from

1
? from ?
lip or lingual vs. lip lingual
James M Scobbie 2nd Ultrasound Workshop UBC
Vancouver April 2004
2
Why ultrasound?

• Approximants involve open constrictions
• EPG is limited to anterior constrictions
• Multiple articulations
• Complex articulatory/acoustic relationships
• EMA is limited to analysis of anterior coils
• Ultrasound can show all of tongue
• Acoustics, EPG and video (lips) can be aligned
• Slow moving articulations
• No need for high sample rate?
• Non-invasive, good for vernacular speech

3
Why English approximants?

• /r/ sandhi is perhaps the most regular
  alternation in Southern British English (SBE)
• Phonological/phonetic status of this segmental
  alternation between something and nothing
• Labiodentalisation of /r/ is one of a set of
  far-reaching changes in contemporary SBE
• Nature of phonetic processes involved
• Phonological mergers and phonotactic changes
• along with vocalisation of /l/ and /l/ sandhi
• Phonetic nature of vocalisation/labialisation
• Relationship to rich systems of light/dark
  allophony

4
Why not EMA?

• Point-based analysis
• Coil position speed
• Interarticulatory timing
• Physically intrusive
• Stylistically off-putting
• Coils in wrong place

5
Labiodentalisation of /r/ to ?

• Large amounts of ?
• A change in production of (onset) /r/, but what?
• A phonological change to /?/?
• Early stages of merger with /l/ or /w/?
• Anecdotal reports
• Misperceptions of /Tr fr pr br spr/ as /fw pw
• Misperceptions of /tr dr kr gr/ etc. as /tw dw
• Merger of /kr/ and /kw/
• Merger of /r/ and /w/
• Almost 50 of speakers on UK TV had a ?

6
Is ? a vocalisation/loss of /r/?

• English multiconstrictional approximant /r/
• Labial, alveolar and velar/pharyngeal gestures
• Variants bunched, retroflex
• Non-approximant allophones, e.g. affricated /tr/
• Diachronic vocalisation of coda /r/ complete
• Weakened gestures? loss of ? in codas plus
• mergers and rejigging of the vowel system

7
Current variation and change

• An increasingly crowded labial-lingual space
• Onset /r/ labial posterior approx labiodent
  al approx
• /w/ labial posterior approx
• vocalised coda /l/ labial posterior approx
• /v f/ ( /D T/) labiodentals

8
Methodology

• Speakers with ? and speakers with ?
• Pilot stage 2 of former, 1 of latter (variable)
• Materials
• a ree and a raw vs. a vee and a vaw
• In a 32 item varied list with clusters, /l/, /w/
• Analyses
• Acoustic analysis of formant targets and movement
• Ultrasound analysis of lingual constrictions
• Video analysis of labial constriction

9
Methodology

• QMUC Hardware software
• Video mix, Articulate Assistant, helmet
• 25Hz sampling rate (40ms per frame)
• Each frame shows 2 interleaved scans or so
• 120 field of view
• Annotation method
• Tongue shape in frame of maximal labialisation
  for /r/ and for /v/
• Lip tongue are roughly time-aligned 40ms?
• Tongue shape in frame of maximal ?-ness
• Tongue shape for following vowel

10
Methodology

• Hypotheses
• lip or lingual ? has no lingual component
  (like v)
• lip and lingualunlike v, ? differs from
  ? in gestural timing/strength
• Tests
• If /r/ minus /v/ 0, assume hypothesis 1
• Otherwise, favour hypothesis 2
• Expect intertoken variation

11
Results

• Impressionistically
• The control Scots have ?
• The labiodental speaker mostly has ? but is
  variable and in particular the onset to some /r/
  sounds labial

12
/ri/ LQ1, LQ2, vLQ1
13
/wi/ LQ1, LQ2, vLQ1
14
/r?/
LQ1 (reps 1-3) LQ2 (reps 1-3) vLQ1 (reps 1-3)
15
Scottish control speakers with ?

• Frames of maximal labialisation of /w/ /r/ /v/
  (/l/) in two vowel contexts /i/ /?/
• Lingual comments
• /v/ has a fairly neutral tongue shape
• /l/ is uvularised
• /w/ is velarised
• /r/ varies but can be pharyngealised
• Speaker 1 (left) is bunched/tip down?
• Speaker 2 (right) is retroflex/tip up?
• Labial comments
• /w/ more bilabial than /r/, /v/ is labiodental

16
wo
wi
17
ro
ri
18
vo
vi
19
lo
li
20
SBE vLQ1 speaker with variable ?

• Fanned grid
• 3 splines taken from maximal labial frame (r
  v)maximal lingual frame (r only)vowel

21
SBE speaker with variable ?

• Frame of maximal labialisation precedes maximal
  lingual ?-like configuration (by more than 1
  frame)
• Lingual comments
• /r/ is tip down, with two clear constrictions
• Labial comments
• /w/ more bilabial than /r/, /v/ is labiodental

22
/ro/ SBE speakervLQ1 delay max lab to max
r-like 3 frames (80-120ms)21
23
/ri/ SBE speaker vLQ1max lab to max r
221
24
SBE vLQ1 speaker with variable ?

• 15 points at 5 on lingual spline measured from
  transducer centrepoint

25
Consistency of /v/ and vowel

• Mean of n3 /v/ in each, n6 vowel
• Consistent, so individual tokens of /r/ can be
  compared to mean /v/ for that vowel

26
SBE vLQ1 speaker with variable ?

• Subtract average v from maximum labial frame
  and maximum lingual frame of /r/
• Is there zero lingual difference?
• Or is labialisation enhanced in size or timing?

27
Labial/lingual asynchrony in /ri/
Blue at max labialisation, red at max lingual r
28
Labial/lingual asynchrony in /r?/
Blue at max labialisation, red at max lingual r
29
Vowel conditioned changes in /r/

• Mean lingmax of /r/ raw locations

30
Vowel conditioned changes in /r/

• Mean lingmax /r/ minus relevant mean v

31
Conclusions

• The SBE speaker using labiodental /r/
• is variable,
• perhaps due to labial-lingual timing variation
• Need quantitative comparison with controls
• Need numerous labiodental speakers
• Acoustic analysis by Mark Jones (2004) shows
  labiodental /r/ can be very labial in character
• The two control subjects have two types of /r/
• /r/ is tip down, with two clear constrictions
• Ultrasound is a good technique, for this study