysis of vowels in
particular environments follows.
Short Vowel Averages
Figure 12 shows the listings of short vowel lengths in chart form followed
by exact averages. These averages are considered general because they encompass
short vowels in all environments (i.e. in stressed/unstressed syllables, penultimate
and final syllables, and in enclosed and open syllables).
32 32
Figure 12. Short vowel averages.
Though it is common for stress to lengthen vowels in many languages (Fry,
1958), the results above show that stress does not appear to affect the length of a
short vowel, regardless of quality. Martin (2011) found, however, that stress does
have a significant effect on the vowel length of both short and long vowels: “Table
5 shows at least a 20% difference in length between stressed and unstressed
versions of each vowel. The difference is sufficient that a stressed short vowel can
on occasion match the length of an unstressed long vowel.”
It should be noted, however, that the section of the thesis from which this
information was drawn (Martin 2011) states, “Our data are not comprehensive
enough to perform rigorous investigation of stress effects upon vowel length
[sic].” Likewise, the methods of evaluation in (Martin 2011) differed from the
methods described in the methods section above. In (Martin 2011), a stressed
syllable was assumed 100% of the time to be penultimate; moreover vowels,
including stressed ones, were measured in all phonemic environments including
semi-vowels, and only the second token (from a batch of three) was counted.
0
0.02
0.04
0.06
0.08
0.1
0.12
0.14
ɜ e ɪ o u
Short Vowel Averages
Stressed
Unstressed
33 33
These differences in method may likely have led to different conclusions in
regards to the effect of stress on Chukchansi vowels.
Though these findings show that stress has no significant effect on vowel
length, as Figure 12 illustrates, it appears as if quantity (inherent vowel length)
does matter, as the next section will show.
Long Vowel Averages
Since all long vowels in Chukchansi are stressed, the averages below do not
contain unstressed long vowels. When compared to the data in Table 12 the data
below in Table 13 show that on average, long vowels are about twice as long as
short vowels.
Figure 13. Long and short vowel averages.
Table 13 shows that there is a clear dichotomy between short vowels and
long vowels with long vowels being at least .10 seconds longer than their
counterpart short vowels.
0
0.05
0.1
0.15
0.2
0.25
0.3
ɜ e ɪ o u
Long & Short Vowels
Long Vowel
Short Vowel
34 34
Vowels in Penultimate and Final Position
The following section shows the averages of vowel durations in
penultimate and final syllable positions. All averages include both vowels in
penultimate and final syllables that are “open” (i.e., CV) and vowels that are
followed by a coda (i.e., CVC).
As mentioned in chapter 3, long vowels do not occur in word-final position,
which is why they do not appear in the chart (as well as long vowels in
penultimate position for which to compare them) (Figure 14).
Figure 14. Penultimate and final vowel duration measurements.
The mean vowel durations above of penultimate and ultimate vowels show
no significant difference. This finding is consistent with the above findings that
show stress has no significant effect on vowel length (as the majority of stressed
vowels are located in the penultimate syllable). It should be noted that all vowels
in the penultimate position measured and included in Figure 14 are stressed and
short; though many long vowels occur in the penultimate, they, again, are not
included in the figure as there are no unstressed long vowels to measure. A
0
0.02
0.04
0.06
0.08
0.1
0.12
0.14
0.16
ɜ e ɪ o u
Penultimate & Final Vowels
PenultimateFinal
35 35
possible reason for the lack of lengthening in the stressed, penultimate vowels in
Figure 14 may have to do with common lengthening of vowels in final syllables in
strong prosodic boundaries (the prosodic word in this case) (Byrd, 2000).
Vowels in Penultimate and Non-penultimate Position
It is well known that vowels in final syllables tend to be longer than other
preceding vowels even when unstressed. However obtaining vowel length data
comparing a stressed penultimate vowel to a non-final vowel is difficult (i.e., not
many of the elicited words allow for this comparison) for a few reasons. One
reason is the prevalence in many words of long vowels. A second is the prevalence
of bisyllabic words. These two facts together with the prevalence of semi-vowels
and liquids, prevent straightforward vowel measurement, leaves just six words
available for measurement. Nonetheless, Figure 15 illustrates the stressed
penultimate vowels in respect to non-final vowels, or antepenultimate vowels.
Figure 15. Penultimate and non-final vowels.
36 36
Even with the variable of the final syllable isolated, length still does not
appear to be a correlate of stress in Chukchansi. Figure 15 above appears to show
longer stressed penultimate vowels, but the mean difference between these two
sets of vowels is .004 seconds, which is far too close to be statistically significant.
An available selection of six words is too small to make a definitive
statement about vowel lengthening in Chukchansi but a larger selection of words
could, in the future, shed more concrete results.
Discussion
The Consistency of Vowel Length
Again, as was mentioned earlier, length does not appear to be a robust or
even consistent acoustic correlate of stress in Chukchansi. It is important to
emphasize the fact that length is not a consistent correlate; words were recorded
that clearly showed a length difference between the stressed syllable and the
surrounding unstressed syllables. This is not common, however. The vast majority
of stressed (short) vowels show no significant lengthening. As the data earlier
show, some stressed vowels are actually, on average, slightly shorter than
neighboring unstressed vowels. The spectrogram below (Figure 16) illustrates this.
Figure 16. Napash.
37 37
Above is one of many words elicited with a stressed vowel that is (slightly)
shorter than a neighboring unstressed vowel. The stress in napash /ˈnɜ.pɜʃ/ is
clearly on the initial, penultimate syllable with shows greater intensity and higher
pitch as well. The final syllable, which is never stressed, shows less intensity and
lower pitch, both indicative of an unstressed syllable. Nonetheless, the final
syllable, at .118 seconds, is slightly longer than the stressed initial syllable, which
is .090 seconds. This difference was seen in each token of this word. Though this
difference of .028 seconds is virtually imperceptible, it is nonetheless significant
because it illustrates that stressed vowels are not (consistently) lengthened.
The example of napash above, as mentioned earlier, could easily be an
example of some form of prosodic boundary lengthening. To illustrate what would
happen if final vowels were taken out of the statistics for vowel lengthening, the
results, albeit with a selection of only 6 words, shows no significant difference.
No Minimal Pairs
No minimal pairs, differentiated by stress, appear to exist in Chukchansi.
This is no surprise as stress is somewhat fixed and the language has a rich
morphology. Therefore, the next best set of words that can illustrate if vowel
length is a correlate of stress is a set of words with complete vowel harmony and
CVC3 syllable structures. However, no words were elicited that had complete
vowel harmony and CVC syllables. Therefore it was decided to use bi- or tri-
syllabic words that consist of the following syllable structures (Table 8).
3 CVC syllables are more desirable than CV syllables as open-syllable vowels, especially when
word-final, tend to be longer as no coda exists to constrain the vowel length.
38 38
Table 8
Syllable structure Syllable Structure Example Word
CVC.CV.CVC Sumk’unut /sum. ˈk’u.nut/ CV.CV.CV Ugugu /ʔu. ˈgu.gu/ CV.CVC Napash /ˈnɜ.pɜʃ/
When these words and others with similar morphophonology are analyzed,
the same results as above are noted. The following spectrograms (Figures 17 and
18) illustrate this:
Figure 17. Sumk’unut.
The stressed penultimate syllable, at .078 seconds is slightly shorter than
the antepenultimate vowel which is .088 seconds and slightly longer than the final
vowel which measures .065 seconds.
Figure 18. Ugugu.
39 39
Again noted is Ugugu. In this example, the stressed vowel, measuring .094
seconds, is clearly longer than the initial unstressed syllable and only slightly
longer than the final syllable, which measures .090 seconds. In this example, as in
other examples, an unstressed vowel tends to be significantly shorter than other
syllables in a word. In the words that illustrate this shortening on a spectrogram,
the syllable that contains the shortened vowel can be ruled as being out as being
stressed (Figure 19).
Figure 19. Napash.
The preceding spectrogram of Napash /ˈnɜ.pɜʃ/ again clearly has stress on
the initial, penultimate syllable. However, the stressed /ɜ/ is about .090 seconds
while the unstressed /ɜ/ is slightly longer at about .110 seconds. As with the
preceding two examples, all three tokens of this word are similar in terms of vowel
length.
Conclusion
The preceding section is intended to illuminate the nature of vowel length
in stressed Chukchansi words as well as to illustrate the dichotomy between short
and long vowels. The above data illustrate that on average long vowels are about
twice as long as short vowels and that there appears to be no significant effect on
40 40
length between stressed and non-stressed vowels. This result appears to hold true
even if stressed penultimate vowels are measured against non-final vowels.
CHAPTER 5: INTENSITY AS THE MOST RELIABLE ACOUSTIC CORRELATE OF STRESS
Introduction
In this chapter, intensity is shown to be a reliable acoustic correlate of stress
in Chukchansi. Every stressed syllable in Chukchansi has greater intensity than
neighboring unstressed syllables. More surprisingly, however, is that all words
that were studied that contained inherently less intense (or less loud) vowels, when
stressed, become at least1 as intense as neighboring inherently louder vowels, a
phenomenon herein labeled as intensification.
The specific phenomenon of intensification appears to not be documented
much in the literature. In general, when intensity is documented as an acoustic
correlate, for example as in Jessen et al (1995), it is seen as a general
intensification over the entire stressed syllable regardless of vowel quality (or the
inherent vowel loudness). Therefore in most languages that have stress, a stressed
syllable will tend to show a slightly greater intensity than a non-stressed syllable.
For example, in European Portuguese, stressed syllables do not always show
greater intensity. Often times, intensity is more indicative of inherent vowel
loudness. The following spectrogram (Figure 20) shows this (in the following
spectrograms in this chapter, the yellow line signifies intensity).
Figure 20 is a spectrogram of the word moída /mu.ˈwi.da/ “ground” with
primary stress on the penultimate syllable that contains a high-front vowel /i/ (as
the accent mark implies). Nonetheless, the syllable that shows the greatest
intensity is the ultimate syllable with the inherently louder low-vowel /a/.
1 It should be noted that only three examples (6%) illustrate intensification of a higher vowel that
is of equal intensity of at least one neighboring low central vowel. The remaining examples (94%) show
intensification that is greater than (a) neighboring lower vowel(s).
42
Likewise, in other Portuguese words like caço /ˈka.su/ “I hunt,” intensity is
greatest over the initial syllable which has both primary stress and an inherently
louder low-vowel /a/ (Figure 21).
Figure 20. Moída.
Figure 21. Caço.
The two spectrograms above are typical of many languages with stress
whenever stress is used as an acoustic correlate of stress.
The reason intensity generally follows behind other correlates (especially
length and F0 movement) as a reliable correlate of stress is because of the
important variable of vowel quality. As was discussed above, and will be shown
below in more detail, not all vowels (which are the loudest sounds in a language)
are of the same intensity level. Therefore in some languages an unstressed syllable
with a low vowel, which is inherently louder than a high or mid vowel, will
43
sometimes have greater intensity than the stressed syllable (if it contains a mid or
high vowel).
Method
Intensity was measured via Praat software. A total of 141 words were
analyzed. However, it soon became apparent that the strongest evidence of
intensity as a correlate came from the fact that vowels with relatively less intensity
(RLI vowels) were more intense than neighboring relatively more intense vowels
(RMI vowels). From this point on, transcribed words were searched to see if high-
vowels where in stressed positions in relation to low(er)-vowels for the sake of
comparison.2 Because of the robustness of vowel harmony in the language, a total
of 18 words exemplified intensification (i.e. 18 words of the total 141 words have
RLI vowels in the immediate environment of RMI vowels). These 18 words were
measured and comparisons were then made within words and across words to
show more thoroughly the role of stress in intensifying vowels.
When looking at intensity in a word, it can be measured across the syllable
(i.e., over the course of milliseconds) or at its highest intensity peak (i.e.,
measured in decibels (dBs)). For this study, intensity was measured at its peak.
Therefore, the syllables with “higher peaks” in the spectrogram were deemed
louder. Figure 22 illustrates this:
Notice that the first syllable is stressed and has greater intensity, which is
measured as having more dBs (72.3), though the second, unstressed syllable has
greater intensity across the peak. Also notice that the above example of the word
2 This could result in two possible scenarios. One scenario is a stressed /ɪ/ or /u/ in the immediate
environment of an unstressed /e/, /o/ or /ɜ/. The second is a stressed /e/ or /o/ in th e environment of /ɜ/.
44
Shuto /ˈʃu.to/ is an example of a high, stressed vowel having greater intensity than
a non-stressed mid-vowel.
Figure 22. Shuto.
After the words were isolated and analyzed, intensity measurements were
then taken of the contrasting syllables in question. This was done for each word
that exhibited intensification. For more accurate intensity measurements, similar to
the method used for acquiring pitch, two measurements were taken: at 25% and
75% of the syllable. These two measurements were then averaged, producing a
more accurate measurement of intensity that better represents the overall intensity
across the syllable.3 The averaged difference (in dBs) between the syllables was
then noted and listed. After this was done for each word, the totals (for both the
stressed vowel and the neighboring unstressed vowels) were added up and
averaged.
Each of the 18 words available to be analyzed was spoken three times by
the informants. Therefore the above process was done for each token of the word.
This too allowed for a more accurate analysis.
3 This method is opposed to measuring just one point of a syllable. Measuring one point leaves the
door open to “outliers,” or measurements that do not truly represent the intensity across the syllable.
45
Results
Intensity can be seen as a reliable correlate of stress in Chukchansi. Before
the phenomenon of intensification is discussed, it is important to note that every
stressed syllable, irrespective of vowel quality, contains the highest intensity for
the word in which that syllable occurs. But most concretely, the data below will
show that stress makes both high and mid vowels louder than neighboring lower
vowels. This occurs with both short and long vowels. Again, because of the
robustness of vowel harmony in Chukchansi, out of the 141 words recorded, 18
words, that do not have vowel harmony, exhibit intensification. However, it is
important to note that every Chukchansi word that has a high or mid vowel in a
stressed position shows at least as must intensity as the neighboring low(er) vowel.
Stress therefore has a 100% “success” rate in intensification.4
Short & Long Vowels
The high- and mid- short vowels of Chukchansi, /ɪ/ /u/ /e/ /o/ and their long
counterparts /iː/ /uː/ /eː/ /oː/, all seem to be affected by stress. As noted above,
stressed high- and mid- short and long vowels become at least as loud as
neighboring mid- or low- short and long vowels. Figure 23 below shows the
increase of stressed short and long vowels:
Figure 23 clearly shows stressed short and long vowels as being louder than
unstressed vowels. Differences in intensity measurements range from 2.5 dBs with
/ɜ/ to 6 dBs with /e/. Also seen is the inherent loudness of the low vowels and
inherent quietness of the high vowels (with /ɜ/ unsurprisingly being the loudest
sound and /ɪ/ and /u/ being the quietest).
4 As was mentioned earlier, three tokens of each word were recorded and analyzed. Therefore it is
important to note that each token of the 18 words showed intensification (as opposed to just 1 or 2 tokens).
This is significant because it rules out the possible variable of (an) “outlying token(s).”
46
Figure 23. Short and Long vowel intensification.
Since long vowels differ from short vowels for purposes of stress, no
research has been found claiming or showing that long vowels are inherently more
or less intense than short vowels. This is the reason for their inclusion in Figure
23. Therefore we can rewrite the Chukchansi vowel chart to take into account this
difference:
Quiet /ɪ/ /i:/ /u/ /u:/
Less Loud5 /o/ /o:/ /e/ /e:/
Loud /ɜ/ /ɜ:/
5 The terms “loud,” “less loud,” and “quiet” are used here comparatively. It should be noted that a
“quiet” /u/ or /ɪ/ is still a fairly loud sound as vowels in general are louder than consonants. The three terms
are also used for lack of better terms.
64
66
68
70
72
74
76
78
80
ɜ e o ɪ u
Stressed
Unstressed
General Stressed and Unstressed Vowels (dBs)
47
Noting the similarity in intensity between short and long vowels is
important. As many Chukchansi words have a long vowel, the elimination of
vowel length as a possible variable is key to the study.
Intra-word Comparisons
Vowel intensity differs from word to word so it is important to make intra-
word comparisons. For example, a stressed /ɪ/ may have an intensity measurement
of 72 dBs, which is relatively low according to Figure 23 above, but may still be
more intense than neighboring unstressed RMI vowels. Figure 24 below shows
general intra-word comparisons between short and long RLI vowels and short and
long RMI vowels.
Figure 24. Intra-word vowel intensification.
Figure 24 shows, in general, short and long RLI vowels in all words used
for this study as being more intense than neighboring RMI vowels. A 3 dBs for
stressed RLI vowels over unstressed RMI vowels is a significant increase.
71
71.5
72
72.5
73
73.5
74
74.5
75
75.5
76
RLI Vowel Intensification (dBs)
Stressed RLI Vowels
Unstressed RMI Vowels
48
Discussion
The data above illustrate quite clearly the effect stress has on vowels
generally and specifically on relatively quiet vowels in Chukchansi. The effect of
stress on RLI vowels is reliable enough (at a success rate of 100% of the words
with stressed RLI vowels in the environment of unstressed RMI vowels) to be
considered an acoustic correlate of stress. Therefore, stressed vowels have greater
intensity, even to the point of “intensifying” beyond the inherent intensity of
normally more intense vowels.
An interesting, but unsurprising finding is that words with the same vowel
quality (i.e., Chukchansi words with vowel harmony) still show a difference in
intensity. The stressed syllable is unsurprisingly more intense. Therefore of the
141 words available for analysis, every stressed syllable shows at least equal
intensity to a non-stressed vowel (with the vast majority, however, showing
greater intensity). Figure 25shows this.
Figure 25. Tootono.
The example above of Tootono /ˈto:.to.no/ shows the initial, stressed vowel
having greater intensity than the neighboring vowels of the same quality6.
6 Though the stressed vowel with greater intensity is a long vowel, as opposed to the other short
vowels, I again have found no evidence that long vowels are inherently more (or less) intense than short
vowels of the same quality.
49
Another interesting finding is that in every word without one of three or
more syllables that exhibited intensification, the final syllable always showed a
significantly lower intensity level. The following example of xatwishta’ illustrates
this7:
xatwishta’ /xɜt.ˈwıʃ.tɜʔ/ xɜ- (78, 78) (77, 78) (75, 77) (77.1) -.2
-ˈwɪʃ- (76, 76) (77,78) (78, 79) (77.3) +2.35
-tɜ (75, 72) (71, 70) (77, 72) (72.8) -4.5
The final syllable is, on average, 4.5 dBs less than the stressed syllable in
the word xatwishta’. The initial syllable is only .2 dBs less than the following
stressed syllable. The only example that does not follow this pattern is the word
hihiina, which is illustrated below:
hihiina /hɪ.ˈhi:.nɜ/ hi- (74, 73) (66, 68) (68, 69) (69.6) -4.9
-ˈhi:- (74, 75) (75, 77) (72, 74) (74.5) +4.7
-nɜ (69, 70) (69, 70) (71, 71) (70) -4.5
The difference between the two unstressed syllables (in relation to the stressed
penultimate syllable) is .4 dBs, which is of very little significance.
A cursory look at this finding could explain this phenomenon as being a
result of a weakly stressed syllable. However, more investigation into this matter
will need to take place in order to confidently diagnosis this issue.
Conclusion
The above evidence convincingly shows that intensity is a reliable correlate
of stress in Chukchansi. It is important to note that the correlation is with stress as
7 The numbers in the brackets represent the dB levels for each vowel at 25% and 75% of the
vowel, hence two measurements for each vowel. The last bracketed number for each syllable represents the
average of the three tokens. The red listings at the end of the initial and final syllables illustrate the
difference in intensity compared to the stressed penultimate syllable listing (in blue).
50
opposed to vowel quality. The main evidence for this stems from the
intensification of relatively less intense (RLI) vowels when stressed. With all
possible variables isolated, stress appears to be the only culprit for this
intensification.
Possible Avenues for Further Research
The preceding study is by no means a definitive understanding of acoustic
intensity in Chukchansi, much less intensity as a correlate of stress. Further
research with a larger word sampling would be beneficial in future intensification
studies of Chukchansi. Out of a complete list of 141 words, only 18 words were
available to analyze (i.e., only 18 words contained an RLI vowel in a stressed
position and were in the immediate environment of RMI vowels). Future research
could utilize a larger pool of words for an even more accurate picture. “Tes t
words” in Chukchansi (fabricated words that conform to Chukchansi phonology
and morphology) can also be utilized to further test and support the phenomena of
intensification. Furthermore, this study was only concerned with primary stress.
The finding discussed earlier in regards to final syllable intensification, or lack of,
could be evidence for secondary or tertiary stress.
CHAPTER 6: CONCLUSION
The present thesis has attempted to shed light on an area of deficiency in
Chukchansi: lexical stress and its acoustic correlates. In chapter 2, I showed that
stress in Chukchansi could be described in a simple two-step hierarchy:
1) stress a long vowel if present.
2) if no long vowel is present, by default stress the penultimate syllable.
The attraction to long vowels, though not necessarily heavy syllables (i.e.
CVV and CVC), explains why stress shifts from the penultimate syllable to the
antepenultimate syllable.
Following the description of stress, in chapters 3 through 5 I illustrated the
acoustic correlates of stress in Chukchansi. In chapter 3, pitch is shown to be a
reliable correlate of stress while vowel quality is not a correlate. Pitch functions as
a correlate of stressed syllables with stressed syllables having, on average, an 18.5
Hz increase over unstressed, non-final syllables. Vowel quality, however, does not
appear to be affected by stress, or lack thereof. Ther