Pr

1

Segmental anchoring or segmental anchorage?Tonal
alignment and scaling of the French late rise and
implications for models of intonational structure
Pauline Welby welbyp_at_tcd.ie Coláiste na
Tríonóide, Baile Átha Cliath Irlanda
2

Research carried out atInstitut de la
Communication Parlée (CNRS)(Département Parole
et Cognition de GIPSA-lab (CNRS))Grenoble,
France Joint work with Hélène Lœvenbruck Welby,
Pauline and Hélène Lœvenbruck. (2006). Anchored
down in Anchorage Syllable structure and
segmental anchoring in French. Rivista di
Linguistica/ Italian Journal of Linguistics 18
74124. Current Issues in Tonal Alignment.
Mariapaola D'Imperio (ed.). Welby, Pauline and
Hélène Lœvenbruck. (2005) Interspeech.
3
Introduction

• Aims
• Examine language universal claims of the
  segmental anchoring hypothesis (Ladd et al.,
  1999)
• Examine influence of syllabic structure on
  tonal alignment in French
• Learn about French intonation and intonation
  structure more generally
• Approach
• Record two read speech corpora
• Examine tonal alignment and tonal scaling

4
Segmental anchoring

• Hypothesis about the nature of tonal alignment
  and tonal scaling, cross-linguistically
  (Arvaniti et al., 1998, Ladd et al., 1999)
• Languages examined for hypothesis include
  Catalan Dutch English German
  Greek Russian Spanish

5
Segmental anchoring and tonal alignment
Tonal alignment timing of peaks and valleys in
the fundamental frequency (F0) curve with
segmental landmarks Segmental anchoring claim
when we look more closely at the differences
of detail that have been documented between one
syllable structure and another or between one
language and another, we find that both the
beginning and the end of the f0 movement can and
must be precisely specified with respect to
segmental landmarks (Ladd et al. 2000,
emphasis added).
6
Segmental anchoring and tonal scaling
Tonal scaling relative position of peaks and
valleys in the tonal space Segmental
anchoring claim Strictly speaking, the
segmental anchoring hypothesis makes no
predictions about the size of the F0 excursion,
yet the idea of independent L and H targets leads
to the expectation that rate will have little
effect on excursion size. (Ladd et al. 1999)
7
A prediction
Steeper slope as rate increases
F0 excursion size
rise time
normal
fast
8
Research question 1

• Do French patterns support the claims of
  segmental anchoring?
• anchors for start and end of F0 rise?
• little variation in position of start and end?
• (within and across rates)
• fairly constant tonal scaling across rates?
• steeper slopes in fast rate utterances?

9
Basics of French prosody
late rise
Les vitaminesAP quils recèlent
naturellement The vitamins that they naturally
contain.
(AP accentual phrase)
10
Basics of French prosody
late rise
Les vitaminesAP quils recèlent naturellement
late rise
early rise
Les vitaminesAP quils recèlent naturellement
(AP accentual phrase)
11
Basics of French prosody
late rise
H
L
L
Les vitaminesAP quils recèlent naturellement
late rise
early rise
H
H
L
L
Les vitaminesAP quils recèlent naturellement
(AP accentual phrase)
12
Basics of French prosody
late rise
H
L
L
Les vitaminesAP quils recèlent naturellement
late rise
early rise
H
H
L
L
Les vitaminesAP quils recèlent naturellement
Models using tonal targets Di Cristo Hirst
(1996), Di Cristo (1999, 2000) Post (2000) Jun
Fougeron (2000, 2002) Mertens (1987, 2002)
Welby (2002, 2003, 2004) Welby Lœvenbruck
(2006)
13
Example from corpus
H2
H1
Fundamental frequency (Hz)
L2
L1
Time (s)
Les vitamines quils recèlent naturellement sont
un prétexte pour les gourmands. The vitamins
that they naturally contain are just an excuse
for people who like to eat.
14
Alignment of L2 and H2

• H2
• close to the end of the last syllable of the
  AP(Welby 2003)
• L2
• sometimes in penultimate syllable
• sometimes in last syllable
• association to (Jun Fougeron, 2000)/
  alignment to (Post, 2002) penultimate syllable

15
Research question 2

• What are the precise anchors for L2 and H2?

16
H2 anchor point end of vowel
salami
mi
vitamine
min
mid
pyramide
17
H2 anchor point end of sonorant rhyme
salami
mi
vitamine
min
mid
pyramide
18
H2 anchor point end of syllable
salami
mi
vitamine
min
mid
pyramide
19
L2 anchor point

• not possible to posit specific anchor points a
  priori
• alignment with respect to penultimate syllable
  (Jun Fougeron, 2000 Post, 2002) or last
  syllable?

20
Methods Participants

• 6 native speakers of Parisian French
• Students or researchers in Grenoble
• Average age 27

21
Methods Corpus

• 6 noun triplets C1V1.C2V2.C3V3(C3)

• Targets embedded in carrier sentences
• le, la, les target qui, que, quils,
  quelles VP VP
• the target that
• Sentences placed in paragraphs

22
La cuisine de Marie-Noëlle regorge de
victuailles. Un salami qui a été rapporté de
Parme suinte d'une graisse très prometteuse. Sur
la desserte, un immense compotier déborde de
fruits de toutes sortes. Les vitamines qu'ils
recèlent naturellement sont un prétexte pour les
gourmands. Une pièce-montée en choux et
nougatine trône sur le buffet. La pyramide qui
dégouline de caramel fera sûrement des
amateurs. Marie-Noëlle's kitchen is overflowing
with provisions. A salami that was brought back
from Parma is dripping with very promising
grease. On the side table, a giant bowl is
brimming with all kinds of fruit. The vitamins
that they naturally contain are just an excuse
for people who like to eat. A layer cake with
puff pastries and nougatine is prominently
displayed on the cabinet. The pyramid, dripping
with caramel, will surely be appreciated.
23
Methods Procedures

• 6 paragraphs read at 2 rates
• normal
• fast
• Recorded to DAT (44.1 kHz), transferred to PC,
  downsampled (22.05 kHz)
• Target sentences saved to separate files
• Labelled (Praat) words, syllables, segments,
  L2, H2
• 133 items with LHLH in critical APs
• Duration, latency, rate, etc. extracted/
  calculated (Praat)

24
Results Rate


Rate, target word (syllable/ms)
Speaker
25
Results Rise time


26
Results F0 excursion size

27
Resultscorrelation rise time/ F0 excursion size
28
Results rise time, f0 excursion
F0 excursion size
F0 excursion size
rise time
rise time
29
Results L2 alignment

• L2 rarely in penultimate syllable
• Generally toward the beginning of last syllable.
• Cf. Jun Fougeron (2000), Post (2002), Welby
  (2003a/b)

30
Results L2 alignment

• Two potential anchors examined
• Rate effects for 2 speakers

31
Results L2 alignment
H2
normal
H1
L2
L1
the sentinel/ watchman
32
Results L2 alignment
start of onset consonant
H2
fast
H1
L2
L1
The man from Milan
33
H2 anchor point end of vowel
salami
mi
vitamine
min
mid
pyramide
34
H2 anchor point end of sonorant rhyme
salami
mi
vitamine
min
mid
pyramide
35
H2 anchor point end of syllable
salami
mi
vitamine
min
mid
pyramide
36
Potential H2 anchor points
mi
min
mid
37
Potential H2 anchor points
min
38
Observed H2 alignment
salami
mi
mid
pyramide
39
Observed H2 alignment
salami
mi
vitamine
min
mid
pyramide
40
Results H2 alignment
mi
mid
min
min
41
Results H2 alignment
H1
H2
L2
L1
les vitamines the vitamins
42
Results H2 alignment
H2
H1
L2
L1
ce matinal this morning person
43
Alignment results summary

• L2
• Differences from previous studies
• No stable anchor point
• Rate effects for 2/5 speakers
• H2
• Effect of syllable structure on alignment
• CV CVCobs vs. CVCson
• No simple anchor point for CVCson great
  alignment variability (60ms)

44
Anchor vs. Anchorage
anchor (n) a mechanical device that prevents a
vessel from moving
45
Anchor vs. Anchorage
anchor (n) a mechanical device that prevents a
vessel from moving anchorage (n) a place where
vessels anchor a place suitable for anchoring
46
segmental anchor
AP
47
segmental anchorage
AP
48
Conclusions

• Segmental anchoring not supported for French
• No clear anchor point for L2
• Rate effects on L2 alignment
• Rate effects on F0 excursion size
• No stable anchor point for H2
• Segmental anchoring claims cannot be spoken
  language universals. For some languages
• Only one end of a rise is anchored
• Rate effects on F0 excursion size
• Anchorages rather than anchors
• Segmental anchorage is not claimed to be a spoken
  language universal. Rather, it is one of the
  range of alignment possibilties.

49
Conclusions

• Alignment of French H2
• Surface alignment differences across syllables
  structures
• Unified account provided by the notion of
• segmental anchorage
• For H2, region stretching from 25 ms before the
  end of the last full vowel of the accentual
  phrase to the end of the phrase.
• Few peaks in voiced obstruents
• physical constraints on maintaining voicing
• preserving of cues to consonant identity

50
Conclusions

• Danger of arguing for language universals based
  on data from a few languages

51
Directions for future research

• Are segmental anchorages found in other
  languages?
• Articulatory alignment of H2?
• Similarities in other languages with LHLH
  patterns (Korean, Occitan)?

52
Acknowledgments
Allison Blodgett Christian Bulfone Barbara Gili
Fivela Nino Medves Pascal Perrier Cristel
Portes Experiment participants This research
was supported by a Marie Curie International
Fellowship within the 6th European Community
Framework Programme.
53

• Gràcies!
• Merci !
• Go raibh maith agaibh!