Witsuwit

1
Witsuwiten final glottalization and voice quality

• Sharon Hargus
• sharon_at_u.washington.edu
• University of Washington
• SSILA, Oakland CA, January 8, 2005

2
Athabaskan tonogenesis

• Proto-Athabaskan tSha? beaver (Leer 87)
• Sekani tshà/ (low-marked language)
• Slave tshá/ (high-marked language)
• Ahtna tsha/ (toneless)

Background
3
Distribution of tonal and toneless languages
(Krauss to appear)
Background
4
Deriving low or high tone from final
glottalization

• Kingston (to appear) 2 different
  Proto-Athabaskan dialects with different
  glottalic consonants

creaky voice, slack ejectives tense voice, stiff ejectives
VOT short long
pitch lowered raised
spectrum increased energy in higher frequencies increased energy in higher frequencies
rise time slower faster
variability of glottal cycle increased ?
Background
5
Voice quality in Athabaskan languages

• Kaska (Morice 1902-3 528) the ...voice must
  also be raised with a sort of constrained effort
  when one pronounces the words khon fire, nehn
  land, tze gum, etc., though many other
  monosyllables lack this distinguishing feature
• Hupa (Gordon 1995) creaky voice accompanies
  final glottalized sonorants
• Tanacross (Holton 2000) high tone syllables
  have up-tilted spectrum

Background
6
Witsuwiten

• Dialect of Babine-Witsuwiten
• Not a tone language
• Impressionistic higher pitch on /-final syllables
• Much historical loss of final glottalization
• tsha beaver lt tSha?
• two types of final glottalic consonant / n,
  m
• Closely related Chilcotin and Carrier are
  high-marked (more uncertainty re Carrier)

Background
7
Babine-Witsuwiten language area
speakers who participated in current study
Background
8
Research questions

• How does final glottalization affect the voice
  quality of the preceding vowel?
• Are there differences between glottalized nasals
  and glottal stop?

9
Methods

• Word list recordings. Sample set
• je louse
• je/ boy (vocative)
• njen across
• jen bridge
• 8 speakers (2 male, 6 female)
• 4-6 sets/speaker
• 4 repetitions/token

10
Measures

• 30 ms. window at vowel midpoint and endpoint
• Pitch
• Jitter (Koike 1973)
• Energy
• Spectral tilt (h1-h2) (only oral tokens measured
  for spectral tilt)
• Normalization
• Measureperturbed Measureendpoint -
  Measuremidpoint

Methods
11
A /-final token
e e? ? ? ?
en ee???
Methods
12
Spectral tilt perturbation

• positive number decrease in creaky voice
• negative number increase in creaky voice

Results
13
Effect of glottal stop on spectral tilt
perturbation (across speakers)
F1,7 6.365, p .0396 (repeated measures
ANOVA)
14
Energy perturbation

• negative number decrease in overall energy
• positive number increase in overall energy

Results
15
Effects of nasality, glottalization on energy
perturbation (across speakers)
Effect of glottalization F1,7 48.574,
p  .0002 Effect of nasality n.s. Interaction
of glottalization, nasality F1,7 32.019, p
.0008
16
Jitter perturbation

• negative number decrease in jitter
• positive number increase in jitter

Results
17
Effects of nasality, glottalization on jitter
perturbation (across speakers)
Effect of glottalization F1,7 34.488, p
.0006 Effect of nasality n.s. No interaction
effect
18
Pitch perturbation

• negative number decrease in pitch
• positive number increase in pitch

Results
19
Effects of glottalization and nasality on pitch
perturbation (across speakers)
Effect of glottalization n.s. Effect of
nasality n.s. No interaction effect
20
Effects of glottalization and nasality on pitch
perturbation (individuals)

• Pitch lowerers HM, LM, MA, MF
• Pitch raisers AJ, KN, (SM)
• Mixed BM

Results
21
Effects of glottalization and nasality on pitch
perturbation for MA, a pitch lowerer
Effect of glottalization F1,61 74.996, p lt
.0001 (factorial ANOVA) Effect of nasality
n.s. No interaction effect
je/ boy (voc.)
MF, HM results similar to MA
22
Effects of glottalization and nasality on pitch
perturbation for LM, a pitch lowerer
Effect of glottalization F1,60 36.450, p lt
.0001 Effect of nasality F1,60 45.048, p lt
.0001 Interaction effect F1,60 24.259, p lt
.0001
je/ boy (voc.)
23
Effects of glottalization and nasality on pitch
perturbation for AJ, a pitch raiser
Effect of glottalization F1,62 165.396, p lt
.0001 Effect of nasality n.s. Interaction
effect F1,62 9.196, p .0035
je/ boy (voc.)
24
Effects of glottalization and nasality on pitch
perturbation for KN, a pitch raiser
je/ boy (voc.)
Effect of glottalization F1,75 28.828, p lt
.0001 Effect of nasality 4.375, p .0399 No
interaction effect
25
Effects of glottalization and nasality on pitch
perturbation for SM, a pitch raiser
Effect of glottalization F1,94 3.949, p
.0498 Effect of nasality n.s. No interaction
effect
je/ boy (voc.)
26
Effects of glottalization and nasality on pitch
perturbation for BM, a pitch raiser/lowerer
Effect of glottalization n.s. Effect of
nasality F1,59 8.908, p .0041 Interaction
effect F1,59 13.731, p .0005
je/ boy (voc.) jen bridge
27
Pitch perturbation before glottalic consonants
Results
28
How does final glottalization affect the voice
quality of the preceding vowel?

• increased energy in h2
• decrease in overall energy
• increase in jitter
• pitch lowering or raising

Discussion
29
Are there differences between glottalized nasals
and glottal stop?

• Pitch effects generally uniform for segment
  types (except BM)
• / has more extreme effect on pitch than n
  (AJ, LM)

Discussion
30
2 types of glottalic consonants?
Correlation matrix
pitch perturb. jitter perturb. spectral tilt perturb. energy perturb.
pitch perturb. 1.000 -.802 (p .0132) .441 -.624
jitter perturb. 1.000 -.118 .692 (p .0570)
spectral tilt perturb. 1.000 .141
energy perturb. 1.000
Discussion
31
Pitch perturbation x jitter perturbation
AJ
SM
KN
BM
MF
LM
MA
HM
Discussion
32
Jitter perturbation x energy perturbation
MA
MF
HM
BM
SM
LM
KN
AJ
Discussion
33
Effects of initial vs. final glottalization

• Initial t (Wright, Hargus and Davis 2002) no
  significant correlations between voice onset
  time, pitch perturbation, jitter perturbation, or
  rise time
• 5 speakers in both initial, final glottalization
  studies
• Significant correlations
• only initial, final pitch perturbation
• not initial rise time, final energy perturbation
• not initial, final jitter perturbation

Discussion
34
Initial vs. final pitch perturbation
significantly correlated (r .888, p .0459)
AJ
SM
MF
LM
MA
Discussion
35
Conclusions

• Witsuwiten a microcosm of Athabaskan?
• final glottalic consonants have both pitch
  raising, lowering effects
• support for Kingston (to appear)
• Pitch raising vs. lowering characteristic of
  speakers in initial, final position
• only shared characteristic of glottalization?

36
Acknowledgements

• Thanks to Witsuwiten speakers for their
  participation
• Thanks for useful advice and comments from
• Michael Krauss, Richard Wright, Laura McGarrity