More Incrementality I.Pronoun Reference II.Language Production.
-
Upload
james-carpenter -
Category
Documents
-
view
218 -
download
0
Transcript of More Incrementality I.Pronoun Reference II.Language Production.
More Incrementality
I. Pronoun Reference
II. Language Production
• Condition B study designs…
Backward Anaphora Studies
Japanese
Japanese
which of his children (DAT) the man (NOM) …
which of his children (NOM) the man (DAT) …
Japanese pronouns and their antecedents
Verb
the man-nom
NP-dat
which of his children
which of his children (DAT) the man (NOM) …
Japanese pronouns and their antecedents
Verb
the man-nom
NP-dat
which of his children
NP-dat
which of his children
which of his children (DAT) the man (NOM) …
Japanese pronouns and their antecedents
Verb
the man-nom
NP-dat
which of his children
NP-dat
which of his children
Verb
NP-nom
which of his children
which of his children (DAT) the man (NOM) …
the man-dat
which of his children (NOM) the man (DAT) …
** ??
which of his children (DAT) the man (NOM) …
which of his children (NOM) the man (DAT) …
Gender Mismatch
the woman
the woman
Gender Mismatch paradigm: Carreiras et al. (1996); Osterhout et al. (1997); Sturt (2003)
Conditionsa. Scrambled - Gender Mismatch
Adverb / [his / which NP]-dat / Adverb / NP FEMALE-nom / Adverb / NP-acc /
verb-Q / NPMALE-top / verb
b. Scrambled - Gender Match
Adverb / [his / which NP]-dat / Adverb / NP MALE-nom / Adverb / NP-acc /
verb-Q / NPFEMALE-top / verb
c. Non-scrambled - Gender Mismatch
Adverb / [his / which NP]-nom / Adverb / NP FEMALE-dat / Adverb / NP-acc /
verb-Q / NPMALE-top / verb
d. Non-scrambled - Gender Match
Adverb / [his / which NP]-nom / Adverb / NP MALE-dat / Adverb / NP-acc /
verb-Q / NPMALE-top / verb.
Results: Scrambled conditions
Slowdown at mismatching NP is observed.
500
600
700
800
900
1000
1100
1 2 3 4 5 6 7 8 9 10
Region
Scrambled, match
Scrambled, mismatch
F1(1, 39) = 8.6, p<.01;F2(1,23)=7.4, p<.01
± Match
his/her
Results: Non-scrambled conditions
Slowdown at mismatching NP only when NP is possible antecedent.
500
600
700
800
900
1000
1100
1 2 3 4 5 6 7 8 9 10
Region
Unscrambled, match
Unscrambled, mismatch
Fs<1± Match
his/her
English
Immediate Constraint Application
While she was taking classes full-time, Jessica was working two jobs to pay the bills.While she was taking classes full-time, Russell was working two jobs to pay the bills.
While she …Jessica …
Russell …
Self-Paced Reading, Gender Mismatch Paradigm
(Kazanina, Lau, Lieberman, Phillips, & Yoshida, 2004)
Immediate Constraint Application
While she was taking classes full-time, Jessica was working two jobs to pay the bills.While she was taking classes full-time, Russell was working two jobs to pay the bills.
She was taking classes full-time while Jessica was working two jobs to pay the bills.She was taking classes full-time while Russell was working two jobs to pay the bills.
While she …
She …
Jessica …
Russell …
while Jessica …
while Russell …
Self-Paced Reading, Gender Mismatch Paradigm
(Kazanina, Lau, Lieberman, Phillips, & Yoshida, 2004)
-60
-40
-20
0
20
40
60
80
100
120
because lastsemester
while-cd SHE wastaking
classes while-ab NAME wasworking
full-time to…
Residual Reading Times
nonPrC GM
nonPrc GMM
PrC GM
PrC GMM
Results
GME at the 2nd NP in non-PrC pair
while while Jessica
Russell
(Kazanina et al., 2004)
-60
-40
-20
0
20
40
60
80
100
120
because lastsemester
while-cd SHE wastaking
classes while-ab NAME wasworking
full-time to…
Residual Reading Times
nonPrC GM
nonPrc GMM
PrC GM
PrC GMM
Results
GME at the 2nd NP in non-PrC pair
NO GME at the 2nd NP in PrC pairCondition C – immediate
while while Jessica
Russell
(Kazanina et al., 2004)
Experiment 2 (check back soon!)
• It seemed to him that John …
• It seemed to him that Mary …
• It seemed to his mother that John …
• It seemed to his mother that Mary …
Condition B & Active Search
• Hajime– Mary made fun of him in a very harsh manner [while Bob was having an
intimate conversation with the host.]
– Sam made fun of him in a very harsh manner [while Bob was having an intimate conversation with the host.]
– Mary made fun of him in a very harsh manner [while Sue was having an intimate conversation with the host.]
– Sam made fun of him in a very harsh manner [while Sue was having an intimate conversation with the host.]
– Rationale: if binding relation is not completed at pronoun, due to Condition B, search for antecedent will continue
Condition B & Active Search
• Takuya– Although John claimed that Mary insulted him, Bill didn’t tell that to Ken.
– Although John claimed that Mary insulted him, Sue didn’t tell that to Ken.
– Although Mary claimed that John insulted him, Bill didn’t tell that to Ken.
– Although Mary claimed that John insulted him, Sue didn’t tell that to Ken.
– Rationale: if binding relation is not completed at pronoun, due to Condition B, search for antecedent will continue
Time-Course Information
• Backwards anaphora studies: critical region is potential antecedent position, directly regulated by binding constraint
• Condition B studies: – Critical region provides information about what happened previously
– Critical region is not directly regulated by binding constraint
– Presumes active search for binder
Where to look for antecedents
a. Structurally
b. Representation-type
What do Pronouns Reactivate?
• Love & Swinney (1995)
– Jeff had read about problems with savings and loan institutions, so he went to the bank to ask about the safety that it provided with respect to CD investments.
Sturt 2003Experiment 2
Accessible-mismatch/Inaccessible-match
Jonathan was pretty worried at the City Hospital.
The surgeon [RC who treated Jonathan] had pricked herself with a used syringe needle. There should be an investigation soon.
Accessible-mismatch/Inaccessible-mismatch
Jennifer was pretty worried at the City Hospital.
The surgeon [RC who treated Jennifer] had pricked herself with a used syringe needle. There should be an investigation soon.
Discourse Interference?
• Lisa– {Sue, Joe} is a very enthusiastic fan of football. {She, he} likes to add
{her,his} own acrobatics to those already being performed. Last game, however, this led to a rather embarrassing result…
– Sue was painfully aware that a cheerleader noticed when Sue kicked her during a failed cartwheel attempt. [acc-match, inacc-match]
– Joe was painfully aware that a cheerleader noticed when Joe kicked her during a failed cartwheel attempt. [acc-match, inacc-mismatch]
– Joe was painfully aware that a cheerleader noticed when Joe kicked him during a failed cartwheel attempt. [acc-mismatch, inacc-match]
– Sue was painfully aware that a cheerleader noticed when Sue kicked him during a failed cartwheel attempt. [acc-mismatch, inacc-mismatch]
Discourse Interference?
• Clare– When their small business had financial troubles and there wasn’t enough
money for both their salaries
– Jane was generous and told Bob to pay him first.
– Jane was generous and told Bob to pay her first.
– Jane was generous and told Mary to pay her first.
– Jane was generous and told Mary to pay him first.
– Rationale: biases in favor of inaccessible antecedent, (a) distance, (b) potentially reflexive predicate, (c) plausibility
Antecedent Type Effects
• Fiorentino/Minai conjecture…
– On Principle B studies, “we observed the following. The results from the experiments using proper names showed effects of binding-theory incompatible antecedents (Badecker & Straub, 2002; Kennison, 2003). However, with full NPs (Nicol & Swinney, 1989; Clifton et al., 1997) and when the accessible proper name is introduced in a lead-in context sentence (Kennison, 2003; Runner, 2003) the apparent violations of Principle B were not evident.”
– “If we assume that the early filter works on discourse representations, then it might be possible to propose a split among full NPs and proper names in terms of the richness of their discourse representation.”
Principle B-as-initial-filter
• Nicol (1988), Nicol & Swinney (1989): cross-modal priming study in which subjects had to make a lexical decision to a visually presented word while listening to sentences
– The boxer told the skier that the doctor for the team would blame him for the recent injury.
punch – facilitationslope – facilitationnurse - no effect
Principle B-as-initial-filter
• Clifton, Kennison & Albrecht (1997): self-paced reading task. The supervisor(s) is a binding-accessible antecedent for his in (c-d) (but there is a number-match only in (d)), but not for him in (a-b).
a) The supervisors paid him yesterday to finish typing the manuscript.b) The supervisor paid him yesterday to finish typing the manuscript.
c) The supervisors paid his assistant yesterday to finish typing the manuscript.d) The supervisor paid his assistant yesterday to finish typing the manuscript.
• A number mismatch/match effect found in (c) vs. (d), but not in (a) vs. (b) => support for PrB as initial filter hypothesis
fast
slow
Principle B-as-a-late-filter
• Badecker & Straub (2002)
a) John thought that Bill owed him another opportunity to solve the problem.
b) John thought that Beth owed him another opportunity to solve the problem.
• The two conditions are different only in the gender of the inaccessible antecedent of him; yet reading times at the two words following him were faster in (a) than in (b) => binding constraints did not immediately rule out binding-inaccessible positions from the consideration.
How to Restrict Antecedents
How to Restrict Antecedents
• If the early filter works off discourse representations (see F/M conjecture), then how is a constraint like Principle B implemented?
How to Restrict Antecedents
• [± reflexive] marking on predicate?
– Self-anaphor marks predicate as reflexive, excludes all else
– Pronoun marks predicate as non-reflexive, excludes co-arguments
How to Restrict Antecedents
• Matt– I convinced the author to praise him because of the work’s merits.
– I convinced the authors to praise him because of the work’s merits.
– I convinced the author to praise his manuscript because of the work’s merits.
– I convinced the authors to praise his manuscript because of the work’s merits.
– Rationale: “the pronominal expression may not corefer with the matrix antecedent, even though it is within a separate clause […] We are thus increasing the level of resolution the parser must have to implement Principle B as an initial filter - it can’t rely on a simpler strategy like ‘clausemates’.
How to Restrict Antecedents
• Ivan– Pedro no quiere que el venga.
P not want that he come-subj.
– Pedro no cree que el venga.P not believe that he come-subj.
– Pedro no quiere que ella venga.P not want that she come-subj.
– Pedro no cree que ella venga.P not believe that she come-subj.
Incrementality in Comprehension
• Overall picture…?
Incrementality in Production
Incrementality in Production
• Different domains
– Speech errors
– Flexibility and incrementality
– Look-ahead in planning
V. Ferreira 1996
• Incremental models predict easier production with syntactic flexibility for two reasons– All structures are freely available to be filled
– Strict incremental construction permits the most active lexical representation (rather than syntactic competition) to determine structural decisions.
+
I gave
toyschildren
to
250ms
500ms
1500ms
250ms
Until button press
I gave toys to the children.I gave the children toys.I donated toys to the children.*I donated the children toys.
She gave it to the child.*She gave the child it.She gave the box to him.She gave him the box.
confused
himstory
250ms
1500ms
1000ms
Until button press
The story confused John.John was confused by the story.The story confused him.*Him was confused by the story.
confused
rejected