Givenness and the dative alternation in Danish

1
Givenness and the dative alternation in Danish

• Johannes Kizach, University of Aarhus, English
  Degree Programme

2
The dative alternation

•  
• Skolelæreren gav eleven et æble NP-construc
  tion
• teacher.the gave student.the an apple
• The teacher gave the student an apple
•  
• Skolelæreren gav et æble til eleven PP-cons
  truction
• teacher.the gave an apple to student.the
• The teacher gave an apple to the student
•  

3
Information structure

• What decides how we order the theme and the
  recipient?
• Theme/rheme
• Given/new
• Topic/comment
• - but is this actually the case?

4
Givenness in corpus studies

• Bresnan et al. (2007) show that the construction
  type is correlated with givenness. It is far more
  likely to observe the NP-construction in cases
  where the recipient is given, than in cases where
  the recipient is new.
• The bar plot shows number of occurrences in the
  corpus. The NP-construction shows a clear bias
  for a given recipient. The PP-construction shows
  no bias.

5
Givenness in corpus studies

• Bresnan et al. (2007). The NP-construction is
  mostly found with given-new order. The
  PP-construction is less discriminate.
• The bar plot shows how the new-given, given-new
  and neutral orders are distributed in the NP- and
  PP-constructions.

6
Manipulating definiteness/givenness

• a. President Clausen promised the man a job
  NPdef-indef
• b. President Clausen promised a man the
  job NPindef-def
• c. President Clausen promised the job to a man
  PPdef-indef
• d. President Clausen promised a job to the man
  PPindef-def

7
Definiteness/givenness(Bresnan et al. 2007)

• Theme

• Recipient

8
Givenness in psycholinguistic experiments

• Clifton Frazier (2004) report a series of
  speeded acceptability judgment experiments,
  showing that processing is facilitated when a
  definite NP precedes an indefinite NP (faster
  RTs), but only in the NP-construction, not in the
  PP-construction where no such effect is found.
• Bar plot shows mean reaction times. The
  difference between the two NP-constructions is
  significant. The difference between the two
  PP-constructions is not.

9
Givenness in psycholinguistic experiments

• Clifton Frazier (2004) experiment 2. Same as
  experiment 1, but now a one-sentence context
  establish the definite argument as given.
• Bar plot shows mean reaction times. The
  difference between the two NP-constructions is
  significant. The difference between the two
  PP-constructions is not.

10
Givenness in psycholinguistic experiments

• Brown, Savova Gibson (2012) report a self-paced
  reading experiment, showing the same result as
  Clifton Frazier (2004) reached. The given-new
  order is preferred for the NP-construction, but
  no preference is found in the PP-construction.
• Bar plot shows mean reading times in milliseconds
  for the second argument. The difference between
  the two NP-constructions is significant. The
  difference between the PP-constructions is not.

11
Experiment 1This work was done in collaboration
with Laura Winther Balling (CBS).

• Do we see the same structural conditioning of
  givenness-effects in Danish?

• Specifically, do we find a reaction time
  difference in the NP-construction, but not in the
  PP-construction?

12
Experiment 1

• Materials. 14 sentences were constructed using
  the 10 most frequent dative-alternating verbs
  (based on a search in KorpusDK, Bergenholtz
  1992).
•  
• a. Direktør Clausen lovede manden et
  arbejde NPdef-indef
• president Clausen promised man.the a
  job
• President Clausen promised the man a job
• b. Direktør Clausen lovede en mand
  arbejdet NPindef-def
• President Clausen promised a man the job
• c. Direktør Clausen lovede arbejdet til en
  mand PPdef-indef
• President Clausen promised the job to a
  man
• d. Direktør Clausen lovede et arbejde til
  manden PPindef-def
• President Clausen promised a job to the
  man
•  
• The materials also included 64 fillers, 40 of
  which were sentences structurally similar to the
  target sentences but with semantic, syntactic or
  orthographic mistakes. The remaining 24 sentences
  were materials from an unrelated experiment.

13
Experiment 1

• Procedure. The stimuli were presented in a
  pseudo-random order one sentence at a time in the
  middle of the screen following a fixation point
  ().
• The subjects were instructed to accept or reject
  the sentences by pressing either a red X
  (rejection) or a green ? (acceptance) marked on
  the keyboard.
• A training session with four items were run first
  to familiarize subjects with the task. Reaction
  times (RT) and answers were recorded.

• Stimulus presentation was done using the free
  DMASTR software (DMDX, version 4.0.4.8) developed
  at Monash University and University of Arizona by
  K.I. Forster and J.C. Forster.
• Subjects. 30, 9 males, 21 females.

14
Statistical methods (following Baayen 2008)

• We analyzed the data using a linear mixed-effects
  regression model. The mixed-part is because it
  includes both fixed and random factors. This
  means that the variance due to differences
  between subjects and differences between items
  can be statistically controlled. In other words,
  some of the noise from people and sentences can
  be filtered out.
• A model was fitted to the dependent variable log
  RT using the software R (R Development Core Team,
  2009) and the lme4 package for R (Bates, Maechler
  Bolker 2009).

15
Table shows the fixed factors in the regression
model fitted to log RT with the NP-construction,
and given-new as reference level.
Estimate MCMCmean HPD95lower HPD95upper p
(Intercept) 79.742 79.670 78.454 80.773 0.0001
Givenness new-given 0.1709 0.1698 0.0775 0.2607 0.0008
Construction PP 0.0898 0.0901 -0.0182 0.1936 0.0972
Error 0.1189 0.1235 0.0537 0.1905 0.0012
Repetition -0.0079 -0.0078 -0.0120 -0.0039 0.0002
Previous RT 0.0000 0.0000 0.0000 0.0000 0.0001
Givenness new-given Construction PP -0.2597 -0.2588 -0.3853 -0.1264 0.0001
16
Results

• Previous RT. When a subject has responded slowly,
  he will answer slowly on the next item too. Fast
  subjects will correspondingly answer fast.
  Notice, that this effect is there despite the
  fact that item and subject variance has been
  filtered out.
• Repetition. If a subject sees an specific
  construction multiple times, he will respond
  faster and faster (cf. Luka Barsalou 2005 and
  Sprouse 2007).
• The crucial question The constructiongivenness
  interaction. Yes, the RT is significantly higher
  when the order is new-given, but only in the
  NP-construction.
• Error. Subjects are slower when they make a wrong
  answer.

17

• The interaction between construction and
  givenness. The difference between the two
  NP-constructions is significant. The difference
  between the two PP-constructions is not
  (ascertained by means of likelihood ratio tests).

18
Error/rejected sentences

• Few cases were subjects have rejected the
  sentences, because all sentences are considered
  grammatical.

19
Conclusion

• The effect of information structure (the
  discourse variable givenness) is structurally
  conditioned in the sense that it only has an
  effect in the NP-construction, not in the
  PP-construction.
• Syntactic representations can include
  information-structural constraints on their
  arguments
• Brown, Savova Gibson (2012194)

20
Experiment 2 (currently running)

• Does the givenness-effect in the NP-construction
  persist, when the complexities of the NPs are
  manipulated?
• a. Ib gav en gammel klog man med rød hat æblet
• Ib gave an old wise man with a red hat the
  apple
• b. Ib gav den gamle kloge man med rød hat et æble
• Ib gave the old wise man with a red hat an
  apple

21
Experiment 2 (currently running)

• Conditions (appear in a def-indef and an
  indef-def version, 5x2)
• a. Ib gave the man an apple
• b. Ib gave the old, wise man an apple
• c. Ib gave the old wise man with a hat an apple
• d. Ib gave the man an apple from Brazil
• e. Ib gave the man a very big, red apple from
  Brazil

22
References (the end)

• Baayen, R.H. (2008) Analyzing linguistic data a
  practical introduction to statistics using R,
  Cambridge University Press, Cambridge.
• Baayen, R. H. (2011) languageR Data sets and
  functions with "Analyzing Linguistic Data A
  practical introduction to statistics", R package
  version 1.2. http//CRAN.R-project.org/packagelan
  guageR.
• Bates, Douglas, Martin Maechler Ben Bolker
  (2011). lme4 Linear mixed-effects models using
  S4 classes. R package version 0.999375-42.
  http//CRAN.R-project.org/packagelme4.
• Bergenholtz, Henning (1992) Dansk frekvensordbog
  baseret på danske romaner, ugeblade og aviser,
  1987-1990, G. E. C. Gad, København.
• Bresnan, Joan, Anna Cueni, Tatiana Nikitina R.
  Harald Baayen (2007) Predicting the Dative
  Alternation, Cognitive Foundations of
  Interpretation, G. Bouma, I. Kraemer J. Zwarts
  (eds), Royal Netherlands Academy of Arts and
  Sciences, Amsterdam.
• Brown, Meredith, Virginia Savova Edward Gibson
  (2012) Syntax encodes information structure
  evidence from on-line reading comprehension,
  Journal of Memory and Language, 66, pp. 194-209.
• Clifton, Jr., Charles Lyn Frazier (2004)
  Should given information appear before new? Yes
  and no, Memory and Cognition, 32, pp. 886-895.
• Luka, B.J. and Barsalou, L.W. (2005). Structural
  facilitation Mere exposure effects for
  grammatical acceptability as evidence for
  syntactic priming in comprehension. Journal of
  Memory and Language, 52, pp. 436-459.
• R Development Core Team (2009) R A language and
  environment for statistical computing, R
  Foundation for Statistical Computing, Vienna,
  Austria. ISBN 3-900051-07-0, URL
  http//www.R-project.org.
• Sprouse, Jon (2007). Continuous acceptability,
  categorical grammaticality, and experimental
  syntax. Biolinguistics, 1, pp. 123-134.