The document describes two experiments that investigate how people represent and compare stories about events they have not directly experienced. In Experiment 1, participants read either a realistic story or a fantastical story and then compared the two. Those who read a realistic story first judged the fantastical story as less similar and shorter than those who read the fantastical story first. In Experiment 2, participants read two realistic stories and showed greater similarity judgments and shorter duration comparisons than those in the fantastical condition in Experiment 1. The findings suggest that coherence of the first story influences representation of the second story, with more coherent first stories facilitating easier mapping between events. Even a single example story can provide grounding to represent a novel story.
6. Grounding can be based on previous examples.
Coherence of base domain influences representations.
Road Map
7. Grounding can be based on previous examples.
Coherence of base domain influences representations.
Lack of difference between events leads to shorter duration.
Road Map
8. Grounding can be based on previous examples.
Coherence of base domain influences representations.
Lack of difference between events leads to shorter duration.
“Novel” stories can be grounded with a single example.
Road Map
9. Grounding can be based on previous examples.
Previous specific examples matter
– Priority of the Specific (Brooks, 1978, Medin & Schaffer,
1978)
– Case-based reasoning (Kolodner, 1992, Leake, 1996)
– Transference (Andersen, Chen & Miranda, 2002)
Can a specific story influence our representation of
another specific story?
10. Coherency of base domain influences representations.
Coherence of base domain matters
– Coherence causes asymmetries (Bowdle & Gentner,
1997)
Coherence manipulated with our stories
– Realistic story (e.g., wolf attacking a chicken)
more coherent then fantastical story (e.g., book
attacking a pencil).
– Fantastical is perhaps more difficult to construct
initial representation.
11. Lack of difference between events leads to shorter duration.
Less differences = shorter duration estimates
– Less segmentation = shorter duration (Poynter,
1983)
Will degree of difficulty in mapping between two
stories be reflected in duration comparison
estimates?
12. What is a story?
• Events driven by intentional behavior
parallels real world experience.
• Contrast with expository text which outlines
an argument or explanation.
• Structure Building Framework (Gernsbacher
1997)
– Readers shift and create new foundations when
incoming information is less coherent with
previous structures.
13. Experiment 1
Fantasy vs. Reality
Between Subjects Design
– Subjects read stories consistent with reality:
• i.e., a wolf attacking a chicken
– Or inconsistent with reality, and therefore
fantastical:
• i.e., a book attacking a pencil.
– We assessed participants' judgments of similarity
and differences, duration and event structure for
the two story types.
15. Methodology
Subjects first read two parallel plot stories then:
• Rate relative similarity judgment
• Gave a relative duration comparison
16. Methodology
Subjects first read two parallel plot stories then:
• Rate relative similarity judgment
• Gave a relative duration comparison
• Wrote similarity/differences
17. Methodology
Subjects first read two parallel plot stories then:
• Rate relative similarity judgment
• Gave a relative duration comparison
• Wrote similarity/differences
• Broke stories into parallel events
18. Methodology
Subjects first read two parallel plot stories then:
• Rate relative similarity judgment
• Gave a relative duration comparison
• Wrote similarity/differences
• Broke stories into parallel events
• Gave a relative duration comparison for those parallel events
19. Methodology
Subjects first read two parallel plot stories then:
• Rated relative similarity judgment
• Gave a relative duration comparison
• Wrote similarity/differences
• Broke stories into parallel events
• Gave a relative duration comparison for those parallel events
• Gave a relative duration comparison for experimenter created
events
20. Subjects Read Two Parallel
Plot Stories
Subjects read two stories
• Realistic, e.g., a wolf attacking a chicken.
• Fantastical, e.g., a book attacking a pencil.
Order varied between subjects
• Subjects read either a fantasy story then a
realistic story or the reverse.
21. .
Subjects Read Two Parallel
Plot Stories
• Reality, e.g., wolf
attacking a chicken.
“Mary observed as
the wolf soared over
the pile of tools and
ran after the
chicken.”
• Fantasy, e.g., book
attacking a pencil.
“Mark watched as the
Math book leaped
over the homework
and chased after the
pencil.”
22. Story Controls
• Parallel plots
• Amount of dialog was the same for each story.
• Matched for word count
– Wolf attacking chicken story = 464 words
– Book attacking pencil story = 468 words
• Number of sentences
– Wolf attacking chicken story = 28
– Book attacking pencil story = 29
• Word frequency
• Verbs
– Type - i.e., action vs. state
– Valence positive vs. negative
23. Overall Judgments of
Similarity and Duration
Overall similarity comparison
• Similarity of the second story in relation to the first story.
Overall duration comparison
• Center vertical line = duration of the first story.
• Draw a vertical line to = duration of the second story in
relation to the first story.
• Greater distance from center vertical line = greater difference
in duration between two stories.
25. Duration Comparison Coding
An Example
• Measure distance between existing vertical line and
subject vertical line
12 mm
26. Duration Comparison Coding
An Example
• Measure distance between existing vertical line and
subject vertical line
• Subject distance/total distance + 1
• E.g., 12/120 + 1 = second story 90% duration of the first.
• Two coders for each packet
12 mm
120 mm
27. Breaking Stories
into Smaller Chunks
Similarity and differences statements
Breaking stories into parallel events
– Stories placed side by side
– Circle sentences that make up an event
– Duration comparison for events
Stories broken into 8 parallel events
– Duration comparison for events
28. Predictions
• Coherence of 1st story will influence representation
of 2nd story
• Because the reality story is more coherent than the
fantasy story
– Duration estimate for second story relative to first:
• Reality First shorter than Fantasy First
– Similarity judgment for second story relative to first:
• Reality First more similar than Fantasy First
– Similarity and differences statements
• More statements for Reality First than Fantasy First
34. “Novel” stories can stories grounded with a single example.
• Reality first influenced representation of
subsequent fantasy story
– Reduced sense of duration for the fantasy
story.
– Increased similarity between the two
stories.
– Increased the number of similarity and
differences statements
35. Experiment 2
Reality vs. Reality
Between Subjects Design
– Subjects read stories consistent with reality:
• e.g., a wolf attacking a chicken
• e.g., a friend attacking a teammate
– We assessed participants' judgments of similarity
and differences, duration and event structure for
the two story types.
36. Methodology
Subjects first read two parallel plot stories then:
• Rated similarity judgment
• Gave a relative duration comparison
• Wrote similarity/differences
• Broke stories into parallel events
• Gave a relative duration comparison for those parallel events
• Gave a relative duration comparison for experimenter created
events
37. .
Subjects Read Two Parallel
Plot Stories
• Reality 1, e.g., wolf
attacking a chicken.
“Mary observed as the
wolf soared over the
pile of tools and ran
after the chicken.”
• Reality 2, e.g., friend
attacking a teammate.
“Mark watched as the
catcher leapt over a
stack of weights and
chased after the pitcher.”
38. Story Controls
• Parallel plots
• Amount of dialog was the same for each story
• Matched for word count
– Wolf attacking chicken story 1 = 464 words
– Friend attacking teammate story 2 = 471 words
• Number of sentences
– Wolf attacking chicken story = 28
– Friend attacking teammate story = 29
• Word frequency
• Verbs
– Type - i.e., action vs. state
– Valence positive vs. negative
39. Predictions
• Coherence of 1st story will influence representation
of 2nd story
• Since both stories are based on reality they are both
coherent.
– Duration estimate for second story relative to first:
• For both conditions the second story will be shorter than first.
• The results for both conditions will be comparable to the relative
duration of the Reality First condition of Experiment 1
– Similarity judgment for second story relative to first:
• Similarity judgments for both conditions will be greater than the
Fantasy First condition of Experiment 1
42. Conclusions
• Coherence of first story influences
representation of second story.
– This coherence influences the mapping
between the two stories.
– And relative duration comparison can be used
to assess the ease of that mapping.
– And similarity judgments can be used to further
assess that mapping.
– And all of this happens with a single example.
43. Conclusions
Or Why this project is cool
• Created a novel dependent variable.
– Relative Duration Comparison
• Extended duration estimation literature to
discourse processing.
• Understand better the processing differences of
fantasy and reality stories.
• Extending analogy literature to larger structures.
• Perhaps constrained schema theories.
44. Future Directions
• Experiments on the near horizon
– Vary similarity question - compare 1st to 2nd story.
– Vary the strength of analogy between stories.
– Differences in duration a reflection of reading time
differences.
• Extend findings to elementary-school-aged
children
– Child’s difficulty at distinguishing reality/fantasy
• Constrain theories of discourse processing
45. Thanks!
Serge Block, Matt Jones, Lisa Narvaez, Tatjiana Feinstein,
Grant Baldwin, Jeff Laux, Micah Goldwater, Jon Rein:
For helpful discussions and
for listening to this talk over, and over and over and over again.
Robin Edwards, Fran Acuna-Neely, Kristen Wike
for their awesome help with coding and their helpful discussions.
Leora Orent:
for managing the Similarity and Cognition Lab
Leland Lockhart:
for not managing the Similarity and Cognition Lab
Nevin Pecorelli:
for always encouraging me.
Ewean Dennis:
for just being.