Focus Particles and Information Structure in German

1
Corpus Data vs. Experimental Results as
Prosodic Evidence On the Case of Stressed auch
in German Denisa Lenertová Stefan
Sudhoff University of Leipzig Linguistic
Evidence University of Tübingen February 4th,
2006
2

• topic prosodic marking of constituents
  associated with the stressed focus particle
  auch also
• associated constituent (AC)
• located to the left of the particle, in the
  prefield (PF) or middlefield (MF) of the
  clause
• represents the element added to a contextually
  given set
• (1) MartinAC hat AUCH Hunger.
• Martin has also hunger Martin is
  hungry, too.
• syntactic, semantic, and information structural
  properties
• Reis Rosengren (1997)
• Krifka (1999)
• Dimroth (2004)

3

• Krifka (1999) - Contrastive Topic Hypothesis
• The associated constituent of stressed
  postposed particles is the contrastive topic of
  the clause in which they occur. (p. 113)
• but ACs are not necessarily identified by the
  prosodic pattern characteristic of contrastive
  topics (p. 126)
• Féry (1993)
• German contrastive topic constructions involve
  the so-called hat pattern intonation
• Frascarelli Hinterhölzl (to appear)
• contrastive topics are marked by LH accents in
  German
• Braun Ladd (2003), Braun (2005)
• importance of gradual phonetic parameters
  instead of categorical distinctions between
  accent types

4

• issues
• specific prosodic realization of ACs
  (categorical vs. gradual distinctions)
• generalizations about the optionality of the
  prosodic marking
• perceptive relevance
• evidence from two different sources
• spoken language corpus
• production and perception experiments

5

• 225 utterances with stressed auch, extracted from
  9 movies and 12 episodes of a TV series
• determination of the ACs on the basis of context
  information
• various functions, e.g., subjects, objects,
  adverbials, predicatives
• positions PF (47), MF (32), unclear cases,
  e.g., ellipses
• accent annotation using GToBI (Grice et al.
  2005)
• 52.4 high or rising accents
• 25.7 LH
• 16.7 LH
• 10.0 H
• 45.7 no accent
• 1.9 covert ACs

6
23.3 non-pronominal ACs 76.7 pronominal
ACs non-pronominal ACs always
accented optionality of AC-accents only found
for pronouns
7
(2) Aber ein Auto hab ich auch nicht
gehört. (Buffy2) but a car have I
also not heard
8
(3) Bist du auch bei Mr Chomsky in Geschichte?
(Buffy1) are you also at Mr Chomsky in
history_class
9

• ACs not characterized by a specific accent
• pronominal ACs frequently unaccented
  (anaphoricity/deixis)
• can be dropped in sentence initial position
  (topic drop)
• in pro-drop languages, these elements are
  standardly dropped
• (4) A Und was ist mit Monica? And what about
  Monica?
• B Ø Wird AUCH da sein.
• will also there be She will be there,
  too. (Lügen)
• accented ACs bear accents of different types
  LH, LH, H
• no evidence for influence of semantic/information
  structural properties
• possible influence of phonetic and/or
  paralinguistic factors

10

• cases with conjoined ACs (N 15)
• (5) Wir brauchen dich, und du brauchst uns
  doch auch. (Gloomy)
• we need you and you need us
  PRT also
• cases with an accent on a non-AC (N 5)
• non-associated element bears rising accent, AC
  unaccented
• AC only identified contextually
• accent does not induce alternatives, but signals
  paralinguistic meaning
• in an appropriate context, the prosodic pattern
  allows the interpretation of the accented
  element as AC

11
(6) Den Eindruck hab ich auch. (Paarungszeit
) this impression have I also I have this
impression, too.
12

• ACs marked by high or rising accents or
  unaccented
• optionality limited to pronominal ACs
• no 11 correspondence between accent and
  association status
• various types of accents appropriate for ACs
• sometimes even accents only on non-ACs ?
  identification by context
• corpus study detailed examination of the
  gradual phonetic parameters impossible
• ? experiment with balanced materials

13

• constructions with two potential ACs to the left
  of auch
• potential ACs comparable both between and within
  utterances
• (7) context 1 Can you tell me who of the PhD
  students gave a talk in June? I heard that
  only Martin gave one at that time.
• a. Der Rudi hat im Juni wahrscheinlich auch
  einen Vortrag gehalten.
• the R. has in June probably also
  a talk given
•  
• b. Im Juni hat der Rudi wahrscheinlich auch
  einen Vortrag gehalten.
• context 2 Can you tell me when Rudi gave talks
  in this term? I only know of the one in May.
• c. Der Rudi hat im Juni wahrscheinlich auch
  einen Vortrag gehalten.
•  
• d. Im Juni hat der Rudi wahrscheinlich auch
  einen Vortrag gehalten.

14

• independent variables
• position of the intended AC (prefield vs.
  middlefield)
• function of the intended AC (subject vs.
  temporal adverbial)
• 5 lexicalizations in 4 conditions ? 20 critical
  items
• 20 fillers
• randomized presentation
• 7 female speakers
• 107 utterances entered the analysis

15

• ACs marked by rising accents LH (81.3), LH
• non-ACs frequently accented, too
• PF 80.8 MF 29.1
• LH, LH,H
• ACs and non-ACs frequently characterized by the
  same accent type
• mean contours
• syntactic function does not influence the
  prosodic realization
• but different contour shapes for PF and MF
  association

16
prefield association
middlefield association
17

• comparisons ...
• between and within utterances (paired t-tests)
• significant differences for
• f0-maximum (H)
• f0-minimum (L)
• f0-rise
• peak alignment
• duration
• within utterances
• differences between ACs and non-ACs
  significantly greater for PF association than
  for MF association
• could be tested only between utterances

18

• identification of ACs cannot be based on their
  accents
• instead, continuous prosodic parameters decisive
• position of the AC relevant for the magnitude of
  the prosodic marking
• but differences significant for both
  association positions
• perceptive relevance tested in the subsequent
  experiments

19
completion task (8) via headphones Der Rudi
hat im Juni wahrscheinlich auch einen Vortrag
gehalten. the R. has in June probably
also a talk given on the
screen ? ... und nicht nur der Martin ? ...
und nicht nur im Mai. (and not only
Martin.) (and not only in May.) perception
study 1 original utterances perception study
2 manipulated stimuli
20

• 91 critical utterances from the production study
• 69 fillers
• determination of 4 parameters for each utterance,
  expressing the clearness of the prosodic marking
• calculation of the differences between AC and
  non-AC wrt. f0- maximum, f0-rise, duration, and
  peak alignment
• transformation to scales ranging from 0
  (smallest difference between AC and non-AC) to 1
  (biggest difference)
• intermediate values proportionally mapped onto
  real numbers between 0 and 1
• separate scales for PF and MF association
• ? listerners judgements can be related both to
  the GToBI annotation and the gradient phonetic
  properties of the utterances

21

• 2944 responses, 72.7 correct
• PF association 84.5 correct
• MF association 59.2 correct
• relation between judgements and accent
  distribution
• utterances with accented AC and deaccented non-AC
  were assigned the intended interpretation more
  often than utterances with accented ACs and
  non-ACs
• but unexpected results in the group with
  accented ACs and non-ACs
• PF matching results above chance level (75.9)
  for the cases with identical accents on AC and
  non-AC (LH or LH)
• MF matching results above chance level (66.4)
  for the cases with LH accents on non-AC and LH
  accent on AC

22
both for PF and MF association significant
correlations (various degrees) between precentage
of correct responses and the parameters for
f0-maximum, f0-rise, and duration PF
association MF association r .555
(p lt .001) r .557 (p lt .001)
23
joint manipulation of f0-minimum, f0-maximum,
duration, and peak alignment on both potential
ACs in an utterance of (9) (9) Der Wiener hat um
sieben wahrscheinlich auch einen Anruf bekommen.
The Viennese probably received a call at
seven, too. 11 versions stimulus 1 clear
case of PF association stimulus 11 clear
case of MF association
24
two groups of subjects group A (N
15) group B (N 29) insensitive to the
prosodic sensitive to the prosodic variation
variation group B stimulus 1 -
highest perc. PF association (87.4) stimulus
11 - highest perc. MF association (84.5)
25

• production study gradual differences between ACs
  and non-ACs rather than consistent differences in
  accent type
• perception studies interpretive preferences
  depend on prosodic realization
• correlation between the strength of the prosodic
  marking and the percentage of decisions for a
  particular association position
• no preference for association with the prefield
  element if prosody is used as a clue
• identification of the ACs governed by the
  relative magnitude of the prosodic parameters
  characterizing the candidates

26

• two perspectives with specific data types
• accent labels as discrete phonological
  categories
• gradual phonetic parameters
• marking of ACs involves continuous parameters
  that are not captured by GToBI
• accented ACs bear high or rising accents
• but non-ACs frequently characterized by the
  same accent type
• comparison of non-categorical properties
  necessary
• no 11 mapping, but clear interaction between
  association status and prosodic realization -
  importance of context information
• may neutralize possible interpretive preferences
  induced by prosody (corpus study)
• may favour the use of prosodic marking
  (experimental studies)

27
Thank you!
28
Braun, Bettina (2005) Production and Perception
of Thematic Contrast in German. Bern etc.
Peter Lang Braun, Bettina D. Robert Ladd
(2003) Prosodic Correlates of Contrastive and
Non- Contrastive Themes in German. In
Proceedings of Eurospeech, September 2003,
Geneva, 789 792 Dimroth, Christine (2004)
Fokuspartikeln und Informationsgliederung im
Deutschen. Tübingen Stauffenburg Féry, Caroline
(1993) German Intonational Patterns. Tübingen
Niemeyer Frascarelli, Mara Roland Hinterhölzl
(to appear) Types of Topics in German and
Italian. In Winkler, Susanne Kerstin Schwabe
(eds.) Information Structure and the
Architecture of Grammar. A Typological
Perspective. Amsterdam, Philadelphia
Benjamins Grice, Martine, Stefan Baumann Ralf
Benzmüller (2005). German Intonation in
Autosegmental-Metrical Phonology. In Jun,
Sun-Ah (ed.) Prosodic Typology The Phonology
of Intonation and Phrasing. Oxford Oxford
University Press, 55 83 Krifka, Manfred
(1999) Additive Particles under Stress. In
Proceedings of SALT 8. Cornell CLC
Publications, 111 128 Reis, Marga Inger
Rosengren (1997) A Modular Approach to the
Grammar of Additive Particles the Case of
German Auch. In Journal of Semantics, 14, 237
309
29
(10) A Ich rede mit ihr, ja? B Ich zufällig
auch. (Midsummer) I talk to her
PRT I accidentally also
30
(11) Einen Sprachfehler haben sie auch
(Haie) a speech_defect have you also
31
(12) ... mit Himbeergeschmack. Kirsch wär
auch ok. (Paarungszeit) ... with
raspberry_taste cherry would_be also ok
32
(13) A Und was ist mit Monica? And what about
Monica? B Ø Wird auch da sein. (Lügen)
will also there be
33
(14) A Where are your other clothes? B Auf
diese Frage hätt ich auch gern ne
Antwort. (Buffy10) to this question would I
also like_to_have an answer
34
(15) A Willow sucht dich überall. Willow is
looking for you. B Gut, ich such sie
auch schon. (Buffy8) good I look
for_her also already
35
(16) Wir brauchen dich, und du brauchst uns
doch auch. (Gloomy) we need you and
you need us PRT also
36
(17) Du hast deine Freundinnen auch nach
Hause gebracht. (Lügen) you have your
girlfriends also home taken
37
(18) Im Winter hat der Günther wahrscheinlich
auch Urlaub gemacht. in winter has the
Günther probably also holidays
made (SK-3-av)
38
(19) Im Winter hat der Günther wahrscheinlich
auch Urlaub gemacht. in winter has the
Günther probably also holidays
made (UG-3-av)