On the production of contrastive accents in
German
Frank Kügler & Anja Gollrad
Potsdam University
Contrastiveness
Concept of contrastiveness has received much attention in the
pycholinguistic research (Kügler & Drenhaus 2006; Toepel 2005; Carlson 2001;
Alter et al. 2001).
Many studies compare for instance the processing of wide focus
accents with contrastive focus accents and investigate whether these
different accent types are accociated with different neurophysiological
correlates (e.g. Toepel 2005).
However, the phonetic and/or phonological nature of a contrastive
accents compared to a neutral accentuation is still a matter of debate.
There is also a diversity of interpretations with respect to the
phonological category both accent types belong to.
Backround
Two different approaches to model nuclear falling accents exist
(cf. Féry 1993, Grabe 1998 vs. Grice et al. 2005, GToBI).
Féry/Grabe
Grice et al.
left-headed pitch accents
left- and right headedness
falling accent (H*L)
(H* L-) and (L+H* L-)
no difference in intonational function
difference in intonational function
H*
L
H*
L
H*
L
L+H*
L
Research Question
How are nuclear falling accents
represented phonologically in German?
Does a contrastive compared to a neutral sentence focus cause
a different phonological category (H*L vs. L+H*L)?
If so, does a low turning point in pitch prior to the accentual
peak phonetically differ as a function of focus?
If so, is the difference expressed in terms of tonal alignment,
and/or tonal scaling and/or excursion?
Production Study
Aim:
a comparison of the production of wide vs. narrow focused
sentences
Subjects:
8 female native speakers of German
Task:
participants were asked to produce visually presented
target sentences as response to the context questions
Context:
context questions were auditorily and visually presented
and elicited either wide focus sentences or sentences
containing a contrastive focus
Targets:
a total of 432 sentences have been recorded
Material
Wide focus
Contrastive focus
(1a)
Erzähl mir bitte, was passiert ist.
Martin hat den Wal gesehen.
'Martin has seen the whale.'
(1b)
Hat Martin den Frosch gesehen?
Nein, Martin hat den Wal gesehen.
'No, Martin has seen the whale.'
(2a)
Erzähl mir bitte, was passiert ist.
Ich habe Ramona gesehen.
'I have seen Ramona.'
(2b)
Hast du Susanne gesehen?
Nein, ich habe Ramona gesehen.
'No, I have seen Ramona.'
Material manipulation
Sentence length:
- Length 1: sentences as in (1) or (2)
- Length 2: gradually lengthened by adding adverb gestern ‘yesterday’
- Length 3: adding glücklicherweise ‘luckily’
- Length 4: adding gestern glücklicherweise
Number of accents in the sentence:
- sentences with only one accent (2)
- sentences with more than one accent (1).
Number of syllables of the target word:
- mono (Wal ‘whale’)
- disyllabic (Roman ‘novel’)
- triyllabic (Admiral ‘admiral’), all with word stress in the ultima
- three syllables, word stress in the penultima (Ramona)
Phonetic measurements
(1) The pitch peak of the target words in Hertz (h);
corresponding time of the peak (th)
(1)
den Ro
man
Phonetic measurements
(1) The pitch peak of the target words in Hertz (h);
corresponding time of the peak (th)
(2) A low turning point in pitch prior to the peak in Hertz (l);
corresponding time of the low point (tl)
(1)
(2)
den Ro
man
Phonetic measurements
(1) The pitch peak of the target words in Hertz (h);
corresponding time of the peak (th)
(2) A low turning point in pitch prior to the peak in Hertz (l);
corresponding time of the low point (tl)
(3) The beginning and the end
of the accented syllable
(tbeg; tend)
(1)
(2)
(3)
den Ro
(3)
man
Calculations
The excursion between the low turning E(Hz) = h-l
point and the peak
The velocity of the pitch rise
V(Hz/s) = (h-l) / th-tl
The alignment of the F0-minimum and
F0-maximum in relation to the
accented syllable
Al = tl – tbeg
Ah = th – tend
The duration of the accented syllable
D(ms) = tend – tbeg
Perceptual inspection
speakers use different phonological contours producing the target answers
(a) non-downstepped nuclear accents
299
(b) downstepped nuclear accents
20
(c) hat patterns, and
33
(d) variations in prenuclear accents
80
Results – main effects
Measurement
Wide focus
Contrastive focus
T-value
low (Hz)
188.07
182.45
2.73
high (Hz)
232.87
233.90
-2.93
E (Hz)
44.79
51.45
-4.592
V (Hz/s)
254.28
297.80
-5.190
Al (ms)
0.008
0.018
-1.08
Ah (ms)
-0.055
-0.065
2.08
D (ms)
0.236
0.256
-3.85
A multilevel model (Gelman & Hill, 2007; Bates & Sarker, 2007; Pinheiro & Bates, 2000) was fit,
using crossed random factors subject and item, and focus condition (WF, CF) as fixed factor.
Factor – Sentence length
This factor allows for controlling the distance between the
sentence initial accent and target accent (cf. Ladd & Shepman 2003).
Hypothesis: A longer distance causes more F0 minima prior to
the final H* accent.
(1) L*H
H*L
Martin hat den Wal gesehen.
(2) L*H
(T*T) H*L
Martin hat gestern den Wal gesehen.
(3) L*H
(T*T)
H*L
Martin hat glücklicherweise den Wal gesehen.
(4) L*H
(T*T) (T*T)
H*L
Martin hat gestern glücklicherweise den Wal gesehen.
Results: Sentence length & F0-minima
Results: Sentence length & Accent peaks
Results: Sentence length & Pitch rise velocity
Results: Sentence length & Pitch excursion
Results: Sentence length & F0-min alignment
Results: Sentence length & H* alignment
Results: Sentence length & Duration
Factor – Number of syllables
This factor allows for controlling the influence of a word boundary
on the realisation of the F0-minimum (cf. Ladd & Shepman 2003).
Hypothesis: A longer distance from the word boundary to the
accented syllable results in more F0 minima.
(1)
H*L
... den Wal
(2)
H*L
... den Roman
(3)
H*L
... den Admiral
Results: Number of syllables & F0-minimum
Results: Number of syllables & F0-min alignment
Results: Number of syllables
No effect on accent peak (H*).
No effect on H* alignment.
Significant difference in pitch rise velocity and pitch
excursion in monosyllables.
Factor – Number of accents
This factor allows for controlling the influence of a previous
accent.
Hypothesis: The F0-minimum should not be affected by
previous accents if being a phonological target.
(1)
H*L
Ich habe Ramona gesehen
(2) L*H
H*L
Martin hat den Roman gelesen.
Results: Number of accents & F0-minimum
Results: Number of accents & F0-min alignment
Results: Number of accents
Number of accents does not affect H* as a function of
focus.
Peak height is about 15 Hz on average lower in
sentences with more than one accent (cf. downtrend).
Rise velocity and excursion differs as a function of
focus, yet significantly only in sentences with more
accents.
Summary
Sentence Length:
• many peak accent contours (vs. hat patterns)
• clear focus difference in longer sentences
Number of syllables:
• virtually no difference for F0-minimum
Number of accents:
• no difference of focus condition w r t alignment
• significant difference of focus condition w r t scaling
– only in sentences with more than one accent !
Discussion
Does a contrastive compared to a neutral sentence focus cause
a different phonological category (H*L vs. L+H*L)?
Yes, given the significant difference in F0-minimum
No, given the similar alignment
If so, does a low turning point in pitch prior to the accentual peak
phonetically differ as a function of focus?
A difference is found in terms of tonal scaling, i.e., a lower F0minimum in contrastive focus
Later H* alignment and the lower F0-minimum cause a
significantly higher excursion and pitch rise velocity
Discussion
F0-min
F0-max
Discussion
The acoustic data suggest a difference in phonetic form as
a function of focus.
Higher and later pitch peaks are well known in Germanic
languages to indicate focus differences
(e.g. Ladd & Morton 1997, Gussenhoven 2002, 2008).
Lower tonal scaling of an F0-minimum prior to an accent
appears to correlate with contrastive focus in German
– at least in sentences with more than one accent.
Perceptually, we will test whether the lower scaling really
manifests an expression of focus (e.g. semantic congruency
task).
Thank you !