Interpreting Pronouns Referring to the Arguments of Experiencer

Interpreting Pronouns Referring to the
Arguments of Experiencer/Stimulus Verbs:
Reversed Antecedent Preferences for Causal
and Consecutive Connectives
Berry Claus and Kalliopi Vozikaki
Psycholinguistics Group
Saarland University
Linguistic Evidence, Tübingen, February 2010
Outline
Interpretation of ambiguous pronouns: Effect of connective
Implicit causality
Implicit consequentiality
Empirical Study: two experiments
Linguistic Evidence, Tübingen, February 2010
Implicit Causality
Verbs that denote interpersonal states or events, such as confess, blame,
impress, admire, can exhibit implicit causality (cf. Garvey & Caramazza, 1974)
implicitly convey information as to which of the verb’s arguments is
the underlying cause
Empirical findings implicit-causality bias:
implicit causality affects the resolution of pronouns in causal subordinate
clauses (e.g., Caramazza et al., 1977)
Examples
Jim confessed to Joe because he …
Joe blamed Jim because he …
(NP1 verb)
(NP2 verb)
preference to resolve the pronoun with the implicated cause (e.g., Jim)
Linguistic Evidence, Tübingen, February 2010
Implicit-Causality Bias
Evidence for implicit-causality bias (i.a.)
from sentence-completion studies: proportion of reference in
continuations of sentence fragments
from reading-time studies: e.g., faster reading times with biascongruent than with bias-incongruent endings (Jim confessed to Joe
because he wanted a reduced sentence / because he offered a reduced
sentence)
Controversy: Are implicit-causality effects due to immediate focussing
or due to clausal integration (of explicit cause)?
Early or late effect?
Linguistic Evidence, Tübingen, February 2010
Implicit Causality: Early Effect
Evidence for early effect
Koornneef & van Berkum (2006): Individuals with different gender
unambiguous pronoun that indicates congruence/incongruence
Example
David and Linda were both driving pretty fast. At a busy intersection they crashed
hard into each other. David apologized to Linda because he according to the
witnesses was the one to blame. [congruent] / Linda apologized to David because
he according to the witnesses was not the one to blame. [incongruent]
Reading-time data (word-by-word) and eye movement data: Effect of
congruency at pronoun/next word
Event-related potential study by van Berkum et al. (2007): replication of
the early effect of congruence
Findings are inconsistent with causal-integration account (late effect)
But: findings do not provide evidence for strong version of focusing
account already at verb
Linguistic Evidence, Tübingen, February 2010
Implicit Consequentiality
Stewart et al. (1998): Implicit-consequentiality bias?
Material: Implicit-causality verbs in causal subordinate clauses
denoted events/states are described as cause
Example Because John annoyed Bill, … (implicit causality: NP1 verb)
Sentence-completion task: more references to non-cause argument
(e.g., Bill), i.e., the bearer of the consequences
reversed preference: implicit-consequentiality bias
Additional experiment: reading times of clauses expressing consequences
either congruent or incongruent with implicit-consequentiality bias
Because John annoyed Bill,
he complained to the art teacher. [congruent]
he was punished by the art teacher. [incongruent]
longer reading times for incongruent versions
Linguistic Evidence, Tübingen, February 2010
Role of Connective
(causal/consecutive)
Further evidence for reversed bias from sentence-completion studies that
manipulated the connective: causal because vs. consecutive so
(Au, 1986; Stevenson et al., 1994; Stevenson et al., 2000)
Stevenson et al. (2000): Clear reversed pattern for mental state verbs that
involve the thematic roles of Experiencer and Stimulus
Examples
Ken[Stimulus] impressed Geoff[Experiencer] because / so he … [SE verb]
Geoff[Experiencer] admired Ken[Stimulus] because / so he … [ES verb]
Causal because: continuations more Stimulus interpretations
Consecutive so: continuations more Experiencer interpretations
reserved pattern pronoun-interpretation bias may depend on connective
However, finding is based on a sentence-completion task that involves
language production components
Linguistic Evidence, Tübingen, February 2010
This Study
Does the effect of connective type on the interpretation of ambiguous
pronouns also obtain in tasks that focus on comprehension?
Two experiments (in German): Experiencer/Stimulus verbs (SE + ES)
Linguistic Evidence, Tübingen, February 2010
Experiment 1: Antecedent-Choice
Antecedent-choice task
Participants had to choose between two possible antecedents of an
ambiguous pronoun
Material:
Fragments of complex sentences: main clause + uncompleted subordinate
clause that ended with an ambiguous pronoun following the connective
Experimental items:
Manipulation of connective causal / consecutive
Main clause: Stimulus-Experiencer- (SE) or Experiencer-Stimulus-Verb (ES)
Examples
SE: The gardener irritates the golfer because / so he …
[Der Gärtner irritiert den Golfspieler, weil / so dass er …]
ES: The acrobat envies the juggler because / so he …
[Der Akrobat beneidet den Jongleur, weil / so dass er…]
Linguistic Evidence, Tübingen, February 2010
Experiment 1: Antecedent-Choice
Procedure:
Each fragment was presented on a separate screen
Below each segment: two alternatives
The gardener irritates the golfer because / so he …
gardener
golfer
Participants had to judge to which individual the pronoun referred to by
clicking on either of the two nouns
24 experimental items (12 SE-verbs + 12 ES-verbs) and 24 filler items
32 participants
Does the manipulation of the connective affect the antecedent choices?
Linguistic Evidence, Tübingen, February 2010
Experiment 1: Results
Antecedent Choices (in %)
90
80
70
60
50
40
30
20
10
because
so
Experiencer
because
so
Stimulus
Linguistic Evidence, Tübingen, February 2010
Experiment 2: Visual World
Does the effect of connective type also occur with more natural
language comprehension, i.e., in an online sentence-processing task
without explicitly drawing participants’ attention to the interpretation of
the pronoun?
Visual-World Paradigm
Look and listen; Participants were presented with pictures and
simultaneously listened to sentences
Linguistic Evidence, Tübingen, February 2010
Excursus: Visual-World Paradigm
Monitoring participant’s eye movements while they listen to spoken text
and view pictures of individuals and objects
Language mediated eye movements in the visual world
Looks to entities that are referred to (e.g., Tanenhaus et al., 1995)
Anticipatory looks towards probably upcoming objects (e.g., Kamide et al., 2003)
Looks triggered by pronouns (e.g., Järvikivi et al., 2005)
Example (from Altmann & Kamide, 1999, modified)
Viewing: scene showing a boy, a cake, and some toys
Hearing: The boy will eat the cake. He is hungry.
boy looks to depicted boy
eat looks to depicted cake
he looks to depicted boy
Linguistic Evidence, Tübingen, February 2010
Turning back to Experiment 2
Spoken sentences:
Complex sentences: main clause + uncompleted subordinate clause that
ended with an ambiguous pronoun following the connective
Experimental sentences:
Completed versions of the sentences used in Experiment 1,
each in two versions: because / so
Examples
SE: The gardener irritates the golfer because / so he constantly
walks across the golf course.
hits behind the ball.
[Der Gärtner irritiert den Golfspieler, weil / so dass er ständig
über den Golfplatz läuft / neben den Ball schlägt.]
ES: The acrobat envies the juggler because / so he always
gets more applause.
is very begrudging.
[Der Akrobat beneidet den Jongleur, weil / so dass er immer
mehr Applaus bekommt / sehr missgünstig ist.]
Linguistic Evidence, Tübingen, February 2010
Experiment 2: Material (cont)
Visually presented displays:
Pictures of four entities: two individuals mentioned in main clause + two
distractor objects
Experimental items:
Pictures of Experiencer and Stimulus (+ two distractor objects)
The gardener irritates the golfer
because / so he constantly
walks across the golf course.
hits behind the ball.
The acrobat envies the juggler
because / so he always
gets more applause.
is very begrudging.
Linguistic Evidence, Tübingen, February 2010
Experiment 2: Looks
Procedure:
Each trial started with the presentation of the visual display;
after a preview of 1000ms, the spoken sentence was presented (while the
depicted objects remained onscreen)
comprehension questions after 1/3 of the trials
24 experimental items (12 SE-verbs + 12 ES-verbs) and 28 filler items
36 participants
Does the manipulation of the connective affect the proportion of looks
to the depicted Experiencer/Stimulus and when does this effect occur?
Data were analyzed in different temporal regions:
pronoun, adverb, and the 500ms-interval following the adverb
The gardener irritates the golfer because / so he constantly […]500ms
The acrobat envies the juggler because / so he always […]500ms
Linguistic Evidence, Tübingen, February 2010
Experiment 2: Results
During pronoun and adverb: no effect of connective
500ms following adverb offset: interaction of connective and thematic role
Proportion of looks during the 500ms following the adverb offset
.35
.30
.25
.20
.15
.10
.05
because
so
Experiencer
because
so
Stimulus
Linguistic Evidence, Tübingen, February 2010
Summary
Experiment 1 – Antecedent Choice:
Effect of connective on antecedent choices causal: preference for
Stimulus, consecutive: preference for Experiencer
Experiment 2 – Visual World:
Effect of connective on looks causal: more looks to Stimulus,
consecutive: more looks to Experiencer
Evidence for reversed antecedent preferences for causal/consecutive
connectives (in line with Stevenson et al., 2000; Stewart et al., 1998)
May indicate that the implicit-causality bias hinges on the causal
connective rather than being due to an immediate focusing on the
implicated cause driven by the verb
Linguistic Evidence, Tübingen, February 2010
Late Effect in Experiment 2?
Experiment 2: Effect of connective was not an immediate effect; it did not
occur before adverb offset
“late” effect could be due to (a methodological flaw in) the material:
connective + pronoun immediately followed NP2
Currently: experiment with intervening additional words in between NP2
and connective (reference to an object on a scene that shows also Experiencer
+ Stimulus who are introduced by proper names, rather than role descriptions)
Other approach wrt time course issue: following Koornneef & van Berkum, 2006:
two characters with different gender: preference-congruent vs. incongruent pronoun
eye movements during reading
Linguistic Evidence, Tübingen, February 2010
Other Implicit-Causality-Verbs?
Current study: only Experiencer/Stimulus verbs
?Reversed antecedent preferences for the two connectives for other
examples of implicit-causality verbs, e.g., those denoting actions?
Experiencer/Stimulus verbs may be special
- strong implicit-causality bias on antecedent preferences
- clear reversed pattern for the two connectives in the Stevenson et al. study
- no clear preference in the absence of a causal/consecutive connective
(Source/Goal-verbs and Agent/Patient verbs: preference for role associated with
consequences [Goal, Patient])
Stewart et al. (1998; sentence completion): no reversed antecedent
preferences for action verbs such as thank, accuse (Rudolph & Försterling:
Agent-Evocator verbs)
e.g., preference for NP2 with Bill accused Ted because he … and with
Because Bill accused Ted, he …
Linguistic Evidence, Tübingen, February 2010
That’s it
Thanks to
Regine Bader and Emilia Ellsiepen for their assistance
in conducting Experiment 2
and to you for your attention
Linguistic Evidence, Tübingen, February 2010