constructing agency: the role of language a dissertation …sk474gw9067/fauseydissertati… · in...
TRANSCRIPT
CONSTRUCTING AGENCY: THE ROLE OF LANGUAGE
A DISSERTATION
SUBMITTED TO THE DEPARTMENT OF PSYCHOLOGY
AND THE COMMITTEE ON GRADUATE STUDIES
OF STANFORD UNIVERSITY
IN PARTIAL FULFILLMENT OF THE REQUIREMENTS
FOR THE DEGREE OF
DOCTOR OF PHILOSOPHY
Caitlin Marie Fausey June 2010
http://creativecommons.org/licenses/by-nc/3.0/us/
This dissertation is online at: http://purl.stanford.edu/sk474gw9067
© 2010 by Caitlin Marie Fausey. All Rights Reserved.
Re-distributed by Stanford University under license with the author.
This work is licensed under a Creative Commons Attribution-Noncommercial 3.0 United States License.
ii
I certify that I have read this dissertation and that, in my opinion, it is fully adequatein scope and quality as a dissertation for the degree of Doctor of Philosophy.
Lera Boroditsky, Primary Adviser
I certify that I have read this dissertation and that, in my opinion, it is fully adequatein scope and quality as a dissertation for the degree of Doctor of Philosophy.
Herbert Clark
I certify that I have read this dissertation and that, in my opinion, it is fully adequatein scope and quality as a dissertation for the degree of Doctor of Philosophy.
James McClelland
Approved for the Stanford University Committee on Graduate Studies.
Patricia J. Gumport, Vice Provost Graduate Education
This signature page was generated electronically upon submission of this dissertation in electronic format. An original signed hard copy of the signature page is on file inUniversity Archives.
iii
iv
Abstract
Speakers of the world's languages differ in how they typically describe the
same events. For example, to describe the same physical event in some languages it
would be natural to say "He broke the vase" while in others one would say "The vase
broke itself." Do such patterns in language matter for how people construe and
remember the same events? Do patterns in language shape whether we construe
someone as being an agent, whether we attend to and remember who was involved,
and how much we blame and punish those involved? Evidence from several
populations - speakers of English, Spanish, and Japanese; adults and children –
suggests that the answer to these questions is "Yes". There are cross-linguistic
differences in eye-witness memory for the same events, and language influences
judgments of blame and punishment. The effects of language appear to be strong:
Patterns in one's linguistic environment affect thinking even when people are not
required to use language in a task and even when other rich sources of non-linguistic
information are available.
v
Acknowledgements
A great many people have helped make this dissertation possible and deserve
more than these brief acknowledgements. Please accept these words as a sparse
representation of truly rich gratitude.
Thanks to many faculty members for providing inspiration, resources and
feedback about scholarly pursuits. Thanks to my advisor, Lera Boroditsky, for
supporting all of this research, and to my committee members for their feedback: Eve
Clark, Herb Clark, Ellen Markman and Jay McClelland. Thanks also to my friend and
colleague, Teenie Matlock, for being there every step of the way.
Graduate school at Stanford has introduced me to awesome friends and
colleagues. Special thanks to the following people who routinely listen to incipient
ideas, look at the latest graphs, and brainstorm about interpretations and next steps:
Casey Lew-Williams, Rosanna Olsen, Alexia Toskos, Adriana Weisleder, and Nathan
Witthoft. Someone once said, “Everyone needs someone to inspire them to do better
than they know how”. This cast of characters has inspired me.
Fun and productive conversations with so many folks over the past many years
have improved my thinking and research. I have learned so much from everyone.
Among faculty I haven’t yet mentioned, Gordon Bower, Carol Dweck, Anne Fernald,
Dedre Gentner, Florian Jaeger, Sotaro Kita, Beth Levin, Kristina Olson, Michael
Ramscar, Daniel Richardson, Ewart Thomas, Jyotsna Vaid, and Phillip Wolff have
spent time discussing matters large and small with me. Among students I haven’t yet
mentioned, Chris Bryan, Janice Chen, Nick Davidenko, Katia Dilkina, Orly Fuhrman,
Steve Flusberg, Jeremy Glick, Sarah Gripshover, Tania Henetz, Alex Jordan, Angela
vi
Kessell, Fred Leach, Jamie Leach, Adam November, Yula Paluy, Neal Snider, Nola
Stephens, Daniel Sternberg, Paul Thibodeau, Hal Tily, Jon Winawer, Dan Yarlett,
Davie Yoon and Emily Zitek have lent kind ears and helping hands whenever I asked.
The generosity of spirit in these terrific people is something I will never forget.
Implementing research is a team effort, especially when stimuli require live
acting talent and when data are collected on multiple continents. This research would
not have been possible without the direct contributions of many talented people.
Thanks to Travis Korenaga, Fred Leach, Casey Lew-Williams, Jordan Otomo, David
Remus, and Scott Seki for acting in videos used for experimental stimuli. Thanks to
Rosa Bahamondes and María Paz Zúñiga for help with data collection in Santiago,
Chile. Thanks to Noburo Saji and Hal Tily for help with data collection in Japan and
Dr. Mutsumi Imai for lending use of her lab space at Keio University in Tokyo, Japan.
Thanks to Jennifer Winters, Chia-wa Yeh, and many teachers for their help in
conducting research at Bing Nursery School. Thanks to Louie Gularte, Kaveh
Moghbeli and Alexia Toskos for help programming experiments. Special thanks to
Kelly McCormick, the Cognation lab manager, for her heroic feat of maintaining a
happy and organized lab.
A wonderful group of undergraduates has contributed to many aspects of this
research. I especially thank Bria Long and Aya Inamori for their work with Japanese
speakers, EB Meade for her work with children, and Rayden Llano for ongoing
projects with Spanish-English bilinguals. I’m also thankful to many others I’ve had
the pleasure to work with: Jessica Alderman, Adam Clare, Evan Cox, Jenilee Deal,
Crystal Espinosa, Louie Gularte, Martha Gutierrez, Annelise Han, Lauren Hay,
vii
Nathalie Heitz, Gavin Jenkins, Tamar Kreps, Maricela Montoy, Dean Park, Tiffany
Shih, Veronica Trejo, Vijay Vanchinathan. I consider myself very fortunate to have
worked with and learned from such motivated and engaged students.
Some truly outstanding people have made everyone’s lives easier in the
Stanford Psychology Department (and wherever else they may be now). Priscilla
Fiden is a gem of a person who solves problems with grace, ease and understanding,
and I’m very grateful for her dedication to students. Thanks also to Harry Bahlman,
Tamara Danoyan, Roz Grayson, Beth McKeown, and Peter Smith for helping us all to
stay paid, housed, fed and generally supported.
I am honored to call many people I’ve met in graduate school close friends. To
Adriana, Alexia, Beth, Casey, Fred, Jamie, and Rosanna, thank you for everything.
To Karen and Lynn Green, thanks for being there always.
Special thanks to my dearest friend, Patrick Macdonald, without whom I may
not have made it to this point in my life intact. His words have steadied me, his
presence grounded me.
Most of all, thanks to my family. Thank you to my sisters, Darcie and Claire,
who’ve earned the right to call me “Cait” and expect a response. Thank you to my
parents, Bill and Mary Kay, for giving me freedom to figure out what I want to do and
for supporting my pursuit of those passions.
I dedicate this thesis to those who take the next step.
viii
Table of Contents
1. Introduction. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1 2. Language and eye-witness memory in causal events. . . . . . . . . . . . . . 7 2.1. Descriptions and eye-witness memory in English and Spanish speakers. 7 2.2. Descriptions and eye-witness memory in English and Japanese speakers. 17 2.3. Eye-witness memory in primed English speakers. . . . . . . . . . . . . 27 3. Language and attributions of blame, financial liability, and moral goodness. . 35 3.1. Language and attributions of blame and financial liability in adults. . . . 35 3.2. Language and judgments of moral goodness in children. . . . . . . . . . 49 4. Conclusions. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 61 5. References. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 63 6. Footnotes. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 73 7. Tables. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 77 8. Figure Captions. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 83 9. Figures. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 85 10. Appendix. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 93
ix
List of Tables Table 1. Causal event stimuli (English and Spanish speakers) . . . . . . . . . . . 77 Table 2. Causal event stimuli (English and Japanese speakers) . . . . . . . . . . . 78 Table 3. Prime sentences (English speakers) . . . . . . . . . . . . . . . . . . . . . 79 Table 4. Restaurant fire reports and questions . . . . . . . . . . . . . . . . . . . . 80 Table 5. Wardrobe malfunction reports and questions . . . . . . . . . . . . . . . . 81 Table 6. Causal event stimuli (English children) . . . . . . . . . . . . . . . . . . . 82
x
List of Figures Figure 1. Distributions of causal event descriptions in English and Spanish . . . . 85 Figure 2. Describing and remembering agents in English and Spanish . . . . . . . 86 Figure 3. Distributions of causal event descriptions in English and Japanese. . . . 87 Figure 4. Describing and remembering agents in English and Japanese. . . . . . . 88 Figure 5. Remembering agents after agentive or non-agentive linguistic primes . . 89 Figure 6. Independent contributions of guilt and linguistic framing to financial liability sentences. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 90 Figure 7. Language changes punishment of an observed individual. . . . . . . . . 91 Figure 8. Language changes how children judge accidental agents. . . . . . . . . 92
1
Chapter 1. Introduction.
Throughout life, we act on the world around us. We move and shake things, we
build and break things. We also routinely make inferences about others’ actions and
observed outcomes, like deciding who to blame for what. Mundane, everyday life may
lead us to think that causal agency is a natural, straightforward, and universal feature
of human experience. Further reflection and careful study, however, show that it is
anything but.
Consider this scenario: A woman rushes into a café, collapses into a chair at
the nearest table, and a glass atop the table falls to the floor and breaks. Was the
woman or fragile glass the cause? Was falling or breaking the effect? Did something
cause or enable the effect? The way that people perceive and remember this event may
depend on what they pay attention to in-the-moment as well as how they typically
attend to agency in causal events. The cues that guide one-time and habitual causal
event construal are not currently well understood. The research in this dissertation
builds on previous research about how visual context and social norms shape people’s
notions of agency, and suggests that language should also be considered among these
influential context cues. Everyday linguistic descriptions, such as whether someone
says “She broke the glass” or “The glass broke” may be pervasive and powerful cues
to agency.
Insightful investigations into the psychology of agency have revealed that
“causal agent” is a context-dependent construct. In vision, the perception of physical
causality is easily altered by minor changes in context. For example, when a certain
2
kind of Michotte-like event is presented in isolation, people perceive it as a non-causal
“pass” (Ball A moves, spatially overlaps with Ball B, Ball B moves). However, simply
adding another causal event to the scene (Ball A moves, contacts B, Ball B moves)
leads people to see the “pass” event as a causal launch (Scholl & Nakayama, 2002).
Whether a moving disc is seen as a causal agent – in the exact same event – depends
on the visual context.
In the social domain, societies across the world promote different concepts of
the self, with East Asian societies being more interdependent and Western societies
being more independent (e.g., Markus & Kitayama, 1991; 2004). Compared to people
in interdependent societies, people in independent societies are more likely to select a
single proximal cause for an event (e.g., Chiu, Morris, Hong, & Menon, 2000; Choi,
Dalal, Kim-Prieto, & Park, 2003), are less aware of distal consequences of events
(e.g., Maddux & Yuki, 2006), are more susceptible to correspondence bias (e.g., Choi,
Nisbett, & Norenzayan, 1999), and are more motivated by personal choice (e.g.,
Iyengar & Lepper, 1999).
“Agent” does not appear to be a stable, universal property of events in the
world. What people see and believe to be an agent is constructed in context. What
kinds of context cues do people use to construct agency? Any pervasive, systematic
information that is part of how people experience events is likely to play a role. In
addition to visual regularities and social norms, language is one such pervasive part of
human experience. We talk about the events of our days, often discussing who did
what to whom. We talk to our children, we update our bosses, we negotiate with
3
friends and colleagues. We learn from other people’s stories. Our everyday
conversations are filled with regularities about how to communicate “what happened”.
Could patterns in language shape whether we construe someone as being the agent of
an event, whether we attend to and remember who was involved, and how much we
blame and punish those involved?
Previous research supports a role for language in guiding how people construe
their world. Using language changes how people perceive emotion (e.g., Barrett,
Lindquist, & Gendron, 2007), represent objects (e.g., Lupyan, 2008) and remember
events (e.g., Billman & Krych, 1998; Gentner & Loftus, 1979; Loftus & Palmer,
1974). People are sensitive to how often certain expressions are used within their
linguistic community (Gahl & Garnsey, 2004; Landauer & Dumais, 1997; Saffran,
Aslin, & Newport, 1996;) and speakers of different languages talk about events
differently (e.g., Gentner & Goldin-Meadow, 2003). Language directs attention (e.g.,
Reali, Spivey, Tyler, & Terranova, 2006; Richardson & Matlock, 2007) and through
repeated use may guide habitual event construals (e.g., Slobin, 1996). Indeed, research
in the tradition of linguistic relativity has suggested that habitual ways of talking
influence how people think about colors (e.g., Roberson & Hanley, 2007; Winawer,
Witthoft, Frank, Wu, Wade, & Boroditsky, 2007), space (e.g., Levinson, Kita, Haun,
& Rasch, 2002), objects (e.g., Boroditsky, Schmidt, & Phillips, 2003; Dilkina,
McClelland, & Boroditsky, 2007; Imai & Gentner, 1997; Lucy, 1992) and events
(Finkbeiner et al., 2002; Gennari et al., 2002; Oh, 2003; Papafragou et al., 2008;
4
Slobin, 1996; 2003; Trueswell & Papafragou, 2010; Wolff, Jeon, & Li, 2009). Could
language impact how people construct agency?
In this dissertation, a series of studies examines this question in detail. In
Chapter 2, the relationship between language and eye-witness memory is examined.
Four experiments test the hypothesis that people in linguistic communities with more
agentive language for certain kinds of events are more likely to pay attention to and
remember causal agents of these events than people in communities with less agentive
language. In Chapter 3, the relationship between language and attributions of blame,
financial liability and moral goodness is examined. Three experiments test whether
agentive descriptions lead adults to blame and punish agents more harshly, compared
to non-agentive descriptions of the very same events. Two experiments test whether
children are also sensitive to event descriptions when judging agents of causal events,
such that agentive descriptions lead children to judge agents as more morally bad,
compared to non-agentive descriptions of the very same events. Because language is
such a pervasive context cue in people’s experience of causal events, it is important to
understand whether patterns in language affect how people construe agency.
Language could influence how people construct agency in several ways. Of
course, one possibility is that language does not affect eye-witness memory or blame
judgments at all. If language does matter, there are at least two ways it could influence
thinking in each domain.
In the case of eye-witness memory, a distinction may be made about whether
people must use language when they witness an event in order for language to
5
influence what they remember about the event. Language might influence memory
only when people actively use language. For example, when observers know that they
will need to talk about an event that they observe, they may use language as a tool to
encode and retrieve details about the event. People who use different language might
remember different details. Alternatively, patterns in one’s linguistic environment
might affect thinking when people are not required to use language as they witness
and recall events. For example, people who speak languages with different
conventions for talking about causal events may remember events differently even
when they do not communicate about the events that they witness.
In the case of blame, a distinction may be made about whether language is the
only information available to reasoners or whether they also witness the event and
know something more about it. Linguistic framing might guide people’s judgments if
language is the only information available. This situation is common in court
proceedings, when linguistic testimony is all that jurors have to go on. In this case,
different language may prompt people to imagine different events and therefore reason
differently about the individuals in the event. Alternatively, linguistic framing might
influence blame attribution when people not only get a linguistic description of an
event, but also have the opportunity to represent the event in non-linguistic ways. For
example, linguistic framing might guide people’s judgments even when they also
observe a video of the event, and when they know other details about the event that are
not described.
6
Thus, language could matter in either a “medium” way or a “strong” way in
two domains for which people must construct agency. The experiments presented in
this dissertation allow us to discover how strongly language may influence eye-
witness memory and attributions of blame. Do patterns in language shape whether we
construe someone as being an agent, whether we attend to and remember who was
involved, and how much we blame and punish those involved? Do patterns
in one's linguistic environment affect thinking even when people are not required to
use language in a task and even when other rich sources of non-linguistic information
are available?
7
Chapter 2. Language and eye-witness memory in causal events.
Are there cross-linguistic differences in eye-witness memory? Can patterns in
our linguistic environment influence what we remember about the events we witness?
In this chapter, we identify cross-linguistic differences in how English, Spanish and
Japanese speakers describe the same events, and find that there are corresponding
cross-linguistic differences in eye-witness memory.
2.1. Descriptions and eye-witness memory in English and Spanish speakers
You see someone accidentally brush against a flower vase and the vase ends up
in pieces on the floor. When asked about what happened, you might say, “She broke
the vase”. In English, agentive descriptions like this are typical and appropriate even
for clearly accidental events. By contrast, non-agentive language often sounds evasive
(e.g., Reagan’s famous “mistakes were made” in the 1987 State of the Union Address).
Linguistic analyses suggest that in other languages, non-agentive expressions are more
frequent and used to distinguish accidental from intentional actions (Dorfman, 2004;
Filipovic, 2007; Maldonado, 1992; Martinez, 2000; Slobin & Bocaz, 1988). For
example, in Spanish non-agentive expressions with the clitic se are often used to
describe accidents (e.g., “Se rompió el florero”, translating roughly as “The vase broke
itself”).1
8
Previous work in English has shown that alternations between agentive and
non-agentive descriptions can have important consequences for how people reason
about events (Fausey & Boroditsky, in press). For example, English speakers who read
a report about Justin Timberlake and Janet Jackson’s wardrobe malfunction containing
the agentive expression “tore the bodice” not only blamed Timberlake more, but also
levied 53% more in fines than those who read the non-agentive “the bodice tore”.
Further, this linguistic framing had a big effect on blame and punishment even when
people watched a video of the event and were able to witness the tearing with their
own eyes.
In the following experiments, we investigate whether agentivity in event
descriptions also affects eye-witness memory. If events would normally be described
less agentively in your linguistic community, would you be less likely to pay attention
to and remember the agents of those events than if they were normally described more
agentively? Previous work has examined the role of linguistic framing in eye-witness
memory within a language by presenting participants with different descriptions of the
same event, for example varying the vividness of verbs, and measuring effects on
memory (e.g., Gentner & Loftus, 1979; Loftus & Palmer, 1974). The studies here
extend this work to the cross-linguistic domain and examine memory for agents.
Instead of giving participants different descriptions of the same event, we ask whether
speakers of two different languages that would typically describe an event differently
would naturally pay attention to, encode, and remember different aspects of the same
event. That is, are speakers of different languages habitually operating in different
9
linguistic framing conditions as a function of how events are normally described in
their linguistic community?
Much previous cross-linguistic work on the role of language in event cognition
has focused on cross-linguistic differences in encoding the manner and path of motion
(e.g., Billman & Krych, 1998; Finkbeiner, Nicol, Greth, & Nakamura, 2002; Gennari,
Sloman, Malt, & Fitch, 2002; Oh, 2003; Papafragou, Hulbert, & Trueswell, 2008;
Papafragou, Massey, & Gleitman, 2002; Slobin, 2003; Trueswell & Papafragou,
2010). Many of these studies have found cross-linguistic differences in how people
encode and reason about motion events (e.g., Finkbeiner et al., 2002; Gennari et al.,
2002; Oh, 2003; Papafragou et al., 2008; Slobin, 2003; Trueswell & Papafragou,
2010), though some find such differences only when people are explicitly instructed to
describe the events during the task (e.g., Gennari et al., 2002; Papafragou et al., 2008).
Observing a cross-linguistic difference on a test of cognitive performance even when
people are not required to use language in the task has become the gold standard for
establishing basic cross-linguistic differences in cognition. Here, we specifically test
for cross-linguistic differences in memory for causal events in a task where
participants are not asked to describe the events at any time before or during the
memory task.
In Study 1 we establish that there is indeed a difference between Spanish and
English speakers’ descriptions of the same causal events, and in Study 2 we test
whether these differences in language have consequences for people’s eye-witness
10
memory. We hypothesized that more agentive language in one’s linguistic
environment would lead to better memory for the agents of events.
Study 1
Method
Participants. 68 English speakers (mean age = 31.49 years) and 29 Spanish
speakers (mean age = 28.69 years) participated. Participants were monolinguals who
completed the study via Amazon’s Mechanical Turk service
(https://www.mturk.com/mturk/welcome). All participants reported that their native
language, and over 80% of their current daily language use (mean = 99.98%), is the
target language and that they did not start learning any other language until after age
12.2
Materials. Participants read instructions in either English or Spanish.
Instructions in the two languages were developed simultaneously and verified by an
independent Spanish-English bilingual.
Videos of intentional and accidental versions of 16 unique events were used
(Table 1). In all events, a man physically interacted with an object. The man’s reaction
differed between the intentional and accidental versions of the event. For example, in
the intentional version of the pencil breaking event, a man who was seated at a desk
picked up a pencil, deliberately broke it in half and looked satisfied. In the accidental
version of this event, a man was writing, and while writing the pencil broke in half. In
this case, the man showed a startle response and threw his hands up in surprise. Thus,
the accidental events were characterized by a “whoops!” reaction such as a startle
11
response, surprised facial expressions and/or surprised hand gestures. Videos of eight
events (both intentional and accidental versions) featured an actor in a blue shirt and
videos of another eight events (both intentional and accidental versions) featured a
different actor in a yellow shirt.
Procedure. Participants watched 16 videos and were asked to provide a
linguistic description for each video. In each description trial, participants viewed a
video and then answered the question “What happened?” (“¿Qué pasó?”). Each video
showed a different event; half featured the actor in blue and half the actor in yellow,
half were intentional actions and half accidental. Whether an event was presented in
its intentional or accidental version was counterbalanced across participants. Videos
were presented in one of two pseudo-random orders that ensured that no more than
three videos of the same agent or the same intention appeared in a row.
Results
Descriptions were coded as agentive if the sentence described the change-of-
state event using a transitive expression. A canonical agentive description would be
“He popped the balloon”. Descriptions were coded as non-agentive if the change-of-
state event was described intransitively. A canonical non-agentive description would
be “The balloon popped”. Some non-agentive descriptions took the form “Someone
was doing X and then Y happened”, in which the agent was linguistically separated
from a change-of-state event that was described intransitively. Most (84.21%) of the
Spanish non-agentive sentences were marked by the clitic se.3 Across all participants,
7.54% of descriptions did not describe the event and were excluded from analyses. All
12
descriptions were coded by the first author and an independent rater, with high point-
to-point reliability (97.15% English, 96.12% Spanish). Disagreements were resolved
upon discussion.4
Results are shown in Figures 1 and 2a. Intentional events were described
equally agentively by both English and Spanish speakers (English M = 95.50, SE =
.95; Spanish M = 92.46, SE = 1.69, t(92) = 1.65, n.s.). Accidental events, on the other
hand, were more often described agentively by English speakers than by Spanish
speakers (English M = 74.55, SE = 2.48; Spanish M = 59.61, SE = 3.56, t(92)= 3.31, p
= .001, d = .69).
To compare how strongly speakers distinguished between intentional and
accidental events in their descriptions, we computed a difference score for each
participant as the proportion of intentional events described using agentive language
minus the proportion of accidental events described using agentive language. This
distinction was more pronounced for Spanish speakers (M = 32.85, SE = 3.51) than for
English speakers (M = 20.87, SE = 2.54), t(92) = 2.61, p = .01, d = .55. This cross-
linguistic difference was also consistent across events, t(15) = 3.05, p = .008.5
English and Spanish speakers described intentional events similarly but
differed in their descriptions of accidental events. In Study 2, we investigated whether
these differences in description may yield corresponding differences in memory.
13
Study 2
Method
Participants. 113 English speakers (Stanford University, mean age = 19.13
years) and 109 Spanish speakers (Universidad de Chile, mean age = 20.85 years)
participated. Participants were selected to be under 25 years old to ensure a
homogenous sample for memory performance. All participants were monolingual, by
the same criteria used in Study 1 (mean target language use = 98.84%). None of the
participants had taken part in Study 1.
General design. Participants read instructions in either English or Spanish. All
participants completed two tasks, first the Object-orientation memory task and then
the Agent memory task. The first task was designed to be a measure of memory
performance that was not predicted to vary across language communities. The second
task was designed to test for differences in non-linguistic eye-witness memory
(memory for the agents of events) between English and Spanish speakers. The two
tasks were non-linguistic measures of memory – participants never described any of
the images or events during these two tasks, nor were they provided with any
linguistic descriptions.
Object-orientation memory
During encoding, participants saw pictures of 15 objects presented on a
computer screen one at a time for two seconds each (images courtesy of Michael J.
Tarr, Brown University, http://www.tarrlab.org/). Each object appeared in one of three
14
possible orientations, counterbalanced across participants. Participants were instructed
to pay attention to the images and were told that their memory would be tested.
After the encoding phase, participants were given a brief distracter task
(counting the number of white squares on a grid of black and white squares), followed
by the memory test. For the memory test, participants were shown the three possible
orientations of each object and asked to indicate which one they had seen previously.
Agent memory
Video materials. For the encoding phase, the same videos were used as in
Study 1.6 For the test phase, we used an additional set of videos showing all the same
events but with actions performed by a new, third actor. The same silent videos served
as stimuli for both English and Spanish speakers.
Encoding. During the encoding phase, participants viewed 16 videos,
following the same counterbalancing scheme as Study 1. Each video showed a
different event (half featured the actor in blue and half the actor in yellow, half were
intentional actions and half accidental). Participants were instructed to pay attention to
the videos and were told that their memory would be tested, but were not given any
further clues. After viewing all 16 videos (once each, with a one second pause
between videos), participants were instructed to count to 10 as a brief distracter task.
Test. Test trials consisted of a probe video followed by still photos of the two
agents from the encoding phase. In the probe videos, a third actor appeared as the
agent of the same events participants saw during encoding. For example, if a
participant had seen the “accidental balloon popping” event during encoding, they
15
would see this same event acted by the new agent in the test phase. After each probe
video, participants were asked, “Who did it the first time?” (“¿Quién lo hizo la
primera vez?”) and responded by clicking on either the blue-shirt man or the yellow-
shirt man. Participants were tested only on the events they had seen during encoding,
presented in a different pseudo-random order from the encoding phase, and received
no feedback.
Results
Twelve participants were excluded from analyses for one of the following
reasons: (a) chance performance on the object-orientation memory task (1 English, 3
Spanish), or (b) a z-score greater than |2| (relative to language group) on the Memory
Difference Score (Intentional Memory minus Accidental Memory) (2 English, 6
Spanish). The Memory Difference Score was the basis for the analysis of interest in
this study and we wanted to be sure that outliers did not drive any observed
differences.7
Results are shown in Figure 2b. Intentional agents were remembered well by
both English (M = 78.18, SE = 1.66) and Spanish (M = 78.00, SE = 1.57) speakers,
t(208) = .08, n.s. However, as predicted, accidental agents were better remembered by
English speakers (M = 78.98, SE = 1.61) than by Spanish speakers (M = 73.75, SE =
1.67), t(208) = 2.25, p = .01, d = .31. As predicted by patterns in language (Study 1),
the distinction between memory for individuals involved in intentional and accidental
events was more pronounced for Spanish speakers (M = 4.25, SE = 1.65) than for
16
English speakers (M = -.80, SE = 1.74), t(208) = 2.09, p =.02, d = .29. This cross-
linguistic difference was also consistent across events, t(15) = 2.02, p = .03.
In addition to equivalent memory for agents of intentional events, English and
Spanish speakers did not differ in their memory for the orientation of objects in the
object-orientation memory task (M = 75.09, SE = 1.42 and M = 73.53, SE = 1.42,
respectively), t(208) = .77, n.s. This helps ensure that the cross-linguistic differences
in eye-witness memory for accidents are not due to more general differences in
memory capacity between the two groups.8
Discussion
English and Spanish speakers remembered agents in a pattern consistent with
event descriptions in their respective linguistic communities. Both groups described
intentional events agentively and had similarly strong memory for the agents of these
events. When it came to accidental events, however, English speakers used more
agentive descriptions than did Spanish speakers and also remembered the agents of
these events better than did Spanish speakers. It is not the case that Spanish speakers
had poorer memory than English speakers more generally. The two groups showed
similar memory for agents of intentional events as well as for object-orientation. Only
accidental events were described and remembered differently across communities.
These findings show a close coupling between the way events are typically talked
about in a linguistic community and what people encode and remember about these
events even when they are not talking.
17
2.2. Descriptions and eye-witness memory in English and Japanese speakers
In the next experiment, we examined language and eye-witness memory for
intentional and accidental events in speakers of English and Japanese. Linguistic
analyses of Japanese have suggested that the frequency of non-agentive expressions
may be higher in Japanese than in English, making Japanese an interesting extension
of the findings with English and Spanish speakers.
In Japanese, two different verbs are often used for the transitive and
intransitive description of the same action. These two verbs often share the same stem.
One example is waru/wareru 割る/ 割れる (to break). An agentive use would be
卵を割った (Tamago-wo watta / (I) broke the egg). A non-agentive use would be
卵が割れた (Tamago-ga wareta / (The) egg broke). Other verbs in Japanese have the
same form for both transitive and intransitive uses, and the presence of the particle
“ga” attached to the affected object marks the non-agentive expression. One example
is hiraku 開く (to open). An agentive use would be 彼がドアを開いた (Kare-ga
DOA-wo hiraita / He opened the door) and a non-agentive use would be
ドアが開いた (DOA-ga hiraita/ (The) door opened).
Verbs are thought to be especially salient in Japanese and typical verb forms in
Japanese may differ from typical verb forms in English. For example, Teramura
(1976) noted that even when an event involves someone who could be described as a
causal agent (e.g., “He dropped the pen”), it is often more natural in Japanese to
18
describe such events using non-agentive expressions like “PEN-ga ochiteshimatta”
(“The pen dropped, unfortunately”), or even sentences that include only a verb like
“ochiteshimatta” (“dropped, unfortunately”). In a recent study, Fukuda and Choi
(2006) reported that the intransitive usage bias in Japanese appears to be strong
enough to influence early language learning such that Japanese speaking children start
producing intransitive verbs before transitive verbs, which contrasts with patterns seen
in English speaking children. Verbs may be especially salient in Japanese because
nouns and pronouns in Japanese are often optional and inferred from context (Fernald
& Morikawa, 1993). The form of the verb may therefore be a potent cue for how to
frame an event.
In Study 3, we aimed to find out how English and Japanese speakers talk about
and remember intentional and accidental events. We used a similar paradigm as with
English and Spanish speakers, with new video stimuli.
Videos in this study featured Japanese actors, in contrast to the Caucasian
actors featured in the videos used with Spanish speakers. In cross-cultural research
about attention to human agents, one necessarily confronts potential challenges in
interpreting memory patterns due to cross-race recognition effects (e.g., Malpass &
Kravitz, 1969) – in many cases, either the exact stimuli or the “same race” status is
held constant across the two groups, but not both at the same time. In the current
paradigm, such concerns may be minimal because all participants attempt to
remember agents for two kinds of events and the relationship between these kinds of
events within each community is of interest. In this study, we sought to extend our
19
understanding of English speakers’ description and memory patterns by testing them
using different video stimuli than used in the previous studies and also to examine
Japanese description and eye-witness memory patterns using stimuli most likely to
invoke natural processing.
We compared English and Japanese speakers’ descriptions and memory for
intentional and accidental events. Participants first completed a simple memory task
that was not predicted to vary across language groups. We then showed English and
Japanese speakers videos of intentional and accidental events. After viewing the
events, participants were tested on their memory for the agents of these events. After
the memory test, participants viewed the videos again and provided a verbal
description for each video.
Here, we empirically confirm a difference between Japanese and English
speakers’ descriptions of the same events, and test whether these differences in
language have consequences for people’s eye-witness memory. We hypothesized that
more agentive language in one’s linguistic environment would lead to better memory
for the agents of events.
Study 3
Participants. 62 English speakers (Stanford University; Mean age = 19.29
years) and 70 Japanese speakers (Keio University, Jochi University, Tokyo Kogyo
University, Surugadai Law School, all in Tokyo, Japan; Mean age = 20.94 years)
received course credit or were paid for their participation. Participants were selected to
be under 25 years old to ensure a homogenous sample for memory performance.
20
Participants were selected to be functionally monolingual. English speakers
reported learning only English before age 12 and did not currently use another
language. Exposure to English in Japan is almost inevitable, including in school before
age 12. Thus, we selected Japanese speakers based on their self-rated proficiency
speaking and understanding English. Using a 5-point scale in which 5 indicated
“native-like”, Japanese speakers who rated themselves as 3 or lower for an English
proficiency measure were included.
General design. Participants read instructions in either English or Japanese.
English and Japanese texts were developed simultaneously and verified by a native
Japanese-English bilingual. All participants completed three tasks, in the following
order: (a) Object-orientation memory, (b) Agent memory, (c) Event descriptions. The
first two tasks were non-linguistic – participants never described any of the images or
events during these two tasks, nor were they provided with any linguistic descriptions.
Only the final description task was linguistic. The first task was designed to be a
measure of memory performance that was not predicted to vary across language
communities. The second task was designed to test for differences in non-linguistic
eye-witness memory (memory for the agents of events) between English and Japanese
speakers. The third task was designed to document usage-based evidence of a
difference in linguistic descriptions in English and Japanese. Importantly, participants
did not describe any events until after the agent memory task.
21
Object-orientation memory
During encoding, participants saw pictures of 15 objects presented on a
computer screen one at a time for two seconds each (images courtesy of Michael J.
Tarr, Brown University, http://www.tarrlab.org/). Each object appeared in one of three
possible orientations, counterbalanced across participants. Participants were instructed
to pay attention to the images and were told that their memory would be tested.
After the encoding phase, participants were given a brief distracter task
(counting the number of white squares on a grid of black and white squares), followed
by the memory test. For the memory test, participants were shown the three possible
orientations of each object and asked to indicate which one they had seen previously.
Agent memory
Video materials. Intentional and accidental versions of 16 unique events
were videotaped (Table 2), one set used for the encoding phase and another set used
for the test phase. Videos were similar to those used with English and Spanish
speakers, except with Japanese actors. For the encoding phase, videos of eight events
(both intentional and accidental versions) featured one actor in a white shirt and videos
of another eight events (both intentional and accidental versions) featured a different
actor in a black shirt. For the test phase, videos featured a third actor in both the
intentional and accidental versions of all 16 events. The same silent videos served as
stimuli for both English and Japanese speakers.
Encoding. During the encoding phase, participants viewed 16 videos, each
showing a different event (half featured the actor in white and half the actor in black,
22
half were intentional actions and half accidental). Whether an event was presented in
its intentional or accidental version was counterbalanced across participants. Videos
were presented in one of two pseudo-random orders that ensured that no more than
three videos of the same agent or the same intention appeared in a row.
Participants were instructed to pay attention to the videos and were told that
their memory would be tested, but were not given any further clues. After viewing all
16 videos (once each, with a 1200 ms pause between videos), participants were
instructed to count to 10 as a brief distracter task.
Test. Test trials consisted of a probe video followed by still photos of the two
agents from the encoding phase. In the probe videos, a third actor appeared as the
agent of the same events participants saw during encoding. For example, if a
participant had seen the “accidental balloon popping” event during encoding, they
would see this same event acted by the new agent in the test phase. After each probe
video, participants were asked, “Who did it the first time?”
(「最初に誰がそれをしましたか?」)9 and responded by pressing a key associated
with the side of the screen of either the white-shirt man or the black-shirt man.
Participants were tested only on the events they had seen during encoding, presented
in a different pseudo-random order from the encoding phase, and received no
feedback.
Event descriptions
After participants completed the agent memory task, they were again shown
the same 16 videos they had seen during encoding and this time were asked to provide
23
a linguistic description for each video. In each description trial, participants viewed a
video and then answered the question “What happened?”
(「何がおこりましたか?」). English speakers typed their responses and Japanese
speakers either typed or wrote their responses at their own pace and received no
feedback.
Results
Eleven participants were excluded from analyses for one of the following
reasons: (a) chance performance on the object-orientation memory task (3 Japanese),
(b) a z-score greater than |2| (relative to language group) on the Memory Difference
Score (Intentional Memory minus Accidental Memory) (4 English, 4 Japanese). The
Memory Difference Score was the analysis of interest in this study, and we wanted to
be sure that outliers did not drive any observed cross-linguistic differences. Results
from the description task are presented first; note that participants produced
descriptions only after having already completed both of the non-linguistic memory
tasks.
Results: Event Descriptions in English and Japanese
Descriptions were coded as agentive if the sentence mentioned the causal agent
in a transitive sentence that described the change-of-state event. A canonical agentive
description would be “He popped the balloon”. Descriptions were coded as non-
agentive if the change-of-state event was described intransitively. A canonical non-
agentive description would be “The balloon popped”. In Japanese, non-agentive
24
descriptions were characterized by an intransitive verb as well as the particle “ga” with
the affected object (e.g., 「風船が割れてびっくりした。」, Balloon-ga popped-
intransitive was surprised). Some non-agentive descriptions in each language took the
form “Someone was doing X and then Y happened”, in which the agent was
linguistically separated from a change-of-state event that was described intransitively.
Across all participants, 2.89% of the descriptions did not describe the event
and were excluded from analyses. All descriptions were coded by two independent
raters, with high point-to-point reliability (95.91% English, 93.25% Japanese).
Disagreements were resolved upon discussion.
Results are shown in Figures 3 and 4a. Intentional events were described
equally agentively by both English and Japanese speakers (English M = 96.70, SE =
.74; Japanese M = 95.70, SE = 1.48, t(119) = .59, n.s.). Accidental events, on the other
hand, were more often described agentively by English speakers than by Japanese
speakers (English M = 70.60, SE = 2.55; Japanese M = 57.98, SE = 3.14, t(119) =
3.09, p = .003).
To compare how strongly speakers distinguished between intentional and
accidental events in their descriptions, we computed a difference score for each
participant as the proportion of intentional events described using agentive language
minus the proportion of accidental events described using agentive language. This
distinction was more pronounced for Japanese speakers (M = 37.71, SE = 2.94) than
for English speakers (M = 26.10, SE = 2.58), t(119) = 2.95, p = .004. This cross-
linguistic difference was also consistent across events, t(15) = 3.49, p = .003.
25
Results: Eye-witness memory in English and Japanese
Results are shown in Figure 4b.10 Intentional agents were remembered well by
both English (M = 71.55, SE = 2.57) and Japanese (M = 70.44, SE = 2.32) speakers,
t(119) = .32, n.s. However, as predicted, accidental agents were better remembered by
English speakers (M = 73.06, SE = 2.42) than by Japanese speakers (M = 66.07, SE =
2.67), t(119) = 1.93, p = .028, d = .35. As predicted by patterns in language, the
distinction between memory for individuals involved in intentional and accidental
events was more pronounced for Japanese speakers (M = 4.37, SE = 2.06) than for
English speakers (M = -1.51, SE = 1.97), t(119) = 2.05, p =.02, d = .38. This
difference remains robust when participants’ memory performance on the object
orientation task is added as a covariate in the analysis (p = .023) ensuring that this
observed cross-linguistic difference in agent memory is not simply due to more
general differences in memory performance. Importantly, the agent memory task itself
serves as the crucial control that guards against concerns about overall memory
differences between the two groups: English and Japanese speakers remembered
intentional agents equally well. They only differed in their memory for the agents of
accidental events.
Discussion
In this study, English speakers and Japanese speakers used agentive
expressions to talk about intentional events and remembered intentional agents equally
well. When it came to accidents, however, cross-linguistic differences in both
language and eye-witness memory were observed. English speakers described
26
accidents using more agentive language than Japanese speakers did and also
remembered agents of accidents better than Japanese speakers did. Importantly, these
memory patterns were observed in a task that participants completed before they had
used any language to describe the events.
Cross-linguistic differences in memory patterns were localized to a particular
kind of event (accidents). Given other findings about cross-cultural differences in
attention (e.g., Masuda & Nisbett, 2001), other patterns of results might have been
predicted. For example, global differences in Japanese and English speakers’ attention
– such as relative attention to context versus focal objects – might have led to overall
lower memory for causal agents in Japanese speakers compared to English speakers.
Cross-linguistic differences in noun and pronoun use (with lower frequency in
Japanese compared to English) might also have resulted in overall lower agent
memory in Japanese speakers. Instead, we found evidence for memory differences
only for those events in which patterns of action descriptions also differed. Thus, this
study refines our understanding of cross-cultural differences in attention to events and
suggests that patterns in verb use may be one mechanism that drives these differences.
This study with English and Japanese speakers (a) replicates patterns of
memory for agents of intentional and accidental events in English speakers with a new
stimulus set, (b) extends evidence for cross-linguistic variation in non-agentive
language use, adding to our understanding of usage biases in causal event descriptions
across the worlds’ languages, and (c) replicates previous findings that accident
27
descriptions covary with memory for accidents, extending evidence for this pattern to
another linguistic community.
2.3. Eye-witness memory in primed English speakers
The cross-linguistic findings leave open an important question: can language
per se shape people’s attention to agents? There are many differences between the
cultural experiences of English, Spanish and Japanese speakers, so it is possible that
other extra-linguistic cultural differences created the memory differences we observed
in the previous studies. To more directly test whether patterns in language per se can
cause people to pay more or less attention to agents, we conducted a fourth study.
Study 4
In Study 4, we primed English speakers with either agentive or non-agentive
language in a separate task before they completed the same agent memory task as used
in the previous studies.
Presumably, any “chronic” influence of language on attention and memory
results from a combination of many shorter-term episodes of linguistic descriptions of
events. It is therefore useful to examine the role of a relatively “temporary”
manipulation of the language environment. This study is one step toward establishing
a more direct link between agentivity in language and eye-witness memory. If patterns
in one’s linguistic environment can bias eye-witness memory, then directly
manipulating the frequency of agentive expressions in the local linguistic environment
should modulate English speakers’ memory for agents.
28
Method
Participants. 65 English speakers (33 agentive prime, 32 non-agentive prime)
participated. Inclusion criteria were the same as in the previous studies.
General design. Participants completed three tasks: first the same object-
orientation memory task as in Studies 2 and 3, then the linguistic priming task
described below, and finally the same agent memory task as in Study 2 (Caucasian
actors). All of the procedures and materials were identical to the cross-linguistic eye-
witness memory paradigm except for the inserted linguistic priming task.
Linguistic priming task. Participants in each condition listened to 24 sentences,
either all agentive (e.g., He burned the toast) or all non-agentive (e.g., The toast
burned) (Table 3). As participants heard each sentence, two related images (e.g., a
piece of bread and a burned piece of bread) were presented on a computer screen. The
participants’ task was to click on the image described by the sentence (e.g., the burned
bread). The same response was correct regardless of priming condition. Stimuli were
presented in random order. All sentences were recorded by a female native English
speaker and were played to participants via computer speakers. Importantly, no verbs
that could describe actions in the agent memory task were used in the priming task.
After completing these trials, participants were given a surprise recall test and
asked to write down as many sentences as they could remember.11 They then
continued on to complete the agent memory task used in the previous studies. The
priming manipulation was designed to change the overall frequency of agentive versus
non-agentive expressions in the participants’ linguistic environment and to produce a
29
“main effect” of linguistic environment: exposure to agentive language should lead
people to have better overall memory for agents than exposure to non-agentive
language.
Results
Five participants were excluded from analyses for one of the following
reasons: (a) chance performance on the object-orientation memory task (1 agentive, 2
non-agentive), (b) ungrammatical or infelicitous descriptions in the prime recall or
description task (1 agentive, 1 non-agentive).
As predicted, participants primed with agentive language showed better
memory for agents (M = 78.02, SE = 2.93) than those primed with non-agentive
language (M = 70.69, SE = 3.07), t(58) = 1.73, p < .05 (one-tailed) (Figure 5). This
was not simply a function of the non-agentive group being worse at memory overall.
Participants in the two conditions remembered object orientation equally well
(agentive M = 71.40, SE = 2.38; non-agentive M = 75.63 , SE =2.50), t(58) = 1.23, n.s.
There was no main effect of event type or interaction of prime by event type. The
effect of prime condition on memory was also reliable across items, t(15) = 3.04, p <
.01.
Discussion
It appears that the local linguistic environment can influence how well people
remember who did what: English speakers who were exposed to agentive language in
the priming task remembered agents better than those who were exposed to non-
agentive language. This was true even though the particular verbs used in the priming
30
task were unrelated to the actions participants observed in agent memory task. This
result suggests a causal role for the linguistic environment in guiding eye-witness
memory.
Language and Eye-witness Memory: General Discussion
Our results demonstrate that our eye-witness memories for even such
momentary events as popping a balloon or breaking a pencil may be susceptible to
influence from linguistic patterns that differ across natural languages. These results
extend previous work on the role of language in eye-witness memory (e.g., Gentner &
Loftus, 1979; Loftus & Palmer, 1974) to the cross-linguistic domain. In this case,
typical ways of talking in one’s linguistic community predict patterns in people’s eye-
witness memory for who did what. We find cross-linguistic differences in memory
even though participants were not asked to describe the events at any time during the
memory task. It appears that an explicit requirement for linguistic description is not
necessary to observe cross-linguistic differences in the case of causal events (see also
Wolff et al., 2009).
The results from these studies suggest that language may influence eye-witness
memory in both “temporary” and “chronic” ways. Study 4 shows that temporary
patterns in our local linguistic context can shift how we attend to and remember
events. Studies 2 and 3 show that speaking a language can also have some “chronic”
cognitive consequences; typical ways of talking in one’s linguistic community appear
to train habits in attention and memory. As a result, even when no language is used in
the context of observing events, speakers of different languages remember the same
31
events differently. Thus, both long-term exposure to distributions of agentive and non-
agentive event descriptions and short-term perturbations in the distributional statistics
of event descriptions appear to influence how well people remember individuals
involved in events.
These “temporary” and “chronic” effects of language are not mutually
exclusive. Indeed, the sum of many temporary episodes may add up to chronic biases
– that is, many consistent experiences may train attention in consistent ways (as has
been demonstrated in other domains such as perceptual and category learning; see e.g.,
Kruschke, 2010; Smith, Jones, Landau, Gershkoff-Stowe, & Samuelson, 2002;
Yoshida & Smith, 2005). Thus, there is likely a continuity between short- and long-
term effects of language on eye-witness memory.
An interesting direction for future research would be to examine how
temporary and chronic effects of language may interact to bias eye-witness memory.
For example, how long lasting would the “temporary” effects of language on memory
be in adult English and Spanish speakers? How many temporary episodes would be
required to reliably shift chronic attention patterns in adults? For children who are
learning linguistic conventions, at what point might they show adult-like attention
biases? And, how susceptible would they be to temporary language manipulations at
different points in the developmental trajectory? Future research using multiple lab
sessions and/or research using longitudinal designs with children may provide
additional insight into continuities and interactions among temporary and chronic
effects of language on eye-witness memory.
32
Another way to explore the roles of short- and long-term linguistic cues in this
case would be to examine the memory patterns of bilinguals. Would factors like age of
exposure to the two languages or the relative amount that an individual uses each
language impact the results? Would memory be affected by short-term manipulations
like whether a bilingual was tested in a Spanish (or Japanese) versus an English
linguistic context? It is possible that patterns in language create chronic biases in
attention, and also that particular linguistic environments can amplify or cue a suite of
learned attentional habits. The rich linguistic contexts of bilingual experience may
provide a unique window into how language guides attention and memory over
multiple timescales.
What are the mechanisms by which language may modulate memory in the
case of causal events? One possibility is that (even in situations when they are not
required to describe) people spontaneously generate sub-vocal descriptions of events,
and these descriptions encode specific details that happen to be useful in later memory
tasks. For example, if one (subvocally) generated and stored a description like “The
guy in the yellow shirt popped the balloon,” this stored description could then be
useful in later reporting whether it was the guy in yellow or in blue that popped the
balloon. That is, such specific descriptions could serve as a secondary code, explicitly
encoding information that would turn out to be relevant in a memory test. The way
that English and Spanish speakers described events in Study 1 suggests that this
mechanism is unlikely to be the source of the memory difference we observed
between the two language groups. Very few descriptions (less than 3% in either
33
language community) included the kind of identifying information that could help
distinguish between the two actors. People most often referred to the actors as simply
“a man” or “a guy” (“un hombre”, “un joven”, in Spanish), descriptions that are not
specific enough to be diagnostic in distinguishing the two actors.
Another possibility is that rather than serving as a specific secondary code
(e.g., by explicitly encoding test-relevant information like shirt color), language may
modulate memory by directing people’s visual attention as they witness events. For
example, if one would most often hear (and produce) agentive descriptions like “He
popped the balloon”, this may orient visual attention to the agent and make one more
likely to represent who that “he” is. There are (at least) two interesting possibilities for
how language could shape attention in this way: an in-the-moment effect or a more
general effect. If people automatically generate internal linguistic descriptions of
events (even in situations when they are not required to speak), it could be these
internal descriptions that then bias people’s attention in the moment. Alternatively,
exposure to more or less agentive language in one’s linguistic environment may create
general attentional biases that do not require access to linguistic processes or linguistic
descriptions in the course of the task. Speakers of languages that rely heavily on
agentive language may become more likely to pay attention to and encode details
about agents whether or not they generate any kind of internal linguistic description in
the moment.
One direction for future studies would be to disrupt people’s access to fluent
linguistic processing during encoding (e.g., with verbal interference). Whether or not
34
verbal interference affects memory outcomes may help us distinguish whether patterns
in language shape attention in the moment (because linguistic descriptions are
automatically generated and meddle in cognition even when people do not plan to
speak) or by training general attentional biases. Both of these mechanisms would serve
as interesting examples of how patterns in a linguistic environment can importantly
shape what people encode and remember about the events they witness.
Conclusion
These findings suggest that our eye-witness memories for events may be influenced by
the languages we speak. Speakers of different languages remember different things
about the same events. Whether or not we are likely to remember who did what
appears to pattern with how such events are normally described in our language
community as well as on the patterns in our local linguistic environment.
35
Chapter 3. Language and attributions of blame, financial liability, and moral goodness.
Can subtle differences in language lead to big differences in attributions of
blame and punishment? In this chapter, we examine the effects of agentive and non-
agentive linguistic frames on important real-world decisions about blame and
punishment. We examine whether linguistic framing matters in situations where
people also have non-linguistic ways to represent events, and find that linguistic
framing strongly affects judgments about causal agents in English-speaking adults and
children.
3.1. Language and attributions of blame and financial liability in adults
When bad things happen, how do we decide who is to blame and how much
they should be punished? Linguistic and contextual framing has been shown to affect
people’s reasoning in a variety of domains (e.g., Lee, Frederick, & Ariely, 2006;
Levin, 1987; Levin & Gaeth, 1988; Loftus, Miller, & Burns, 1978; Loftus & Palmer,
1974; Shiv, Carmon, & Ariely 2005; Tversky & Kahneman, 1973; Tversky &
Kahneman, 1981), including causal attribution (see Pickering & Majid, 2007, for a
recent review). In this series of studies we build on this work by exploring the effects
of linguistic framing in a domain of paramount real-world importance: blame and
punishment.
Linguistic descriptions are of course ubiquitous in legal disputes. People
linguistically frame incidents right from the very moment they occur and later in
police reports, legal statements, court testimony and public discourse. Could the
36
linguistic descriptions of an event influence how much we blame the people involved?
Could language also influence how financially liable we think a person is for any
resulting damage? Could linguistic framing shape construal even for well-known
events (ones for which we already have rich knowledge and established mental
representations) and even when we can witness the event with our own eyes?
The particular linguistic contrast of interest here is between transitive agentive
descriptions and intransitive non-agentive descriptions. A canonical agentive
description (e.g., Timberlake ripped the costume) includes a person as the subject in a
transitive expression describing a change of state (in this case, ripping). A canonical
non-agentive description (e.g., The costume ripped) is intransitive and does not place
the person as the subject for the change of state event.12 Previous work has shown that
people are sensitive to this distinction between agentive and non-agentive frames. For
example, people are more likely to remember the agent of an event when primed with
agentive language than with non-agentive language (e.g., Fausey & Boroditsky, 2010).
The attributional consequences of these linguistic frames, however, are not well
understood.
The linguistic contrast between agentive and non-agentive frames has the
potential to have serious real-world consequences, especially in legal contexts. For
example, in the 197,745 trials held between 1674 and 1913 at London's central
criminal court (Old Bailey Proceedings Online, 2009), cases with the agentive phrase
“broke it” in the court records resulted in a guilty verdict more often than cases with
the non-agentive phrase “it broke” (76% and 70% guilty, respectively), with similar
37
patterns for other consequential actions such as “burned it” versus “it burned” (77%
and 57% guilty, respectively), χ2(1, N = 2748) = 11.04, p < .05. In the most serious of
cases (when the charge was “killing”), the transitive/intransitive contrast as marked by
different verbs also predicted verdicts. Saying “killed” resulted in more guilty verdicts
than saying “died” (65% and 56% guilty, respectively), χ2(1, N = 3814) = 21.34, p <
.05. These examples suggest that agentivity may be part of a suite of linguistic cues
that are influential in legal reasoning.
In a correlational analysis like this, however, it is impossible to determine
whether different linguistic forms actually caused a difference in verdicts. It could be
that agentive descriptions indeed led the court more often to guilty verdicts. But it is
also possible that people were simply more likely to use agentive language in cases
where the defendant was actually more guilty. While the attributional consequences of
transitivity have not been directly explored in the empirical literature, the question has
been debated (and adjudicated!) in court. For example, in a case petitioning to change
the title of a ballot measure (California’s high-profile Proposition 8 in the 2008
election titled “Eliminates right of same-sex couples to marry”), the judge rejected the
petitioners’ claim, ruling that “There is nothing inherently argumentative or
prejudicial about transitive verbs” (Jansson v. Bowen, 2008). Few other questions in
psycholinguistics have risen to a sufficient level of civic importance to be ruled on in
high court.
With the high stakes of guilt, innocence and the legality of constitutional
amendments on the line, it is important to empirically establish whether agentive and
38
non-agentive frames indeed have any attributional consequences. In the next three
studies we examine the effects of agentive and non-agentive linguistic frames on
important real-world decisions about blame and punishment.
Study 5
In this study, participants read about an accidental restaurant fire that resulted
in property damage. They then made judgments about the person involved in the
accident. The survey was one of many unrelated surveys in a packet presented to
participants.
Method
Participants. 236 students at Stanford University (96 male; mean age = 19.22
years) completed one survey in partial fulfillment of a course requirement. 116 read
the agentive version of the story and 120 read the non-agentive version of the story.
Materials. Participants read either the agentive or the non-agentive account
about an individual – Mrs. Smith – involved in a restaurant fire, and then answered
two questions (Table 4). The two accounts contain all of the same content words (all
of the same nouns, verbs and adjectives are used), involve the same individual and
describe the same outcomes. The accounts differ only in the frames used to describe
the accidental events (underlined sections of Table 4): transitive frames are used in the
agentive account and intransitive frames in the non-agentive account.
Results and Discussion
Linguistic framing influenced both people’s judgments of blame and financial
liability. Participants who read the agentive account (M = 4.83, SE = .14) blamed Mrs.
39
Smith more than did participants who read the non-agentive account (M = 4.01, SE =
.15), t(234) = 4.04, p < .001, d = .53. Impressively, a subtle difference in language
caused a big difference in dollars: people who got the agentive report ruled that Mrs.
Smith should pay $247, or 36%, more in fines (M = $935.17, SE = $43.48) than
participants who got the non-agentive report (M = $688.75, SE = $43.64), t(234) =
3.99, p < .001, d = .52.
In Study 5, linguistic framing influenced people’s judgments of financial
liability. One explanation for this result could be that Mrs. Smith was punished more
harshly because she was also blamed more harshly. That is, the effect of language on
financial liability might be indirect, such that language influences blame, which then
determines punishment. Could language directly impact judgments of financial
liability? This question is important because of the somewhat flexible sentencing
process that occurs after guilt judgments in legal decision-making. A direct impact of
language on sentencing would be an important applied result. Study 6 was designed to
address this question.
Study 6
In Study 6, participants got an agentive or non-agentive accident description
and also learned of a blame attribution generated by an independent review panel. This
panel attributed either low, middle, or high blame to the person involved in the
accident. After learning how blameworthy other people judged the person to be,
participants determined the person’s financial liability for the property damage. This
paradigm allows us to target the independent role of language on financial liability
40
sentences. People’s decisions about financial liability may be guided by
blameworthiness, language, or both.
Method
Participants. 179 students at Stanford University (59 male; mean age = 19.01
years) completed one survey in partial fulfillment of a course requirement. 91 read the
agentive account of the restaurant fire accident (33 low-blame, 30 mid-blame, 28 high-
blame) and 88 read the non-agentive account (33 low-blame, 28 mid-blame, 27 high-
blame).
Materials. As in Study 5, participants read either the agentive or the non-
agentive narrative and then answered the financial liability question shown in Table 4.
Thus, participants in this study answered only the financial liability question, after
learning that an independent panel judged the person to be either a “one” (low), a
“four” (mid) or a “seven” (high) in terms of blame.
Results
The level of blame assigned by the independent panel influenced participants’
judgments of financial liability (Figure 6). Overall, people judged that Mrs. Smith
should pay more in damages when the independent panel ruled her to be highly to
blame (M = $974.19, SE = $61.97) than when the panel assigned her a middle level of
blame (M = $615.00, SE = $56.27) than when she was ruled to be of low blame (M =
$425.63, SE = $50.89).
Interestingly, language also influenced financial liability judgments. As in
Study 5, a subtle change in language led to a substantial change in financial liability:
41
Mrs. Smith was held responsible for $153, or 26%, more in damages by people who
got the agentive report (M = $730.75, SE = $49.57) than by those who got the non-
agentive report (M = $577.77, SE = $52.35).
A 3 (Blame: Low, Mid, High) by 2 (Language: Agentive, Non-agentive)
factorial ANOVA revealed reliable main effects of assigned blame level (F (2, 173) =
25.23, p < .001, η2 = .22) and of language (F(1, 173) = 5.53, p = .02, η2 = .03).
Assigned blame level and language did not interact, F (2, 173) = 1.40, n.s.
Discussion
Guilt and linguistic framing independently influenced how much someone was
required to pay for accidental property damage. Increasing assigned blame led to
greater financial liability and agentive framing led to greater financial liability than
non-agentive framing. This finding replicates the result from Study 5. Further,
sentencing itself appears to be susceptible to linguistic framing effects.
Results from the first two studies suggest that agentive and non-agentive
language can shape how people attribute blame and financial liability to individuals
involved in accidents. Of course, in these two studies the only information that
reasoners had about the accident was linguistic. Were people inevitably swayed by
language because it was the only thing that guided what they imagined about the
event? Perhaps people who received differently phrased reports imagined substantially
different scenarios of what happened? In many real-life situations, the information we
have about an event is purely linguistic – in court arguments, insurance claims, news
accounts. But in other situations we may also have visual evidence, either as eye-
42
witnesses or on videotape. Would linguistic framing still have an effect even if people
were able to see the event with their own eyes? Further, the restaurant fire described in
Studies 5 and 6 was a novel event, one for which participants had no other previous
information. Would people be so easily influenced by linguistic framing if they were
reasoning about an event that they already knew something about, for which they
already had a rich set of mental representations?
To address these questions, we capitalized on a widely known, much
discussed, well-publicized and video-recorded event: the “wardrobe malfunction” of
Super Bowl 2004 when a performance by Justin Timberlake and Janet Jackson ended
with Janet Jackson’s breast being exposed on national television. Post-experiment
questioning confirmed that this is indeed a well-known event; nearly all of our
participants (96.9%) had heard about it and many had also seen the video (67.9%)
before the experiment. With prior knowledge, and current visual evidence, could
linguistic framing still influence blame and punishment?
Study 7
In Study 7, participants reasoned about the wardrobe malfunction incident
under one of three conditions: (a) they read about the incident, (b) they first read about
the incident and then watched the video, or (c) they first watched the video and then
read about it. In each condition, people read either an agentive or non-agentive
account of the incident.
43
Method
Participants. 589 participants (188 male; mean age = 31.17 years) were paid
for completing one survey online. Participants were recruited from the pool of English
speakers who use Amazon’s Mechanical Turk
(https://www.mturk.com/mturk/welcome). 306 read the agentive account of the event
(116 read-only; 88 read-then-watch; 102 watch-then-read) and 283 read the non-
agentive account of the event (93 read-only; 106 read-then-watch; 84 watch-then-
read).
Materials and Design. Participants read either the agentive or non-agentive
account of the “wardrobe malfunction” incident (Table 5). In two conditions
participants viewed a video of the final six seconds of the performance, which
included the infamous malfunction (http://www.youtube.com/watch?v=O6j-OkvydPI).
After reading about the incident (and in two of the conditions also watching it
on video), participants answered the questions shown in Table 5. The order of the
three response options was randomized and the particular order presented to each
participant was the same for the blame and financial liability judgments. Because
Timberlake initiated movement right before the “wardrobe malfunction” and also
because of his prominent apology to Super Bowl viewers (in which he coined the very
phrase “wardrobe malfunction”, Timberlake, 2004), our narratives focused on the
actions of Timberlake. As a result, we expected that any effects of linguistic framing
should be strongest for judging the guilt and financial liability of Timberlake. Also,
44
because the FCC tried to fine CBS for broadcasting the incident, CBS was included
among the possible targets for financial liability.
Results
In brief, linguistic framing affected people’s judgments of blame and financial
liability in all conditions: language mattered whether it was presented before, after, or
without video evidence. The main results of interest are shown in Figure 7.
Conclusions from these data are the same whether all three framing contexts
are considered (as reported below) or whether only the two multimodal contexts are
considered. Conclusions are also supported by nonparametric analyses (see
Appendix).
Effects of language on blame and financial liability
Blame and financial liability attributions were analyzed using a 2 (Language:
Agentive, Non-agentive) by 3 (Task context: Read-only, Read-then-watch, Watch-
then-read) factorial ANOVA for each dependent measure. For clarity of presentation,
we focus on effects of language here (see Appendix for effects of task context).
Language and task context never interacted.
Blame. Linguistic framing influenced people’s blame attributions (Figure 7a).
Overall, people blamed Timberlake more after reading agentive language (M =
38.76%, SE = 1.59%) than after reading non-agentive language (M = 30.49%, SE =
1.43%), F(1, 583) = 17.94, p < .001, η2 = .03. The effect of language was seen across
the three conditions, with no interaction of the effect of language by condition, F(2,
583) = .15, n.s.
45
Language also affected attributions to chance. Overall, people attributed the
outcome to chance more after reading non-agentive language (M = 42.87%, SE =
2.40%) than after reading agentive language (M = 33.92%, SE = 2.26%), F(1, 583) =
8.99, p = .003, η2 = .01. Again this effect of language was seen across the three
conditions, with no interaction of the effect of language by condition, F(2, 583) = .20,
n.s.
Financial liability. The modal response for financial liability was $0 (57.2% of
all data). This is likely because the sentence “Eventually the fine was dismissed in
court” appeared in the liability question. Nevertheless, the linguistic framing of the
event influenced people’s judgments about financial liability. Overall, the proportion
of people who gave any non-zero amount of financial liability to Timberlake depended
on linguistic framing. 46.7% assigned a non-zero fine after reading agentive language,
while only 38.5% did so after reading non-agentive language, χ2(1, N = 589) = 4.05, p
= .044.
The amount of money for which Timberlake was held liable likewise depended
on linguistic framing (Figure 7b). Participants who got the agentive report asked that
Timberlake pay an extra $30,828.69, or 53%, more in fines than those who got the
non-agentive report (Agentive M = $88,818.12, SE = $8,115.75; Non-agentive M =
$57, 989.43, SE = $6,465.34), F(1, 575) = 10.31, p = .001, η2 = .02.13, 14, 15 Again there
was no interaction of the effect of language by condition, F (2, 575) = 1.22, n.s.
Agentive and non-agentive linguistic framing did not affect people’s
attributions of blame or financial liability to Janet Jackson or CBS (see Appendix).
46
In an additional set of analyses, all of the reported contrasts were conducted
with an additional factor: whether or not the participant reported having seen the video
of this incident prior to the experiment. This factor was not a reliable main effect nor
did it interact with effects of linguistic framing in any of the analyses.
Discussion
Linguistic framing influenced how much people punished an individual
involved in an event, even when they witnessed the event with their own eyes, and
even though the event was one our participants already knew about. Agentive
language led to harsher punishment than non-agentive language. Replicating results
from the first two studies, linguistic framing not only influenced attributions of blame
but also financial liability. In the case of the wardrobe malfunction incident, an
agentive report led people to think that Justin Timberlake owed more than $30,000
more (an extra 53%) in fines compared to a non-agentive report. In real-world
contexts, visual evidence of accidents is rarely presented in the absence of linguistic
framing. These results suggest that the form of this framing guides punishment.
Conclusion
In three studies, linguistic framing influenced participants’ judgments about
blame and punishment. Financial liability judgments in particular were strongly
affected by linguistic framing: agentive descriptions led to 30-50% more in requested
financial damages than non-agentive descriptions. Judgments of financial liability
were affected by linguistic frame even when blame was held constant. This finding
suggests that linguistic framing can have an influence not only on verdicts of guilt and
47
innocence, but also on the sentencing process. Impressively, linguistic framing
influenced reasoning even about an event that people knew a lot about, had seen
before, and witnessed (again) right before judging the individual involved.
Previous inquiries into effects of language on attribution have examined the
role of verbs, voice, and word order in guiding how people determine the cause of an
event (e.g., Brown & Fish, 1983; Garvey, Caramazza, & Yates, 1976; Kasof & Lee,
1993; Kassin & Lowe, 1979; Pryor & Kriss, 1977; Schmid & Fiedler, 1988; Semin,
Rubini, & Fiedler, 1995). Here, we provide the first report on the impact of transitivity
on both people’s attributions of blame and also on the real-world outcomes of these
attributions (punishment). These studies extend previous research in several important
ways. First, we probed people’s decisions about a concrete form of punishment –
financial liability, freely estimated in dollars – in addition to more abstract ratings of
blame. Second, we examined effects of linguistic framing in the presence of previous
knowledge as well as with current visual evidence – a condition that is absent from
many previous attribution framing studies but present in many real-world reasoning
contexts. Finally, we considered the transitive/intransitive alternation, a property of
event description that both has important real-world consequences and differs
interestingly across languages.
Previous work has shown that languages differ from one another in their
preference for agentive versus non-agentive frames (e.g., Fausey & Boroditsky, 2010;
Fausey, Long, & Boroditsky, 2009). The present findings raise the possibility that
speakers of different languages may prescribe more or less severe punishment as a
48
function of the frequency of particular grammatical frames in their language. While
there have been many demonstrations showing the power of linguistic frames in
shaping people’s decisions, there has not been much contact between such findings
and the literature investigating cross-linguistic differences in cognition. Establishing
that linguistic framing has psychological consequences in a domain where languages
naturally differ from one another opens the possibility for connecting these two rich
bodies of knowledge.
In particular, as Sher and McKenzie (2006) have pointed out, the linguistic
frames typically provided in framing studies often are not informationally equivalent.
Each linguistic description is situated in a set of pragmatic norms within a language,
and participants may be responding to the pragmatic cues implied by the choice of
frame. The possibility of cross-linguistic comparisons offers an exciting extension to
the framing literature: rather than having frames provided by an experimenter, in the
cross-linguistic case, speakers of different languages may self-generate different
frames for the same events because of the prevalent patterns in their respective
languages (e.g., Maass, Karasawa, Politi, & Suga, 2006). In this way, cross-linguistic
comparisons may allow us to investigate conceptual framing not just as a phenomenon
in the communicative context (where participants may use pragmatic information to
infer what the experimenter must mean by their choice of frame), but also in contexts
where the participant naturally frames the event for themselves.
The linguistic (and cross-linguistic) framing of agentivity is of particular
importance in court proceedings. Filipovic (2007) highlights a case from Northern
49
California, in which a Spanish-speaking suspect’s non-agentive (and appropriate in
Spanish) description of events (“se me cayó”, roughly “to me it happened that she
fell”) was translated into English for the broader court into the agentive (and
appropriate in English) “I dropped her.” Do these two descriptions mean the same
thing? Or does this change in framing have serious attributional consequences? Our
results raise the possibility that speakers of different languages may arrive at rather
different conclusions regarding blame and punishment for the same events.
In three studies we find that agentive descriptions of events invite more blame
and more severe punishment than do non-agentive descriptions. These results
demonstrate that even when people have knowledge and visual information about
events, linguistic framing can significantly shape how they construe and reason about
what happened. In the case of agentive and non-agentive language, subtle differences
in linguistic framing can have important real-world consequences. Deciding how
much to blame an individual, and how much to hold them financially liable, appears to
be broadly susceptible to linguistic framing.
3.2. Language and judgments of moral goodness in children
Any hypothesis about the way that language influences memory and reasoning
is fundamentally developmental. Presumably, everybody starts out with the same basic
resources as an infant, and by virtue of experience in a certain environment they come
to use the linguistic conventions of their particular community. To better understand
how relationships between language and cognition develop, it is important to learn
50
how people integrate linguistic and non-linguistic sources of information at many
stages of development. In the next two studies, we take a step toward this goal by
examining whether linguistic descriptions of causal events influence how four- and
five- year old children judge individuals in these events.
In these studies, we examine how children reason about individuals in
intentional and accidental events. First, we establish that children can indeed
distinguish between these kinds of events. Second, we show that language may
influence children’s reasoning more powerfully when non-linguistic cues are
ambiguous (accidents) compared to when they are more straightforward (intentional
events).
Study 8
In Study 8, we investigated how children judge individuals in intentional and
accidental events. Children watched videos of events and made a simple judgment
about the individual in each event. In this task, children received only non-linguistic
information about the events. The events were never described. Most previous studies
about the development of moral reasoning have presented children with static pictures
accompanied by verbal narratives, and so this study expands our knowledge about
how young children interpret dynamic, naturalistic events that they see.
How and when children come to understand the difference between intentional
and accidental events is an interesting question in itself. Classic theories suggested that
such understanding was not fully developed until late childhood (e.g., ages 8-9, Piaget,
1932) while more recent findings suggest that with natural scenarios even very young
51
children are sensitive to intentionality (e.g., Vaish, Carpenter, & Tomasello, 2009).
Here, we examine whether young children distinguish between intentional and
accidental events that they simply observe. Finding that children do distinguish
between these types of events opens the door to examine whether linguistic
descriptions of each kind of event might guide children’s reasoning in similar or
different ways.
Method
Participants. The participants were four- and five-year-old children. 40
watched intentional events (Mean age = 4;7; range 4;0-5;2) and 40 watched accidental
events (Mean age = 4;9; range 4;0 to 5;5). Data from four additional participants were
excluded due to inattention (N = 1) or experimenter error (N = 3). Participants were
recruited and tested at a university preschool.
Materials. Twelve events, each involving one person and at least one object,
were designed (Table 6). An intentional version and an accidental version of each
event were filmed. In intentional events, an individual purposefully acted on an object.
In accidental events, an individual accidentally made contact with an object, causing
an outcome. For example, in the intentional version of the “water” event, an individual
looked angry at his toy animal, picked up a cup of water, and spilled it on the toy
animal, looking satisfied. In the accidental version of this event, the individual was
playing with his toy animal, turned to look at something behind him, and while turning
his arm knocked a cup of water, causing it to spill on his toy animal. In this case, the
individual looked startled by the outcome.
52
Individuals in all events were male adults. To guard against the possibility that
children’s judgments might be due to qualities of particular actors, each version of
each event was filmed using two different actors and the actor-event pairing was
counterbalanced across participants. Four practice videos were also filmed, featuring
two men who did not appear in any of the experimental videos.
Children made judgments by pointing to a face on a four-point smiley-face
scale. The four faces represented the options “Really Good,” “A Little Good,” “A
Little Bad,” and “Really Bad” (Table 6).
Design. All children saw the same four practice videos. They then saw 12
experimental videos. Children saw a different individual in each video.
The experimental videos were presented in one of two different presentation
sets, designed to counterbalance actor-event pairings. For example, if in Set 1 Actor A
was the individual in the cartoon event and Actor B was the individual in the spilling
event, in Set 2 Actor B would be the individual in the cartoon event and Actor A
would be the individual in the spilling event. Each set was presented in one of two
random orders counterbalanced across participants. Children saw either 12 intentional
videos or 12 accidental videos.
Procedure. On each trial of the practice phase and the main experiment,
children watched a video and answered a question.
The practice phase was designed in order to promote children’s understanding
of the response scale. The scale was presented incrementally, such that children first
53
saw only the “Really Good” and “Really Bad” faces for the first two videos. For the
final two videos, they saw all four faces. Each of the practice videos was designed for
a particular rating on the scale, so that children would have the experience of pointing
to each of the faces before beginning the experimental trials.
After children had shown that they understood the scale, they watched 12
experimental videos. After each video, children were asked, “Is he good or bad?” and
then, “Is he really ___ or a little ___?” If children responded with a specific answer
after the first question, either verbally or by pointing to a specific face on the scale,
they were not asked the follow-up question. If the child was distracted during the
video, or responded, “I don’t know,” the experimenter replayed the video and repeated
the questions.
Thus, children saw 12 events (either all intentional or all accidental) and
judged the individual in each event.
Results
Children judged individuals on a scale of one (“Really Good”) to four (“Really
Bad”). Children judged individuals in intentional events (M = 2.86, SE = .07) more
harshly than individuals in accidental events (M = 2.45, SE = .07), t(78) = 4.23, p <
.001. This contrast was also reliable across events, t(11) = 4.03, p = .002.
Discussion
In this study, four- and five- year-old children judged individuals in intentional
events as “more bad” than individuals in accidental events. These results show that
young children can distinguish between intentional and accidental events represented
54
in naturalistic videos unaccompanied by language. Children appear to understand
clues from individuals’ behavior about whether actions are intentional or accidental,
and they use this information when judging people.
Previous research suggests that linguistic frames for accidents influence how
adults judge individuals involved in accidents. Would linguistic framing also change
how children reason about an individual in an event that they witness? Would hearing
“He spilled the water” prompt harsher judgment of the individual than hearing “The
water spilled” paired with the exact same accidental event?
Previous research suggests that children are sensitive to linguistic framing. For
example, children infer that traits about someone are more stable over time and
contexts when the trait is described as a noun (e.g., “carrot-eater”) than when it is
described as a verb (“eats carrots”) (Gelman & Heyman, 1999). Language can also
influence children’s motivation. For example, how children react to failure depends on
the kind of praise that they had previously received. After making a drawing mistake,
children who had received entity praise like “You are a good drawer” displayed more
helpless reactions, wanting to stop drawing and return to their classrooms, compared
to children who had heard effort praise like “You did a good job drawing” (Cimpian,
Arce, Markman, & Dweck, 2007).
Does language matter when children judge the goodness or badness of other
people? Do children use event descriptions to construct agency and guide their
judgments of individuals in causal events that they witness? Are linguistic descriptions
55
equally influential in guiding judgments about agents of intentional events and
accidental events?
Study 9
In this study, children witnessed causal events that were linguistically
described using either agentive or non-agentive language, and then made judgments
about the individuals in the events. We hypothesized that children would judge
individuals involved in accidents more harshly when described with agentive language
than when described with non-agentive language (as we found with English-speaking
adults in Studies 5-7). The non-linguistic representations of intentional events may be
less ambiguous and more robust than accidents and so may be less susceptible to
linguistic influence (e.g., Muentener & Lakusta, 2009; Rosset, 2008). How do children
integrate linguistic and non-linguistic information when reasoning about people
involved in causal events?
Method
Participants. The participants were four- and five- year old children. 40
watched intentional events (Mean age = 4;7; range 4;0 to 5;4) and 40 watched
accidental events (Mean age = 4;7; range 4;0 to 5;3). Data from six additional
participants were excluded due to inattention (N = 3) or experimenter error (N = 3).
Participants were recruited and tested at a university preschool.
Materials. Materials in this study were the same as in Study 8, except that a
pre-recorded sentence was added to each video and played during a black screen at the
end of each video. Each sentence was spoken by a female native English speaker. For
56
each event, one agentive and one non-agentive linguistic description were recorded,
making a total of 24 sentences (Table 6). Four practice sentences were also recorded
and accompanied the practice videos.
Design and procedure. The design and procedure were the same as Study 8,
except for the creation of four presentation sets in order to counterbalance language-
event pairings. From each set used in Study 8, two new sets were created. Within each
set, six events were described with agentive language and six events were described
with non-agentive language. Whether an event was described using agentive or non-
agentive language was counterbalanced across participants. Videos were presented in
pseudo-random order such that no more than three videos in a row were described
using the same linguistic frame.
Children watched 12 videos (either all intentional or all accidental), each
followed by a pre-recorded description of the event (half described using agentive
language, half using non-agentive language), and then judged the individual in the
event.
Results
Results are shown in Figure 8. Language did not influence how children
judged individuals in intentional events: children rated individuals in intentional
events described by agentive language (M = 2.80, SE = .08) as harshly as individuals
in intentional events described by non-agentive language (M = 2.78, SE = .07), t(39) =
.41, n.s. However, language did influence how children judged individuals in
accidents. Children rated individuals in accidental events described by agentive
57
language (M = 2.88, SE = .07) more harshly than individuals in accidental events
described by non-agentive language (M = 2.63, SE = .08), t(39) = 3.01, p < .01. A 2
(Language: Agentive, Non-agentive) x 2 (Event Kind: Intentional, Accidental)
repeated-measures ANOVA confirmed the interaction between language and event
kind, F(1, 78) = 3.96, p = .05 (by-events: F(1,11) = 4.8, p =.05). By-events analyses
revealed the same pattern of results (intentional: t(11) = .27, n.s.; accidental: t(11) =
2.57, p < .05).
Discussion
Children judged individuals involved in accidents more harshly when they
heard agentive descriptions of the accidents compared to when they heard non-
agentive descriptions of these same events. Children witnessed the accidents with their
own eyes, and the linguistic description that they heard about the event guided how
children construed the agent of the accident. Linguistic framing did not influence how
children judged agents of intentional events. These results suggest that children
selectively use language to guide moral judgments about people they observe.
Conclusion
In this pair of studies, we first demonstrated that children distinguish between
individuals involved in intentional and accidental events when witnessing dynamic,
naturalistic events. We then showed that children’s judgments about individuals in
causal events are influenced by linguistic descriptions when the events are accidents,
but not when they are intentional.
58
These findings expand previous research that has examined how children
integrate information from linguistic and non-linguistic sources. In the case of causal
events, children appear to selectively use language to help them interpret ambiguous
situations like accidents. Accidents may be more ambiguous than intentional events
because the physical behavioral cues (an individual touches an object and physically
causes an outcome) and the social cues (the individual reacts as if surprised by the
outcome) may not support the same inference about whether it was a goal-directed
action. Such ambiguity does not exist for intentional events. In fact, children who
heard non-agentive language paired with intentional events in our study sometimes
rejected the linguistic framing, saying things like “Water spilled ‘cuz he spilled it” or
“Not telling truth. He knocked it off”. This suggests that intentional events have a
clearer interpretation than accidents, making certain kinds of linguistic description
infelicitous. In the case of accidents, linguistic frames may direct how children resolve
any ambiguities and therefore guide how children evaluate the person they observed.
Four- and five-year-olds who watched dynamic videos of causal events
discriminated between intentional and accidental events. They blamed individuals
involved intentional events more harshly than agents of accidental events. When
children also heard linguistic descriptions of these events, the form of the description
influenced how they judged individuals involved in accidents. Like adults, children
blamed individuals involved in accidents more harshly after hearing agentive
descriptions than after hearing non-agentive descriptions. Language is part of young
59
learners’ environments and these studies suggest that language may play a role in the
development of moral reasoning.
Language and Blame: General Discussion
In this chapter, we examined whether agentive and non-agentive linguistic
frames influence important decisions about blame, punishment and moral goodness. In
short, linguistic framing strongly affected how adults and children judged individuals
involved in accidents. Agentive language led people to blame individuals more, hold
them more financially liable, and judge them to be “more bad”. Linguistic framing
mattered even in situations where people also had non-linguistic ways to represent
events, such as witnessing the actions with their own eyes. These results demonstrate
that even when people have rich established knowledge and visual information about
events, linguistic framing can shape event construal, with important real-world
consequences. Subtle differences in linguistic descriptions can change the way people
construe what happened and how they attribute blame and dole out punishment.
In these studies, events were described by neutral, third-party observers. An
interesting question for future research will be to examine whether the consequences
of linguistic framing depend on who does the framing. For example, jurors may
interpret the same sentence differently depending on whether the prosecutor or the
defense lawyer says it. Also, people may interpret language differently depending on
whether an agent describes his own actions or whether somebody else does. In some
cases, saying “I did it” may prompt more positive evaluations than saying “It
60
happened” because the agent takes responsibility instead of being evasive. Who
frames an event may moderate how a description influences listeners’ judgments.
Another interesting question for future research will be to examine the
consequences of agentive and non-agentive linguistic framing on crediting people for
positive outcomes. Would agentive descriptions lead to more positive evaluations of
individuals who intentionally or accidentally cause something positive? Could
agentive event descriptions heighten people’s sense of self-efficacy? Understanding
how language might promote an agentive mindset could suggest useful interventions
for populations struggling with learned helplessness and how to effectively increase
control over one’s life (e.g., Bandura, 1997).
The studies reported in this chapter have shown that linguistic framing
influences how people construe agents of accidents, punishing them more harshly after
agentive event descriptions than after non-agentive event descriptions. Even when
people witness events with their own eyes, subtle differences in language strongly
affect judgments about individuals involved in accidents. Language matters for how
children and adults construe and judge agents.
61
Chapter 4. Conclusions.
Consider this scenario: A woman rushes into a café, collapses into a chair at
the nearest table, and a glass atop the table falls to the floor and breaks. The woman
freezes and looks surprised. What happened? Who or what is to blame?
Evidence presented in this dissertation suggests that the answer depends on
what language you speak, what kind of language you have heard recently (in unrelated
conversation), and what kind of language you hear from another observer (perhaps the
barista’s exclamation). English speakers would be likely to construe this situation as
“She broke the glass” while Spanish and Japanese speakers might say “The glass
broke”. And the next weekend, return visitors who speak English might be especially
likely to identify the woman. Café-goers who had recently gossiped about a friend’s
latest adventures (e.g., “He built a great new table, re-painted his deck and then
moved the table onto the deck…”) would be more likely to pay attention to the woman
compared to people who had just listened to a friend recount his unfortunate start to
the day (e.g., “The alarm went off late and as I rushed out, my shirt caught the edge of
the door and ripped, and I realized later that a button had fallen off, too…”). A
hurried barista shouting “Someone broke a glass!” would invite more punishing looks
from the crowd than “A glass broke!”.
Results from studies of eye-witness memory and blame suggest that language
strongly influences how people construct agency. In the case of memory, language
influenced what people remembered even though they did not describe the events they
witnessed. In the case of blame, linguistic framing influenced punishment even when
62
other rich sources of non-linguistic information were also available. Important
decisions like remembering who did what, and holding someone financially liable, are
guided by linguistic cues to agency.
People talk about the events of their lives. How we communicate “what
happened” shapes whether we construe someone as being the agent of an event,
whether we attend to and remember who was involved, and how much we blame and
punish those involved. Like visual context and social norms, language appears to be a
pervasive and important cue for how to construct agency.
63
References
Bandura, A. (1997). Self-efficacy: The exercise of control. New York: Freeman.
Barrett, L.F., Lindquist, K., & Gendron, M. (2007). Language as a context for emotion
perception. Trends in Cognitive Sciences, 11, 327-332.
Billman, D. & Krych, M. (1998). Path and manner verbs in action: Effects of
“skipping” or “exiting” on event memory. Proceedings of the 20th annual
meeting of the Cognitive Science Society. Mahwah, NJ: Lawrence Erlbaum
Associates.
Boroditsky, L., Schmidt, L., & Phillips, W. (2003). Sex, syntax, and semantics. In D.
Gentner & S. Goldin-Meadow (Eds.), Language in Mind: Advances in the
study of Language and Cognition.
Brown, R., & Fish, D. (1983). The psychological causality implicit in language.
Cognition, 14, 237-273.
Chiu, C., Morris, M.W., Hong, Y., & Menon, T. (2000). Motivated cultural cognition:
The impact of implicit cultural theories on dispositional attribution varies as a
function of need for closure. Journal of Personality and Social Psychology, 78,
247-259.
Choi, I. Dalal, R., Kim-Prieto, C., & Park, H. (2003). Culture and judgment of causal
relevance. Journal of Personality and Social Psychology, 84, 46-59.
Choi, I., Nisbett, R. E., & Norenzayan, A. (1999). Causal attribution across cultures:
Variation and universality. Psychological Bulletin, 125, 47-63.
64
Cimpian, A., Arce, H.C., Markman, E.M, & Dweck, C.S. (2007). Subtle linguistic
cues affect children’s motivation. Psychological Science, 18, 314-316.
Dilkina, K., McClelland, J. L., & Boroditsky, L. (2007). How language affects thought
in a connectionist model. Proceeding of the 29th annual meeting of the
Cognitive Science Society, Nashville, TN.
Dorfman, A. (2004). Footnotes to a double life. In W. Lesser (Ed.), The Genius of
Language (pp. 206-217). New York: Random House.
Fausey, C.M., & Boroditsky, L. (2010). Whodunnit? Cross-linguistic differences in
eyewitness memory. Manuscript submitted for publication.
Fausey, C.M. & Boroditsky, L. (in press). Subtle linguistic cues influence perceived
blame and financial liability. Psychonomic Bulletin & Review.
Fausey, C.M., Long, B.L., & Boroditsky, L. (2009). The role of language in eye-
witness memory: Remembering who did it in English and Japanese. In N.A.
Taatgen & H. van Rijn (Eds.), Proceedings of the 31st Annual Conference of
the Cognitive Science Society, Austin, TX: Cognitive Science Society.
Fernald, A. & Morikawa, H. (1993). Common themes and cultural variations in
Japanese and American mothers’ speech to infants. Child Development, 64,
637-56.
Filipovic, L. (2007). Language as a witness: Insights from cognitive linguistics. The
International Journal of Speech, Language and the Law, 14(2), 245-267.
Finkbeiner, M., Nicol, J., Greth, D., & Nakamura, K. (2002). The role of language in
memory for actions. Journal of Psycholinguistic Research, 31, 447–457.
65
Fukuda, S., & Choi, S. (2006). Early Verbs in Japanese and Korean Children: Toward
a Transitive or Intransitive Bias? Paper presented at the San Diego Linguistics
Circle, SDSU.
Gahl, S., & Garnsey, S. (2004). Knowledge of grammar, knowledge of usage:
Syntactic probabilities affect pronunciation variation. Language, 80, 748– 775.
Garvey, C., Caramazza, A., & Yates, J. (1976). Factors influencing assignments of
pronouns and antecedents. Cognition, 3, 227-243.
Gelman, S.A., & Heyman, G.D. (1999). Carrot-eaters and creature-believers: The
effects of lexicalization on children’s inferences about social categories.
Psychological Science, 10, 489-493.
Gennari, S.P., Sloman, S.A., Malt, B. C., & Fitch, W.T. (2002). Motion events in
language and cognition. Cognition, 83(1), 49-79.
Gentner, D., & Goldin-Meadow, S. (Eds.). (2003). Language in mind: Advances in the
study of language and thought. Cambridge, MA: MIT Press.
Gentner, D., & Loftus, E.F. (1979). Integration of verbal and visual information as
evidenced by distortions in picture memory. American Journal of Psychology,
92(2), 363-375.
Grimshaw, J. (1982). On the lexical representation of romance reflexive clitics. In J.
Bresnan (Ed.), Mental representations of grammatical relations. Cambridge:
MIT Press.
Imai, M., & Gentner, D. (1997). A cross-linguistic study of early word meaning:
universal ontology and linguistic influence. Cognition, 62, 169-200.
66
Iyengar, S. S., & Lepper, M. R. (1999). Rethinking the value of choice: A cultural
perspective on intrinsic motivation. Journal of Personality and Social
Psychology,76, 349-366.
Jansson v. Bowen, No. 34-2008-00017351 (Sacramento Superior Court August 7,
2008).
Kasof, L., & Lee, J.Y. (1993). Implicit causality as implicit salience. Journal of
Personality and Social Psychology, 65, 877-891.
Kassin, S.M., & Lowe, C.A. (1979). On the use of single sentence descriptions of
behavior in attribution research. Social Behavior and Personality, 7(1), 1-8.
Kruschke, J. K. (2010). Models of attentional learning. In E. M. Pothos and A. Wills
(Eds.), Models of Learning and Categorization. Cambridge University Press.
Landauer, T.K., & Dumais, S.T. (1997). A solution to Plato’s problem: the latent
semantic analysis theory of acquisition, induction, and representations of
knowledge. Psychological Review, 104(2), 211-240.
Lee, L., Frederick, S., & Ariely, D. (2006). Try it, you’ll like it: The influence of
expectation, consumption, and revelation on preferences for beer.
Psychological Science, 17(12), 1054-1058.
Levin, I.P. (1987). Associative effects of information framing. Bulletin of the
Psychonomic Society, 25, 85-86.
Levin, I.P., & Gaeth, G.J. (1988). How consumers are affected by the framing of
attribute information before and after consuming the product. Journal of
Consumer Research, 15, 374-378.
67
Levinson, S.C., Kita, S., Haun, D. & Rasch, B. (2002). Returning the tables: Language
affects spatial reasoning. Cognition, 84, 155-188.
Loftus, E.F., Miller, D.G., & Burns, H.J. (1978). Semantic integration of verbal
information into a visual memory. Journal of Experimental Psychology:
Human Learning and Memory, 4, 19-31.
Loftus, E.F. & Palmer, J.C. (1974). Reconstruction of automobile destruction: An
example of the interaction between language and memory. Journal of Verbal
Learning and Verbal Behavior, 13, 585-589.
Lucy, J.A. (1992). Grammatical Categories and Cognition: A Case Study of the
Linguistic Relativity Hypothesis. Cambridge: Cambridge University Press.
Lupyan, G. (2008). From Chair To "Chair:" A Representational Shift Account Of
Object Labeling Effects On Memory. Journal of Experimental Psychology:
General, 137(2), 348-369.
Maass, A., Karasawa, M., Politi, F., & Suga, S. (2006). Do verbs and adjectives play
different roles in different cultures? A cross-linguistic analysis of person
representation. Journal of Personality and Social Psychology, 90, 734-750.
Maddux, W.W., & Yuki, M. (2006). The “ripple effect”: Cultural differences in
perceptions of the consequences of events. Personality and Social Psychology
Bulletin, 32(5), 669-683.
Maldonado, R. (1992). Middle voice: The Case of Spanish se. Dissertation Abstracts
International, 53(07), 2351A. (UMI No. 9235939)
68
Malpass, R.S., & Kravitz, J. (1969). Recognition for faces of own and other race.
Journal of Personality and Social Psychology, 13, 330-334.
Martinez, I. (2000). The effects of language on children’s understanding of agency and
causation. Dissertation Abstracts International, 61(10), 3976A. (UMI No.
9990936)
Markus, H., & Kitayama, S. (1991). Culture and the self: Implications for cognition,
emotion, and motivation. Psychological Review, 98, 224-253.
Markus, H. R., & Kitayama, S. (2004). Models of agency: Sociocultural diversity in
the construction of action. In V. Murphy-Berman & J. Berman (Eds.) The 49th
Annual Nebraska symposium on motivation: Cross-cultural differences in
perspectives on self (pp. 1-57). Lincoln: University of Nebraska Press.
Masuda, T. & Nisbett, R. E. (2001). Attending holistically vs. analytically: Comparing
the context sensitivity of Japanese and Americans. Journal of Personality and
Social Psychology, 81, 922-934.
Muentener, P., & Lakusta, L. (2009, April). The language of causal events: Exploring
the link between representations of intentional action and causal events. Paper
presented at the 2009 biennial meeting of the Society for Research in Child
Development, Denver, CO.
Oh, K. (2003). Language, cognition, and development: Motion events in English and
Korean. Dissertation Abstracts International, 64(09), 4655B. (UMI No.
3105326)
69
Old Bailey Proceedings Online (2009). Retrieved November 3, 2009, from
www.oldbaileyonline.org.
Papafragou, A., Hulbert, J., & Trueswell, J. (2008). Does language guide event
perception? Evidence from eye movements. Cognition, 108, 155-184.
Papafragou, A., Massey, C., & Gleitman, L. (2002). Shake, rattle, ‘n’ roll: The
representation of motion in language and cognition. Cognition, 84(2), 189-219.
Piaget, J. (1932). The moral judgment of the child. New York: Free Press of Glencoe.
Pickering, M. J., & Majid, A. (2007). What are implicit causality and
consequentiality? Language and Cognitive Processes, 22(5), 780-788.
Pryor, J.B., & Kriss, M. (1977). The cognitive dynamics of salience in the attribution
process. Journal of Personality and Social Psychology, 35(1), 49-55.
Reali, F., Spivey, M., Tyler, M., & Terranova, J. (2006). Inefficient conjunction search
made efficient by concurrent spoken delivery of target identity. Perception and
Psychophysics, 68, 959-974.
Richardson, D.C., & Matlock, T. (2007). The integration of figurative language and
static depictions: An eye movement study of fictive motion. Cognition, 102,
129-138.
Roberson, D., & Hanley, J.R. (2007). Color categories vary with language after all.
Current Biology, 17, 605-606.
Rosset, E. (2008). It’s no accident: Our bias for intentional explanations. Cognition,
108, 771-780.
70
Saffran, J. R., Aslin, R. N., & Newport, E. L. (1996). Statistical learning by 8-month-
old infants. Science, 274, 1926–1928.
Schmid, J., & Fiedler, K. (1998). The backbone of closing speeches: The impact of
prosecution versus defense language on judicial attributions. Journal of
Applied Social Psychology, 28(13), 1140-1172.
Scholl, B. J., & Nakayama, K. (2002). Causal capture: Contextual effects on the
perception of collision events. Psychological Science, 13(6), 493 - 498.
Semin, G.R., Rubini, M., & Fiedler, K. (1995). The answer is in the question: The
effect of verb causality upon locus of explanation. Personality and Social
Psychology Bulletin, 21, 834-842.
Sher, S., & McKenzie, C.R.M. (2006). Information leakage from logically equivalent
frames. Cognition, 101, 467-494.
Shiv, B., Carmon, Z., & Ariely, D. (2005). Placebo Effects of Marketing Actions:
Consumers May Get What They Pay For. Journal of Marketing Research,
42(4), 383-393.
Slobin, D.I. (1996). From “thought and language” to “thinking for speaking”. In J.J.
Gumperz & S.C. Levinson (Eds.), Rethinking linguistic relativity. Cambridge:
Cambridge University Press.
Slobin, D.I. (2003). Language and thought online: Cognitive consequences of
linguistic relativity. In D. Gentner & S. Goldin-Meadow (Eds.), Language in
mind: Advances in the study of language and thought (pp. 157-191).
Cambridge, MA: MIT Press.
71
Slobin, D.I., & Bocaz, A. (1988). Learning to talk about movement through time and
space: The development of narrative abilities in Spanish and English. Lenguas
Modernas, 15, 5-24.
Smith, L.B., Jones, S.S., Landau, B., Gershkoff-Stowe, L., & Samuelson, L. (2002).
Object name learning provides on-the-job training for attention. Psychological
Science (13), 13-19.
Teramura, H. (1976) Naru’hyoogen to ‘suru’ hyoogen: nichiei ‘tai’ hyoogen no
hikaku. Nihongoto Nihongo Kyooiku Mojihyoogenhen, National Language
Center.
Timberlake, J. (February 1 2004). "Statement From Justin Timberlake," PR Newswire.
Trueswell, J.C., & Papafragou, A. (2010). Perceiving and remembering events cross-
linguistically: Evidence from dual-task paradigms. Journal of Memory and
Language, 63(1), 64-82.
Tversky, A., & Kahneman, D. (1973). Availability: A heuristic for judging frequency
and probability. Cognitive Psychology, 5, 207-232.
Tversky, A., & Kahneman, D. (1981). The framing of decisions and the psychology of
choice. Science, 211, 453-458.
Vaish, A., Carpenter, M., & Tomasello, M. (2009). Young children’s moral
evaluations mediate their prosocial behavior. Poster presented at the 2009
biennial meeting of the Society for Research in Child Development, Denver,
CO.
72
White, P.A. (2003). Effects of wording and stimulus format on the use of contingency
information in causal judgment. Memory & Cognition, 31(2), 231-242.
Winawer, J., Witthoft, N., Frank, M., Wu, L., Wade, A., & Boroditsky, L. (2007). The
Russian Blues reveal effects of language on color discrimination. Proceedings
of the National Academy of Science, 104, 7780-7785.
Wolff, P., Jeon, G., & Li, Y. (2009). Causers in English, Korean and Chinese and the
individuation of events. Language and Cognition, 1(2).
Yoshida, H., & Smith, L.B. (2005). Linguistic cues enhance the learning of perceptual
cues. Psychological Science, 16, 90-95.
73
Footnotes
1 The clitic se serves a variety of functions in Spanish. Among other uses,
Spanish se expressions may be considered reflexive (Se peinó; He combed his hair),
impersonal (Aquí se habla español; Spanish is spoken here), passive (Se vendieron los
coches; The cars were sold), reciprocal (Se abrazaron; They hugged each other), and a
marker of accidentality (La taza se rompió; The cup broke). Some semantically
oriented analyses have suggested that different se usages are connected in a network of
related meanings. On one proposal, se expressions encourage event perceptions in
which the causal initiator is schematic and underspecified, and the results of events are
highlighted (e.g., Maldonado, 1992). Some syntactically oriented analyses have
proposed that se expressions are derived from transitive expressions, “losing” an
argument in the process (e.g., Grimshaw, 1982; see also Maldonado, 1992 for a
thorough discussion of several approaches). For example, the se expression “El florero
se rompió” would be derived from the transitive expression “Jon rompió el florero”,
deleting the argument “Jon”. Though terminology and conclusions about the syntax
and semantics of Spanish se vary across theoretical approaches, the main thrust is that
many uses of Spanish se appear to highlight outcomes more than causes. Here, we
offer a usage-based approach to examining Spanish se and test whether its use as an
“accidentality marker” impacts memory for causal events.
74
2 A total of 152 people completed the study. Data from participants who
provided ungrammatical or infelicitous descriptions (N = 7) or failed to meet the
language background criteria (N = 48) were not analyzed.
3 Spanish verbs may be used intransitively without se, and were sometimes used
this way in our data (e.g., caer (to fall), salir (to leave), reventar (to burst)).
4 Three participants (1 English, 2 Spanish) were excluded from subsequent
analyses because over a third of their descriptions did not describe the event.
5 All conclusions are also supported by non-parametric analyses (Mann-Whitney
U and Wilcoxon Signed Ranks).
6 One of the 16 events – a crumpling scene – was filmed in two versions and due
to a presentation error roughly half of participants in Study 2 saw a crumple video
using a plastic cup and others saw a crumple video using a soda can (cup: N = 46
Spanish, N = 65 English). No reliable differences between these two groups were
observed and so data were pooled. All participants in Study 1 viewed the soda can
stimulus.
7 Results from Study 1 motivate directed predictions and so one-tailed planned
contrasts are reported.
8 Patterns revealed by analyses of memory for agents remain the same when
object-orientation memory performance is included as a covariate.
9 45 participants received this wording; 25 participants received the wording
「第一回目に誰がそれをしましたか?」We used the second phrasing after
discussions with several native Japanese speakers revealed differing opinions about
75
the best translation for the English question. Question wording did not interact with
patterns of memory for intentional versus accidental agents and so data were pooled.
10 As with the analyses of memory in Spanish and English speakers, one-tailed
planned contrasts are reported.
11 Participants recalled about half of the prime sentences that they had heard. The
amount recalled did not differ for those primed with agentive language (M = 41.53, SE
= 1.65) and those primed with non-agentive language (M = 45.83, SE = 2.24), t(58) =
1.56, n.s.
12 Please note that the agentive/non-agentive distinction we draw here is different
from the distinction between active and passive voice (e.g., He ripped the costume
versus The costume was ripped by him). The active/passive distinction has been shown
to shift focus to or away from the agent (e.g., Garvey, Caramazza, & Yates, 1976;
Kassin & Lowe, 1979; White, 2003). Here we focus on transitivity and investigate not
just the attributional consequences of transitivity (blame) but also the concrete real-
world outcomes of these attributions (punishment).
13 Eight participants whose financial liability responses exceeded $550,000 were
excluded from this analysis.
14 These conclusions are the same when analyses consider just those participants
who assigned Timberlake a non-zero fine (N = 244). Among these participants, those
who got the agentive report assigned more fines (M = $193,726.47 , SE = $12,893.53)
than those who got the non-agentive report (M = $153,179.61, SE = $12,430.78 ),
t(242) = 2.22 , p = .028.
76
15 These data show some heteroscedasticity, but our main conclusions remain the
same after appropriate corrections. A t-test which does not assume equal variances
confirms a reliable difference between the financial liability assigned by participants
who got agentive versus non-agentive reports, t(559.36) = 2.97, p = .003. The main
effect of task context (see Appendix) was similarly confirmed by a Welch ANOVA
test, F(2, 371.55) = 3.24, p = .04.
77
Table 1
Causal event stimuli (English and Spanish speakers)
Action
Crumple can (Crumple cup) Knock box Knock cups Close book Rip paper Turn off light Spill water Crack egg Close drawer Pop balloon Open umbrella Open door Drop keys Break pencil Stick sticker Release balloon
Intentional
Crumples can on floor by stepping on it (Picks up cup from table and crumples it) Faces table, knocks box off table Faces cup tower, swipes, knocks down tower Reading book, then turns head, closes book Sits at table, rips page from notebook Using hand, hits switch and turns off light Spills water by an outdoor plant Takes egg from carton, cracks it against bowl Faces table with open drawer, closes with knee Pops balloon using tack Stands with closed umbrella, then opens it By turning doorknob, opens door Drops keys onto table Sits at table, breaks pencil in half Places nametag sticker on shirt Sits among balloons, releases one that is untied
Accidental
Turns to walk and crumples can on floor by stepping on it (Reaches to move cup, grabs too hard and crumples it) While gesturing, knocks box off table, reaches to grab it Faces cup tower, reaches for a cup, knocks down tower Reading book, then turns to look at something, closes book Sits at table, turns page in notebook and rips it By leaning against wall, hits switch and turns off light While watering outdoor plant, spills water As picking up egg from carton, cracks it against bowl Turning away from table with open drawer, closes with knee Reaches to put tack in container, pops balloon during reach Stands with closed umbrella, jumps back as opens it By leaning too hard against door, opens it and stumbles Attempts to put keys on table, but drops them on floor Sits at table, breaks pencil in half while writing Flops arm onto table without looking, then sticker is on arm Sits among balloons, releases one, reaches to grab it
78
Table 2
Causal event stimuli (English and Japanese speakers)
Action
Crumple can Knock box Knock cups Close book Rip paper Turn off light Spill rice Crack egg Close drawer Pop balloon Open umbrella Open door Drop keys Break pencil Stick sticker Release balloon
Intentional
Crumples can on floor by stepping on it Faces table, knocks box off table Faces cup tower, swipes, knocks down tower Reading book, then turns head, closes book Sits at table, rips page from notebook Using hand, hits switch and turns off light Pours rice into a measuring cup Takes egg from carton, cracks it against bowl Sits at desk with open drawer, closes with knee Pops balloon using tack Stands with closed umbrella, then opens it By turning doorknob, opens door Drops keys onto table Sits at table, breaks pencil in half Places nametag sticker on shirt Sits among balloons, releases one that is untied
Accidental
Turns to walk and crumples can on floor by stepping on it While gesturing, knocks box off table, reaches to grab it Faces cup tower, reaches for a cup, knocks down tower Reading book, then turns to look at something, closes book Sits at table, turns page in notebook and rips it By leaning against wall, hits switch and turns off light While pouring rice into a measuring cup, spills rice As picking up egg from carton, cracks it against bowl Turning toward desk with open drawer, closes with knee Reaches to put tack in container, pops balloon during reach Stands with closed umbrella, jumps back as opens it By leaning too hard against door, opens it and stumbles Attempts to put keys on table, but drops them on floor Sits at table, breaks pencil in half while writing Flops arm onto table without looking, then sticker is on arm Sits among balloons, releases one, reaches to grab it
79
Table 3
Prime sentences (English speakers)
Agentive Non-agentive He wore out the shoe. He shrunk the shirt. He ignited the grill. He unfastened the necklace. He crashed the car. He squirted the ketchup. He cooked the chicken. He dried the flowers. He burned the toast. He bent the clip. He started up the computer. He loosened the hinge. He unbuttoned the jeans. He scattered the cards. He shut down the laptop. He splattered the paint. He melted the ice cream. He boiled the water. He unwound the yoyo. He straightened the slinky. He lowered the chair. He crumbled the cookie. He unfolded the lawn chair. He blew out the match.
The shoe wore out. The shirt shrunk. The grill ignited. The necklace unfastened. The car crashed. The ketchup squirted. The chicken cooked. The flowers dried. The toast burned. The clip bent. The computer started up. The hinge loosened. The jeans unbuttoned. The cards scattered. The laptop shut down. The paint splattered. The ice cream melted. The water boiled. The yoyo unwound. The slinky straightened. The chair lowered. The cookie crumbled. The lawn chair unfolded. The match blew out.
80
Table 4
Restaurant fire reports and questions
Reports Agentive Report Non-agentive Report Mrs. Smith and her friends were finishing a lovely dinner at their favorite restaurant. After they settled the bill, they decided to head to a nearby café for coffee and dessert. Mrs. Smith followed her friends and as she stood up, she flopped her napkin on the centerpiece candle. She had ignited the napkin! As Mrs. Smith reached to grab the napkin, she toppled the candle and ignited the whole tablecloth too! As she jumped back, she overturned the table and ignited the carpet, as well. Hearing her desperate cries, the restaurant staff hurried over and heroically managed to put the fire out before anyone got hurt.
Mrs. Smith and her friends were finishing a lovely dinner at their favorite restaurant. After they settled the bill, they decided to head to a nearby café for coffee and dessert. Mrs. Smith followed her friends and as she stood up, her napkin flopped on the centerpiece candle. The napkin had ignited! As Mrs. Smith reached to grab the napkin, the candle toppled and the whole tablecloth ignited too! As she jumped back, the table overturned and the carpet ignited, as well. Hearing her desperate cries, the restaurant staff hurried over and heroically managed to put the fire out before anyone got hurt.
Questions for Study 5 Blame Mrs. Smith is discussing the damage with the restaurant. How much should she be blamed for the fire? (Likert scale from 1 to 7, anchored by “Not at all to blame” and “Completely to blame”.) Financial Liability The restaurant’s insurance policy does not cover minor fires. The restaurant has sought legal action to require Mrs. Smith to pay for the damage. Total costs to the restaurant were $1500. How much should Mrs. Smith be required to pay? Question for Study 6 Financial Liability The restaurant’s insurance policy does not cover minor fires and so the restaurant has sought legal action to require Mrs. Smith to pay for the damage. An independent review panel used their standard blame assessment scale in reviewing this case. On this scale, 0 means “not at all to blame” and 8 means “completely to blame”. The panel gave Mrs. Smith a {1,4,7}. The total costs to the restaurant were $1500. How much should Mrs. Smith be required to pay?
81
Table 5
Wardrobe malfunction reports and questions
Reports Agentive Report Non-agentive Report Justin Timberlake and Janet Jackson performed during the 2004 Superbowl Half-time Show. Toward the end of the song, Timberlake followed Jackson across the stage and stood beside her. As they sang the last line, Timberlake reached across the front of Jackson’s body. In this final dance move, he unfastened a snap and tore part of the bodice! He slid the cover right off Jackson’s chest! This incident made for a lot of controversy.
Justin Timberlake and Janet Jackson performed during the 2004 Superbowl Half-time Show. Toward the end of the song, Timberlake followed Jackson across the stage and stood beside her. As they sang the last line, Timberlake reached across the front of Jackson’s body. In this final dance move, a snap unfastened and part of the bodice tore! The cover slid right off Jackson’s chest! This incident made for a lot of controversy.
Questions Blame In your opinion, was someone to blame or was it just chance? Please allocate the percentage of blame. Be sure your numbers add up to 100%! (Response options: Justin Timberlake, Janet Jackson, Chance) Financial Liability The FCC (Federal Communications Commission) tried to fine CBS $550,000 for this incident. Eventually the fine was dismissed in court. How much do you think each of the parties below should have been fined for this incident? (Response options: Justin Timberlake, Janet Jackson, CBS)
82
Table 6 Causal event stimuli (English children)
Practice videos and descriptions Event Valence Description Hugging Really good They hugged each other. Kicking Really bad They kicked each other. Putting toy away A little good His toy is on the shelf. Not washing hands A little bad His hands are dirty. Experimental videos and linguistic descriptions Event Agentive Description Non-agentive Description Candy He took out the candy. The candy came out. Spices He dropped the spices. The spices fell. Cartoon He turned on the cartoon. The cartoon came on. Paint He splattered the paint. The paint splattered. Radio He turned off the radio. The radio turned off. Hose He sprayed the hose. The hose sprayed. Ball He rolled the ball. The ball rolled. Water He spilled the water. The water spilled. Egg He cracked the egg. The egg cracked. Box He knocked over the box. The box fell. Rip He ripped the paper. The paper ripped. Balloon He popped the balloon. The balloon popped. Response Scale
83
Figure Captions Figure 1. Distributions of causal event descriptions in English and Spanish:
(a) Intentional, (b) Accidental, (c) Difference (Intentional minus Accidental).
Histograms (with proportion of the sample on the y-axis) of the proportion of agentive
language use in each language community are plotted.
Figure 2. Describing and remembering agents in English and Spanish: (a) Causal
event descriptions, with the mean proportion of agentive descriptions plotted on the y-
axis, (b) Causal agent memory, with mean proportion correct plotted on the y-axis.
Error bars are +/- 1 SEM.
Figure 3. Distributions of causal event descriptions in English and Japanese:
(a) Intentional, (b) Accidental, (c) Difference (Intentional minus Accidental).
Histograms (with proportion of the sample on the y-axis) of the proportion of agentive
language use in each language community are plotted.
Figure 4. Describing and remembering agents in English and Japanese: (a) Causal
event descriptions, with the mean proportion of agentive descriptions plotted on the y-
axis, (b) Causal agent memory, with mean proportion correct plotted on the y-axis.
Error bars are +/- 1 SEM.
Figure 5. Remembering agents after agentive or non-agentive linguistic primes. Mean
proportion correct is plotted on the y-axis. Error bars are +/- 1 SEM.
Figure 6. Independent contributions of guilt and linguistic framing to financial
liability sentences. Mean values are plotted on the y-axis, with whiskers representing
+/- 1 SEM.
84
Figure 7. Language changes punishment of an observed individual. (a) Blame
attribution to Timberlake, (b) Financial liability to Timberlake. Mean values are
plotted on the y-axis, with whiskers representing +/- 1 SEM.
Figure 8. Language changes how children judge accidental agents. Mean badness
judgment is plotted on the y-axis. Error bars are +/-1 SEM.
85
!English
Spanish
(a)
!
(b)
!
(c)
!
86
Intentional Accidental Event Type
Intentional Accidental Event Type
*
n.s. * English Spanish
*
*
n.s.
(a) (b)
87
!English
Japanese
(a)
!
(b)
!
(c)
!
English Japanese
88
* *
0.5
0.6
0.7
0.8
0.9
1
!"
Pro
po
rtio
n A
gen
tiv
e
0.55
0.6
0.65
0.7
0.75
0.8
!"P
rop
ort
ion
Co
rrect
Intentional Accidental Event Type
Intentional Accidental Event Type
*
n.s. n.s. *
(a) (b)
English Japanese
89
0.65
0.7
0.75
0.8
!"
Pro
port
ion C
orr
ect
Agentive Nonagentive Prime
*
90
Agentive Non-agentive
Low Mid High Overall Blame
0
300
600
900
1200
1500
Fina
ncia
l Lia
bilit
y (d
olla
rs)
91
0
10
20
30
40
50
60
Bla
me
0
25000
50000
75000
100000
125000
150000
Fina
ncia
l Lia
bilit
y (d
olla
rs)
(a)
Frame Frame Frame Frame Only First Second Overall
(b)
Agentive Non-agentive
Frame Frame Frame Frame Only First Second Overall
92
2
2.2
2.4
2.6
2.8
3
Intentional Accidental
Mea
n ba
dnes
s jud
gmen
t
Event Kind
n.s.
*
Agentive Non-agentive
93
Appendix
Supplementary Materials for Restaurant Fire and Wardrobe Malfunction
Wardrobe Malfunction: Descriptive statistics
Mean blame attribution.
Agentive Non-agentive Timberlake Jackson Chance Timberlake Jackson Chance
Read only 34.04 27.43 38.53 25.56 28.70 45.73 Read-then-watch 45.85 28.61 25.53 35.17 27.20 37.63 Watch-then-read 38.00 26.09 35.91 30.04 23.67 46.30
Mean financial liability (dollars).
Agentive Non-agentive Timberlake Jackson CBS Timberlake Jackson CBS
Read only 70,481.70 69,416.91 102,341.84 46,797.15 66,597.12 58,210.61 Read-then-watch 123,741.38 74,176.14 76,835.23 66,980.95 55,933.33 74,983.33 Watch-then-read 79,427.86 61,320.38 42,585.03 59,020.49 50,680.73 49,048.19
94
Wardrobe Malfunction:
Effects of task context on judgments of blame and financial liability
The factorial analysis used in Study 3 also revealed a main effect of task
context for blaming Timberlake and chance, as well as for Timberlake’s financial
liability. The degree to which people blamed Timberlake varied across task contexts
(F(2, 583) = 8.70, p < .001), as did the degree to which people blamed chance (F(2,
583) = 4.15, p < .016). Participants who “read-then-watched” blamed Timberlake (M
= 40.02%, SE = 1.97%) more than those who “watched-then-read” (M = 34.40%, SE =
1.89%) more than those who only read (M = 30.27%, SE = 1.73%) about the event.
For chance, participants who “read-then-watched” blamed chance (M = 32.14%, SE =
2.74%) less than those who “watched-then-read” (M = 40.60%, SE = 2.97) and those
who only read (M = 41.73%, SE = 2.85) about the event.
Task context influenced judgments about Timberlake’s financial liability, F(2,
575) = 4.42, p = .012. Participants who “read-then-watched” (M = $92,700.52, SE =
$10,525.82) judged Timberlake to be more liable than those who “watched-then-read”
(M = $70,121.20, SE = $9302.59) than those who only read (M = $59,955.23, SE =
$7,386.42).
For all analyses, data in “read-then-watched” differed reliably from the other
two contexts. As described in the main text, this effect of task context did not interact
with effects of linguistic framing. Rather, reading before watching appeared to elevate
overall blame and financial liability to Timberlake compared to the other two contexts.
Task context also appeared to influence people’s judgments about the financial
liability of CBS, χ2(2, N = 589) = 6.28, p = .043. People were least likely to assign any
95
fine to CBS when they had first watched then read about the incident (25.8%)
compared to the other two task contexts (Read-only: 35.9%; Read-then-watch:
36.6%). Watching the event prior to any linguistic construal appeared to dampen the
perceived liability of the network that had broadcast the event.
Task context did not influence blame or financial liability attributions to
Jackson.
Restaurant Fire and Wardrobe Malfunction: Non-parametric analyses
To be sure that inferences based on parametric statistical analyses were not
unduly biased by violations of normality in responses, non-parametric analyses were
conducted. All comparisons are two-tailed. Results concur with parametric analyses.
Restaurant fire (blame and financial liability).
Participants who got the agentive report (Median = 5.00) blamed the individual
involved in the accident more than participants who got the non-agentive report
(Median = 4.00), U (N1 = 116, N2 = 120) = 4914.50, p < .001. Participants who got the
agentive report (Median = $1,000.00) were more punitive toward the individual than
were participants who got the non-agentive report (Median= $750.00), U (N1 = 116,
N2 = 120 )= 4916.00, p < .001.
Restaurant fire (just financial liability).
Participants judged the high-blame individual (Median = $1,000.00) to be
more financially liable than the mid-blame individual (Median = $725.00) than the
low-blame individual (Median = $225.00), χ2(2, N = 179) = 35.39, p < .001.
Participants who got the agentive report (Median = $750.00) were more punitive
96
toward the individual than were participants who got the non-agentive report
(Median= $500.00), U (N1 = 91, N2 = 88) = 3239.00, p = .027.
Wardrobe malfunction.
Participants blamed Timberlake more after the agentive report (Median = 50%)
than after the non-agentive report (Median = 30% ), U (N1 = 306 , N2 = 283) =
36224.50, p < .001. Participants blamed chance more after the non-agentive report
(Median = 40%) than after the agentive report (Median = 12.5%), U (N1 = 306 , N2 =
283) = 38126.00, p = .009. Participants held Timberlake more financially liable after
the agentive report than after the non-agentive report, U (N1 = 301 , N2 = 280) =
37724.50, p = .015. Because a large number of participants assigned no fine to
Timberlake, the median in each framing condition was $0 but the tail ends of each
distribution differed. Among participants who assigned a non-zero fine to Timberlake,
those who got the agentive report (Median = $184,166.50) assigned more fines than
those who got the non-agentive report (Median = $112,500.00 ), U (N1 = 138 , N2 =
106) = 6265.50, p = .054.
The degree to which people blamed Timberlake varied across task contexts
(Kruskal-Wallis Test, χ2 (2, N = 589) = 11.37, p = .003), as did the degree to which
people blamed chance (Kruskal-Wallis Test, χ2 (2, N = 589) = 6.26, p = .044). For
Timberlake, participants who “read-then-watched” blamed him (Median = 50%) more
than those who “watched-then-read” (Median = 45%) more than those who only read
(Median = 30%) about the incident. For chance, participants who “read-then-watched”
blamed chance (Median = 10%) less than those who “watched-then-read” (Median =
25%) and those who only read (Median = 25%) about the incident. Non-parametric
97
analyses of task context effects on Timberlake’s financial liability did not reach
significance, Kruskal-Wallis Test, χ2 (2, N = 581) = 3.11, p = .21.