Linguistic Phonetics in the UCLA Phonetics Lab

1
Linguistic Phonetics in the UCLA Phonetics Lab

• Pat Keating
• Sound to Sense / June 11, 2004

2
I. Language description

• Archives of recordings
• Korean
• Intonation
• Phonation

3
Intonation

• ToBI Tones and Break Indices
• Intonation in 14 languages Prosodic Typology
  The Phonology of Intonation and Phrasing (Sun-Ah
  Jun, ed.)
• Phonology and phonetics of intonation/ ToBI
  models of Korean (Seoul, Chonnam, Kyungsang),
  French, Greek, Argentinian Spanish, Farsi

4
Phonation

• Contrastive phonation types (voice qualities) in
  languages
• Modal, breathy, creaky
• e.g. Zapotec languages of Oaxaca, Mexico

5
a Zapotec language(San Lucas Quiavini)
modal
gets bitter
rdaa
gets ripe
breathy
rah
lets go of
creaky
rdààà
(M. Epstein)
6
Esposito (2003) Santa Ana del Valle Zapotec
H1-F3
Modal can lat
Breathy place la?t
Creaky field la?ts
7
Effect of f0 on phonation Contrast is minimal
with high f0
(C. Esposito)
8
II. Prosody

• (Intonation description)
• Prosody and voice quality
• Phrasing and articulation

9
Prosody

• the organization of speech into a hierarchy of
  units or domains
• grouping function
• some units are more prominent than others
• prominence-marking function

10
Prosody and voice quality

• Epstein (2002, 2003) Voice quality variation in
  English as a function of position and accent
• 2 kinds of voice quality variation
• Modal vs. non-modal (breathy, creaky)
• Variation within modal (laxer, tenser)

11
English phrase-final non-modal phonation

• Low boundary tones (but not low f0 in general)
  have more non-modal phonation

(M. Epstein)
12
English prominence and non-modal phonation

• Unaccented words have
• more non-modal phonation

(M. Epstein)
13
Phrasing and articulation

• Prosody (grouping, prominence) affects segmental
  articulatory properties
• How each segments phonological properties are
  realized phonetically depends in part on the
  segments position in prosodic structure

14
Prosodic strengthening

• Some prosodic positions are stronger than others,
  and segments there are stronger
• Articulatory strengthening more extreme
  articulations
• Stronger positions derived from a prosodic
  hierarchy
• Domain initial is a strong position

15
(partial) prosodic hierarchy across languages
16
Electropalatography studies

• Compare peak linguopalatal contact of segments
  across prosodic positions, e.g. different initial
  positions
• Several languages
• English (Fougeron Keating 1997)
• Korean (Cho Keating 2001 Kim 2001)
• French (Fougeron 1998, 2001)
• Taiwanese (Keating, Cho, Fougeron, Hsu 2003)

17
Pseudo-palate for EPG(Kay Elemetrics)
18
Sample frame showing contact Korean
word-initial /n/
front
42 contacted
Circles are electrodes filled ones are contacted
19
Sample contacts French /n/
Tata / Nadia
Tata Nadia
(C. Fougeron)
20
4 Korean consonantsin 4 initial positions
21
Korean fricatives in 3 positions (Kim 2001,2003)
IPi APi Wi
IPi APi Wi
22
Bigger pictureProsody and production planning

• Each phonetic segment - with its features - is a
  terminal node in a prosodic tree
• So each segment has a position in the tree
  relative to the domains and prominences
• Pronunciation of each feature depends in part on
  this prosodic position

23
Features in a prosodic tree

• IP
• wp
• ip ip
• egi
• Wd Wd Wd
• that new ei
• s s s s
• ð pro pa gan da
• continuant
• p p
• -continuant
• -voice

24
III. Coarticulation

• Initial strengthening
• Lexicon

25
Coarticulation and initial strengthening

• Cho (2002, 2004)
• Coarticulation interaction effects between
  neighboring segments, generally due to
  articulatory overlap
• How does prosodic strengthening affect overlap
  and thus coarticulation? Does a strong segment
  resist coarticulation?

26
Vowel-to-vowel coarticulationacross different
boundaries

And each vowel pitch-accented or not
(T. Cho)
27
EMA Carstens Articulograph Receivers on
articulators
L1
T2
T1
T3
L2
Jaw
(T. Cho)
28
Less effect of V1 /i/ on V2 /a/ across a larger
boundary
/a/ pulled towards /i/
(T. Cho)
29
Coarticulation and the lexicon

• Brown Scarborough (2001, 2004)
• Are words from dense lexical neighborhoods, with
  many lexical competitors, produced with more or
  less coarticulation than other words?

30
Lexical competitors
Low Relative Frequency Low-R
High Relative Frequency High-R
(R. Scarborough)
31
Production of nasal coarticulation

• Compared hard and easy CVN and NVC words on
  nasal coarticulation during the vowel
• using the Chen (1996) measure A1-P0

Sample CVN words
32
CVN result
less nasal
more nasal
hard words
easy words
Answer more coarticulation for hard words
(R. Scarborough)
33
IV. Production and Perception

• Optical prosody
• Heritage language ability

34
Optical prosodyphrasal stress-accent

• Extents, durations, and velocities of movements
  of lips, chin, head, and eyebrows are all
  potentially visible to perceivers
• Production-perception comparison Which of the
  optical correlates of stress account for visual
  intelligibility?

35
Production of phrasal stress

• So TOMMY gave Timmy a song from Debby.
• So Tommy gave TIMMY a song from Debby.
• So Tommy gave Timmy a song from DEBBY.
• So Tommy gave Timmy a song from Debby.

36
Facepoint markers locations and 11 measurements

• Left eyebrow displacement
• Head displacement
• Interlip maximum distance
• Interlip opening displacement
• Interlip closing displacement
• Lower lip opening peak velocity
• Lower lip closing peak velocity
• Chin opening displacement
• Chin opening peak velocity
• Chin closing displacement
• Chin closing peak velocity

eyebrow marker
head marker
lip markers
chin marker
37
Correlates of phrasal stress

• from all 11 measures, e.g.
• Chin and eyebrow measures are most consistent
  across speakers

Chin Closing Peak Velocity
accented
unaccented
38
Perception of phrasal stress

• 72 sentences from this corpus, video presentation
  (no sound)
• 16 hearing perceivers (not screened for
  lipreading ability)
• Task See written sentence, click on the name
  perceived as stressed, or on NoStress

39
Visual perception above chance
By perceiver
By talker
Line shows significantly above chance performance
40
Production-perception comparisonCorrelational
analysis

• Chin opening measures (opening displacement, peak
  opening velocity) account for most variance in
  perception
• Not chin closing, lips, or head or eyebrow
  movements, even though these cues are available

41
Heritage Language ability

• Jun Au with students, e.g. Oh et al. (2003)
• compared 4 groups of adults
• Lifelong native Korean speakers
• Childhood-only speakers (stopped by 7)
• Childhood-and-later overhearers
• Control group (novices)

42
Adult production of Korean VOT

• Childhood-only speakers as good as native
  speakers
• Childhood hearers show no advantage (nor on
  overall accent rating, not shown)

(Oh et al.)
43
Adult perception of Korean VOT

• Childhood-only speakers as good as native
  speakers
• Childhood hearers also as good as native speakers

(Oh et al.)
44
ConclusionUCLA Phonetics Lab

• Language description
• Prosody
• Coarticulation
• Production and perception
• And much more!