Issues in the Theory of Lexical Tone: On the phonological modelling of the tones of Tamang in light of a comparison with level tones (Naxi) and pitch-plus-voice-quality tones (Vietnamese)

1
Issues in the Theory of Lexical Tone On the
phonological modellingof the tones of Tamang in
light of a comparison with level tones (Naxi)and
pitch-plus-voice-quality tones (Vietnamese)
Fourteenth Manchester Phonology Meeting (14mfm),
May 25-27, 2006.

• Martine Mazaudon Alexis Michaud
• CNRS, Paris LPP-University Paris 3
• mazaudon_at_vjf.cnrs.fr alexis.michaud_at_univ-paris3.
  fr

1
2
Plan

• 1. The issue Between segmental and tonal
  contrasts
• 2. Experimental data (Tamang, Naxi, Vietnamese)
• 3. Phonological discussion

3
(No Transcript)
4
(No Transcript)
5
1.2. Localisation of the languages
Tamang
Naxi
Vietnamese
5
6
The 3 languages of our study

• Typologically
• Tamang 4 contour tones on the word, with use of
  voice quality (breathiness)
• Naxi 3 basic level tones 1 recent rising tone
• Vietnamese 6 contour tones, with use of voice
  quality (glottalisation)
• Historically
• Tamang and Vietnamese 2-way tonal split
• Naxi no tonal split

7
Tamang a textbook case of an Asian two-way tonal
split
voiceless aspirated voiceless unaspirated voiced
p?, t?... p, t... b, d...
p, t...
voiceless voiced
?m, ?n, ?l... m, n, l...
m, n, l
voiceless voiced
s z
s
8
From two tones (A, B) to four (1, 2, 3, 4)
A B
p?..., p..., ?m gt m, ?n... gt n, ?l gt l... s A high gt tone 1 B high gt tone 2
b gt p... m, n..., l... z gt s A low gt tone 3 B low gt tone 4
9
Some of the old features linger on

• as redundant features realised
• weakly
• or intermittently
• or on some phonemes only
• Voicing
• m on high tone lt?m fully voiced
• s on low tone lt z fully devoiced
• p on low tone ltb occasionally weakly voiced
• Breathiness
• is a usual feature of the 2 low tones

10
How should we analyse breathiness?

• The phonetic pitch of the four tones is variable
• Breathiness is definitely a clue
• It was noted on both low tones in the 1970sin
  several dialects of Tamang and other languages of
  its group
• Is it conditioned by the low tone?
• Is it a feature independent of tone?
• 2 tones x 2 voice qualities 4 prosodies?

11
Comparative evidence the transphonologisation of
the old voicing contrast in different Asian
languages
1 2 3 4
split of vocalic system
Reconstructed/ conservative stage tones
p t k high series
b d g low series
Open vowels
Closed vowels
Proto-Mon-Khmer Ancient Chinese, Old Thai
Khamou
Vietnamese, Chinese, Thai, Tamang
Modern Khmer (Cambodian)
Modern Mon
11
12
2. Experimental approach

• Data were available on Naxi (level tones) and
  Vietnamese (pitch-plus-voice-quality tones)
• Recordings of Tamang in Nepal (Summer 2005)
  audio electroglottographic recordings.

Naxi 5 speakers, 120 words Vietnamese 4
speakers, 126 words Tamang 5 speakers, 143
words Comparable experimental setup, materials,
and carrier sentences.
12
13
Method rests on comparison with simultaneously
recorded images.
Using the EGG signal to obtain a clue to voice
quality (glottalised, modal, breathy) Rothenberg
Mahshie 1989
glottis

• Transillumination data
• kindly communicated by Nathalie Henrich, Institut
  de la Communication Parlée, CNRS

EGG signal
DEGG signal
13
14
Oq provides an indication on whether voice is
pressed (low Oq) or relaxed/breathy (high Oq).
(Rothenberg Mahsie 1988)
cycle length
Open quotient, Oq open phase divided by cycle
length (duration between two closings)
14
15
Calculation of average curves

• 1. Each curve is normalized for duration and
  interpolated at 100 equally spaced time points.
  Creation of average curve for each subset
• 2. Creation of a syllable simulation.

15
16
Vietnamese Pitch-plus-voice-quality
tones.Voice quality characteristicsconstant.
(red emphatic/ impatient)
17
Naxi tones defined by pitch levels (H, M,
L...). Speaker-specific strategies in the use of
voice quality. speaker M7 H more lax than M
and Lspeaker M5 H more tense than M and L
H
H
M
M
L
L
Speaker M5
Speaker M7
18
Tamang Realisation of tone is highly variable.

• Considerable variability
• Difference in mean pitch for 1, 2, 3/4
• Tone 3 rising, lax/breathy voice quality
• Difference clearer on first syllable.

Here disyllabic phonological words, made up of a
verb non-tonal suffix pa
19
For 2 of the 5 speakers differences of open
quotient minute or contrary to prediction
20
For 2 of the 5 speakers differences of open
quotient minute or contrary to prediction
21
Synthesis of first results of open quotient
measurements Challenge for statistical analysis,
and for modelling
Open quotient averaged over all syllables, for
each of the five Tamang speakers.
Much less regularity than in Vietnamese.Yet
cross-speaker tendency on tones 3 and 4, unlike
in Naxi.
22
Sorting out the factors involved in variability
seeking modes of representation other than
average curves, to bring out a regularity
Relations of compensation between F0 and Oq?
Expectation points scattered along
diagonal. Not verified so far.
23
Taking the realisation of consonants into account

• 2cu-ri 3pa-pa

2cu-ri 1pa-pa
24
Conclusion of part 2

• Tone 3 appears more breathy than tone 4
• Unlike the speakers' grandparents
• Breathiness is timed all along the syllable,
  maybe the word
• Speaker M1 who has clearest pitch has (almost) no
  breathiness
• Oq is not the whole story of breathiness. More
  adequate as a measurement of glottalisation than
  as a measurement of breathiness.

24
25
Part 3. Phonological analysis and conclusion
3.1. Analysis of the prosodic system of Tamang

• Concomitant use of pitch and breathiness not
  phased
• an exception to Silvermans generalisation?
• OR rather an argument (additional) against an
  analysis as 2 tones x 2 laryngeal qualities
• In favor of 4 tones with breathiness as a feature
  of tone
• Tones as bundles of features spanning several
  segments
• T1 High ( TENSE SHORT)T2 MidT3 Mid-low
  ( Non-Falling BREATHY some voicing of
  Ci)T4 Low ( Falling BREATHY for some
  speakers ? some voicing of Ci)

26
3.2. Breathiness and Low as mutual enhancers?

• diachronally probably true in Tamang
• Loans from Nepali tend to show a historical
  evolution
• words in voiced aspirated initials (breathy)
  gt tone 4
• words in voiced initials gt tone 3
• synchronically not anymore
• tone-3 is more breathy than tone-4
• tone 4 has lower endpoint than tone 3
• once tone has been phonologised, its various
  features are dissociated gt multiple cues that
  can be used independently, ultimately in a
  compensatory manner (trading cues).

27
3.3. For further testing breathiness and tone a
relation of compensation?

• Hypothesis
• pitch and breathiness in Tamang tones 3 and 4
  compensation (either the one or the other is
  stronger/more salient). Prototype/complex
  template?
• Parallels
• Tibetan final stop and falling pitch as
  allophones ( allotones )
• Test
• Across speakers yes
• Individual strategies currently being
  investigated

Acknowledgments Many thanks to N. Henrich, J.
Vaissière, Vu-Ngoc T.
28
More?

• A four-page conference paper (Speech Prosody
  2006)
• http//www.cavi.univ-paris3.fr/ilpga/ED/student/st
  am/SP2006_MazaudonMichaud.pdf
• Short sound files, and some figures
• http//ed268.univ-paris3.fr/lpp/pages/EQUIPE/micha
  ud/TAMANG/index.htm
• Contact us
• mazaudon_at_vjf.cnrs.fr
• alexis.michaud_at_univ-paris3.fr