Transcript Tooley et al 2009 Qi Presentation

Syntactic Priming
in Sentence Comprehension
(Tooley, Traxler & Swaab, 2009)
Zhenghan Qi
How do comprehenders organize and
utilize their grammatical knowledge?


Autonomous syntax models: structural
decision without reference to lexical,
semantic or other nongrammatical
information. (Chomsky, 1965; Frazier, 1979)
Lexicalist parser: a constraint-based neural
network architecture, in which syntactic
parsing processes are closely intertwined
with lexical processing. ( Traxler & Tooley,
2007)
Syntactic priming: test of the two theories

Two predictions:


Autonomous syntax models: syntactic priming
occur whether or not particular content words
are repeated across the prime and the target.
Lexicalist parser: syntactic priming increase
when the same word appear in prime and target,
but may not occur if there is no overlapping
lexical material. ??
Structural Priming in Production
(Bock et al., 2007)
Structural Priming in Comprehension
(Ladoux et al, 2007)
•RR Prime – RR Target: Significantly
reducing P600;
•MC Prime – RR Target: no
difference
N400
P600
How to exclude semantic effects?

Hypothesis:



Autonomous syntax models: modulation of the
P600 component for both repeated verb primes
and primes with synonymous verbs
Lexicalist parser model: P600 reduction
observed in repeated verb primes only.
The priming effects result from semantic match:
no P600 modulation in either priming condition.
(Tooley, Traxler and Swaab, 2009)
Experiment 1: Participants and Stimuli


20 students from UC-Davis
160 reduced relative clause sentences




R Prime: The man watched by the woman was
tall and handsome.
S Prime: The man observed by the woman was
tall and handsome.
Target: The child watched by the parent was
playing quietly.
Central presentation, 2 words/second
ERP Results
The child watched by …
N400 limited at posterior sites
ERP Results
No N400 effects here: semantic priming effects are not always observed when the
proceding sentence or discourse context is highly constraining (Couson et al, 2005).
ERP results
Some questions about Exp. 1

Any potential artifacts produced by
developed strategy across the trials?



The change in the P600 amplitude across
primes and targets did not differ b/w the first
and the second halves of the experiment.
Meta-analysis on 12 eye-tracking experiments
Slower reading rate in ERP recording

Natural reading in Experiment 2
Experiment 2: Participants and Stimuli


37 students from UC-Davis
7 prime-target pairs in the repeated
condition and 7 prime-target pairs in the
synonym condition.



The spy caught/captured by the FBI agent
disappear forever.
The criminal captured/caught by the detective
was in a state of panic.
Careful description of the counterbalancing
design
Eye-tracking results
The spy caught by the FBI agent disappear forever.
Verb
Region
PP Region
Summary



Robust priming effects of the disambiguating
region in the reduced relative sentences,
when the verb is repeated across prime and
target.
No priming in synonym condition.
No evidence supporting strategic prediction.
Conclusion



Syntactic priming occurs in comprehension.
Syntactic priming (largely) depends on the
presence of lexical overlap b/w prime and
target sentences with reduced relative
clauses.
Structural representations are bound to
individual words, and are not constructed on
the basis of generic elements.
Some of my confusions…



Will we see priming effects in synonym
condition if the prime and target sentences
are in main clause structure, given the
overwhelming subcategorization preference
of these verbs to main clause interpretation?
Is verb repetition necessary or sufficient for
structural facilitation?
What is the benefits of syntactic priming in
comprehension, given the semantic format
of memory storage?
Verb repetition is not sufficient for
structural facilitation



The doctor warned the patient would need
an operation. (SC Prime)
The family doctor warned the patient about
smoking. (DO Prime)
The security staff warned the spectators
would get rowdy...(Ambig. SC Target)
Kim and Mauner, 2006
Verb repetition is not sufficient for
structural facilitation


1st Half: Structural Priming. SC targets
read faster following SC primes
2nd Half: No Priming! SC targets facilitated
by SC AND DO primes.
With more distracters…


The prosecutor tried the defendant on a
murder charge.
Lorraine tried to get good grades all
semester long.
Results: No Facilitation! Robust ambiguity
effects in both halves.
Conclusion


Verb repetition is necessary but not
sufficient for structural facilitation (learning
instead of priming).
Activation of representations with more
surface details than verb lemmas (e.g. short
or long passives also matters)
(Mauner and Reeves, 2008)