Person. indiuid. DijJ Vol. 23, No. 4, pp.

619-629, 1997
Q 1997 Elwier Science Ltd. All rights reserved
Pergamon Printed in Great Britain
PII: so191-886!I(!I7poo74-3 0191-8869/97 $17.00+0.00

ATTENTION TO ANGER-RELEVANT AND IRRELEVANT

STIMULI FOLLOWING NATURALISTIC INSULT

Christopher I. Eckhardt* and Dale J. Cohen

Department of Psychology, University of North Carolina at Wilmington.
Wilmington, NC 28403-3297, U.S.A.

(Received 8 December 1996)

Summnry-Attentional biases toward mood-congruent, task-irrelevant stimuli have previously been dem-
onstrated in anxious and depressed subjects using the emotional Stroop procedure. We used a variation of
this method to investigate basic information processing mechanisms underlying anger arousal. Eighty-
eight undergraduates classified as high or low trait anger according to the Trait Anger Scale were randomly
assigned to insult or no-insult conditions. Subjects response latencies to name the colors of computer-
presented anger words, positive emotion words, and neutral words were then assessed. An interaction
between anger classification, word, and insult was found. Specifically, anger words interfered with the
performance of high trait anger-insulted subjects, but not with performance of low trait anger-insulted
subjects. There was no effect of word type for all non-insulted subjects. This finding is consistent with
Persons and Mirandas mood-state hypothesis. 0 1997 Elsevier Science Ltd

INTRODUCTION

A substantial number of articles investigating the information processing styles of emotionally

disordered individuals have accumulated in the past 15 years (for a review see MacLeod & Mathews,
1991; Williams, Mathews & MacLeod, 1996). A frequently cited starting point for this area of
research is Bowers (198 1) seminal article on mood and memory. While not specifically addressing
the area of experimental psychopathology in his review, Bower stated that associative memory links
exist between each distinct emotion and other information, concepts, ideas, etc. stored in long-term
memory. Activation of one emotion node in this associative network has the effect of activating
associated or related nodes throughout the memory structures to which it is connected. Additionally,
Bower (1981) explicitly hypothesized that (a) an emotion will enhance the salience of mood-
congruent material for selective attention and learning (p. 142) and (b) an emotion should cause
[mood] congruent words to pop out at the perceiver (p. 142) since these stimuli will be attended
to in greater detail because of the activation of the relevant emotion node. This tendency to allocate
attentional resources toward emotionally relevant stimuli is known as a mood congruent bias. As
noted by Blaney (1986), mood-congruence findings . .. are numerous and have been obtained across
varied tasks, indices of memory access, stimuli, and subject samples (Blaney, 1986, p. 233).
Various modifications of the Stroop (1935) color naming procedure have been used to study
Bowers hypotheses regarding selective attention, emotion, and perceptual processes. In the
emotional Stroop task, subjects who differ on an affective state or trait are asked to name the
colors of both affectively and nonaffectively valenced words. Color naming response latency is the
dependent variable. Bowers hypothesis predicts that subjects who score high on an affective state
or trait will engage in task irrelevant processing when presented with affectively-valenced words.
This interference will produce longer response latencies for color-naming. Data accumulated during
the past decade have indicated that, relative to control subjects, emotionally disordered subjects
demonstrate longer color naming latencies to emotion/threat related words. This effect has been
found in samples of spider phobics (Lavy & van den Hout, 1993; Watts, McKenna, Sharrock &
Trezise, 1986), generalized anxiety disorder patients (Mathews & MacLeod, 1985; Mogg, Mathews
& Weinman, 1989), panic disorder patients (Ehlers, Margraf, Davies & Roth, 1988; McNally,
Riemann & Kim, 1990), rape victims (Foa, Feske, Murdock, Kozak & McCarthy, 1991), combat

*To whom all correspondence should be addressed. Now at: Department of Psychology, Southern Methodist University,
Dallas, TX 75275, U.S.A.

619
620 Christopher I. Eckhardt and Dale J. Cohen

veterans with post-traumatic stress disorder (Kaspi, McNally & Amir, 1995), obsessive compulsive
disorder patients (Foa, Ilai, McCarthy, Shoyer & Murdock, 1993), depressed undergraduates (Gotlib
& McCann, 1984) depressed outpatients (Segal, Gemar, Truchon, Guirguis & Horowitz, 1995)
and depressed inpatients (Gotlib & Cane, 1987) (for a complete review see Williams et al., 1996).
A recent extension of the mood-congruency phenomenon has been articulated by Persons and
Miranda (1992). They noted that research investigating cognitive factors in depression is often
incongruent with cognitive theory. Specifically, cognitive differences between depressed and non-
depressed individuals seem to disappear following the remittance of the depressive episode. However,
when a negative mood state is induced in currently nondepressed individuals with a history of
depression, the cognitive distortions become apparent (Miranda & Persons, 1988; Miranda, Persons
& Byers, 1990). Persons and Mirandas (1992) mood-state hypothesis posits that activation of
negative mood increases the likelihood of the activation and accessibility of mood-relevant cognitive
processes. For example, patients in the midst of a depressive episode (high levels of state depression)
may be more likely to access depressogenic cognitive distortions relative to when they are nonde-
pressed. Therefore, while previous research indicated that individuals scoring high on an affective
trait evidence the mood congruency effect, the mood state hypothesis suggests that manipulation of
the individuals emotional state at the time of assessment may be needed to activate latent cognitive
processes.
The data have been inconsistent concerning the effects of inducing specific moods on subsequent
information processing. While some researchers have found that trait levels of anxiety interact with
increases in state anxiety to produce mood congruency effects (Broadbent & Broadbent, 1988;
MacLeod & Mathews, 1988; Richards, French, Johnson, Naparstek &Williams, 1992) others have
found the effect only in high trait anxiety subjects exposed to a prolonged (rather than laboratory-
induced) stressor (Mogg, Bradley & Hallowell, 1994) and in both high and low trait anxiety subjects
exposed to a high stress condition (Mogg, Mathews, Bird & Macgregor-Morris, 1990). In contrast
to these data, Mathews and Sebastian (1993) demonstrated that traditional emotional Stroop effects
can be nulfzjied if a relevant mood is induced. Williams et al. (1996) have noted that these inconsistent
results are perhaps due to the type of subjects under investigation (clinical vs non-clinical), differing
methods of anxiety induction, and different modes of assessing mood congruent allocation of
attention (e.g. emotional Stroop vs dot probe tasks). The lack of clarity in this area certainly suggests
the need for continued research to further specify the parameters under which mood induction
affects emotional Stroop performance.
Thus far, little attention has been directed toward basic information processing mechanisms
involved with the emotion of anger. The lack of anger related research may stem from the difficulty
of selecting relevant subject samples. As anger problems are not formally recognized in official
diagnostic nomenclature, there exist no accepted criteria for identifying individuals with anger
disorders (Eckhardt & Deffenbacher, 1995; Novaco, 1985). While several official diagnostic cat-
egories include symptoms of anger or irritability (e.g. antisocial personality disorder, borderline
personality disorder, post-traumatic stress disorder), anger is not a necessary criterion. The lack of
an official anger diagnostic category, however, does not mean that significant and measurable
individual differences in anger do not exist or that they are better described by existing classifications.
Given the recent attention afforded to anger-related public health concerns (e.g. interpersonal
aggression and cardiovascular diseases), it is clear that a more systematic study of the emotion of
anger is needed (DiGiuseppe, Tafrate & Eckhardt, 1994; Tavris, 1989).
From a clinical cognitive perspective, inappropriate levels of anger have been theoretically (Beck,
1976; Berkowitz, 1993; Eckhardt & Deffenbacher, 1995; Ellis, 1962, 1977; Huesmann, 1988) and
empirically (see below) associated with numerous cognitive distortions. Researchers have conducted
correlational research using self-report measures of anger and questionnaires assessing Ellis (1962)
irrational beliefs construct with a diverse array of subjects, including college undergraduates in the
U.S. (Hazaleus & Deffenbacher, 1985; Hogg & Deffenbacher, 1986; Mizes, Morgan & Buder, 1990;
Zwemer & Deffenbacher, 1984) and abroad (e.g., Kassinove & Eckhardt, 1994), violent prisoners
(Ford, 1990), clinical outpatients (Deffenbacher, 1992; Deffenbacher & Stark, 1992) and maritally
violent men (Lohr, Hamberger & Bonge, 1988). These studies generally conclude that there exists a
moderate, but significant, overlap between ones self-reported level of anger and irrational ideation.
Although these studies have advanced our understanding of the cognitive distortions associated
 Attention allocation and anger 621

with anger, the attention and perceptual processes involved in anger arousal have not been inves-
tigated. We used a modified version of the emotional Stroop task to assess the attentional processes
of undergraduates scoring high and low on the Trait Anger Scale (Spielberger, 1988). To investigate
whether the interference effects were due to differences at the trait or state level, we randomly
assigned subjects into a naturalistic insult condition or a no insult condition previously implemented
by Nisbett (1993; Cohen, Nisbett, Bowdle & Schwarz, 1993). As discussed earlier, priming manipu-
lations (such as an anger inducing insult) are designed to activate associative networks involved in
processing affect-related information (Anderson & Bower, 1973; Broadbent & Broadbent, 1988;
Persons & Miranda, 1992). This activation may produce differential response outcomes relative to
subjects who have not been primed. We hypothesized that (a) subjects high in trait anger who were
primed would show longer color-naming latencies for anger-related words relative to both high trait
anger subjects who were not primed and all low trait anger subjects; and (b) insulted subjects,
regardless of whether they were high or low in the trait of anger, would demonstrate longer color
naming latencies.

METHOD

Participants
Eighty-eight subjects participated. Subject selection began with a large-scale questionnaire dis-
tribution to undergraduates enrolled in Introductory Psychology during the Spring semester, 1995,
at a medium-sized public university in the Southeast. Of the 167 subjects who completed the
questionnaire packet, those scoring in the upper and lower third of the Trait Anger Scale (TAS) of
the State Trait Anger Expression Inventory (STAXI; Spielberger, 1988) were selected as high anger
and low trait anger subjects for the second phase of the study (high anger = raw score of 21 or
higher; low trait anger = raw score of 17 or lower). National college student norms were used to
identify cut scores (Spielberger, 1988). The STAXI is a well-validated measure of the experience
and expression of anger that possesses excellent psychometric properties (Fuqua, Leonard, Masters,
Smith, Campbell & Fischer, 1991; Spielberger, 1988; Spielberger & Sydeman, 1994). The IO-item
TAS has been frequently used to define client analog samples by Deffenbacher and colleagues (for
a review see Deffenbacher, 1992) and has consistently been shown to possess excellent construct
and criterion validity.
Of the 111 high and low trait anger subjects selected for follow-up, 88 subjects (34 males, 54
females) were located, scheduled, and completed all requirements of the study. The remaining 23
subjects were not assessed because they could not be located. High and low trait anger subjects were
randomly assigned into anger-insult or control (no insult) groups. Thus, four groups of subjects
participated: (a) High Trait Anger, Insult; (b) High Trait Anger, No Insult; (c) Low Trait Anger,
Insult; and (d) Low Trait Anger, No Insult. Demographic information and TAS scores for subjects
in each of the four groups are presented in Table 1. To illustrate that the high trait anger samples
were indeed relevant analogs, note that these subjects TAS means exceed norms presented by
Spielberger (1988) for a large sample of male and female prison inmates (male prisoners M= 21.66,
SD = 6.71; female prisoners M= 19.79; SD = 6.44). The four groups were not significantly different
with regard to age, race, or gender. No gender differences were found on the TAS. All subjects had
normal or corrected 20120 vision.

Table I. Characteristics of the four subject groups

Measure High Anger-Insulted High Anger-Not Insulted Low Anger-Insulted Low Anger-Not Insulted

Mean age 19.4 19.3 19.5 19.4

?/. Female 47.4 58.3 63.2 73. I
91.White 100 79.2 94.7 100

Mean STAXl
Trait anger 27.58 26.08 I 3.6Xb I 3.42b
(6.25) (4.91) (1.77) (1.81)

Note. STAXI = State Trait Anger Expression Inventory. SD in parentheses. For variables with a significant omnibus F, post hoc
Fishers LSD tests were calculated to determine group differences.
bWithin a variable row, those groups sharing the same letter did not differ significantly (p>O.O5).
622 Christopher I. Eckhardt and Dale J. Cohen

Apparatus and materials

Stimulus words. One hundred-twenty colored stimulus words of three different varieties were
presented to subjects: (a) Anger Words, (b) Positive Emotion Words, and (c) Neutral Words.
Emotion words were utilized as stimuli in the present study since they serve as important semantic
cues for specific emotions (for a discussion of subjective labeling of anger arousal, see Eckhardt &
Deffenbacher, 1995; Kassinove & Sukhodolsky, 1995). Thus, it would be expected that these words
may activate associated emotion-cognition complexes. Forty Anger words were selected using
empirical and rational criteria. We began with a list of 80 rationally-derived words that were
reflective of the constructs of anger, hostility and aggression. From this initial list, we utilized Clore,
Ortony, and Foss (Clore, Ortony & Foss, 1987) empirically derived taxonomy of the affective
lexicon and matched 40 words from our original list that were found to represent true affective
conditions (see Clore et al., 1987, pp. 752-754 for a definition of this term). The average word length
of Anger words was 6.9 characters. The same procedure was used to identify 40 Positive Emotion
words, which had an average word length of 8.0 characters. The third set of 40 words were affectively
Neutral words. These words were selected using a random number generator and an unabridged
dictionary. The first non-emotion word after a randomly selected number of dictionary pages was
selected.* Average word length was 7.0 characters.
Presentation of stimulus words. All stimuli were randomly presented on a 15 VGA color monitor
with a 60 Hz refresh rate controlled by an 80486 micro computer using the DOS operating system.
The resolution of the monitor was 640 x 480. The stimuli were centered on the monitor in IBM
default text at twice the standard size (the visual angle subtended about 2 degrees). The words were
presented in one of five colors: red, purple, blue, green, and yellow (IBM palette numbers 4, 5, 1, 2,
and 14 respectively). The screen background was always black.
In the present study, we used a modified version of the emotional Stroop task to assess anger-
related attentional biases. This modification was made to assess whether attentional biases can be
found using a paradigm theoretically similar but not identical to the emotional Stroop task. Recall
that in the original Stroop (1935) task, the semantic content of the color word interfered with the
response of naming the ink color of the presented word. This interference is termed response
competition. In the emotional Stroop task, the subject is required to name the ink color of an
emotionally valenced word. Because the semantic content of the word and the required response
are unrelated, there is no possibility for response competition. The interference often found in the
emotional Stroop task is hypothesized to be the result of a variety of extraneous cognitive processes
initiated by the semantic content of the word (MacLeod, 1991). These processes are hypothesized
to interfere with the subjects color identification decision. It is untenable to propose that these
extraneous cognitive processes onZy interfere with the response required by the emotional Stroop
task; one would expect that extraneous, task-irrelevant cognitive processes would interfere with the
completion of any task requiring shared resources. Therefore, we tested subjects using a modified
version of the emotional Stroop task. In this version, subjects are given a target color before each
trial and are then presented with an emotionally valenced word in colored ink. Subjects indicate
whether the target color is the same as the color of the presented words ink. We hypothesize that
high anger subjects may automatically attend to the content of the anger words, which would initiate
extraneous cognitive processes that interfere with the decision of whether the target color matches
the emotion word color. Although this task superficially differs from the standard emotional Stroop
task, both tasks susceptibility to inhibition by extraneous cognitive processes are theoretically
identical.
A trial consisted of a fixation cross presented for 500 ms, followed by the name of the target color
printed in white letters (e.g., blue). The target color name remained on the screen for 800ms,
followed by a 200 ms blank screen, followed by the stimulus word (e.g. annoy). The subjects task
was to judge whether the stimulus word was presented in the same color as the target color name.
The stimulus word remained on the screen until the subject pressed either the d or k key on the
computer keyboard to indicate their response.? The key indicating a positive response (i.e. target

*We should note that Neutral words were not associated with a single category.
tMacLeod (1991) notes that while interference effects can be lowered in tasks requiring a manual response, the effect cannot
be explained by response mode alone.
 Attention allocation and anger 623

color and word color were the same) was counterbalanced between subjects within groups. Negative
trials (i.e. target color and word color were different) were randomly presented with a probability
of 0.50. Subjects reaction time (RT) was recorded by the computer.

Procedure
The nature of the anger induction for those subjects in the Anger-Insult conditions was adapted
from a script provided by R. Nisbett (Cohen et al., 1993). High and low trait anger subjects were
contacted by phone to schedule their assessment times. Upon arrival to the Psychology department,
subjects were greeted by the experimenter and escorted down a hallway. Midway down the hall, the
experimenter instructed the subject to complete a consent form at a desk situated at the end of the
hallway, and the experimenter left the area. Eight feet prior to arriving at the desk, the subjects
path was impeded by an accomplice of the experimenter who worked at a file cabinet with the file
drawer pulled out as far as possible. The accomplice was a 21 year old male of average height and
build. Subjects typically excused themselves to walk by the accomplice. The accomplice would then
move forward and gently push the file drawer in to allow the subject to pass behind. When the
subject attempted to pass by the accomplice a second time to return to the laboratory, the accomplice
stared into the eyes of the subject, slammed the file drawer into the cabinet, and said Asshole! as
he brushed past the subject and entered a room at the end of the hallway with a self-locking door.
Fifteen seconds after the insult was delivered, the experimenter came down the hallway and
instructed the subject to enter the 6 x 6 laboratory room.
It should be noted that of 38 subjects insulted in the present study, no subject attempted to follow
the accomplice or otherwise retaliate following the insult. Results from Nisbetts laboratory are
quite similar (see Nisbett, 1993). As Cohen et al. (1993) has reported on the validity of this procedure
as an anger induction manipulation (especially for college-aged students from the south), and since
a direct assessment of anger following this apparently natural insult may have suggested the studys
purpose to the subject, post-insult anger was not assessed.
Once the experimenter and subject were in the laboratory room, the emotional Stroop task was
explained using both verbal and written instructions. The experiment began when the subject
indicated understanding of the task. Each subject was individually assessed for a single 15 minute
session. The sessions consisted of 12 practice trials and 120 experimental trials. The experimenter
left the laboratory room during the experimental trials. Upon completion of the Stroop task, the
experimenter entered the room, announced that the experiment was over, and inquired about the
perceived purpose of the study. Upon learning of the deception, subjects were immediately reunited
with the accomplice, who entered the laboratory room and shook hands and interacted with the
subject. No subject reported feeling upset at the deception. The experimenter proceeded to review
the studys hypotheses, the necessity for employing the deception, the random nature of the anger-
insult assignment, and the feelings and concerns of the subjects. All subjects were asked to sign a
statement indicating that they would not reveal the nature of the experiment to anyone. The 88
subjects received course credit and $10 for their participation.

RESULTS

Separate 3 (anger, positive, or neutral content words) by 2 (high trait anger vs low trait anger) by
2 (insult vs no insult) by 2 (male vs female) analyses of variance (ANOVAs) were run on positive
trials (target color and word color are the same) and negative trials (target color and word color
are different). Negative trials were analyzed separately since subjects reaction times to negative
trials are often contaminated by irrelevant second guessing strategies (Sternberg, 1970). These
strategies often increase reaction time and mask most effects. Furthermore, since reaction time data
are very susceptible to positive outliers, all trials in which the subjects response exceeded two
seconds (over three SDs away from the mean) were deleted (Bowerman & OConnell, 1990). Finally,
since reaction time data are only valid if the subject does the task correctly, only correct responses
were included in the analysis (Weltford, 1980). Six subjects were excluded in the analysis because
they had error rates over 15% (three low-anger no insult subjects, two high-anger no insult subjects,
624 Christopher I. Eckhardt and Dale J. Cohen

Table 2. Mean color naming latencies (in msec) to positive trials by anger status, word type, and insult condition

Insulted subjects Non-insulted subjects

n Positive Neutral Anger n Positive Neutral Anger

High anger 19 680 663 711 22 735 750 747

(307) (323) (337) (303) (321) (317)
Low anger 18 615 641 23 712 705 743
(265) (284) (320) (318) (341)

Note. SD in parentheses.

and one low-anger insult subject). The average subjects error rate was 5% and did not vary between
groups.
Positive trials
Reaction time means and standard deviations are presented in Table 2. There was a significant
main effect of word, F(2,146) = 5.8 1, p = 0.004 (q = 0.27). Fishers LSD (alpha =0.05) indicated that
subjects took significantly longer to identify the color of anger words (M= 715, SD= 326) than
either neutral (M= 693, SD = 315) or positive words(M= 688, SD = 304). There was a significant
main effect of insult, F( 1,73) = 6.58, p = 0.0 1 (q = 0.29), such that those who were insulted (M= 660,
SD = 304) took less time to respond than those who were not insulted (M= 73 1, SD = 321). There
was also a significant interaction between gender and insult, F( 1,73) = 6.34, p = 0.01 (q = 0.20), such
that female subjects RTs were unaffected by the insult (insult: M= 698, SD = 3 19; no insult: M= 702,
SD = 306), while male subjects RTs were faster in the insulted condition (M= 606, SD = 273) than
the no insult condition (M= 802, SD = 343).
Importantly, there was a significant 3-way interaction between word type, anger classification, and
insult, F(2,146) = 3.49, p = 0.03 (q = 0.28). Fishers LSD (alpha = 0.05) indicated that the response
latencies to anger words were not significantly different from those to neutral words for all low trait
anger subjects and high trait anger non-insulted subjects. However, high trait anger insulted subjects
had significantly longer response latencies to anger words than neutral words. For all subject groups,
response latencies to positive words were not significantly different from those to neutral words.
With the important exception of the longer response latencies to anger words from high trait anger
insulted subjects, the response latencies were facilitated by the insult (i.e. RTs were shorter for
insulted subjects than those for non-insulted subjects).
There were no other significant effects (all Fs < 2.18).
Negative trials
There was a significant main effect of insult, F( 1,73) = 6.58, p = 0.01 (q = 0.32), such that those
who were insulted (M= 719, SD = 303) took less time to respond than those who were not insulted
(M=804, SD= 326). There was also a trend towards an interaction between gender and insult,
F(1,73)= 3.81, p=O.O54 (q =0.22), such that female subjects RTs were less affected by the insult
(insult: M= 754, SD = 3 13; no insult: M= 789, SD = 326) than male subjects RTs (insult: M= 670,
SD = 282; no insult: M= 849, SD = 322). Finally, there was a 3-way interaction between gender,
word type, and anger classification, F(2,143) = 3.45, p = 0.03 (q = 0.21). Fishers LSD (alpha = 0.05)
indicated that female high-anger subjects RTs to anger words were significantly greater than their
RTs to neutral words but no different than positive words (see Table 3). There were no significant
effects of word type for all low-anger subjects and male high-anger subjects.

Table 3. Mean color naming latencies (in msec) to negative trials by anger status. word type, and gender

Male subjects Female subjects

Positive Neutral Anger n Positive Neutral Anger

High anger 17 794 778 755 24 813 783 817

(337) (328) (322) (340) (319) (344)
Low anger 12 714 716 738 29 751 763 740
(274 (2871 (310) (300) (321) 13011

Note. SD in parentheses.
 Attention allocation and anger 625

There were no other significant effects (all Fs ~2.34).

DISCUSSION

Let us return to Bowers (1981) hypothesis of selective attention to mood congruent material,
and Persons and Mirandas (1992) hypothesis of mood state-dependent cognitive accessibility. Based
on these models, we hypothesized that high trait anger subjects experiencing an episode of state
anger would be affected differently by the anger related semantic content of the colored word than
those subjects classified as low trait anger. This prediction was supported by the data.* In the
primed condition, high trait anger subjects had longer color decision latencies for anger words
relative to both positive and neutral words. Regardless of priming, low trait anger subjects did not
show this pattern of response inhibition.
Contrary to our predictions anger induction priming facilitated all decision times. These data are
in contrast to emotional Stroop findings by Richards et al. (1992) and Mogg et al. (1990), who
reported increases in decision latencies following anxiety induction, but consistent with findings of
Mathews and Sebastian (1993), who reported decreased color naming latencies when snake avoidant
subjects color-named in the presence of a snake or spider. While Williams et al. (1996) explained
the latter finding in terms of the ability for nonclinical subjects to override the emotion-relevant
aspects of the stressful situation, allowing them to focus more intently on task demands, we
hypothesize that our subjects responding may have been a result of the energizing function of
anger (Novaco, 1985). As adrenal hormones are released during anger arousal, one may be more
vigilant, prepared for threat, and attack-ready (as opposed to the escape/avoidance readiness present
during anxiety). Anger-related individual differences may thus be unobservable in the absence of
this physiological arousal. For example, Suarez and Williams (1989) found that subjects classified
as high or low in hostility on the basis of the Cook-Medley Hostility Inventory did not differ on
various physiological measures at rest. When angered during a frustration task, however, high
hostility subjects had significantly higher systolic and diastolic blood pressure and a higher heart
rate. Among the subjects assessed in the present study, dispositional tendencies may have interacted
with the cognitive, affective, and physiological changes that accompanied anger elicitation to
produce faster reaction times. While all groups color naming latencies were reduced during the
priming condition, high trait anger individuals--those who experience more frequent episodes of
state anger-nevertheless attended to task-irrelevant anger words longer after anger induction than
anger induced subjects low in trait anger. Thus, in the context of the generalized response facilitation
for insulted subjects, high trait anger insulted subjects demonstrated the expected finding of longer
decision latencies for anger related stimuli.
When primed, high trait anger subjects had longer color decision latencies for anger words relative
to both positive and neutral words. This effect could not have been the result of stimulus factors
alone (e.g. average word length of lexical stimulit) since (a) high trait anger subjects did not
demonstrate the effect when they were not primed, and (b) low trait anger subjects did not dem-
onstrate the effect in either priming condition. Persons and Mirandas (1992) variation of Bowers
(1981) theory would explain these data by hypothesizing that high and low trait anger subjects
differentially activated attentional resources to anger-relevant stimuli. High trait anger subjects,
when primed, are more likely to selectively attend to anger stimuli because of complex and diffuse
interconnections between anger activation and information processing operations. Importantly,
when high trait anger subjects are not primed, these interconnections are hypothesized to remain
latent. From a theoretical standpoint, our data support these predictions.
Because we did not assess subjects emotional states after the insult, we cannot definitively
conclude that anger was indeed induced. Therefore, our results may appear equivocal. However,
low trait anger insulted subjects performed similar to that of low and high trait anger noninsulted

*All conclusions are based upon data from positive trials. Subjects RT to negative trials has been shown to be contaminated
by second guessing strategies (Stemberg, 1970). The time associated with these strategies often masks the effects of the
manipulation.
tA hypothesis of reaction time latencies based solely on word length would predict faster responses to the anger words by
all groups given that the anger words had fewer characters on average.
626 Christopher 1. Eckhardt and Dale J. Cohen

subjects. Therefore, any emotion induced in low trait anger insulted subjects had no effect on their
reaction times. The effect of word type on high trait anger insulted subjects reaction times indicates
that the emotion induced by the insult, which Cohen et al. (1993) have reliably found to be anger,
triggered a tendency for high trait anger subjects to allocate attentional resources to the anger
words. The question therefore remains as to what specific emotion was induced by the insult for the
high trait anger subjects. Although this is a theoretically important issue, it is nonetheless of little
practical importance. The data show that when high trait anger subjects are insulted they engage in
task-irrelevant processing of anger related stimuli. It is the situation in which the task-irrelevant
processing occurs that is of practical importance. Future experiments should explore the emotion
associated with this important finding.
Although gender did not interact with word type, we are intrigued by the unexpected finding that
insulted males were more likely than insulted females to show faster decision latencies. Researchers
have indicated that men and women are angered by different classes of events (Frodi, 1977, 1978).
However, Harris (1993) demonstrated that while different classes of stimuli triggered anger for
males and females, obscenities that came from males produced similar levels of anger in males and
females. Thus, Harris data indicate that the insult used in the present study should have elicited
similar levels of anger in male and female targets. It is important to distinguish between questions
regarding the dimensions of anger experience (i.e. what cognitive, physiological, and behavioral
dimensions are activated during anger arousal?) and questions concerning the causes of anger
experience (i.e. in the presence of what stimuli is anger elicited?). While ample evidence exists
demonstrating that males and females are prompted to feel angry by different stimuli, a wealth of
data indicate that, in general, males and females experience similar levels of anger (Eckhardt,
Kassinove, Tsytsarev & Sukhodolsky, 1995; Goldman & Haaga, 1995; Hazaleus & Deffenbacher,
1985; Hogg & Deffenbacher, 1986; Mizes et al., 1990; Spielberger, 1988; Zwemer & Deffenbacher,
1984). Future researchers can shed light on the relation between anger eliciting events and anger
experience by employing in viz~oanger induction techniques with male and female research par-
ticipants while carefully assessing how anger is experienced by the two genders.
As the present investigation represents the first attempt to delineate basic processing mechanisms
involved in anger arousal, emerging theory and data from the childhood aggression literature may
relate to our findings. Huesmanns (1988) information processing model of aggression has likened
the connections between aggression and social information processing to a script which the
individual accesses to: (a) select and perceive events, (b) guide responses to these events, and (c)
predict the likely outcomes following response enactment. Anger elicitation fosters attention to
salient script-relevant cues, often at the expense of other contextual cues that may redirect script
retrieval (Huesmann & Eron, 1989). The aggressive script is likely to be reinforced since others tend
to reciprocate with anger and aggression. Similarly, Dodge (e.g. Dodge & Coie, 1987; Dodge, Pettit,
Bates & Valente, 1995) has demonstrated that aggressive children characterized by intense anger
evidence a hostile attribution bias, whereby ambiguous events or actions by others are automatically
perceived as being hostilely motivated and worthy of an aggressive response. Similar findings have
been found for antisocial adolescents (e.g. Slaby & Guerra, 1988).
The present data, while not directly confirming the existence of a hostile attribution bias in angry
adults, are suggestive of attentional attraction toward anger-related stimuli. It is important to
understand the utility of such a bias. Research investigating the reciprocal nature of marital
violence using marital interaction paradigms (But-man, Margolin & John, 1993; Cordova, Jacobson,
Gottman, Rushe & Cox, 1993; Jacobson, Gottman, Waltz, Rushe, Babcock & Holtzworth-Munroe,
1994) has indicated that maritally violent men engage in a variety of negative communication
patterns (e.g., anger, belligerence, aggression) which tend to be reciprocated by their partners. These
men are also more likely to attribute the cause of marital conflict situations to the hostile intentions
of their wives (Holtzworth-Munroe & Hutchinson, 1993). If husband negative behavior is assumed
to follow attribution of negative wife intent, these biases are then verified and reinforced by the
subsequent negative responses of their wives (Cordova et al., 1993). This cognitive bias favoring
anger-relevant contextual information may explain why the anger-primed, high trait anger subjects
in the present study engaged in task-irrelevant information processing; these processes are perhaps
the first part of the anger activation sequence. Future studies would be wise to examine whether
preliminary attentional biases, such as those demonstrated in the present analog sample of insulted
 Attention allocation and anger 6?1

high trait anger subjects, are also present in conjunction with hostile attributional biases in a sample
of maritally violent men.
An important limitation to this study deserves mention. While the two high trait anger samples
in the present study had mean TAS scores that were (a) in the 90th percentile according to
Spielbergers (1988) college student norms and (b) were approximately one standard deviation above
the mean of a sample of prisoners, they are nevertheless an analog sample. As was mentioned at the
outset of this paper, little progress has been made in clinical research on angry patients. Recently,
the first author has delineated a series of diagnostic criteria to describe angry patients and has
constructed interview and self-report measures to facilitate identification of angry subjects for
clinical research purposes (Eckhardt, 1993, 1995; Eckhardt & Deffenbacher, 1995). Until criteria
such as these are considered a valid way to identify clinically angry individuals, anger research must
continue to use available assessment and screening techniques such as the TAS. It is worth restating
that Deffenbacher and colleagues (for a review see Deffenbacher, 1992) have also used TAS-defined
high trait anger college student samples in their series of anger treatment-outcome studies and have
found that high trait anger subjects demonstrate a remarkably high level of symptom distress and
anger-related functional impairment.
In conclusion, the quantity and quality of clinical research on anger has lagged behind the
substantial advancements that have occurred in clinical research on anxiety and depression. One
way of reducing this research gap is to apply paradigms used in the latter research to study the
variables involved in the experience of anger. In our experiment, we have demonstrated that high
trait anger subjects who are insulted tend to process irrelevant anger information, while both high
trait anger subjects who were not insulted and low trait anger subjects in all conditions did not
evidence such processing biases. We attribute this effect to the tendency for high trait anger
individuals who are primed to overapply angry scripts and selectively attend to potential anger-
relevant stimuli in ambiguous situations (i.e. perceiving a hostile threat when none is present).
Although both Bowers (1981) and Persons and Mirandas (1992) theories predicted these results,
they cannot be validated by one particular method. Innovative applications of traditional cognitive
tasks, such as the method used in the present study as well as visual search tasks and word/non-
word priming manipulations, are needed to further investigate the many important questions that
remain regarding the role of information processing in the experience and expression of anger.

Acknolcledgemenfs-This research was supported by a Cahill Faculty Research and Development Grant awarded to the first
author. Portions of this research were presented at the November 1996 convention of the Association for Advancement of
Behavior Therapy. Special thanks to: Richard Nisbett for providing details on his anger induction paradigm, Anita Herring
for assisting in data collection, Shawn Prevatte for delivering the insults, and Julian Keith and Ellen Dennehy for helpful
comments on earlier versions of this manuscript.

REFERENCES
Anderson. J. R. & Bower, G. H. (1973). Human associatice memory. Hillsdale. NJ: Lawrence Erlbaum.
Beck, A. T. (1976). Cognitire therapy and the emotional disorders. New York: International Universities Press.
Berkowitz, L. (1993). Aggression: its causes, consequences, and control. New York: McGraw-Hill.
Blaney, P. H. (1986). Affect and memory: a review. Psychological Bulletin, 99. 2299246.
Bower, G. H. (1981). Mood and memory. American Psychologist, 36, 129-148.
Bowerman, B. L. & OConnell, R. T. (1990). Linear statistics model: an applied approach. Boston, MA: PWS-Kent.
Broadbent, D. E. & Broadbent, M. H. P. (1988). Anxiety and attentional bias: state and trait. Cognition and Emotion. 2.
165-183.
Burman, B., Matgolin, G. &John, R. S. (1993). Americas angriest home videos: behavioral contingencies observed in home
reenactments of marital conflict. Journal of Consulting and Clinical Psychology. 61, 28839.
Clore, G. L., Ottony. A. & Foss, M. A. (1987). The psychological foundations of the affective lexicon. Journal qfPersonalir>
and Social Psychology, 53,751-X6.
Cohen. D., Nisbett, R. E., Bowdle, B. F. & Schwatz. N. (1993). Insult. aggression, and the southern culture ofhonor: an
experimental ethnography. Unpublished manuscript.
Cordova, J. V., Jacobson, N. S., Gottman, J. M., Rushe, R. & Cox. G. (1993). Negative reciprocity and communication in
couples with a violent husband. Journal of Abnormal Psychology, 102. 559-564.
Deffenbacher, J. L. (1992). Trait anger: theory, findings, and implications. In J. N. Butcher and C. D. Spielbetger (Eds).
Advances in personality assessment (Vol. 9, pp. 177-201). Hillsdale, NJ: Lawrence Erlbaum.
Deffenbacher. J. L. & Stark, R. S. (1992). Relaxation and cognitive-relaxation treatments of general anger. Journal of
Counseling Psychology, 39, 158-l 61.
DiGiuseppe, R.. Tafrate. R. & Eckhatdt, C. I. (1994). Critical issues in the treatment of anger. Cognitive and Behaaioral
Practice, I, 111-132.
628 Christopher I. Eckhardt and Dale J. Cohen

Dodge, K. A. & Coie, J. D. (1987). Social information-processing factors in reactive and proactive aggression in childrens
peer groups. Journal of Personality and Social Psychology, 53, 1146-l 158.
Dodge, K. A., Pettit, G. S., Bates, J. E. & Valente, E. (1995). Social information-processing patterns partially mediate the
effect of early physical abuse on later conduct problems. Journal of Abnormal Psychology, 104.632643.
Eckhardt, C. I. (1993). A diagnostic category for dysfunctional anger and hostility. Paper presented at the annual convention
of the American Psychological Association, Toronto, Ont., Canada.
Eckhardt, C. I. (1995). Preliminary validation of the Structured Interviewfor Anger Disorders. Paper presented at the annual
convention of the American Psychological Association, New York.
Eckhardt, C. I. SCDeffenbacher, J. (1995). Assessment of anger disorders. In H. Kassinove (Ed.), Anger disorders: definition,
diagnosis, and treatment. Washington, DC: Taylor & Francis.
Eckhardt, C. I., Kassinove, H., Tsytsarev, S. & Sukhodolsky, D. (1995). A Russian version of the State-Trait Anger
Expression Inventory: preliminary data. Journal of Personality Assessment, 64,44&W.
Ehlers, A., Margraf, J., Davies, S. & Roth, W. T. (1988). Selective processing of threat cues in subjects with panic attacks.
Cognition and Emotion, 2,201-220.
Ellis, A. (1962). Reason and emotion in psychotherapy. New York: Lyle Stuart.
Ellis, A. (1977). Anger: how to live with and without it. New York: Readers Digest Press.
Foa, E. B., Feske, U., Murdock, T. B., Kozak, M. J. & McCarthy, P. R. (1991). Processing of threat related information in
rape victims. Journal of Abnormal Psychology, 100. 156-162.
Foa, E. B., Ilai, D., McCarthy, P. R., Shoyer, B. & Murdock, T. (1993). Information processing in obsessive-compulsive
disorder. Cognitive Therapy and Research, 17, 173-189.
Ford, B. D. (1990). Anger and irrational beliefs in violent inmates. Personality and Individual Differences, 12, 21 I-215.
Frodi, A. (1977). Sex differences in perception of a provocation: a survey. Perceptual and Motor Skills, 44, 113-l 14.
Frodi, A. (1978). Experiential and physiological responses associated with anger and aggression in women and men. Journal
of Research in Personality, 44, 335-349.
Fuqua, D. R., Leonard, E., Masters, M. A., Smith, R. J., Campbell, J. L. & Fischer, P. C. (1991). A structural analysis of
the State-Trait Anger Expression Inventory. Educational and Psychological Measurement, 51,439446.
Goldman, L. & Haaga, D. A. F. (1995). Depression and the experience and expression of anger in marital and other
relationships. Journal of Nervous and Mental Disease, 183, 505-509.
Gotlib, I. H. & Cane, D. B. (1987). Construct accessibility and clinical depression: a longitudinal investigation. Journal of
Abnormal Psychology, 96, 199-204.
Gotlib, I. H. & McCann, C. D. (1984). Construct accessibility and depression: an examination of cognitive and affective
factors. Journal of Personality and Social Psychology, 47,427-439.
Harris, M. B. (1993). How provoking! What makes men and women angry? Aggressive Behavior, 19, 199-21 I.
Hazaleus, S. L. & Deffenbacher, J. L. (1985). Irrational beliefs and anger arousal. Journal of College Student Personnel, 26,
47-52.
Hogg, J. A. & Deffenbacher, J. L. (1986). Irrational beliefs, depression, and anger among college students. Journal of College
Student Personnel, 27,349-353.
Holtzworth-Munroe, A. & Hutchinson, G. (1993). Attributing negative intent to wife behavior: the attributions of maritally
violent versus nonviolent men. Journal of Abnormal Psychology, 102.206-2 I 1.
Huesmann, L. R. (1988). An information processing model for the development of aggression. Aggressive Behavior, 14, I3-
24.
Huesmann, L. R. & Eron, L. D. (1989). Individual differences and the trait of aggression. European Journal of Personality,
3,95-106.
Jacobson, N. S., Gottman, J. M., Waltz, J., Rushe, R.. Babcock, J., Holtzworth-Munroe, A. (1994). Affect, verbal content,
and psychophysiology in the arguments of couples with a violent husband. Journal of Consulting and Clinical Psychology,
62,982-988.
Kaspi, S. P., McNally, R. S. & Amir, N. (1995). Cognitive processing of emotional information in post traumatic stress
disorder. Cognitive Therapy and Research, 19.433444.
Kassinove, H. & Eckhardt, C. I. (1994). Irrational beliefs and self-reported affect in Russian and America. Personality and
Individual Differences. 16, 133-142.
Kassinove, H. & Sukhodolsky, D. (1995). Anger disorders: basic science and practice issues. In H. Kassinove (Ed.), Anger
disorders: definition, diagnosis, and treatment (pp. l-26). Washington, DC: Taylor 8t Francis.
Lavy, E. & van den Hout, M. (1993). Selective attention evidenced by pictoral and linguistic Stroop tasks. Behavior Therapy,
24,645657.
Lohr, J. M., Hamberger, L. K. & Bonge, D. (1988). The nature of irrational beliefs in different personality clusters of spouse
abusers. Journal of Rational-Emotive and Cognitive Behavioral Therapy, 6, 273-285.
MacLeod, C. M. (1991). Half a century of research on the Stroop effect: an integrative review. Psychological Bulletin, 109,
163-203.
MacLeod, C. M. & Mathews, A. M. (1988). Anxietv and the allocation of attention to threat. Ouarterlv
- _ Journal of
Experimental Psychology: Human Experimental Psychology. 38,659-670.
MacLeod. C. M. & Mathews, A. M. (1991). Coanitive-exoerimental avvroaches to the emotional disorders. In P. R. Martin
(Ed.), .Handbook of behavior therapy and psychological science: an-integrative approach (pp. 116-150). Elmsford, NY:
Pergamon.
Mathews, A. M. & MacLeod, C. (1985). Selective processing of threat cues in anxiety states. Behaviour Research and Therapy,
23, 563-569.
Mathews, A. M. & Sebastian, S. (1993). Suppression of emotional Stroop effects by fear-arousal. Cognition and Emotion, 7,
517-530.
McNally, R. J., Riemann, B. C. & Kim, E. (1990). Selective processing of threat cues in panic disorder. Behaviour Research
and Therap_v, 28,407-412.
Miranda, J. &Persons, J. B. (1988). Dysfunctional attitudes are mood-state dependent. Journalof Abnormal Psychology, 97,
76-79.
Miranda, J., Persons, J. B. & Byers, C. N. (1990). Endorsement of dysfunctional beliefs depends on current mood state.
Journal of Abnormal Psychology, 99,237-24 I.
 Attention allocation and anger 629

Mizes, J. S., Morgan, G. D. & Buder, J. (1990). The relationship between cognitions. assertion, and anger arousal. Journal
of CognitivePsychotherapy: An International Quarterly, 4, 369-316.
Mogg, K., Bradley, B. P. & Hallowell, N. (1994). Attentional bias to threat: roles of trait anxiety, stressful events. and
awareness. The Quarterly Journal of Experimental Psychology, 47,841-864.
Mogg, K., Mathews, A., Bird, C., Macgregor-Morris, R. (1990). Effects of stress and anxiety on the processing of threat
stimuli. Journal of Personality and Social Psychology. 59, 1230-1231.
Mogg, K.. Mathews, A. & Weinman, J. (1989). Selective processing of threat cues in anxiety states: a replication. Behaviour
Research and Therapy, 27. 311-323.
Nisbett, R. E. (1993). Violence and U.S. regional culture. American Psychologist. 48, 441-449.
Novaco, R. W. (1985). Anger and its therapeutic regulation. In M. A. Chesney and R. H. Rosenman (Eds), Anger and
hostility in cardiovascular and behavioral disorders (pp. 203-226). Washington. DC: Hemisphere.
Persons, J. B. &Miranda, J. (1992). Cognitive theories of vulnerability to depression: reconciling negative evidence. Cognitive
Therapy and Research, 16,485-502.
Richards, A., French, C. C., Johnson, W., Naparstek, J. & Williams, J. (1992). Effects of mood manipulation and anxiety
on performance of an emotional Stroop task. British Journal of Chnical Psychology, 83.479491.
Segal, Z. V., Gemar. M.. Truchon, C.. Guirguis, M. & Horowitz, L. M. (1995). A priming methodology for studying self-
representation in major depressive disorder. Journal of Abnormal Psychology, 104. 205213.
Slaby, R. G. & Guerra. N. G. (1988). Cognitive mediators of aggression in adolescent offenders: I. Assessment. Developmental
Psychology, 24, 58G588.
Spielberger, C. D. (1988). Manualfor the State Trait Anger E.upression Inventory. Odessa, FL: PAR.
Spielberger, C. D. & Sydeman, S. J. (1994). State-Trait Anxiety Inventory and State-Trait Anger Expression Inventory. In
M. E. Maruish (Ed.), The use ofpsychological testsfor treatment planning and outcome assessment (pp. 292-321). Hillsdale,
NJ: Lawrence Erlbaum.
Sternberg, S. (1970). Memory scanning: mental processes revealed by reaction time experiments. In J. S. Antrobus (Ed.).
Cognition and affect (pp. 13-59). Boston, MA: Little, Brown & Co.
Stroop. J. R. (1935). Studies of interference in serial verbal relations. Journal ofbkperimental Psychology, 18, 643-662.
Suarez, E. C. & Williams. R. B. (1989). Situational determinants of cardiovascular and emotional reactivity in high and low
hostile men. Psychosomatic Medicine, 51,404418.
Tavris, C. (1989). Anger: the misunderstood emotion. New York: Touchstone.
Watts, F. N., McKenna. F. T.. Sharrock. R. & Trezise, L. (1986). Colour-naming of phobia related words. British Journal
of Psychology. 77.97-108.
Weltford. A. T. (1980). Reaction times. New York: Academic Press.
Williams. J. M. G., Mathews, A. & MacLeod, C. (1996). The emotional Stroop task and psychopathology. Psychological
Bulletin, 120, 3-24.
Zwemer, W. A. & Deffenbacher, J. L. (1984). Irrational beliefs, anger, and anxiety. Journal of Counseling Psychology. SI,
391-392.