Chewing on It Can Chew You Up: Effects of Rumination on Triggered Displaced Aggression

Journal of Personality and Social Psychology
Copyright 2005 by the American Psychological Association
2005, Vol. 88, No. 6, 969 –983
0022-3514/05/$12.00
DOI: 10.1037/0022-3514.88.6.969
Chewing on It Can Chew You Up: Effects of Rumination on Triggered
Displaced Aggression
Brad J. Bushman
Angelica M. Bonacci
University of Michigan
Iowa State University
William C. Pedersen
Eduardo A. Vasquez and Norman Miller
California State University, Long Beach
University of Southern California
Ruminating about a provocation increases the likelihood of displaced aggression following a minor
annoyance (trigger). In Study 1, provoked participants who ruminated for 25 min were more aggressive
toward a fumbling confederate than were distracted participants. Provocation-induced negative affect was
positively related to aggression but only among those who ruminated. Study 2 conceptually replicated
Study 1 and also found that the more negatively people reacted to the trigger, the more likely the trigger
was to increase displaced aggression. Study 3 replicated the findings of Studies 1 and 2 by using an 8-hr
rumination period. All 3 studies suggest that ruminating about a provocation increases the likelihood that
a minor triggering annoyance will increase displaced aggression.
Keywords: displaced aggression, rumination, triggering event
Frank arrives 15 min late to work for the 3rd day in a row. His
Displaced Aggression
angry boss gives him a stern reprimand, stating that the company
has no place for lazy and unreliable workers. On the way to the
Aggression is behavior intended to harm another person. Ag-
gression is direct when an individual is provoked and, in retalia-
cafeteria for his morning cup of coffee, Frank keeps thinking about
tion, inflicts harm against the provoker. Aggression is displaced
the reprimand and how much he hates his boss. As he waits in line
when the target is innocent of any wrongdoing but is simply in the
to pay, a coworker cuts in front of him. Frank angrily snaps at her
wrong place at the wrong time. Displaced aggression can occur
and orders her to the end of the line in an outburst laced with
when a person cannot aggress or is constrained from aggressing
obscenities.
against a source of provocation. To aggress directly against the
Jill spent weeks preparing a presentation for her college class.
source of the initial provocation may be unfeasible because the
After her presentation, her professor harshly criticizes her perfor-
source is unavailable (e.g., the provoker has left the situation) or
mance, stating that she was unprepared and disorganized. For the
because the source is an intangible entity (e.g., loud noise, foul
entire day, Jill silently fumes about her professor’s critical re-
odor, hot temperature). Fear of retaliation or punishment from the
marks. When she returns to her apartment at the end of the day, Jill
provocateur might also constrain direct aggression (Dollard, Doob,
curses at her roommate for leaving dirty dishes in the sink.
Miller, Mowrer, & Sears, 1939). For example, if the provocateur is
one’s boss, as in the beginning scenario, then one might get fired
for an aggressive response.
Apart from our own work with colleagues (Bushman &
Baumeister, 1998; Bushman, Baumeister, & Stack, 1999; Ped-
ersen, Gonzales, & Miller, 2000), virtually no recent experimental
Brad J. Bushman, Institute for Social Research, University of Michigan;
research has examined displaced aggression. Moreover, content
Angelica M. Bonacci, Department of Psychology, Iowa State University;
William C. Pedersen, Department of Psychology, California State Univer-
analysis of social psychology textbooks shows that contemporary
sity, Long Beach; Eduardo A. Vasquez and Norman Miller, Department of
social psychology has largely ignored this concept (Marcus-
Psychology, University of Southern California.
Newhall, Pedersen, Carlson, & Miller, 2000). Despite this apparent
Study 2 was conducted in partial fulfillment of Angelica M. Bonacci’s
disinterest, meta-analytic results confirm that displaced aggression
master’s degree requirements. Study 1 was supported by a grant to Norman
is alive and well (Marcus-Newhall et al., 2000). Displaced aggres-
Miller from the James H. Zumberge fund, University of Southern Califor-
sion is a reliable effect that is of moderate magnitude (d
0.55;
nia. Preparation of this article was facilitated by National Institute on
see Cohen, 1988).
Alcohol Abuse and Alcoholism Grant R21-AA013343.
Correspondence concerning this article should be addressed to Brad J.
Bushman, Institute for Social Research, University of Michigan, 426
Triggered Displaced Aggression
Thompson Street, Ann Arbor, MI 48106, or to Norman Miller, Department
of Psychology, University of Southern California, Los Angeles, CA 90089-
In the typical displaced-aggression research paradigm, partici-
1061. E-mail: bbushman@umich.edu or nmiller@usc.edu
pants are initially provoked and then allowed to aggress against an
969

---

970
BUSHMAN, BONACCI, PEDERSEN, VASQUEZ, AND MILLER
innocent third party. Such studies, however, do not capture all
negative affect. Negative affect, in turn, prompts two different
instances of displaced aggression. Often, the target of displaced
reactions: fight tendencies, which are associatively linked with
aggression is not an innocent, nonprovoking individual. Rather, as
aggression, and flight tendencies, which are associatively linked
illustrated by our opening anecdotes, the target emits a mildly
with fear. If fight tendencies are activated, then aggressive
annoying act, as when the coworker cut in line or the roommate
thoughts, feelings, and behavioral tendencies are also activated
left dirty dishes in the sink. The term triggered displaced aggres-
because they are part of the same associative network. Similarly,
sion refers to such instances. Following an initial provocation, the
thoughts, feelings, arousal levels, and behavioral intentions are
target commits a minor provocation, the triggering event, which in
interconnected in the general aggression model (GAM; Anderson
turn prompts an aggressive response (Dollard, 1938). The initial
& Bushman, 2002). For example, angry feelings and increased
provocation and the subsequent triggering event can synergisti-
arousal levels might bring to mind aggressive thoughts.
cally combine to elicit a disjunctively augmented aggressive re-
Ruminatively based displaced aggression describes acts of dis-
sponse toward the triggering target. By disjunctively augmented,
placed aggression that occur long after the initial provocation.
we mean that the aggression directed toward the target exceeds
When the temporal gap between the initial provocation and the
that which is predicted by a tit-for-tat matching rule (Axelrod,
subsequent triggering event exceeds 20 min, then displaced ag-
1984). That is, the aggression directed toward the target exceeds
what is expected on the basis of the intensity of the triggering
gression is primarily a result of rumination. Miller et al. (2003)
event or exceeds what is expected by adding the separate
argued that ruminative processes maintain a cognitive representa-
aggression-eliciting effects of the provocation and the triggering
tion of the subjective state generated by the initial provocation,
event (Miller & Marcus-Newhall, 1997).
even though the initial physiological arousal from that provocation
An important caveat is that the intensity of the triggering event
has dissipated. Rumination facilitates displaced aggression be-
must be minor in comparison with the initial provocation in order
cause it maintains the aggression network activated by the initial
for such a synergistically interactive aggressive response to occur
provocation.
(Miller & Marcus-Newhall, 1997; Pedersen et al., 2000; Vasquez,
Denson, Pedersen, Stenstrom, & Miller, 2005). Minor triggering
events are more ambiguous with respect to provocation and inten-
Triggered Ruminative Aggression
tionality than are strong ones (Vasquez et al., 2005). In the absence
Rumination has been defined as self-focused attention toward
of the initial provocation, an individual might judge the triggering
one’s thoughts and feelings (Lyubomirsky & Nolen-Hoeksema,
event as nonprovocative and unworthy of an aggressive retaliation.
However, a strong initial provocation might prime individuals to
1995). Rumination can also be defined more narrowly as
experience triggering events as more intentional and provoking
provocation-focused thought. Ruminative thought can maintain
and deserving of an aggressive retaliation (Duncan, 1976; Higgins
angry feelings (Martin & Tesser, 1989; Rusting & Nolen-
& King, 1981). If the triggering event does not produce an aggres-
Hoeksema, 1998). In previous research, the relationship between
sive response in the absence of a strong initial provocation, then it
an initial provocation and subsequent displaced aggression was
must be relatively minor in comparison with the initial
mediated by how angry people became after experiencing a trig-
provocation.
gering event (Pedersen et al., 2000).1 Among provoked partici-
One problem with the few initial studies on triggered displaced
pants, the triggering event generated angry feelings. These angry
aggression is that the intensity of the triggering event was at least
feelings, in turn, prompted an aggressive retaliation against the
as strong as the intensity of the initial provocation (Pedersen et al.,
mildly annoying target. Among nonprovoked participants, the trig-
2000). Consequently, the hypothesized interaction between prov-
gering event did not augment aggressive behavior. Recall our
ocation and trigger did not occur. However, two recent studies
opening anecdote in which the college student who was harshly
confirmed the moderating effect of a minor triggering event on
criticized by her professor and had ruminated about it during the
displaced aggression (Pedersen et al., 2000). For example, in one
car ride home subsequently cursed at her messy roommate when
of these studies, participants were either provoked or not provoked
she got home. In parallel, any process that maintains an angry
by an experimenter. Later, they interacted with a confederate who
mood after an initial provocation, such as rumination, should
was either incompetent (trigger) or competent (no trigger). As
increase triggered displaced aggression. Likewise, any process that
expected, participants who were initially provoked displayed more
distracts attention away from an angry mood should decrease
displaced aggression in the presence of a triggering event than in
triggered displaced aggression.
its absence. The trigger had no effect on participants who were not
No prior study has examined whether ruminating about a prov-
provoked (see also Vasquez et al., 2005).
ocation augments aggressive responding to a mildly annoying
Theoretical Model of Triggered Displaced Aggression
Miller, Pedersen, Earleywine, and Pollack (2003) proposed a
theoretical model of triggered displaced aggression that is based, in
1 Reanalysis of Pedersen et al.’s (2000) data indicated that provoked par-
part, on Berkowitz’s (1989, 1990, 1993) cognitive neoassociation
ticipants became angrier after experiencing a triggering event than did non-
model of aggression. Berkowitz posited that aggressive thoughts,
provoked participants. This finding held true in both Study 1, t(30)
1.98,
emotions, and behavioral tendencies are linked together in an
p
.06, and Study 2, t(22)
2.23, p
.04, despite differences in how the
associative network. Aversive events (e.g., provocation) prompt
provocation and the trigger were operationally defined in the two studies.

---

RUMINATION AND TRIGGERED DISPLACED AGGRESSION
971
event that occurs long after an initial provocation.2 We propose
placed aggression but only among people who ruminated about the
that ruminating about a provocation will increase aggressive re-
provocation.
sponding to a subsequent triggering event. We expected provoked
individuals who were induced to ruminate for an extended pe-
Method
riod—namely, 25 and 20 min (Studies 1 and 2, respectively) or 8
hr (Study 3)—to exhibit more displaced aggression after a trigger
Participants and Design
than those not induced to ruminate.
How does rumination augment aggressive reaction to a trigger-
Participants were 42 undergraduate students from the University of
ing event? In keeping with Berkowitz’s (1989, 1990, 1993) cog-
Southern California (29 women and 13 men) who received extra course
nitive neoassociation model, which ascribes a key aggression-
credit for voluntarily serving in a 3 (rumination, distraction, positive
instigating role to negative affect, our preceding discussion has
mood)
2 (trigger, no trigger) between-subjects design.
emphasized cognitive representations of negative (angry) affect as
a critical aggression-inducing component of rumination. By con-
Procedure
trast, the GAM (Anderson & Bushman, 2002) emphasizes physi-
ological arousal, affect, attributions, and behavioral intentions as
Participants were tested individually. They were told that the researchers
conceptually distinct factors, any or all of which can function
were conducting a three-part study on various cognitive skills. The first
independently to increase aggression. Thus, although rumination is
part ostensibly investigated the effect of music on problem-solving ability.
necessarily cognitive, the GAM implies that it can consist of
In reality, it served as the induction of provocation. Each participant
cognitive representations and elaborations of any or all four of
received a list of 15 difficult anagrams (e.g., NVTNIMEREON unscram-
these aggression-inducing antecedents: (a) arousal cognitions—
bled to spell ENVIRONMENT) and was told to solve all of them within 4
awareness of one’s prior physiological responses to the provoca-
min. The participant listened to loud and distracting background music
tion, (b) affect cognitions—awareness of specific emotions felt at
(viz., Stravinsky’s Rites of Spring) while working on the anagrams. After
4 min, the experimenter returned, turned off the music, took the anagram
the time of the provocation, (c) attributions or interpretations
sheet (ostensibly to grade it), and gave the participant anagram scores from
concerning the provocateur’s aggressive act—“he thought I was
a sample of engineering students who manifestly performed quite well on
stupid,” and (d) behavioral inclinations—thought about specific
the task. A few minutes later, the experimenter returned with the partici-
acts of harm that one wished one had bestowed on the provocateur.
pant’s test “score.” He stated that the participant’s score was below average
Extending the GAM, to the degree that rumination contains cog-
as compared with the sample of engineering students. He said that the
nitive representations of any or all four of these factors, it will
participant’s performance was very poor and that the anagram portion of
heighten sensitivity and reaction to a mild triggering event, thereby
the experiment should be done over again. The experimenter then added, in
increasing aggressive responding to it.
an exasperated and irritated tone, that it would be a waste of his own time
Here, however, we cannot systematically explore such details
to repeat the test and that they should just proceed to the second part of the
concerning how rumination produces these effects, nor can we test
study. Pedersen et al. (2000, Study 1) have successfully used this procedure
to induce provocation in previous research.
the comparative validity of the neoassociationistic versus the
Participants were told that the second part of the study assessed imag-
GAM implications regarding the critical mediating components of
ination and creativity. They were randomly assigned to one of three groups:
rumination. Thus, although we do examine the aggression-
rumination, distraction, or positive mood. Participants in the rumination
mediating role of the rumination-induced augmentation of angry
and distraction groups were given a packet with a phrase on each page.
affect and angry attributions in response to the trigger, we do not
They were told to think about each phrase, spend 1 or 2 min writing any
examine the other types of cognitive representations that may be
thoughts that came to mind on a pad of paper, and then move on to the next
elicited during rumination (i.e., physiological arousal and behav-
page of the packet and repeat this same process for 25 min. They were told
ioral intentions) and whose cognitive accessibility is seen by the
not to worry about spelling or grammar. In the rumination condition, the
GAM as also mediating aggression. Nor do we explore here the
phrases were internally focused. To avoid making participants suspicious,
temporal positioning of their occurrence.
the phrases did not mention specific emotions (e.g., anger). Examples
included “what kind of a person you are,” “why people treat you the way
they do,” and “how you interact with people.” In the distraction group, the
Study 1
phrases were externally focused. Examples included “the layout of the
local post office,” “a double-decker bus driving down the street,” and
In typical displaced-aggression paradigms, the temporal gap
“clouds floating by in the sky.” In previous research, judges rated both
between provocation and the displaced-aggression opportunity
types of phrases as affectively neutral (Rusting & Nolen-Hoeksema, 1998).
rarely, if ever, exceeds 10 min. Research suggests that without
In the positive mood group, participants recalled a specific time in their
rumination, anger generally dissipates within 10 min (Fridhandler
lives when they were very happy and then wrote about it (see Krauth-
& Averill, 1982; Tyson, 1998). The purpose of Study 1 was to test
Gruber & Ric, 2000, for a similar mood-induction procedure). Participants
whether ruminating about a provocation beyond 10 min would
were told to focus on specific details of the incident and to describe why
increase the likelihood of triggered displaced aggression. In the
they were happy. When they finished writing about one incident, they were
presence of a triggering event, we expected displaced aggression to
be higher for people induced to ruminate about the provocation
2 We could locate only a single study that has examined the effect of
than for those not induced to ruminate. In the absence of a
ruminative activity on aggressive behavior. Konecni (1974) found that pre-
triggering event, we expected little displaced aggression, regard-
venting people from engaging in rumination reduced direct aggression toward
less of whether people ruminated about the provocation. We also
an insulting confederate. No studies, however, have examined the relation
expected the anger produced by the provocation to increase dis-
between rumination and displaced or triggered displaced aggression.

---

972
BUSHMAN, BONACCI, PEDERSEN, VASQUEZ, AND MILLER
told to write about another and to continue this process for 25 min. The
the initial provocation. On an a priori basis, 10 adjectives were used to
positive mood condition served as an additional control.
measure provocation-induced negative affect (i.e., angry, defiant, hostile,
The experimenter informed participants that the third part of the study
disgusted, irritable, scornful, annoyed, loathing, grouchy, and frustrated).
investigated the impact of audio and visual stimuli on problem solving.
Participants were instructed to describe how they felt after performing the
Participants watched a video of an undergraduate research assistant who
anagram task (viz., the provocation). The items were scored on a 4-point
stated trivia questions aloud and displayed a card with the multiple-choice
scale ranging from 0 (definitely did not feel this way) to 3 (definitely did
foils for each question. Participants answered as many questions as they
feel this way). Responses from the 10 items were summed to form a
could. They were told that the research assistant on the tape had applied for
composite score (Cronbach’s
.90). Finally, participants were fully
a coveted position as a paid researcher in a faculty member’s lab and that
debriefed.
the faculty member wanted participants to evaluate applicants. At the
conclusion of the tape, the experimenter reentered the room, retrieved each
Results
participant’s trivia answer sheet, provided a summary sheet indicating the
average score obtained by a group of engineering students on the same
Preliminary Analyses
trivia game from the previous semester, and then left to “score” the test.
Within each of the three groups (i.e., rumination, distraction, and posi-
Sex differences.
There were no main effects or interactions
tive mood), participants were randomly assigned to the trigger or the
involving sex for any of the measures. Thus, the data for men and
no-trigger group. In the trigger group, the research assistant read the trivia
women were combined for subsequent analyses.
questions too quickly, mispronounced some of the words and names (e.g.,
Statistical assumptions.
Tukey’s (1977) box plot was used to
Leonardo da Vinci was pronounced Leon Divinsky), and occasionally
identify extreme outliers, but none were found.
mixed up the multiple-choice responses (e.g., presenting the multiple-
Trigger manipulation checks.
To assess the effectiveness of
choice responses to Question 12 after reading Question 9). In addition,
participants were told that they did poorly in comparison with the average
the trigger manipulation, participants rated their emotional re-
engineering student, but they were not insulted about their performance. In
sponse to the research assistant’s performance (viz., irritated,
the no-trigger group, the research assistant read the trivia questions slowly,
angered or upset, or happy). They also rated the research assis-
made no pronunciation errors, and correctly matched the questions with the
tant’s performance on the following dimensions: (a) read the
appropriate sets of multiple-choice answers. In addition, participants were
questions slowly, (b) spoke clearly, (c) administered the questions
told that their score was about the same as that of the engineering students’
efficiently, and (d) read the questions correctly. After the happy
average score. Pedersen et al. (2000, Study 1) have successfully used this
and task performance items were reverse scored, the seven items
trigger manipulation in previous research.
were averaged to form a composite score (Cronbach’s
.93).
After obtaining feedback from the experimenter about their performance
As expected, composite scores were significantly higher in the
on the trivia game, participants received a four-page packet containing
trigger group (M
7.08, SD
1.72) than in the no-trigger group
dependent measures and manipulation check items. The aggression mea-
(M
3.02, SD
1.38), t(40)
8.45, p
.01, d
2.60.
sure, positioned on the first page, consisted of five items that were rated on
A further analysis using this same composite variable assessed
an 11-point scale ranging from 1 (strongly agree) to 11 (very strongly
disagree). One item assessed how strongly they recommended the research
whether the manipulation of rumination differentially affected
assistant for the paid assistantship position. Four other items assessed their
participants’ reactions to a subsequent triggering event. Among
evaluation of the confederate on the following dimensions: likable,
triggered participants, those in the rumination condition reported a
friendly, competent, and intelligent. Aggression is generally defined as any
more negative reaction to the triggering event than did those in
behavior intended to harm another person (Bushman & Anderson, 2001).
both the distraction and positive mood conditions, t(19)
2.52,
By giving a negative evaluation, participants could harm the research
p
.05.
assistant’s chance to obtain a highly coveted paid job.
Rumination manipulation checks.
To assess the effectiveness
The second page of the packet contained the trigger manipulation check
of the rumination manipulation, participants indicated how often
items. Participants gave their general impression of the research assistant’s
and how strongly they thought about the anagram task while
performance on the trivia game. Three items assessed the participants’
completing the imagination and creativity task. Because of their
emotional reaction to the assistant’s performance (viz., irritated, happy,
high internal consistency (Cronbach’s
.87), the two items
and angered or upset), whereas the remaining four items assessed how well
the research assistant performed the task (viz., read the questions slowly,
were averaged to form a composite score. Analysis of variance
spoke clearly, administered the questions efficiently, and read the questions
(ANOVA) revealed that the three groups (i.e., rumination, distrac-
correctly). Again, all items were rated on an 11-point scale ranging from 1
tion, and positive mood) did not differ on this purely cognitive
(strongly agree) to 11 (very strongly disagree).
measure of rumination, F(2, 38)
0.15, p
.05.
The third page of the packet contained the rumination manipulation
Participants indicated their affective response to the rumination
check items. Two items assessed a cognitive response to the rumination
manipulation (viz., how angry, grouchy, happy, irritated, pleased,
manipulation. Participants rated how often and how strongly they thought
and sad they felt). After the positive adjectives were reverse
about their performance on the anagram task (viz., the initial provocation)
scored, the six items were averaged to form a composite negative
while performing the imagination and creativity task (viz., the 25-min
affect score (Cronbach’s
.88). ANOVA showed a main effect
assignment that served as the manipulation of rumination). These two items
for group, F(2, 30)
4.64, p
.05. To interpret the main effect,
were rated on a 7-point scale ranging from 1 (not at all) to 7 (very often or
two orthogonal contrasts were performed. The first contrast
very strongly). The other six items assessed affective responses to the
showed that participants in the rumination group experienced more
rumination manipulation (i.e., angry, grouchy, happy, irritated, pleased,
and sad). They were rated on an 11-point scale ranging from 1 (strongly
negative affect while completing the imagination and creativity
agree) to 11 (very strongly disagree).
task than did those in the distraction and positive mood groups
The fourth page of the packet contained a modified Mood Adjective
(M
5.86, SD
1.90, and M
3.78, SD
1.84), F(1, 30)
Check List (Nowlis, 1965) to report how the participants were affected by
9.23, p
.01, d
1.18. The second contrast showed that the

---

RUMINATION AND TRIGGERED DISPLACED AGGRESSION
973
distraction and positive mood groups did not differ (M
3.88,
mood groups (rs
.27 and
.35, respectively, ps
.20). The
SD
1.87, and M
3.68, SD
1.80), F(1, 30)
0.06, p
.05.
correlation for the rumination group reliably differed from the
correlations for the distraction and positive mood groups (zs
Primary Analyses
2.03 and 2.26, ps
.05). The correlations for the distraction and
positive mood groups did not differ (z
0.23, p
.80). Thus, not
Aggression.
Evaluations of the research assistant measured
only did rumination augment participants’ negative affect (see the
displaced aggression (i.e., recommendation of the research assis-
rumination manipulation check results), but among those induced
tant for the paid assistantship position and the evaluative ratings of
to ruminate, the magnitude of their negative affective reaction to
liking, friendliness, competence, and intelligence). A negative
the provocation was in turn associated with greater aggressive
evaluation would harm the research assistant’s likelihood of ob-
response to the trigger. Moreover, within the rumination condition,
taining a highly coveted paid assistantship. Scores on the five
other correlations reflecting intermediate steps in a causal chain
items were summed (Cronbach’s
.86) and analyzed using a 3
that links provocation-induced negative affect to aggressive re-
(rumination, distraction, positive mood)
2 (trigger, no trigger)
sponding to the trigger were also significant. Provocation-induced
between-subjects ANOVA.
negative affect correlated with negative reactions to the trigger as
The analysis revealed a main effect for trigger, F(1, 36)
8.79,
assessed by the trigger manipulation check (r
.64, p
.05), and,
p
.01. This main effect, however, was qualified by a Trigger
in turn, these negative reactions to the trigger were associated with
Rumination interaction, F(1, 36)
3.53, p
.05 (see Figure 1).
a stronger aggressive response to it (r
.70, p
.01).
Rumination influenced aggression in the presence of a triggering
event but did not influence aggression in the absence of a trigger-
Discussion
ing event, F(2, 36)
3.25, p
.05, and F(2, 36)
0.96, p
.05,
respectively. In the presence of a triggering event, participants in
In Study 1, we examined whether ruminating about a provoking
the rumination group were more aggressive than were those in
event increased the displaced aggression elicited by a minor trig-
both the distraction and positive mood groups, F(1, 36)
6.11,
gering event. As expected, participants induced to ruminate after
p
.05, d
0.92. The distraction and positive mood groups did
being provoked were more aggressive after a minor triggering
not differ, F(1, 36)
0.39, p
.05.
event than were those who were distracted or induced to think
Provocation-induced negative affect and aggression.
Given
positively. In the absence of a triggering event, rumination did not
the long temporal gap between the initial provocation and the
intensify displaced aggression.
measurement of aggression, we expected to find a link between
Consistent with Berkowitz’s (1989, 1990, 1993) model of ag-
provocation-induced negative affect (assessed by the Mood Ad-
gression, provocation-induced negative affect was associated with
jective Check List final postmeasure items) and aggression only
increased displaced aggression for participants induced to rumi-
among participants induced to ruminate during that time interval.
nate but not for those who were distracted or were induced to think
We did not expect evidence of such a link among participants in
positively. Other correlational evidence suggested that within the
the distraction and positive mood groups.
rumination condition, the stronger the provocation-induced nega-
As we expected, provocation-induced negative affect was asso-
tive affect, the more negative was participants’ reaction to the
ciated with aggression among those in the rumination group (r
trigger. And the more negative the reaction to the trigger, the more
.56, p
.05)3 but not among those in the distraction and positive
aggression it elicited. These data are consistent with a model in
which (a) rumination that follows a provocation augments negative
affect; (b) cognitive representations of that negative affect in turn
prime reactivity to a trigger, thereby augmenting negative reac-
tions to it; and (c) these negative reactions to the trigger in turn
augment aggressive retaliation. By contrast, in the nonrumination
conditions, the minor triggering event was experienced as only
mildly aversive and was not strong enough to elicit an aggressive
response.
Study 2
One limitation of Study 1 is that we did not include no-
provocation conditions. Previous research has shown that a minor
trigger does not increase aggression in the absence of provocation
(Pedersen et al., 2000). Therefore, in Study 1, we assumed that
rumination would not increase triggered displaced aggression in
the absence of provocation. However, such a conclusion cannot be
drawn unless provocation is manipulated in the experimental de-
Figure 1.
Interactive effects of rumination and a triggering event on
aggression (Study 1). Aggression was measured using the evaluation that
3
participants gave an individual for a coveted research assistantship posi-
To reduce the impact of outliers, we calculated this association by
tion. Scores could range from 1 to 11, with higher scores indicating a more
using percentage bend correlations instead of Pearson correlations (Wilcox,
negative evaluation. Vertical bars denote plus or minus one standard error.
1996).

---

974
BUSHMAN, BONACCI, PEDERSEN, VASQUEZ, AND MILLER
sign. In Study 2, therefore, we included no-provocation conditions.
Participants in the no-provocation condition had a more pleasant expe-
Additionally, to provide a conceptual replication of Study 1, we
rience during the anagram task. First, although they also listened to
used different operational definitions of the key variables in
background music, it was softer and less distracting (viz., Handel’s Water
Study 2.
Music). Second, they were given simpler anagrams to solve (e.g., ESTT for
Study 2 also further investigated the process by which rumina-
TEST). Third, the experimenter did not insult them. After the participants
completed the 4th, 8th, and 12th anagrams, the experimenter simply said,
tion increases displaced aggression (Miller et al., 2003). Correla-
“You have just completed the 4th [8th or 12th] anagram,” respectively.
tional evidence from Study 1 suggested that rumination not only
After participants completed the anagrams, the experimenter took their
maintains negative affect over time but also increases the chances
solutions to their partner, ostensibly for evaluation. Participants were told
that the triggering event will be experienced more negatively and
that the next activity involved writing an essay. Participants in the rumi-
thereby will increase displaced aggression. Study 2 more explicitly
nation condition were told that another professor, unrelated to the study,
tested this mediational model. Specifically, we expected that pro-
was interested in learning about participants’ perceptions of the research
voked participants who were induced to ruminate about a provo-
process. The other professor was asking students to write an essay about
cation would experience the triggering event more negatively than
their experiences as a research participant. Participants were instructed to
would those who did not ruminate. In turn, we expected these
write about what they had done from the start of the study until the present
negative reactions to the trigger to mediate displaced aggression.
time as well as the thoughts and feelings they had experienced. They were
We expected no such mediation for those not provoked.
also told to write about any individuals whom they encountered in the
study and their thoughts and feelings toward those individuals. Note that
the only individual with whom participants had interacted was the exper-
Method
imenter. Participants were told to spend 20 min writing their essay.
Participants in the distraction condition were also asked to write an essay
Participants and Design
for another professor conducting research unrelated to the current study.
Participants were 385 undergraduates at Iowa State University (194
The other professor was supposedly studying visual maps and asked
women and 191 men) who received extra course credit for voluntarily
students to write essays about the layout of the college campus. Participants
taking part in a 2 (rumination, distraction)
2 (provocation, no provoca-
were instructed to create a mental map of the campus and describe what
tion)
2 (trigger, no trigger) between-subjects design.
they saw. They were to write about the various buildings, the purpose of
the various buildings, the landscape architecture of the campus, and the
spatial relations (e.g., location of buildings) on campus. Distracted partic-
Procedure
ipants were also told to spend 20 min writing their essay.
After the participants completed their essay, the experimenter returned
Participants arrived at the lab under the ruse that they were taking part
with the anagram solution sheet supposedly completed by each partici-
in an impression formation study with a same-sex partner. They were told
they would engage in several tasks that would help them form an impres-
pant’s partner. The experimenter instructed participants to examine the
sion of their partner. As a rationale for the displaced-aggression measure,
solutions and evaluate their partner’s performance. The ostensible partner
participants were told that one task involved tasting food, which would
had solved the same list of anagrams that the participant had solved, plus
give them an idea of the types of food their partner liked. They were given
three more. If the participant had fewer than three incorrect solutions, then
a “food preference form” and told that they and their partner would each
the partner correctly solved all of the anagrams, which happened 7% of the
taste and rate one of the foods on the list. Participants then rated how much
time. The evaluation form asked participants to rate their partner’s anagram
they liked certain types of food (e.g., dairy food, snack food, seafood, spicy
performance by using the following items: (a) “Taking into account the
food). Ratings (e.g., I like spicy food) were made on a 21-point scale
difficulty level of the task, the other participant’s overall performance on
ranging from
10 (strongly disagree) to 10 (strongly agree).
the anagram test seems _____”; (b) “If you had to guess, the concentration
The second task ostensibly assessed verbal skills. All participants saw a
level used by the other participant on the anagram task appears to be
series of scrambled letters on a computer screen for 5 s. Participants were
_____”; and (c) “Based on the limited information I have, it seems that the
given 7 s to solve the anagram by writing the correct solution on an answer
likelihood of the other participant performing very well in a class at Iowa
sheet and by saying the answer aloud over an intercom. Finally, they were
State University that requires good verbal skills is _____.” Ratings were
given 5 s to speak into the intercom a first-person sentence that used the
made on a 21-point scale ranging from
10 (very poor) to 10 (very good).
word. The experimenter emphasized that participants must loudly and
There was also space for written comments.
clearly state their solution and sentence.
After the participants had evaluated their partner, the experimenter
The actual purpose of the anagram task was to provoke or not provoke
returned with the partner’s evaluation of the participant’s anagram perfor-
participants. Those in the provocation condition were subjected to three
mance. Participants in the trigger condition received scores of
2 for
aggravations. First, they listened to loud and distracting background music
overall performance,
1 for concentration level, and
1 for success in a
(viz., Holst’s Mars, the Bringer of War) while working on the anagrams.
class requiring good verbal skills, along with the following written com-
Second, the anagrams were difficult (e.g., DMMPAIUNNEO unscrambled
ment: “Although the task was difficult, I would have thought a college
to spell PANDEMONIUM). Third, the experimenter interrupted and rudely
[class standing of the participant; e.g., sophomore] would have performed
insulted them three times during the anagram task. After the 4th anagram,
better.” Participants in the no-trigger condition received ratings of 2 for
the experimenter said, via the intercom, “Look, I can barely hear you. I
overall performance, 1 for concentration level, and 1 for success in a class
need you to speak louder please.” After the 8th anagram, the experimenter
requiring good verbal skills, along with the following written comment:
said in a louder and angrier tone, “Hey, I still need you to speak louder
“Although the task was difficult, I thought my partner did a fairly good job
please!” After the 12th anagram, the experimenter said in a loud, frustrated,
for a college [class standing of the participant].” Thus, the trigger evalu-
and exasperated tone, “Look, this is the third time I’ve had to say this!
ations were slightly negative, whereas the no-trigger evaluations were
Can’t you follow directions? Speak louder!” The experimenter’s insults
slightly positive. This trigger procedure has been successfully used in
were prerecorded and played over an intercom system. This provocation
previous research (e.g., Pedersen et al., 2000, Study 2).
procedure has been successfully used in previous research (e.g., Pedersen
The next phase of the study served as the displaced-aggression oppor-
et al., 2000).
tunity (Lieberman, Solomon, Greenberg, & McGregor, 1999). After the

---

RUMINATION AND TRIGGERED DISPLACED AGGRESSION
975
participants finished reading the evaluation, the experimenter returned with
Primary Analyses
the partner’s “food preference form.” Participants were told to examine it
and note what types of food their partner did and did not like to eat. They
We used a 2 (provocation, no provocation)
2 (rumination,
also were told that they would next sample one of the items on the list. For
distraction)
2 (trigger, no trigger)
2 (participant sex) between-
all participants, their partner positively rated most of the foods. However,
subjects factorial design to analyze the data. We obtained signif-
the partner rated spicy foods as
9 on a scale ranging from
10 to 10. The
icant main effects for provocation and trigger and a two-way
partner also provided the written comment “I like most of the foods listed
Rumination
Trigger interaction, Fs(1, 331)
8.12, 8.67, and
above, but I hate spicy foods.”
4.01, respectively, ps
.05. However, these lower order effects
The experimenter then returned with a box containing a 3.5-oz (99.22-g)
were qualified by the predicted three-way Provocation
Rumi-
Dixie cup, a container of hot sauce, a lid, two spoons, and a cup of water.
nation
Trigger interaction, F(1, 331)
4.12, p
.05. To
The hot sauce was a mixture of 5 oz (141.75 g) of Melinda’s original
interpret the three-way interaction, the two-way interactions be-
habanero pepper sauce XXX HOT combined with 12 oz (340.19 g) of
tween rumination and triggering event were examined separately
Heinz chili sauce. Participants were told that they were randomly assigned
to eat pretzels and that their partner was randomly assigned to eat hot
for provoked and nonprovoked participants. For provoked partic-
sauce. Participants were also told that their partner would decide how many
ipants, the Rumination
Trigger interaction was significant, F(1,
pretzels they would consume and that they would decide how much hot
161)
7.46, p
.01. As shown in Figure 2, in the presence of a
sauce their partner would consume. It was emphasized that they, and their
minor triggering event, participants allocated more hot sauce if
partner, would be required to consume the entire amount of food product
they had been induced to ruminate than if they were distracted,
that they were given. Participants then sampled the hot sauce, which was
t(75)
3.21, p
.01, d
0.68. In the absence of a triggering
very spicy. They were given water to drink if it was too spicy. The
event, there was no difference in hot sauce allocation between
experimenter then instructed them to spoon into the cup as much hot sauce
participants induced to ruminate and participants who were dis-
as they wanted their partner to consume. Participants were also told to
tracted, t(86)
0.56, p
.05, d
0.13.
place a lid on the cup so the experimenter did not know how much hot
For nonprovoked participants, the Rumination
Trigger inter-
sauce the participants put in the cup. After the participants finished allo-
action was not significant, F(1, 178)
0.07, p
.05. As shown in
cating the hot sauce, the experimenter removed their cup and returned
Figure 3, participants induced to ruminate and participants who
shortly with a cup containing three pretzels for the participants to eat. The
participants then rated how much they liked eating the pretzels.
were distracted did not differ in how much hot sauce they allocated
Participants also rated how they felt about their partner’s evaluation of
to their partner, regardless of whether there was or was not a minor
their own anagram performance (i.e., the triggering event). We expected
triggering event, t(83)
0.20, p
.05, d
0.04, and t(95)
0.59,
these ratings to mediate the effect of the trigger on displaced aggression for
p
.05, d
0.14, respectively.
participants who ruminated about the provocation. Participants rated how
Other effects less central to the hypotheses being tested were
angry it made them feel, how much it bothered them, whether it was overly
also found. Overall, men were more aggressive than women, F(1,
critical, whether it was harsh, and whether it was nasty. Ratings were made
331)
5.09, p
.05, d
0.80. There was also a significant
on a 10-point scale ranging from 1 (strongly disagree) to 10 (strongly
Rumination
Sex interaction, F(1, 331)
5.85, p
.05. In the
agree). Other items were added as fillers. These key items were standard-
rumination condition, there were no sex differences in aggression,
ized and summed to yield an overall composite score of how negatively
t(176)
0.11, p
.05, d
0.03. In the distraction condition,
participants perceived the evaluation (Cronbach’s
.88). Higher scores
men were more aggressive than women, t(167)
3.35, p
.01,
indicated a more negative reaction to the anagram evaluation. To control
for order effects, half of the participants rated their anagram evaluation
d
0.52. Participants’ sex did not interact with trigger or provo-
before allocating hot sauce to their partner, whereas the other half allocated
cation ( ps
.05).
hot sauce to their partner before rating their anagram evaluation. Finally,
the experimenter probed participants for suspicion and debriefed them.
Results
Preliminary Analyses
Statistical assumptions.
Tukey’s (1977) box plot was used to
identify potential outliers in the primary dependent variable—the
amount of hot sauce allocated to the ostensible partner. Because
extreme outliers can unduly influence least squares estimates, they
were removed from the data set (Barnett & Lewis, 1978). An
extreme outlier was defined as allocating more than 25 g of hot
sauce to the partner. Removing the extreme outliers left a total
sample size of 347 participants (186 men and 161 women). A
Fisher’s exact test showed that the outliers removed did not depend
on experimental condition ( p
.05).
Figure 2.
Effects of triggering event and rumination on aggression after
Order effects.
The order in which participants completed the
an initial provocation (Study 2). Aggression was measured using the
measures did not influence their level of aggression, either alone or
number of grams of hot sauce that participants gave a confederate (who
interacting with other variables ( ps
.05). Thus, the data from the
hated spicy food) to consume. Vertical bars denote plus or minus one
two orders were combined for subsequent analyses.
standard error.

---

976
BUSHMAN, BONACCI, PEDERSEN, VASQUEZ, AND MILLER
examine the effect of constraining the causal paths between neg-
ative reactions to the trigger and aggression to 0 for the other
conditions, another model was run in which these paths were
allowed to vary. Allowing the paths to vary did not significantly
improve the fit of the model over the hypothesized model,
2(3,
N
337)
2.87, p
.05. Moreover, the causal paths between
negative reactions to the trigger and displaced aggression were not
significant ( ps
.05). This suggests that negative reactions to the
trigger and aggression were not related to each other in the other
conditions.
The mediation analyses provide support for the hypothesized
model. However, for provoked participants who ruminated, it is
possible that the triggering event had a direct effect on displaced
aggression. Consequently, we modified the hypothesized model by
adding a direct causal path between the trigger and displaced
aggression and treating that path as a free parameter. Inclusion of
Figure 3.
Effects of triggering event and rumination on aggression with-
this path, however, did not significantly improve the fit of the
out an initial provocation (Study 2). Aggression was measured using the
model, 2(3, N
77)
3.41, p
05. Moreover, the direct path
number of grams of hot sauce that participants gave a confederate (who
between trigger and displaced aggression was not significant (z
hated spicy food) to consume. Vertical bars denote plus or minus one
1.87, p
.05).
standard error.
The mediation observed among participants in the provocation–
rumination condition can be explored by examining the indirect
effects estimate. The indirect effects estimate and its significance
Mediation Analyses
test are based on Sobel’s (1982) method, which divides the indirect
Negative reactions to the triggering event were hypothesized to
effects estimate by the standard error and compares that value with
mediate the relationship between the triggering event and dis-
a standard normal distribution. Because this significance test has
placed aggression among provoked participants who had been
low statistical power, a modified z distribution was used rather
induced to ruminate about the provocation (i.e., the provocation–
than a standard normal distribution (MacKinnon, Lockwood, Hoff-
rumination
condition).
For
the
other
three
conditions
man, West, & Sheets, 2002). For the hypothesized model, the
(provocation– distraction, no provocation–rumination, and no
provocation– distraction), no mediation was hypothesized. Media-
tion analyses were conducted to test these hypotheses by using the
Table 1
LISREL 8.52 computer program (Jo¨reskog & So¨rbom, 2002).
Variance–Covariance Matrices Used for LISREL Mediation
Table 1 presents the variance– covariance matrices used for
Analyses in Study 2
analyses.
For participants in the provocation–rumination condition, a
Measure
1
2
3
causal path was specified linking trigger (1
trigger, 0
no
trigger) to negative reactions to the trigger. A second path was
Provocation-rumination group (n
77)
specified linking these negative reactions to displaced aggression.
1. Trigger
0.25
For participants in the other three conditions, a causal path was
2. Evaluation
1.28
23.01
specified linking negative trigger reactions to aggression. How-
3. Aggression
0.78
7.53
18.36
ever, the path between negative reactions to the trigger and ag-
gression was set to 0. We hypothesized that, regardless of condi-
Provocation-distraction group (n
83)
tion, a mildly negative trigger would be interpreted more
1. Trigger
0.25
negatively than a neutral evaluation in all conditions. The proposed
2. Evaluation
0.83
14.67
model exhibited a good fit to the data, 2(7, N
337)
7.19, p
3. Aggression
0.15
1.42
8.15
.05; goodness-of-fit index
.97; comparative fix index
1.00,
No provocation-rumination group (n
87)
root-mean-square
error
0.14.
For
participants
in
the
provocation–rumination condition, the path between trigger and
1. Trigger
0.25
negative reactions to it was positive and significant (
.46, z
2. Evaluation
0.59
14.60
5.48, p
.05). Also, the path between negative reactions to the
3. Aggression
0.17
1.49
9.36
trigger and aggression was positive and significant (
.37, z
No provocation-distraction group (n
90)
3.44, p
.05). These results suggest that for provoked participants
who were induced to ruminate about the provocation, a minor
1. Trigger
0.25
triggering event produced a negative reaction to the trigger, which
2. Evaluation
1.06
15.44
in turn increased displaced aggression (see Figure 4).
3. Aggression
0.09
0.41
13.77
For the other conditions, the causal paths between trigger and
Note.
Variances are on the diagonal, and covariances are below the
negative reactions to the trigger were positive and significant. To
diagonal.

---

RUMINATION AND TRIGGERED DISPLACED AGGRESSION
977
Figure 4.
Negative reactions to the trigger as a mediator between trigger and displaced aggression for provoked
participants who ruminated (Study 2). *p
.05.
indirect effect of trigger on displaced aggression via negative
actions by others as hostile (e.g., Crick & Dodge, 1994). Numerous
evaluation was significant (
.20, z
2.91, p
.05). More-
studies have shown a strong relationship between hostile attribu-
over, this remained true even after the direct path between trigger
tion of intent and aggressive behavior among both children and
and displaced aggression was added to the model (
.17, z
adults (de Castro, Veerman, Koops, Bosch, & Monshouwer, 2002;
2.97, p
.05).
Epps & Kendall, 1995). Thus, it appears that ruminating about a
Finally, we also examined a model that reversed the causal
provocation increases the likelihood that people will perceive
sequencing and thereby tested instead whether the aggression
minor triggering events in a hostile manner.
elicited by the trigger mediated participants’ negative reaction to
One might hypothesize that it was the triggering event itself, not
the triggering event. Analyses revealed that this reversed media-
rumination, that prompted displaced aggression. Evidence, how-
tion model was not viable. In the full reversed causal model for the
ever, refutes this alternative explanation. First, under conditions of
provocation–rumination group, the path between trigger and ag-
no provocation, the presence or absence of a triggering event had
gression was not significant. Moreover, dropping the direct path
no effect on displaced aggression. This suggests that under normal
between trigger and negative reactions to the trigger reduced the fit
circumstances the triggering event in itself is insufficient to prompt
of the model,
2(3, N
77)
41. 59, p
.05, goodness-of-fit
an aggressive retaliation. In addition, the triggering event had no
index
.88. Thus, the data are not consistent with a model that
effect on displaced aggression either among those nonprovoked or
sees the rumination-induced aggressive responding to the trigger
among those both provoked and distracted. Although nonprovoked
as causing negative reactions to the trigger. In sum, the mediation
and provoked– distracted participants interpreted a minor trigger-
analyses suggest that for provoked participants who ruminated
ing event negatively, that interpretation was not sufficient enough
about the provocation, their negative reaction to the triggering
to prompt aggression.
event mediated its effect on displaced aggression.
Finally, Miller et al. (2003) suggested that the contribution of
rumination to displaced aggression becomes important when a
Discussion
lengthy delay separates the initial provocation and the subsequent
triggering event. In Studies 1 and 2, through the use of a 20 –25-
Study 2 conceptually replicated the findings of Study 1. Pro-
min delay period, the interval between provocation and trigger
voked participants who were induced to ruminate engaged in more
exceeded the 10 –15-min duration over which provocation-induced
triggered displaced aggression than did provoked participants who
arousal typically persists (Fridhandler & Averill, 1982; Tyson,
did not ruminate. The absence of either an initial provocation or a
1998). Moreover, this delay substantially exceeded that used in all
triggering event nullified the effect of rumination on displaced
prior studies of triggered displaced aggression. Yet, it is still
aggression. These results suggest that the triggering event, in itself,
conceivable that some arousal generated by the initial provocation
is indeed minor and not severe enough to elicit an aggressive
lingered and this arousal, not rumination, prompted displaced
retaliation.
aggression. If, however, the delay between provocation and trigger
On the basis of the theoretical model proposed by Miller et al.
is lengthened substantially beyond a 25-min interval, then it seems
(2003), we hypothesized that ruminating about a provocation
much more likely that ensuing displaced aggression is rumina-
maintains an internal state that primes individuals to interpret
tively based rather than arousal-based (Miller et al., 2003). There-
triggering events more negatively than warranted and more de-
fore, in Study 3, we extended the delay between provocation and
serving of an aggressive retaliation. In Study 2, we included a
the triggering event to 8 hr.
measure assessing negative reactions to the triggering event, and
we found support for our hypothesis. Among provoked partici-
Study 3
pants who ruminated, their negative reaction to the triggering event
mediated the relationship between the triggering event and dis-
The goal of Study 3 was to replicate the findings of Studies 1
placed aggression. Participants experienced the minor triggering
and 2 by using a considerably longer rumination period (8 hr as
event negatively (e.g., anger provoking, harsh, and overly critical),
opposed to 20 or 25 min) and using different operational defini-
and this negative interpretation, in turn, prompted an aggressive
tions of the key variables. In the presence of a triggering event, we
retaliation.
expected higher levels of displaced aggression among participants
The fact that people who ruminated about the provocation
who had been induced to ruminate about the provocation than
interpreted the trigger more negatively is related to research on the
among those who had not been induced to ruminate about the
hostile attribution bias, in which one tends to perceive ambiguous
provocation. In the absence of a triggering event, we expected low

---

978
BUSHMAN, BONACCI, PEDERSEN, VASQUEZ, AND MILLER
levels of displaced aggression regardless of whether participants
The next part of the procedure, presented as a competitive reaction time
were induced to ruminate about the provocation. Finally, because
task, was based on a paradigm developed by Taylor (1967). Previous
we showed in Study 2 that in the absence of an initial provocation
research has established the construct validity of Taylor’s paradigm (e.g.,
an induction of rumination had no aggression-augmenting effect
Anderson & Bushman, 1997; Bernstein, Richardson, & Hammock, 1987;
Giancola & Zeichner, 1995). The ostensible purpose of the reaction time
(irrespective of the presence or absence of a trigger), we omitted
t