Big questions in linguistics:
Issues in Information Structure
Part III: my own work on IS and
experimental studies
Natalia Slioussar
Utrecht University and St.Petersburg State University
[email protected]
My own work: what IS notions are
encoded in the grammar?
Background:
- First IS theories: topic and focus. Then: more notions.
- Giveness (D-linkedness). Starting from Chafe (1976):
degrees of accessibility. Chafe (1976): 3, Givón (1983): 9,
Ariel (1990): 15 degrees of accessibility.
- Contrast
- Encoding? Many authors: ambiguous or no grammatical
encoding for topic / focus
My own work: what IS notions are
encoded in the grammar?
Main claim: grammar encodes relative accessibility and
broadly conceived contrast (including emphasis).
My own work: what IS notions are
encoded in the grammar?
Russian: S V IO DO = neutral order, S V DO IO = scrambling
Previous approaches: scrambling = giveness or D-linking.
My own work: what IS notions are
encoded in the grammar?
Russian: S V IO DO = neutral order, S V DO IO = scrambling
Previous approaches: scrambling = giveness or D-linking.
However:
My own work: what IS notions are
encoded in the grammar?
R
My own work: what IS notions are
encoded in the grammar?
R
My own work: what IS notions are
encoded in the grammar?
Interim generalisations:
Accessibility: if A is more accessible than B, A moves over
B or looses the main stress.
Contrast: contrasted constituent must be the most
embedded (then stressed by default) or the main stress
must be shifted to it.
My own work: what IS notions are
encoded in the grammar?
Interim generalisations:
Accessibility: if A is more accessible than B, A moves over
B or looses the main stress.
Contrast: contrasted constituent must be the most
embedded (then stressed by default) or the main stress
must be shifted to it.
Exception: focus fronting. But: no IS-related effects in
Russian + Neeleman and Titov (2009) show that fronted foci
pass through the most embedded position.
My own work: what IS notions are
encoded in the grammar?
Paired contrasts: the first constituent? A phrasal stress is
shifted to it or it is made the most embedded within the
syntactic constituent corresponding to a prosodic phrase.
My own work: what IS notions are
encoded in the grammar?
Advantages: go back to the beginning and show how this
approach allows solving various problems traditionally
associated with topics and foci.
My own work: how are IS notions
encoded in the grammar?
How can relative accessibility and contrast be encoded?
- Choose configurations rather than features (difficult to
encode nested contrasts and impossible to encode relative
notions)
- Relying on the arguments for the primacy of syntax, build
the model around syntax. What can drive IS-related
movement? Chomsky’s edge features (some modifications
are required)
- Explain prosodic phenomena that prima facie do not have
syntactic counterparts
Experimental data
Two experiments:
- a child language experiment studying stress shift (as
opposed to IS-related movement) by Costa and Szendrői
(2006)
- my own self-paced reading time experiment on the
processing of different word orders in context (Slioussar
2007)
Experimental data: Experiment 1
Different approaches to IS: predictions about stress shift.
Child language acquisition:
- children successfully produce stress shift from very early
on (e.g. Baltaxe 1984; Nederstigt 2001)
- but fail to comprehend it until the age of 5 (Solan 1980;
McDaniel and Maxfield 1992; Crain, Ni and Conway 1994;
Halbert et al. 1995; Gualamini, Maciukaite and Crain 2003)
NB! children definitely can hear it and are in general very
sensitive to prosodic distinctions (Morgan 1986; HirschPasek et al. 1987).
No similar data on word order alternations.
Experimental data: Experiment 1
Costa and Szendrői (2006) on European Portuguese:
DP focus => the Tiger did not give the game to anybody
else. VP focus => the Tiger did not do anything else.
Independently known: children tend to choose wide focus.
Experimental technique: truth value judgement task.
Experimental data: Experiment 1
European Portuguese: both IS-related movement and
stress shift.
5 year olds understand the former, but not the latter. I.e.
they opt for the VP focus in (10b), rejecting it in the contexts
where the Tiger also did something else, but not in (10a)).
Experimental data: Experiment 2
First psycholinguistic studies in the middle of the XX
century:
Transformational grammar: surface structure = deep
structure + transformations. Prediction: each transformation
should take some time to process. Detect it!
Final outcome: time spent on many operations is impossible
to detect experimentally + difficult to exclude other factors.
Experimental data: Experiment 2
This approach survived in the experiments studying how
different word orders are processed.
Idea: compare neutral and non-neutral (‘scrambled’) word
orders and see how much time is spent on additional
movements.
Neutral word orders are indeed processed faster in many
languages (Frazier and Flores d’Arcais 1989; Hyönä and
Hujanen 1997; Stojanović 1999; Miyamoto and Takahashi
2002; Vasishth 2002 etc.).
Experimental data: Experiment 2
My idea (also in Kaiser and Trueswell 2004): neutral word
order is felicitous in isolation, while scrambled orders are
not. Maybe, this is the source of the additional slow-down?
Materials: S V IO DO, DO S V IO and DO IO V S Russian
sentences presented in one-sentence appropriate or
inappropriate contexts.
The contexts established two constituents in the target
sentence as given. Appropriate contexts presupposed
Given-Given-New word order, while inappropriate contexts
presupposed New-Given-Given or Given-New-Given word
order.
Experimental data: Experiment 2
Results:
- The context factor was significant, while the word order
factor was not. The less pronounced context effect
evidenced in previous studies (e.g., Kaiser & Trueswell,
2004) might be due to the use of shorter target sentences
and less extensive contexts.
Experimental data: Experiment 2
- The slow-down starts at the first contextually inappropriate
constituent => the information about context requirements is
taken into account immediately.
- However, the slow-down develops faster on preverbal
subjects and postverbal indirect objects (occupying their
canonical positions) than on preverbal indirect objects
(occupying a noncanonical position, or scrambled).
In the second experiment, these findings were replicated for
IO S V DO and IO DO V S orders. S V IO DO orders with a
continuation were used to show that there is no additional
effect of inappropriate context at the end of the sentence.