Bettina Braun Max Planck Institute for Psycholinguistics

1
Effects of dialect and context on the realisation
of German prenuclear accents

• Bettina BraunMax Planck Institute for
  Psycholinguistics
• 01.06.2007

2
Related work effect of dialect on accents

• Peters (1999) speakers from Hamburg placed f0
  peak in nuclear high accents earlier than Berlin
  speakers (diff. 29ms or 79 vs. 57 of overall
  sylldur)
• AttererLadd (2004) Southern German speakers
  align peaks in prenuclear accents later than
  Northern Germans (34 vs 21ms)
• no functional manipulation!

3
General research question

• What happens when speakers from different
  dialectal backgrounds produce different
  functional categories?
• Hyp 1 Dialectal differences (i.e. in alignment)
  persist when producing different categories
• Hyp 2 Dialectal differences collaps when
  producing different functional categories
  sincefunctional demand is stronger

4
Related work contrast in German

• Sentences with a double contrast
• thematic (prenuclear) accents have later and
  higher peaks (Braun, 2006)

What do the twins want for Christmas?
Peter wants a green bicycle
Theme
Rheme
and Johanna a blue one.
5
Braun (2006)

• Pairwise comparison of sentences produced in
  contrastive and non-contrastive paragraphs
• Short sentences with preverbal PPs or NPs (i.e.
  Italiener sind sehr gastfreundlich)
• Different numbers of pre and poststressed
  syllables
• Speakers from all over Germany
• Results contrast affected
• phonetic realization of theme accents (later and
  higher peaks) but no different accent types!
• type of rheme accents (more falling accents (i.e.
  HL) than high ones (HL-)

6
AttererLadd (2004)

• Speakers from Munich and a not further defined
  Northern region
• Long sentences with different syntactic
  structures (i.e. Die Ernennung Meiers zum
  Minister wurde nicht von allen Parteimitgliedern
  begrüßt.)
• Different numbers of prestressed and poststressed
  syllables
• Sentences read out of context from a list
  (possibly non-contrastive reading)
• Results sign. later L alignment and tendency for
  later H alignment for Southern Germans

7
This study

• Highly controlled materials
• Only one syntactic structure (subject-verb-object)
  
• Only one rhythmic structure
• Participants
• Speakers from Munich (Southeast of Germany) and
  Münster (Northwest of Germany)
• Question-answer methodology to elicit standard
  German with a regional touch

8
Elicitation

• Non-contrastive theme accentsWhat did Johanna
  rent? Johanna rented a car.
• Contrastive theme accentsSam rented a truck.
  And Johanna? Johanna rented a car.
• Non-corrective rheme accentsWho rented a car?
  Johanna rented a car
• Corrective rheme accentMartin rented a car?
  Johanna rented a car

9
Specific research questions

• Do Northern and Southern Germans differ in choice
  of (theme and rheme) accent type when signalling
  contrast?
• Do Southerners align all thematic rises later
  than Northerners, irrespective of context?
• Is there an effect of dialect on the use of
  f0-excursion when expressing contrast?

10
Materials

• 10 triysyllabic proper names with stress on
  second syllable
• Five with long stressed vowel, i.e. Marina,
  Ramona
• Five with short stressed vowel (and ambisyllabic
  consonant), i.e. Johanna, Camilla,
• combined with 10 different verb phrases with a
  comparable grammatical and rhythmic structure
  (i.e. knitted an apron) via a pseudo Latin square
• Questions recorded by speakers from the
  respective areas, all with rising intonation

11
Participants

• 9 female speakers from a 50km range around the
  citiy of Münster (recorded at Psychology
  Institute of the University of Münster)
• 9 female speakers from the city of Munich
  (recorded at the Institute for Phonetics of the
  LMU Munich) one had to be excluded because her
  phonemes were not Southern German
• Naïve with respect to the purpose of the
  experiment

12
Procedure

• Participants heard context question by a speaker
  of their region and read sentence that was
  presented on a computer screen
• 10 filler sentences without context question
• 4 randomised lists every subject read two lists
  (50 trials each)
• Only first list analyzed

13
Intonation Analysis

• Both thematic and rhematic pitch accents labelled
  following GToBI (Grice et al, 2005)
• 2 types of theme accents
• LH when stressed syll perceived as low,
• LH when perceived as high
• 2 classes of rheme accents
• falling accents
• high accents

14
Results theme accent types

• No effect of contrast or dialect
• Highly speaker specific
• One Southern German bias for LH
• 3 Northern and 3 Southern speakers bias for LH
• 6 Northern and 4 Southern speakers no bias
• (Bias one accent four times as often as other
  accent)

15
Results theme accent types ctd

• For 6 speakers without a bias towards a theme
  accent in non-contrastive contexts, sign. more
  LH than LH accents

?26.9, df1, plt0.01
16
Results rheme accent types

• No effect of region
• Effect of contrast in contrastive contexts sign.
  more falling than high rheme accents
• (replication of Braun (2006) with different
  elicitation technique)

?224.6, df1, plt0.0001
17
Acoustic phonetic analysis example
18
Dependent measures

• Alignment of the f0-minimum with respect to the
  start of the stressed syllable in ms al(L,C0)
• Alignment of the f0-maximum with respect to the
  poststressed vowel in ms al(H,V1)
• Rise-excursion in semitones

19
Analysis

• Univariate Anovas with fixed factors
• Region
• Contrast
• Vowel length
• Theme accent type
• Rheme accent type

20
L-alignment

• Main effects of
• contrast (for contrast 15ms later than for
  non-contrast)
• theme accent type (for LH 51ms later than for
  LH)

ns
ns
21
H-alignment

• Main effects of
• Region (for north 16ms later than for south)
• Contrast (for contrast 12ms later than for
  non-contrast)
• Vowel length (for short vowels 16ms later than
  for long vowels)
• Theme accent (for LH 14ms later than for LH)
• Rheme accent (for falling rheme accents 15ms
  later than for high rheme accent)
• Interaction
• Between contrast, rheme accent and region (p
  0.005)

22
H-alignment ctd

• No effect of contrast effect of contrast for N

High rheme accent

ns


23
Rise-excursion

• Main effect of
• Contrast (for contrast 0.8st larger than for
  non-contrast)
• Theme accent (for LH 0.7st larger than for LH)
• Rheme accent (for falling rheme 1.3st larger than
  for high rheme)
• Interaction
• between region and contrast (p0.016)

24
Rise-excursion ctd

• Northerners use rise-excursion to mark
  contrast Southerners dont

ns

25
Conclusion

• Contrast does not affect theme accent type
• Contrast affected the phonetic realisation of
  theme accents
• Later L and H
• higher peak
• Contrast affected rheme accent type
• More falling than high rheme accents

26
Conclusion

• Double contrast is realised differently for theme
  and rheme accents
• Phonological modification for rheme
  accents(more falling than high rheme accents in
  contrastive contexts)
• Phonetic modification in theme accents (later
  and higher peak, larger rise)

27
Regional influences

• Atterer and Ladds findings of later peaks for
  Southerners replicated in one condition only
  (non-contrastive context, realized with a high
  rheme accent)
• Same condition as Atterer and Ladd
• Magnitude of H-alignment comparable to Atterer
  and Ladd

28
Regional differences ctd

• Differences between Northern and Southern German
  speakers small
• No difference in accent types
• Difference in H-alignment and rise-excursion

29
Do dialectal differences persist?

• Northern Germans mark contrastiveness more than
  Southern Germans do
• For H-alignment dialectal differences become more
  pronounced in contrastive contexts
• For rise-excursion, dialectal differences
  disappear when producing contrast

L-alignment
H-alignment
rise-excursion
ns
ns
ns
ns
ns
ns
30
Thank you for your attention