This is to certify that the
dissertation entitled

THE RELATIONSHIP BETWEEN A PREDISPOSITION TO
THINK ABOUT KILLING AND MEDIA VIOLENCE
EXPOSURE: EXPLORING A NEW MEASUREMENT MODEL

 

presented by
AARON R. BOYSON

has been accepted towards fulfillment
of the requirements for the

Doctoral degree in Communication

\w t QM

a or rofesSorlsSignature

\\ I 0-: log

Date

 

MSU is an Afiirmative Action/Equal Opportunity Institution

 

 

 

UBRARY i
Michigan State
University

 

 

 

 

 

PLACE IN RETURN BOX to remove this checkout from your record.
TO AVOID FINES return on or before date due.
MAY BE RECALLED with earlier due date if requested.

 

, DATE DUE DATE DUE DATE DUE
DILC O I 203

 

 

1231:} 13

 

 

 

 

 

 

 

 

 

 

 

 

 

2/05 p:/CIRC/DateDue.indd-p.1

THE RELATIONSHIP BETWEEN A PREDISPOSITION TO THINK ABOUT
KILLING AND MEDIA VIOLENCE EXPOSURE: EXPLORING A NEW
MEASUREMENT MODEL

By

Aaron R. Boyson

A DISSERTATION

Submitted to
Michigan State University
in partial fulfillment of the requirements
for the degree of

DOCTOR OF PHILOSOPHY
Department of Communication

2006

ABSTRACT
THE RELATIONSHIP BETWEEN A PREDISPOSITION TO THINK ABOUT
KILLING AND MEDIA VIOLENCE EXPOSURE: EXPLORING A NEW
MEASUREMENT MODEL
By

Aaron R. Boyson

A two part survey (11 = 195) was conducted to replicate and extend previous work
investigating the relationship between media violence exposure and homicidal thinking.
A trait-like, multidimensional inventory of homicidal thought was employed in addition
to retrospective reports. Data were consistent with a 7-factor solution. Six of these
dimensions were related positively to the number of actual thoughts reported. The
finding that media violence exposure predicted recalled homicidal thoughts was
replicated. Media violence also predicted scores on several trait-like factors. After
controlling for aggression, the pattern of data suggests that media violence is most closely
related to reactive thoughts, generated under conditions of negative affect and for whom
attention to homicide in the mass media is most likely. Results are discussed in terms of

fitting homicidal thinking into process models of other aggressive thinking.

To my wife; tu eres mi amor de verdad.

iii

ACKNOWLEDGEMENTS

This dissertation exists because of the mentoring afforded to me by an outstanding
committee. Nearly everyone says that. In my case, it is especially true. To rest well, I
must highlight for posterity each member’s particular dedication to my pursuit.

Professor Stacy L. Smith served as inspiration, desperation, nudger, prodder,
mentor, kicker-in-the-pants, counselor, advisor and committee chair.

Professor Chuck Atkin served as committee member, who was keenly
instrumental in providing an estimate of the number of words one should write per day to
complete a dissertation (i.e., 192). I wish I had listened to him.

Professor Franklin J. Boster served as committee member, occasionally enduring
phone calls in the midst of sleep and nagging on the weekend, one of which included the
4th of July.

Professor Tim R. Levine served as committee member, occasionally willing to
append his summer plans to facilitate my progress. He was and is patient, empathic, and
uncommonly reasonable.

Dean Charles T. Salmon served as the outside member in nearly the truest sense

of the moniker, reading manuscripts from both East Lansing and while on fellowship to

the Middle East.

iv

TABLE OF CONTENTS

LIST OF TABLES .............................................................................. vi
LIST OF FIGURES ............................................................................ vii
INTRODUCTION ............................................................................... 1
CHAPTER 1
REVIEW OF LITERATURE .................................................................. 2
Explaining Homicidal Thinking ...................................................... 2
Conceptual Definition ................................................................... 6
Summary ................................................................................... l 6
Predictions .............................................................................. 1 6
CHAPTER 2
METHOD ............................................................................................. 19
Participants ............................................................................. 19
Instrumentation ........................................................................ 20
Procedure ............................................................................... 26
CHAPTER 3
RESULTS ......................................................................................... 28
The Distribution of the Predictors .................................................... 29
Predictions .............................................................................. 3 1
CHAPTER 4
DISCUSSION ................................................................................... 39
FOOTNOTES ................................................................................... 48
REFERENCES ................................................................................. 50
APPENDIX ...................................................................................... 54

LIST OF TABLES

Table]
Intercorrelations Between Indices of Homicidal Thinking 58

Table 2
Correlations Between Predictor Variables ............................................. 59

Table 3
Summary of Non-trivial Predictors of Homicidal Thought .......................... 60

Table 4
Partial Correlations Between Media Violence & Homicidal
Thought Measures, Controlling for Aggression ................................................... 61

vi

LIST OF FIGURES

Figure 1
The Proposed Structure of a Predisposition to Think about Homicide ............ 55
Figure 2
An Inventory of Homicidal Fantasizing Capacity ..................................... 56

vii

INTRODUCTION

“There ’s this kid & I desperately wanna lace him - I wanna throw him on the
ground - spray the pound - And erase him - But in my heart i be 2 scared 2 face him -
Matta fact this kid goes by the name Jason” — 10th grade student web log, 2004

How often are people moved to have thoughts about homicide? To date, very
little research has been conducted to answer this question. Some research attention has
been devoted to assessing psychiatric populations. The general finding from this group
of studies is that about one-third to one-half of psychosocially impaired individuals report
having thoughts about killing others (Altman, Sletten, Eaton, & Ulett, 1971; Korn,
Plutchik, & Praag, 1997; Schwartz, Peterson, & Skaggs, 2001; Wisner, Peindl, & Hanusa,
1994). Diagnoses for a host of psychiatric disorders often include homicidal thoughts as
serious warning signs that warrant medical attention.

Some research suggests that these thoughts may not be symptomatic of just the
emotionally unhealthy. Three studies have examined the homicidal thoughts of people
with ostensibly no psychiatric impairments. Kenrick and Sheets (1993) asked
undergraduates to detail the last time they had thoughts about killing someone, including
the circumstances triggering such thoughts and their relation to the victim. Overall, a full
68% of the students reported having had at least one homicidal fantasy. Feeling
personally threatened was the most common cause of such thoughts. In a follow up
study, a similarly high percentage (69%) of students said they had fantasized about

homicide.

Using similar methods, Crabb (2000) asked participants to recall their homicidal
fantasies. His instrument included a measure of the reported sources that influenced
participants’ thoughts. Undergraduate students who reported fantasies involving weapons
also were asked if their fantasies were influenced by a model, mass media or otherwise.
Among a sample of 295 participants, he found that 45% recalled a recent homicidal
fantasy. The most common weapons used were the body (24%) followed closely by
firearms (23%). Of those who fantasized with a weapon, 84% said their fantasy had been
modeled by some source. Interestingly, the three most commonly reported modeling
influences were films (83%), news stories (72%), and TV programs (68%).

Does the mass media really have the potential to influence thoughts about
homicide? Crabb’s (2000) study only implies such a relationship because media violence
exposure was not examined. The main aim of this investigation is to provide a more
direct test of the relationship between media violence consumption and thoughts about
homicide. At least three theoretical frameworks bear on homicidal thinking as a routine
cognitive behavior. These perspectives rely on evolutionary psychology, and both short-

and long-term schematic processing to derive key theoretical concepts.
REVIEW OF LITERATURE

Explaining Homicidal Thinking
Findings from both of the previous two investigations were interpreted by the
authors from an evolutionary perspective. Using this lens, homicide itself would have

served many benefits including prevention of premature death and gaining or protecting

life-sustaining resources. Thinking through scenarios involving homicide may have
increased killing ability. If so, homicidal ideation would have been a genetically
preferenced, cognitive trait.

. This perspective works well when considering the environment of evolutionary
adaptation but it has clear limits today. Data from the latter study (Crabb, 2000) suggest
that the mass media may play a role in the development of such thoughts. This invokes
an additional theoretical perspective. Perhaps long-term exposure to violence or
homicide in the mass media helps develop more sophisticated procedural information
about killing (Huesmann, 1998). To the extent that repeated exposures serves as
rehearsal or retention, programs for homicidal behavior may be more easily drawn to
mind in some social situations. Both an evolutionary and a social-learning perspective
may be useful in exploring the generation of homicidal thoughts.

Boyson, Smith, and Rosaen (2004) began to address this question by surveying
undergraduate students about their homicidal thoughts as well as their patterns of
exposure to violent media Results indicated that exposure to media violence was
positively related to homicidal thinking, even afier controlling for aggressiveness,
empathy, and impulsivity. Such findings are consistent with a large and growing body of
research demonstrating that viewing media violence can contribute to aggressive
cognitions and behaviors in both children and adults (for review, see Anderson &
Huesmann, 2003). Thus, homicidal ideation may be a particular type of aggressive
thought that exposure to media violence cultivates.

Or does it? One problem with this conclusion is that aggression and homicidal

thinking were not positively related as both evolutionary and social information

processing theories would predict. This finding was unexpected due to previous research
showing that trait-like aggression moderates the relationship between exposure to violent
stimuli and aggressive thought (Bushman, 1995). Given Bushman’s finding, one would
expect that increases in media violence exposure would enrich the schematic information
available to people prone to react aggressively. The absence of a relationship between
aggression and homicidal thought, if valid, suggests that this type of thinking may be
qualitatively different than other aggressive thinking.

A second finding is somewhat surprising. Unlike aggression, empathy was
related positively to homicidal thinking. Yet studies typically reveal an inverse
relationship between aggression and empathy (Cohen & Strayer, 1996; Miller &
Eisenberg, 1988). Indeed, as expected these two variables correlated negatively with
each other in the only previous investigation (Boyson et al., 2004) to include them as
predictors of homicidal thought. What can explain this finding? Vulnerability to
emotional reactivity is part of the empathy measure (Davis, 1983), and other research
suggests that emotional regulation has a role in both feelings of distress (Eisenberg, et al.,
1994) and aggression (for review, see Caprara & Pastorelli, 1989). Perhaps empathy taps
a kind of emotional reactivity that promotes homicidal thinking.

While these possibilities are intriguing, a substantial conceptual limitation may be
a bigger concern. The way homicidal thinking has been measured may explain both of
these findings. It could be argued that all three of the previous studies examined state-
like thoughts. In these cases, participants first were asked to list any homicidal thought
remembered. A myriad of situational influences were reported. As a result, we know

very little about the cross-situational consistency of thinking about homicide. Only two

studies attempted to measure such a tendency (Boyson et al., 2004; Kenrick & Sheets,
1993). Both used single-item measures with severely limited content validity and
unknown psychometric properties. For both methodological and conceptual reasons,
these data have limited value in terms of building a process model that eventually could
lead to an understanding of 1) the personality profile of a homicidal thinker, 2) the
conditions under which people are most prone to think about killing, and 3) the similarity
or differences between aggressive and homicidal cognitions. A general tendency or
predisposition to think about homicide may be important to a theoretical understanding of
the process leading to homicidal thoughts.

Distinguishing between states and traits has been important in personality
research (Eysenck, 1983). However, Mischel (1968) has argued that trait
conceptualizations of personality are misleading because behavior depends heavily on
situational factors. Contemporary measurement models attempt to account for both states
and traits as they influence cognition or behavior (Steyer, Schmitt & Bid, 1999). Latent
state-trait theory rests on the assumption that human cognition, emotion, and behavior are
dependent upon stable characteristics of individuals. From this perspective, we know
almost nothing about the possibility that some individuals may have a characteristic
tendency that predisposes them to homicidal thinking. A trait-like aspect of homicidal
thinking likely was left unexamined by previous researchers. Thinking about killing, as a
tendency, simply may not be decomposed so easily into self-reported recollections and
single-item measures of thought. This limitation makes interpreting the previous findings

risky.

Given the above limitations of the previous research, it is relatively safe to
conclude that we have no empirical evidence that suggests whether individuals can be
trait-like in their tendency to ruminate about killing. Therefore, the purpose of the
present study is fourfold: l) to provide a conceptual framework for homicidal thinking, 2)
to build a measurement model of a trait-like predisposition toward homicidal thinking, 3)
to examine the relationship between homicidal thinking and media violence exposure,
and 4) to retest the relationships between empathy, aggression, and homicidal thinking
using a more trait-like measurement model.

This new measure relies heavily on the descriptive analysis of previous homicidal
thoughts collected. But it also has been informed by research and theory focusing the
role of aggressive cognitions as antecedents of aggressive behavior (Anderson, 1997;
Bandura et al., 2001; Berkowitz, 1984; Crick & Dodge, 1994; Huesmann, 1998) and the
multidimensional nature of imagined interactions (Honeycutt, 2003; Honeycutt, Zagacki
& Edwards, 1989). Together, these perspectives provide a reasonable framework for
defining homicidal thinking more completely and suggest that cognitive, affective, and
frequency content domains should be used to conceptualize a measurement model.
Figure 1 presents a diagram of seven dimensions extracted from these domains. A
general definition is offered next, followed by an explanation of each domain.
Conceptual Definition

Broadly, homicidal thinking is defined as the act of developing or rehearsing
cognitions about killing another person. Killing another person is defined as the
intentional or willful act of taking another human life by, or on behalf of, the thinker.

Accordingly, a homicidal thought or collection of thoughts may be an imagined scenario

of killing that features real or fictitious people, in real or fictitious settings, and may or
may not be directly perpetrated by the person conjuring the thoughts. This definition
lacks any key distinction between other aggressive thoughts, other than delimiting the
behavior type in imagination (i.e., killing versus harming).

Qualitative analyses from previous homicidal thought data overlaps somewhat
with the literature investigating aggressive cognition. The Buss and Perry (1992)
measure, for example, focuses on the cognitive and emotional components of aggression.
Additionally, Honeycutt’s explication of imagined interactions (Honeycutt, 2003;
Honeycutt et a1 , 1989) suggests that thoughts can be measured as tendencies that vary in
terms of their perceived utility and emotional arousal. As with an aggression
predisposition, these literatures suggest that homicidal thinking may also be examined as
a stable personality feature which is multidimensional. Seven such dimensions are
detailed next.

Reactive Thinking. One dimension of thought that emerges most obviously from
the data gathered is the tendency to conjure thoughts following a threat. A substantial
case has been made that much of aggression and homicidal behavior follows provocation
(for reviews, see Baumeister & Boden, 1998, Berkowitz, 1993). In addition, this research
clearly reveals two different types of both homicidal and aggressive behavior. On the
one hand, some homicide and aggression is goal-directed in the absence of threat
(proactive). On the other hand, the vast majority of homicide and aggressive acts appear
to be the result of provocation and/or threat (reactive) where the goal may simply be

protection or retaliation.

In terms of imagining homicide, reactive thinking is defined as the tendency to
have a homicidal thought caused by aversive situations that are perceived to damage or
potentially damage one 's psychological or physical well-being. This definition would
include both threats to the self and others and may be damaging to the body, personal
property, or the mind. Two examples from Boyson et a1. (2004) are illustrative. In one
thought, a young girl was inspired to have a homicidal fantasy against a third-party who
killed her best friend in a drunk-driving accident. This accident did not threaten her
directly, but did appear to damage her psychological well-being. A second fantasy came
from a young man who apparently was humiliated in public when his hat was repeatedly
flicked by another man while attending a basketball game. In this case, retaliation
apparently took the form of an imagined homicidal interaction.

Proactive Thinking. A less common thought motive may be to imagine killing in
the absence of threat/provocation but in order to satisfy some other physical or
psychological goal. This type of thought is consistent with an instrumental orientation
(Moyer, 1976), where aggression is viewed as a viable way to satisfy needs or wants. A
minority of collected thoughts and fantasies about killing seems to point to this type.
Some homicidal imaginations did not follow any apparent aversive situation. In fact,
these thoughts seem goal-oriented, if only as a way to pass the time or to satisfy some
curiosity. One person wondered “what it would be like” to be in the mob and kill for a
living. Some thoughts about homicide may be a means to some end, regardless of how
large or small the gratification. Thus, proactive thinking is defined as the tendency to
engage in thinking about homicide that, by its act, is self-gratifying (at the individual or

group levels) and is not born from a negative affective state. This motive may include

attainable goals in reality (e.g., power), or psychological goals (e. g., entertainment or
curiosity). The term collective goal is included to allow thoughts of killing in order to
advance some group position or purpose, as in war, for example.

Proactivity is also a dimension of imagined interactions (Honeycutt, 2003). From
this perspective, proactivity is defined as an interpersonal interaction imagined in
advance of the actual interaction often for preparatory reasons. The definition offered
here for proactive homicidal thought is distinctly different. No claim is made that some
homicidal thoughts are proactive in the sense that they might prepare an individual for
some future interaction. Here proactive is intended to indicate a type of homicidal
thought that occurs, without threat, when a person imagines a homicidal act of any kind,
regardless of whether the target is a real person.

Proactive homicidal thinking may well take the form of fantastic engagements
rather than real, although this is not required. For example, a proactive thought could be
one that includes fictitious weapons, characters, and settings found in a video game that is
formed out of boredom while a student is languishing in class. This would be a fantasy-
based (i.e., not possible) imagined homicidal interaction that serves some entertainment
goal. Alternatively, homicidal thoughts could be fantastic, although based in reality (i.e.,
actions. are possible). Thoughts about killing in war would qualify provided that the
killing was unprovoked. For example, thoughts about killing Osama bin Laden after
September 1 1th would be reactive thoughts. However, generalized interest in killing in
battle with a non-specific threat or reason could promote homicidal thought that would
fall under the proactive rubric. These examples have been found in previously gathered

thoughts, and seem to be tied in some way to the mass media.

Media-related Thinking. The thought mentioned above by the person who wanted
to know “what it would be like” was apparently inspired by The Sopranos, a popular
mob-based drama on HBO. Given the vast array of symbols and themes including
violence (Smith et al., 1998) and homicide (Huston et al., 1992), on television alone, it is
not surprising that two previous reports point to the role of mediated models in homicidal
thought (Boyson et al., 2004; Crabb, 2000). These connections are bolstered by theories
indicating the influence of observational modeling via the mass media (Bandura, 2001)
and of the role of repeated exposure to violence in the formation of script-based
information about aggression (Huesmann, 1998). When considering the possibility that
homicidal thinking may be a genetically-based predisposition (Kenrick & Sheets, 1993),
mass-mediated images may be an important source of information for some who have no
direct experience with killing or weapons. Therefore, media-related thinking will be
defined as the tendency to think about homicide as a consequence of consuming mediated
images of violence and death, or the tendency to invoke mediated models of killing when
forming unique thoughts.

The dimensions postulated thus far are derived from the cognitive components of
thought. The next section will outline the emotional component. In data gathered
previously (Boyson et al., 2004), emotion seems to play a key role in forming homicidal
thoughts.

Negative Affective Homicidal Thought. A substantial literature suggests that
anger producing situations can cause greater accessibility of aggressive thoughts.
Anderson and Huesmann (2003) review this body of research and conclude that

provocation “is undoubtedly the strongest situational instigator of human aggression” (p.

10

304). Not only are aggressive thoughts more accessible to people following provocation
(see Berkowitz, 1984), but this accessibility appears to be stronger for people scoring
higher on measures of trait-like aggression (Bushman, 1995). Further, irritable people
who may be more prone to provocation seem to be important to this theorizing (Caprara
& Pastorelli, 1989). For example, trait hostility and negative arousal seem to interact to
make aggressive thoughts more accessible (Anderson, Anderson, Dill, & Deuser, 1998).
In part, this association may be explained by self esteem. People with inflated or
unstable views of self-superiority seem to be the most prone to encountering threats, that
may turn violent (Baumeister, Smart, and Boden, 1996). In other words, a fearful
orientation towards social interaction, one of protection, may promote this type of
thought.

Not surprisingly then, homicidal thinking seems to be connected strongly to
conditions of negative affect. Most of the reports analyzed to date resulted from
frustrating, potentially anger producing situations (Boyson et al., 2004; Crabb, 2000;
Kenrick & Sheets, 1993). Negative aflective fantasizing is defined as the tendency to
formulate homicidal fantasies as a result of cued aversive emotional states. Such states
would include but are not limited to anger, fear, or disgust. But why might negative
affective conditions be separate from other, more positive ones?

Positive A flective Homicidal Thought. Socio-cognitive biases play a mediating
role in aggressive behavior. Dodge and Coie (1987), for example, have found that
aggressiveness in boys can be categorized as reactive or proactive based on a hostile
attribution bias. Reactive aggressors can likely be categorized as emotional, irritable, and

distressed as a result of perceiving others in their environment to be acting with the intent

ll

to harm. By contrast, proactive (i.e., instrumental) aggressors do not have this same bias.
Accordingly, their motivation for aggressive behavior (or thoughts) seems to come from
the belief that aggression is an acceptable tool for gratifying needs or wants. In addition,
some evidence indicates that this profile includes evidence of superior thinking
(Baumeister & Boden, 1998) and a decreased ability to perceive others as equal to the
self (Greco & Cornell, 1992). These distinctions suggest that some people may find it
useful, even pleasurable, to entertain homicidal thoughts, even when they do not evolve
from negative feelings.

This possibility is even more likely considering the goal-directed nature of
selective exposure to mass media (for review, see Rubin, 2002). People use the mass
media as a function of both cognitive and emotional motives. Evidence that people are
attracted to media fare just because it is violent is weak (Diener & de Four, 1978; Diener
& Woody, 1981). However, the popularity of movies, television shows, books, video
games, and even music that discuss or portray homicide suggests that people are willing
to expose themselves to situations involving killing. Homicidal thinking simply may be a
fun, entertaining or positively arousing cognitive activity because of its novelty. Thus,
positive affective thought is defined as the tendency for an individual to develop
homicidal fantasies in the absence of negative aflect, or to gain a positive emotional state
from doing so.

In terms of imagined interactions, (Honeycutt, 2003) affect also has been
conceptualized as its own dimension of thought. In this case, the dimension is called
valence and is defined as the degree of emotional affect produced while having an

imagined interaction. This research shows that affect is important to consider in terms of

12

both the type of interaction, and the subject’s memory for it. Honeycutt’s definition is
directed more toward the type of emotion produced as a result of having the thought. In
the two previous affect dimensions conceptualized here, I am concerned with the role of
emotional affect in producing a homicidal thought. Alternative dimensions of affect
could be advanced that focus on affect produced as a result of thinking, but this inventory
is designed to tap a dispositional tendency to engage in a specific cognitive activity.
Thus, it is most useful at present to focus on emotions as antecedents of thought. In part,
this is because most of the reported thoughts seem to be produced in angry or threatening
situations and thus may be caused by severe emotion. In at least one way, this is
consistent with some research in the imagined interaction literature. Zagacki, Edwards,
and Honeycutt (1992) have found that strong emotional intensity during imagined
interactions is most closely tied to negative and mixed feelings, and that stronger
intensity is more often associated with topics such as relational conflict.

The preceding dimensions of thought encompass the parallels between aggressive
and homicidal thought from research in both domains. However, the limitation of
previous state-like measures would be inadequately addressed without accounting for the
tendency to have repeated thoughts about homicide.

Frequency. The relative number of fantasies a person experiences is an equally

important dimension to tap. Rachman and di Silva (1978) compared the obsessional

thoughts and impulses of eight clinical and 124 non-clinical individuals. They found more
similarities than. differences in the thought content across both groups. Withing this
sample, clinical patients reported more frequent and longer lasting obsessions than their

non-clinical counterparts. Because thoughts about killing were present among both

13

groups, the authors concluded that normal and obsessive people’s general obsessions
differ with respect to frequency and duration. Psychiatric diagnoses including post-
traumatic stress disorder, obsessive compulsive disorder, general anxiety disorder, and
others use the number or duration of thoughts as one symptom of psychological
impairment. Thus, the frequency of homicidal thought may be an important concept to
examine if one aims to distinguish healthy from dysfunctional people.

This reasoning suggests that frequency can be divided into two types that center
on the reason or situation that inspired the thought. Some people may have one recurring
homicidal thought or fantasy that is rehearsed over time. Similarly, these same
individuals may have different homicidal thoughts targeted at one or two individuals
from one or two memorable events. Thoughts such as these seem to be repetitive in
nature. They would tend to be inspired by the same event and focus on killing the same
individual. Alternatively, another type of fiequency is defined by the number of thoughts
a person may have across a variety of situations and that focus on a variety of targets.
These thoughts seem independent of one another and are not necessarily repetitive
cognitions.

The previous research suggests that these frequency types may be useful to
examine. Several reports from college students have included thoughts more than 10
years old (Boyson et al., 2004). One female student’s fantasy included the use of a gun to
shoot off the genitalia of her mother’s boyfriend. Apparently, she was 11 years-old at the
time. The mother had provided the student with the weapon because the boyfi’iend was
suspected to be sexually attracted to the daughter. Memory for such a situation could be

long-lasting due to its extreme negative arousal (Bower, 1981; Brown & Kulik, 1977).

14

This type of thought could be replayed in one’s mind for months, if not even years, and
may explain why it could be retrieved after long periods of time.

As a result, frequency of homicidal thought will be defined as I) the tendency to
have a wide variety of distinct homicidal imaginations, and 2) the tendency to replay
(cognitively) a small number of homicidal thoughts. This definition is intended to allow
for the measurement of the raw frequency of homicidal thought, paying attention to
thoughts that differ by motive, target, setting, method or cause. As suggested, these
frequency tendencies may be the product of a different latent trait or experience. As a
result, each frequency definition will be treated as its own dimension of homicidal
fantasizing.

Summary

A host of theories leading to violent behavior features aggressive cognitions
(Anderson, 1997; Bandura, 1990; Berkowitz, 1984; Crick & Dodge, 1994; Huesmann,
1998). This line of research generally has examined thought accessibility, or attributions
of hostility are inferred following isolated conditions varying in provocation or negative
arousal (Berkowitz, 1984; Dodge, 1985). Still, a predisposition toward aggressiveness
has been shown to affect aggressive cognitions (Bushman, 1995). Previous limitations in
measurement of homicidal thought preclude conclusions that homicidal thinking is
anything more than an irregular or rare phenomenon. Consequently, seven dimensions of
thought have been explicated that should underlie a tendency to think about homicide.

Cognitive components of thought generally include procedural information about
the behavior itself, various situational antecedents, boundary conditions that govern the

display of some behavior and are tapped by the proactive, reactive and media-related

15

dimensions. Research reveals that emotion and emotional control play an important role
in aggressive thoughts (Berkowitz, 2000; Caprara & Pastorelli, 1989). The negative and
positive affect dimensions will estimate the affective tendencies that should precede
homicidal thought. Finally, two types of frequency have been defined. Some people
may be willing and able to think about a variety of homicidal scenarios over time
whereas others may instantiate such a thought infrequently. Two frequency dimensions
distinguish these possibly very different tendencies.

Predictions

Homicidal thought inventory. As mentioned, the development of this instrument
is intended to provide a more complete method of measuring homicidal thinking and to
attempt to examine a trait-like tendency to think about homicide. Seven dimensions are
predicted to differentially underscore a predisposition to think about homicide. While
they all may be correlated with each other, they are not intended to represent a second-
order unidimensional factor structure]. Rather, these dimensions should represent a more
sensitive and meaningful set of interrelated concepts from which to expand our
knowledge of the relationship between exposure to media violence and this type of
thought.

Several dimensions from this inventory may cluster together, however. Two
profiles or types of homicidal thinkers are expected, given the present theorizing. Anger,
frustration, and fear seem to encompass the vast majority of emotional states while
engaging in this type of activity. Because these types of thoughts have differing
antecedents (reaction vs. proaction) they should also be associated with different

emotional states (negative vs. positive). A person prone to hostile reactions to threat may

16

be qualitatively different than a person prone to thoughts that appear to serve self-
gratification in the absence of threat. Thus, it is expected that the reactive and negative
affect dimensions will be more strongly related to each other than to the other factors.
Similarly, the proactive and positive aflect dimension will be more strongly related to
each other than to the other factors.

Social-information processing models of aggressive cognition suggest that those
who elaborate or dwell on homicide in the mass media (media-related thoughts) should
have a greater ability to compose thoughts under any circumstance, regardless of the
purpose for doing so. Higher scores on the media-related thought dimension suggest that
homicide violence is attended to or rehearsed more frequently or more intently. As a
result, it is expected that a predisposition to focus on or identify with homicide in the
mass media should be positively related to both proactive and reactive thoughts about
killing. Thus, this dimension should be substantially related to both conceptualizations of
thinkers outlined thus far.

A fourth finding is predicted with respect to the homicidal thought inventory.
Crabb’s (2000) participants who reported homicidal thoughts (retrospective reports)
involving weapons were apparently influenced by televised and movie images of
violence. Items on the inventory derived here are designed to measure more specifically
the tendency to identify with or rehearse scenes, characters, or genres of media that are
the most likely to feature homicidal behavior. Thus, it is expected that the media-related
dimension will be positively related to the open-ended measure of homicidal thinking.

Media violence exposure. The primary goal of this investigation is to examine

more closely the relationship between media violence exposure and homicidal thinking.

17

Whereas previous reports suggest a link, conceptualizing homicidal thinking differently
may change the nature of these findings. One prediction is warranted. The media-related
dimension of thought is designed to estimate a willingness to think about images and
scenarios involving media-related homicide. Given that these individuals are more likely
to seek out violent media containing homicide, it is expected that media violence
exposure will be positively and substantially related to media-related homicidal thought.
It is unknown how the remainderof the inventory will relate to media violence exposure
at this point. Thus, a research question is advanced: What is the relationship between
media violence exposure and six remaining dimensions of homicidal thought?

A second important goal is to retest earlier findings involving aggression,
empathy, and homicidal thinking (Boyson et al., 2004). It was expected that aggression
and homicidal thinking would be positively correlated but the data were not consistent
with this expectation. Instead, aggression was uncorrelated with either open-ended
reports or the single-item measure of thought frequency. In addition to the measurement
limitations of the previous study, sample limitations may also explain these findings.
Almost three-fourths of the undergraduates previously reported on were female (Boyson
et al., 2004). Given that males tend to score higher on the aggression measure (Buss &
Perry, 1992), restriction in range may have attenuated a true relationship.

In terms of empathy, it was expected this variable would correlate negatively with
homicidal thought, but instead the data suggested that it was positively related to the
amount and type of thoughts about killing. Empathy should promote an ability to
consider others’ feelings (empathic concern) and to contemplate multiple viewpoints in

situations involving conflict (perspective taking). Thoughts about killing should be

18

uncomfortable to people who are better equipped to consider its consequences (Bandura,
2001). Given the method limitations raised so far, these two findings should be
interpreted with skepticism and reexamined. First, what is the relationship between
aggression and homicidal thinking? Second, what is the relationship between empathy

and homicidal thinking?

METHOD

Participants

A non-probability sample of undergraduate students was drawn from classes at a
large Midwestern university (i.e., Michigan State University). These classes were taught
across several departments from several colleges. Students reported at least 17 different
majors ranging from pre-law to communication to civil engineering. One-hundred
ninety-five students completed both surveys. All but one reported their sex. Just less
than 40% of the respondents were male. Half of the participants were either 21 or 22
years old, 25% of the sample were 18-20 years old. The youngest participant was 17 and
the oldest reported 34 years old. Class level distributed somewhat unevenly, skewing
toward upper class students. A full 80% of the sample was either junior or senior level.
In terms of ethnicity, 71% of the sample was Caucasian, while roughly 10% were either
African-American or Asian. As detailed below, exposure to media content was measured
across four dominant outlets — tv, movies, music and video games.

Study participants reported substantial daily media use. The typical student

reported consuming approximately two and one-half hours of music (SD = 190.97

19

minutes, Minimum = 0.00 Maximum = 1440.00), two hours of television (SD = 97.14;
Minimum = 0.00; Maximum = 571.90), 90 minutes of movies (SD = 71.80; Minimum =
0.00; Maximum = 481.71), and 27 minutes of video games (SD = 66.47; Minimum =
0.00; Maximum = 481.71) per day. Summing these outlets, the typical student reported
spending almost 6.7 hours per day with these forms of media content. This media use is
slightly higher than expected, specifically for music and movies, given that college
students typically spend less time with the mass media than other demographic segments
(Condry, 1989). However, it is possible that these use estimates are higher because of
media “multi-tasking,” where someone is exposed to several media outlets
simultaneously.
Instrumentation

Homicidal thoughts. In this study, participants were given both open-ended and
forCed—choice items to measure recollections of homicidal thoughts and a predisposition
for doing so. The open-ended measure follows the form of previously reported methods
(Boyson et al., 2004; Crabb, 2000; Kenrick & Sheets, 1993) with one change. Afier a
primer question, students were asked to bubble yes or no to the statement, “In the spaces
provided below, detail the homicidal thoughts you can remember having at one time or
another in your life.” Three spaces were provided for each participant to write as many
as three thoughts. Previously, a homicidal thought was left up to the researcher’s
interpretation. This method has the advantage of providing some clarity when a memory
is written ambiguously. A participant was given zero, one, two, or three depending on

how many “yes” answers were completed (and, thus, not based on coding of the thoughts

themselves).

20

The trait-like tendency to think about homicide was estimated by an index
developed for this study. In total, this measure featured 59, five-point Likert-type items
designed to measure the willingness to have a homicidal thought, even if one cannot be
remembered at the time (see Appendix B). Tendency is defined as the willingness to
think about homicide (e. g., “if someone threatens me, I may think about killing that
person”), or its probability given certain situations (e.g., “when I’m in my car and
someone angers me, I sometimes think about killing that person”). Each of the seven
dimensions outlined previously were measured with multiple items featuring good
reliability. Confirmatory factor analysis was used to examine the possibility that these
dimensions were tapping a second-order factor. Data were not consistent with this
possibility. Several large errors between the predicted and obtained factor correlations
were obtained. Thus, each dimension is treated as an independent criterion variable in
the analyses. Below is a qualitative description of each factor. Each dimension is
described quantitatively in the next section.

The predisposition to think about homicide only when threatened (reactive
thought) was measured using nine items. Higher scores indicated more agreement with
each statement or an increased tendency to think about homicide under the conditions
specified. For example, students responded to the statement, “Given enough aggravation,
I may think about killing someone.” Eight other similar items were intended to tap this
particular disposition.

The tendency to think about homicide absent of any provocation or threat, or
proactive thought, was measured with 15 items. This dimension was intended to estimate

a participant’s tendency to think about or fantasize killing another person simply as a way

21

to pass the time or because the idea satisfied some curiosity. To illustrate, the items “As I
daydream, I may fantasize about killing someone,” or “I sometimes wonder what it
would be like to be a hit-man” were included in this dimension.

A tendency to have thoughts about homicide that is in some way guided by mass
mediated content was estimated using 10 items such as, “I don’t mind listening to music
that talks about killing,” or “When I’m at the movies I have fantasized about being a
killer.” Generally, this set of items aimed at the tendency to identify with, or elaborate on
characters or scenes involving killing.

Two dimensions of homicidal thought focused on the emotion that may
accompany such cognitive activity. Negative affect and positive affect were conceived as
separate dimensions of thought activity. Six items each measured these two concepts.
Negative affect items assessed anger, rage, fright, and stress that may accompany
homicidal thought. In all cases, these items were based on the assumption that the
affective state was an antecedent of thought. Positive affect items were similar in form
but focused on feelings associated with liking, joy, arousal, and thrill.

Finally, two dimensions were composed to measure a predisposition to think
repetitively about homicide. At least two types of repetition are possible. First, some
thoughts or fantasies may be replayed in one’s mind for days, months or even years. Six
items were written to tap an individual’s tendency to labor over, rehearse or dwell on one
or two thoughts. For example, a person may have a particularly traumatic experience that
conjures a homicidal thought. This trauma and the accompanying set of thoughts may be

rehearsed over time.

22

A second type of frequency is the predisposition to have numerous thoughts
involving different actors and scenarios. In this case, the individual is not repeating one
or two thoughts but rather conjuring many different thoughts over time. The item, “I
could have many, different fantasies about killing” exemplifies the final six items of this
inventory.

Violence exposure. Three exposure measures were used to estimate the amount of
violent content absorbed by each respondent. These measures differ in form as well as
specificity. First, participants were asked to report the amount of time spent with each
medium under study (movies, TV, music, and video games). This is a traditional, albeit
crude estimate of exposure to violent content (Webster & Wakshlag, 1985) because it
does not account for content specific effects. Nonetheless, participants reported the hours
and minutes of exposure to each of the four media included across several days. Nine
parallel questions were used for each medium. All answers were converted to minutes per
day, the final unit of analysis for this measure. Each of these four estimates was added
together to compute an overall measure of time spent with mass media content.

To assess media violence exposure more directly, both objective and subjective
ratings of violence were obtained. This decision is warranted because the context of
violent portrayals varies substantially, particularly on television (Potter & Smith, 2000).
Consumers of violence interpret this violence quite differently (Gunter, 1985; Potter, et
al., 2002). As portrayals vary by interpretation, so do effects outcomes, such as fear
(Gunter & F umham, 1984). To gather these data, students also were asked to list their
five favorite and five most frequently viewed choices by medium (i.e., songs, movies) by

writing them in 10 spaces provided on the questionnaire. Participants then rated the

23

amount of violence in each selection on a five-point semantic scale from “no violence at
all” to “a lot of violence”. This method attempts to quantify violence exposure from the
viewpoint of the media-user, but is limited by variable perceptions of what may qualify
as violence according to a theoretical definition of violence. These ratings were averaged
across medium to provide an overall estimate of violence exposure. Not all participants
listed 10 media selections, however. To account for this missing or unidentified data, the
number of programs, movies, games or musicians listed was multiplied to each
individual’s violence rating average, within medium. Thus, the total violence score takes

into account the number of items listed by the participant.2

To compute a more objective rating, independent coders examined each listing
and provided a rating of violence in exactly the same manner as the participants did.
Two undergraduate students, both female, reviewed a commonly used definition of
violence and several examples before coding the data3. They used the same five-point
rating scale as the actual participants did to compute the subjective ratings. Because of
the amount of coding to be done, the chance that one or both would encounter a movie,
show, game or musician unknown to them was high. In this case, each was instructed to
use one of four web based sources of information about the medium (i.e., tvguide.com,
imdb.com, gamespace.com, and muze.com for tv, movies, video games and music,
respectively). Coder agreement was adequate to good. Bivariate correlations were
computed between each coder’s ratings, and the Spearman-Brown correction formula
was applied. For movies, r = .78 (corrected r = .87); for TV r = .42 (corrected r = .59);
for video games r = .71 (corrected r = .83); for music r = .52 (corrected r = .68). As a

result, the average of the two raters’ scores was computed for each individual. These

24

data were then averaged across media to provide an overall violence exposure estimate
(using the same formula to account for the number of items written within each medium).

This method created two overall violence exposure measures from the viewers as
well as the researcher’s perspectives. The correlation between these two estimates was
high (r = .66; corrected r = .80). As a result, these were treated as alternate indicators of
violence exposure by averaging these two ratings.

Finally, a measure of recall of examples of homicide was used to estimate the
amount of exposure to homicide in specific to which participants had been exposed. Ten
spaces were provided and study participants were asked to briefly list as many scenes,
images or scenarios involving homicide to which they were exposed in the mass media.
The number listed was interpreted as the number recalled. Thus, no answer was expected
to be written in each space. This interpretation precluded the necessity to treat any blank
spaces as missing data. This measure did not relate well to other exposure measures, did
not reveal any non-trivial associations, and is therefore not presented in the results
section.

As a result, three types of measures of violence exposure were generated for 1)
raw exposure (minutes per day) to all media content, 2) a composite of subjective and
objective ratings of violence at the program or media selection level, and 3) a recall-
based measure of homicide exemplars. This method yields three general media violence
exposure estimates in addition to the scores derived for each medium. These three
alternate measures did not correlate with one another and were retained as independent

indicators in analyses.

25

To retest earlier findings involving aggression and empathy, the Buss and Perry
(1992) measure of aggression and the Interpersonal Reactivity Index (Davis, 1983) were
again used. Responses to the items in each scale were submitted to confirmatory factor
analysis to examine the first and second-order factor structure in order to provide
evidence for computing an overall score for both concepts.

Aggression. Buss and Perry’s (1992) 29-item self report measure was used to
estimate an individual’s predisposition to aggression. Four factors comprise this measure
- physical aggression, verbal aggression, hostility, and anger. These four dimensions
have been shown to be both internally consistent and valid measures of a relatively stable
tendency to act aggressively (Buss & Perry, 1992). The scale authors cite some evidence
of a second-order factor that can be estimated by summing responses to all items.

Empathy. The Interpersonal Reactivity Index (1983) developed by Davis was
used to estimate empathy as a personality characteristic. This measure has demonstrated
good internal consistency with each factor. Twenty-eight items are used to examine a
person’s empathic concern, fantasy, perspective taking, and personal distress. The author
of this scale does not present evidence of a second-order factor structure. These factors
are substantially correlated with each other (Davis, 1983), and in the previous study
(Boyson et al., 2004) data from confirmatory factor analysis suggested a unidimensional
structure for all these factors.

Procedure

Two surveys were administered separately in order to reduce the likelihood that

students would recognize the research aims. The first survey was designed to measure 1)

preference for violent media content across television, movies, video games, and music,

26

and 2) personality variables that may be related to homicidal fantasies. It was distributed
in a classroom setting and students spent approximately 30 minutes of class time
completing it. The second survey was distributed approximately two to three days later,
usually the next time the class met. To reduce the likelihood that the surveys would be
connected in the minds of the respondents, a confederate researcher was used whenever
possible. The second instrument was designed to elicit individuals’ recollections of
homicidal thoughts, including estimates of both the frequency and willingness to think
homicidally. Additionally, one alternate measure of homicide exposure from the mass
media was included in the second survey.

The main concern in gathering data to address the questions here is that
participants would perceive the researcher aims and answer “favorably” to the homicidal
thought measures. To address this concern, measures were divided across two
disconnected surveys. In order of appearance, the first survey included media exposure,
personality, and demographic measures. The second instrument solicited open-ended and
forced-choice responses about homicidal memories and thoughts, respectively.

The initial questionnaire was passed-out to students during a variety of classes
taught during a summer academic session. The study was introduced to the class first as
a study investigating media use and some aspects of personality that may or may not be
related to media use. All instructors provided class time to complete the survey. After
about 30 or 40 minutes the surveys were collected. At that time, students were not told
about the follow-up survey or the connection to homicidal drinking.

Usually two to three days later, during the next meeting of each class, the second

instrument was distributed. Again, this set of measures was completed in a classroom

27

setting. This time, the researcher explained that the purpose of the study was to learn
more about the nature of homicidal thinking, given that it has been shown to be quite
normal in several previous reports. No connection to the initial survey was made at this
time. When all participants were finished, students were be debriefed by the researcher

or a research assistant in the group setting.

RESULTS

In addition to the seven dimensions of homicidal thinking that were assessed an
open-ended measure that asked participants to describe as many as three specific
homicidal recollections was obtained yielding eight different measures of homicidal
thought. Second order factor analyses performed on these data resulted in the rejection of
the model that posited a single second order dimension. Moreover, no multi-factor second
order model that fitted the data was found. Thus, each of these eight measures of
homicidal thinking is treated subsequently as a separate dependent variable (see table 1).

Table 2 summarizes the zero-order correlations between the set of variables used
to predict homicidal thinking statistically. As argued previously, three variables (media
violence exposure, aggression, and empathy) were predicted to be important antecedents
of homicidal thinking. Additionally, sex, grade point average (a proxy for intellectual
capacity), and general media exposure were included as control variables. Sex was
included because it has been associated with homicidal thinking in some studies, with
women being less likely than men to report homicidal thoughts (Crabb, 2000; Kenrick &

Sheets, 1993). Although no known literature has demonstrated such an effect, a proxy for

28

intelligence was included to examine the alternative explanation that those with greater
intellectual capacity have less homicidal thoughts. Similarly, no known studies have
demonstrated an effect for general media exposure. Nevertheless, the hypothesis that
those with greater media exposure engage in more homicidal thinking has intuitive
plausibility. The overall pattern across media, particularly television, movies and video
games, is that violence is featured frequently. Consequently, general media use will be
examined.

The Distribution of the Predictors

Of the 195 participants who completed both surveys, 194 (99.50%) responded to
the item asking them to indicate their sex. Of these 194 participants 60.30% indicated that
they were female and 39.70% reported being male. To characterize this sample in terms
of homicidal thinking, 29.90% could recall at least one homicidal memory. This estimate
is a slightly lower estimate than two previous reports (Crabb, 2000; Kenrick & Sheets,
1993) but is similar to previous data (about 33% reported thoughts about killing in
Boyson et al., 2004). With respect to sex, 24% of the female respondents recalled a
homicidal thought vs. 38% of the males (:50, N=194) = 3.67, p = .06; r = .14; OR =
1.92). Thus, this sample is similar to those studied previously in terms of the primary
outcome measure.

When asked to report their grade point average, 96.40% of the participants
responded. The distribution of scores approximated closely the normal distribution. These
scores ranged from two to four with a mean of 2.97 and a standard deviation of .46.

Data allowing the calculation of the general exposure index measure were

available from 99.50% of the participants. Scores on this index ranged from 84.29 to

29

2,060.76 with a mean of 404.55 (SD = 266.57). The distribution of these scores was
skewed positively and was leptokurtic. Notably, the highest score in this distribution is a
number that is highly unlikely for the subject to have experienced because it exceeds the
number of hours in the time period for which the participant was to make the judgment. '
Nevertheless, inclusion of this data point did not influence subsequent parameter
estimates substantially, and therefore, is retained.

The media violence measures were completed by 96.40% of the respondents.
Scores ranged from 1.82 to 4.91 with a mean of 2.89 and a standard deviation of .60. The
distribution of these scores exhibited a slight positive skew, but otherwise approximated
closely the normal distribution. The reliability of this index was estimated by correlating
self and rater measures and applying the Spearman-Brown correction. The result of these
calculations produced a standardized a of .80.

The four dimensions of Davis’ (1983) interpersonal reactivity index, perspective
taking, fantasy, empathic concern, and personal distress, were treated as alternate
indicators of empathy and submitted to confirmatory factor analysis. A one factor
solution fitted the data well. NO error larger than .03 (absolute value) was found.
Perhaps due to the content domain, this second-order dimension featured only modest
reliability (standardized item a =.63). As a result, scores from these four dimensions
were averaged to produce a composite empathy score. All 195 participants reported
scores on this variable. This distribution approximated the standard normal curve with a
mean slightly above the scale midpoint (x = 4.19; SD = .63). Scores on this variable

ranged from 2.46 to 5.89.

30

As with empathy, the four dimensions of Buss and Perry’s (1992) aggression
measure were submitted to confirmatory factor analysis to derive an estimate. Verbal
aggression, physical aggression, hostility and anger were treated as alternate indicators of
the same underlying construct. The data were consistent with this hypothesis.
Differences between the predicted and obtained correlations computed did not differ by
more than .02 (absolute value). The estimate of reliability suggested an acceptable level
of internal consistency (standardized item a = .80). Thus, a composite aggression score
was computed. All participants completed this measure. The resulting scores
approximated closely the normal distribution. The average response was below the scale
midpoint (x = 3.25; SD = .99) and scores ranged from 1.27 to 6.40.

Predictions

Homicidal thought intercorrelations. Taking the research questions and
predictions in the order presented, the first two predictions involve reactive and proactive
thinking. Specifically, it was expected that reactive thinking and negative affect would
relate more strongly together than with any other dimensions of thought. Data were not
consistent with this expectation. The correlation between these dimensions was .74
(corrected r = .85). This was the strongest association among the dimensions for either
the reactive or negative affect dimensions. Ninety-five percent confidence intervals
computed with the remainder of the inventory, for both reactive thinking and negative
affect, resulted in consistent overlap. Thus, the strong, positive correlation between
reactive thinking and negative affect is not significantly different than the other
associations found between the inventory dimensions for either negative affect or reactive

thought. Similar findings were obtained for the hypothesis that proactive thought would

31

be most strongly related to positive affect (r = .67; corrected r = .80). This correlation
was not the largest in the group on these dimensions (see table 1).

In addition, it was expected that the media-related dimension of thought would be
substantially and positively correlated with both proactive and reactive inventory
dimensions. The data were consistent with both expectations. A tendency to think about
mediated homicide increased proportionally with a tendency to think proactively about
homicide (r = .76, p < .001; corrected r = .86) and also an increased disposition to
reactively contemplate homicide (r = .54, p < .001; corrected r = .70).

Finally, given previous findings suggesting that mass mediated models help
individuals form homicidal thoughts (Crabb, 2000), the media-related dimension of
thought was expected to be related positively to thought retrospective reports of
homicidal thinking. Data were consistent with this expectation (r = .23, p < .05;
corrected r = .37). In fact, only the positive affect factor on the inventory was not
significantly related to recalled homicidal thoughts.

Predicting the dimensions of homicidal thinking. To examine whether or not
homicidal thought could be predicted by media violence exposure multivariate regression
analysis was employed. Each dimension of thought was regressed onto sex, grade-point
average, general media exposure, empathy, aggression and media violence exposure
simultaneously. In all cases presented next, trivial predictors were removed and an
abridged model was retested. Together, these equations examine the research questions
involving media violence exposure, empathy, and aggression (see table 3).

The indicators of reactive homicidal thinking were completed by all participants.

Scores ranged from nine to 41 with a mean of 19.02 (SD = 6.90). The distribution of

32

scores approximated closely the normal distribution, albeit with a slight positive skew.
The reliability of this nine item index was estimated by Cronbach’s a and found to be .86
(standardized item a = .87).

When reactive homicidalthinking was regressed on sex, grade point average,
general media exposure, media violence exposure, aggression, and empathy, two
substantial predictors, aggression and exposure to media violence, emerged. Inspection of
scatterplots of the regression of reactive homicidal thoughts on each of these predictors
produced no indication of non-linearity. Those with higher aggression reported a
proportionally higher propensity for reactive homicidal thoughts (,8 = .35, t (179) = 4.95,
p < .001). Similarly, those with more exposure to media violence reported a
proportionally higher propensity for reactive homicidal thoughts (,6 = .20, t (179) = 2.43,
p < .05). The multiple correlation of these six predictors with reactive homicidal droughts
was .48 (.45 corrected for shrinkage, F (6, 179) = 8.80, p < .001). When the extraneous
predictors were removed from the regression equation no predictive ability was lost. The
effect for aggression remained substantial and in the same direction (,6 = .37, t (185) =
5.56, p < .001), as did the effect for exposure to media violence (3 = .22, t (185) = 3.27, p
=.001). The multiple correlation coefficient was .47 (.46 corrected for shrinkage, F (2,
185) = 26.43,p < .001).

The second dimension assessed the predisposition to think about homicide
proactively. All participants completed this 14-item composite. Scores ranged from 15
to 75 with a mean of 24.50 (SD = 9.74). The distribution approximated normality but did
feature a slight positive skew. Reliability analysis provided evidence of high internal

consistency among these items (Cronbach’s a = .91; standardized item a = .92).

33

Two predictors emerged when this factor was regressed onto age, sex, grade-point
average, aggression, empathy and media violence exposure (heretofore, the full model).
As before, the scatterplots for the non-trivial predictors appeared linear. Increases in
aggression were associated with proportional increases in proactive homicidal thought ([3
= .25, t (179) = 3.77, p < .001). A substantial sex association was observed indicating
that men were more likely than women to report a predisposition to think proactively
about homicide (B = -.34, t (179) = -4.24, p < .001). The multiple correlation from this
full model was .56 (.53 corrected for shrinkage, F (6,179) = 13.45, p < .00). Removing
the extraneous predictors resulted in an abridged model that was equally accurate in
predicting the criterion (R = .55 or .54 corrected for shrinkage, F (2,192) = 41.01, p <
.00). By removing age, sex, empathy, and media violence exposure the relationship
between the two non-trivial predictors remained relatively unchanged. Aggression
remained stable (3 = .25, t (192) = 4.00, p < .001), and sex improved its predictive ability
slightly (,8 = -.42, t (192) = -6.63, p < .001).

A propensity to have media-related thoughts about homicide was estimated by
forming a composite score from 10 items. All 195 respondents completed this measure.
Scores on this variable ranged from 11 to 55. This distribution was slightly skewed
positively, influenced by one participant who responded “strongly agree” to all items.
Still, the distribution closely approximated normality and this high score was included in
analyses. The mean of this dimension was 23.70 with a standard deviation of 8.42.
Cronbach’s and standardized item a were both equal to .86.

When this composite was regressed onto the full model the six predictors

produced an ample multiple correlation of .57 (R = .55 corrected for shrinkage, F (6, 188)

34

= 15.30, p < .001). Three variables appeared non-trivial in the prediction of media-
related homicidal thought. As both aggression (,8 = .17, t (188) = 2.71, p < .01) and
media violence exposure (,6 = .22, t (188) = 2.87, p < .01) increased this dimension of
homicidal thinking increased proportionally. Additionally, men reported an increased
tendency to have this type of thought compared to women ([3 = -.38, t (188) = -4.79, p <
.001). Revising the model to include only meaningful predictor variables did not alter
these results substantially. Aggression (,8 = .17, t (191) = 2.67, p < .01), media violence
exposure (fl = .22, t (191) = 2.88,p < .01), and sex (,6 = -.35, t(191) = -.46,p < .001) all
retained their predictive ability. In addition, the multiple correlation for these three
predictors was .57 (.55 corrected for shrinkage, F (3,191) = 29.84, p < .001).

A fourth factor analyzed was the tendency to ruminate about killing under
conditions of negative affect. This dimension of thought was estimated with six items.
Cronbach’s and standardized item a both were .90, providing ample evidence of internal
consistency. The mean of this distribution was 10.74 with scores ranging from 6 to 30.
The distribution was slightly platykurtic and again appeared to skew somewhat positively
relative to the standard normal distribution. On the other hand, a relatively high
proportion of participants reported the minimum value for this dimension. Almost 40%
(n = 76) reported a total score of 6 for this factor.

The. regression equation for the full model yielded a multiple correlation of .40 (R
= .37 corrected for shrinkage, F(6,188) = 6.16, p < .001). Aggression and violence
exposure emerged as non-trivial predictors in this model. Increases in both aggression (,B
= .29, t (188) = 3.96, p < .001) and media violence exposure (3 = .17, t (188) = 1.97, p =

.05) were associated with an increased predisposition to think about killing under

35

conditions of negative affect. The scatterplot data suggested that both of these
relationships were linear. The regression equation was recomputed by removing the
extraneous variables general exposure, grade-point average, empathy, and sex from the
equation. This revised model did not lose predictive power by a meaningful amount (R =
.38, or .37 corrected for shrinkage, F (2,192) = 16.42, p < .001). Aggression appeared
similar in association (,8 = .29, t (192) = 4.15, p < .001) and media violence exposure
predicted this type of thought tendency somewhat more strongly (fl = .20, t(192) = 2.85,
p < .01 ).

The counterpart to the previous dimension, positive affective thought, was
assessed by using 6 items which again demonstrated some evidence of reliability
(Cronbach’s a = .80; standardized item a = .85). All participants responded to these
items. The range and distribution shape of these scores was similar to its counterpart
dimension. A full 42% of the sample “strongly disagreed” that they ever think about
killing in a positive affective state. One participant reported the maximum score of 30.
This score contributed to the distribution’s positive skew. Its mean was 9.00 with a
standard deviation of 3.93.

When positive affect was regressed onto the full model only one predictor
emerged clearly as non-trivial. The full model multiple correlation was .39 (R = .35
corrected for shrinkage, F (6,188) = 5.65, p < .001). However, sex appeared to be the
only substantial predictor (,6 = -.29, t (188) = -3.16, p < .01) of thinking about killing
under conditions of positive affect. Regressing this dimension onto sex produced a

comparable solution. This univariate model produced a similar multiple correlation (,8 = -

36

.38, t (193) = -5.60, p < .001). It appears that men are more predisposed to thinking
about killing while in a positive emotional state.

Finally, two dimensions were composed in order to estimate the frequency of
homicidal thought. A tendency to have repeated homicidal thoughts was estimated using
six items, however two of these items decreased the estimate of this variable’s reliability
(Cronbach’s a = .52; standardized item a = .69). These two items were subsequently
judged to be faulty due to wording ambiguity. Thus, on the basis of their low inter-item
correlations and their lack of clarity, these items were removed from analyses. The
resulting four items formed this composite with an improved Cronbach’s a = .93
(standardized item a was equivalent). This variable was again distributed with a positive
skew. Scores ranged from one to five on this index with a mean of 1.37 and a standard
deviation of .67.

When this outcome variable was regressed onto the full model two non-trivial
predictors emerged. The multiple correlation in this model was .32 (R = .27 corrected for
shrinkage, F(6,188) = 3.66,p < .001). Aggression (fl = .14, t (188) = 1.89, p = .06) and
sex (15’ = -.21, t (188) = -2.17, p =.03) were again substantial predictors of this thought
dimension. Aggression increases were associated with increases in the tendency to
ruminate repetitively about homicide, and men were more prone than women to report
such a tendency. The abridged model was retested with only these two variables. The
relationships did not change substantially. R was .32 (.30 corrected for shrinkage,
F(2,192) = 10.73, p < .001), and both aggression (B = .13, t (192) = 1.86, p = .07) and sex

were relatively unchanged (,6 = -.26, t (192) = -3.54, p < .001) in their predictive ability.

37

The second estimate of frequency, independent thought, was derived by averaging
scores from six items. As with repetitive thoughts, the internal consistency of this
measure was attenuated by one problematic statement that correlated negatively with all
but one of the other five items. Removal of this indicator resulted in a more homogenous
set of items with Cronbach’s a = .75 (standardized item a = .83). This variable’s
distribution approximated normality, albeit with a slight positive skew. The mean was
1.83 with a standard deviation of .76.

When independent thought was regressed onto the full model, aggression and sex
again appeared to be non-trivial predictors in the equation. The six variables produced a
multiple correlation of .53 (R = .51 when corrected for shrinkage, F (6,188) = 12.11, p <
.001). Increases in aggression were associated positively with increases in the tendency
to have multiple or different homicidal thoughts (,6 = .30, t (188) = 4.44, p <.001). The
other non-trivial predictor was sex. Again, men appear more likely than women to report

a predisposition to think about homicide (,6 = -.30, t (188) = -3.60, p <.001). Inspection
of scatterplots indicated no evidence of non-linearity in these data.

By removing the extraneous variables in the equation, the results were robust.
The strength of the association between aggression and independent thought remained
stable (,8 = .29, (t (192) = 4.50, p < .001). The overall predictive ability of the equation
also remained stable (R = .52 or .51 corrected for shrinkage, F (2,192) = 35.41, p < .001).
Notably, the relationship between sex and this type of thought predisposition
strengthened slightly (8 = -.35, t (192) = -5.53,p <.001).

Finally, recall that participants were asked to list and write in some detail up to

three homicidal thoughts from memory. As reported previously, 70.3% of the sample

38

could not recall ever having such thoughts. Of those writing thoughts, 21.5% of
participants detailed one thought, 7.2% wrote two, and one percent (11 = 2) could
remember three. This distribution was positively skewed and leptokurtic. The mean was
.40 with a standard deviation of .67.

Regressing this variable onto all six predictors resulted an equation that produced
a small multiple correlation of .23 (R = .16 corrected for shrinkage, F (6,188) = 1.80, p =
.10). Only one of the six variables emerged as a non-trivial criterion. Increases in media
violence exposure were associated with a proportional increase in the ability to recall
having a homicidal thought (,8 = .17, (t (188) = 2.00, p = .05). By removing the trivial

predictors, this relationship strengthened slightly (,6 = .21, (t (193) = 3.00, p < .01).
DISCUSSION

Overall, the aim of this study was to replicate and extend the Boyson et a1. (2004)
investigation. The results of this survey replicate some of the results from previous
research and extend them by gathering data using two different measures of homicidal
thinking — a retrospective report measure and a scaled dispositional measure. Using an
open-ended measure, 30% of the participants reported one or more homicidal thoughts,
which is only slightly smaller than the Boyson et a1. study (33%). Additionally, the
finding that aggression was unrelated to homicidal thought was replicated as well.
However, this was only found with the same open-ended measure used in previous

research (retrospective reports).

39

Interestingly, scores on the Buss and Perry (1992) measure of aggressiveness
predicted increases in six of the seven dimensions of a new inventory developed in this
study to measure trait-like homicidal thought. Compared to all other predictors,
aggression was the strongest and most consistent predictor of the new trait-like measure
of homicidal thought. The strength of associations and the pattern of findings across

multiple measurements of homicidal thought suggest that this finding is robust.

t-JJ I" -

It appears that an aggressive predisposition is related positively to a predisposition
to think about killing, as one might suspect. The discrepancy between the open-ended
measure and the inventory tapping a predisposition for homicidal thinking is consistent
with the idea that these two formats may be tapping different constructs. Retrospective
reports seem to be measuring situations rather than dispositions. More work is needed to
examine a broader network of associations to learn more about the possibility that these
are not merely alternate indicators of the same construct.

For empathy, the results from the Boyson et a1. (2004) study were not replicated.
As indicated earlier, Boyson et a1. (2004) found that increases in empathy were
associated with increases in homicidal thought. In this study, however, an increase in
empathy was not associated with an increase in either retrospective reports or trait-like
thoughts about killing. It is possible the previous results were explained by sex
differences, given the disproportionate number of females in the Boyson et a1. (2004)
sample. To examine this potential more closely, all regressions involving empathy were
rerun by splitting the file by sex. Empathy did not predict either retrospective reports or
trait-like homicidal thoughts for men or women in this study. This null finding requires

some attention.

40

As mentioned, empathy was included previously because of its potential to
negatively predict aggressive thought in general, and homicidal thought in specific. A
null relation still is somewhat unexpected. In this sample, a negative but weak
relationship was found between aggression and empathy (r = - .12, p = .16). These two
characteristics might be expected to relate inversely more strongly than they did in this
sample.

Relations between each scale’s factors qualify this finding. Personal distress was
positively related to the hostility and anger factors on the aggression scale. This overlap
with an aggressive orientation may be clouding the overall negative relationship empathy
has on this type of thought. In Davis’ previous work (1983) with this empathy measure,
personal distress was found to be associated positively with a “specific emotional
constellation of vulnerability, uncertainty, and fearfulness” (p.121). Personal distress
may be a proxy for a type of emotional regulation (or lack thereof) that promotes
sensitivity to threat or harm. Under certain conditions, this sensitivity could even lead to
more homicidal thinking.

Perhaps emotional vulnerability increases a fearful orientation in social contexts
that in turn promotes homicidal thinking. In fact, emotional susceptibility of this sort has
been studied in the context of aggressive behavior (Caprara & Pastorelli, 1989). This
type of rumination may be a hallmark of only some empathic people and may serve some
type of coping function. At present, it appears that an association exists between
emotional susceptibility and a tendency to ruminate about transgressions in general.
Further work is needed to clarify the relations between empathic tendencies and

homicidal thought.

41

In addition to replicating previous research, this study was designed to extend
previous work by advancing a conceptual framework for homicidal thinking that is driven
by research and theory in aggressive cognitions and imagined interactions. This model
was designed to estimate an individual’s predispositional capacity to think about
homicide by measuring seven dimensions of thought derived from cognitive, affective
and frequency content domains. Although much more work is needed before conclusions
can be drawnabout the validity of this instrument, data were consistent with a seven-
factor solution. All seven factors were strongly, but heterogeneously, related. The
expectation that reactive thinking and negative affect would be uniquely related to each
other was not supported by these data. Similarly, no unique association was found
between proactive and positive affect thought.

At present, one can only speculate about why these profiles did not emerge. It is
possible that the act of having one homicidal thought encourages others. Perhaps a
single, triggering event makes future homicidal cognitions more likely. This would be
true to the degree that homicidal thoughts are rewarding or gratifying. The strong
positive association between proactive and reactive thoughts is consistent with this idea.

Looking across the entire inventory, only one of the dimensions failed to
significantly predict retrospective homicidal thoughts. Positive affect was unrelated to
retrospective thoughts. This finding is consistent with a bias in remembering negative
events (Brown & Kulik, 1977). Pleasantness may negatively facilitate recall of imagined
interactions (Honeycutt, Zagaki, & Edwards, 1992). Perhaps recall of homicidal thoughts
is hampered likewise by positive affect, which could explain why the vast majority of

retrospective thoughts come from negative events. Some evidence suggests that negative

42

stimuli can affect long-term memory better than neutral stimuli (Kensinger & Corkin,
2003). In this study, the positive affect dimension was conceptualized to include both
neutral and positive emotions. This fact could explain the lack of association between
retrospective thoughts. An evolutionary perspective also preferences memory for
negative events because it is assumed that survival largely has depended on an emotional
system that allows for quick retrieval of threatening or disturbing stimuli (Christianson &
Engelberg, 1999).

Generally speaking, however, these results provide some criterion-related
validity. A tendency to think about homicide should be positively related to actual
reports of thoughts themselves. This evidence represents a promising start in building a
measure that more closely fits the theoretical perspective involving homicidal ideation.
Evolutionary psychologists have advanced the argument that a biologically based
tendency toward homicide exists and, as a result, a tendency toward ruminating about the
act. Although this tendency appears measurable, the inventory has several notable
conceptual limitations that may provide some avenues for future study.

First, affect was conceived as two conceptually distinct dimensions. This
distinction is made possible by considering how a person feels before having a homicidal
thought. These dimensions do not examine the valence of emotion following a homicidal
thought. Doing so would more closely match Honeycutt’s (2003) viewpoint of imagined
interactions. Reports suggest that imagining social encounters can produce positive
affect (Kroll-Mensing, 1992). However, when conflict is measured, others have found
that negative affect is more likely to result (Honeycutt et al., 1992). More detailed

descriptive analysis of homicidal thoughts, the conditions that produce them, and how

43

individuals feel about their thoughts after having them could be used to extend the
present conceptualization of homicidal thinking.

A second limitation is related. As mentioned, the act of composing a homicidal
thought may be an important c0ping mechanism, perhaps as a way to reduce anxiety or
uncertainty. Beck’s (1999) findings that some individuals are prone to converting hurtful
feelings to hateful ones as a way to manage self-esteem lend further credence to this
possibility. From either a cognitive or emotional standpoint, the present
conceptualization does not allow for the possibility that people cope with perceptually
severe social exchanges by thinking about killing. It even may be possible that an
imagined homicidal interaction makes one less likely to behave aggressively — either in
terms of homicide or general aggressive behavior. For example, Kroll-Mensing (1992)
also found that imagined interactions were helpful in resolving conflict. The present
inventory is better suited to estimate the motivations or uses for homicidal thinking.
Additional work needs to be done to examine the gratifications received fiom this type of
thought. Research related to imagined interactions clearly supports the idea that
gratifications play a key role in determining the frequency and type of thoughts people
have (Honeycutt, 2003).

A third aim of this investigation was to examine more closely the relationship
between media violence and homicidal thought. Media violence exposure was
considered holistically, summing across content consumed from movies, television, video
games, and music. An attempt was made to account for differing perceptions of violent
content by including violence ratings from both the participant’s and the researcher’s

perspectives. As in Crabb (2000) and Boyson et al. (2004), the pattern of these data

44

suggests that media violence exposure may play an important role in homicidal drinking.
The frequency of retrospective reports of homicidal thoughts was again predicted by
violence exposure. In fact, it was the only non-trivial predictor retained from the full
model (i.e., aggression, sex, grade-point average, empathy and general exposure were
trivial).

This relationship between violence exposure and thinking about killing is less
clear when considering results from the trait-like measure. As expected, a predisposition
to think about media-related homicide was positively related to exposure to media
violence. But overall, media violence was retained in only four of the seven dimensions
of thought. The clearest pattern in the data suggests that violence exposure is related to
reactive thoughts, under conditions of negative affect, and with increases in a tendency to
think about homicide in the mass media. To explore the influence sex may have on these
relationships, the file was split and partial correlations were computing controlling for the
only other variable that appears to play a key role (i.e., aggression; see table 4). It
appears that the pattern indicative of reactive, negative, and media-related thoughts is
even stronger when considering only females. In fact, associations between media
violence and homicidal thought remained significant only for women, and not for men. If
media violence exposure is an antecedent of homicidal thinking, these data suggest that
women may be attending to or processing violence in the media quite differently than
there male counterparts.

At least three limitations in the present design are noteworthy. First, exposure to
media violence was measured in terms of a “drip” orientation rather than from a “drench”

perspective (Greenberg, 1988). The drench hypothesis posits that the same dosage of

45

media images may impact people quite differently. Dramatic portrayals of homicide may
be highly impressionable images for some people. Given their severity, and often
graphic nature (particularly in movies), this hypothesis seems plausible. The potential
effect of one or two high impact scenarios involving killing on homicidal thought is
unaddressed in this study.

Second, homicide exposure in specific has not been measured well. In this study,
participants were asked to retrieve from memory homicide exemplars from the mass
media. This measure was problematic. It did not relate well to overall media exposure or
media violence exposure in specific but this may be an artifact of the way media violence
was measured. Retrieving and writing down homicide exemplars requires high
involvement on the part of the participant. This measure may best be obtained separately,
when students are not fatigued from answering other questions. Several of the students
interviewed following the survey made statements suggesting the instrument was too
long or took too much time to complete.

To be sure, most people have little or no direct experience with homicide,
composing a thought involving murder seems most likely to be derived from comparable
scenes involving similar outcomes. This possibility is more in line with the drench
hypothesis. General media violence exposure over time could be impacting homicidal
thinking by developing more well-developed schemata involving all forms of aggression
which could facilitate the construction of homicidal thoughts indirectly. The present
results are consistent with this possibility, but measures of exposure that account for
identification and involvement with characters and content in the mass media may also be

important.

46

Third, despite the assumptions in the regression models used here, no statements
can yet be made about the causal ordering of media violence and homicidal thinking. It
is plausible that a tendency to think about homicide could predispose an individual to
seek out homicidal scenes and images so often presented in the mass media. No studies
speak directly to this possibility, including this one. Indeed, research in the area of media
violence and aggression suggests that the link between media violence and aggression
may be bi-directional (Eron, Huesmann, Lefltowitz, & Walder, 1972; Fenigstein, 1979).
Given the evolutionary perspective that has been invoked to explain this type of thought,
the likelihood that media violence exposure is a consequent of homicidal thought
tendency is even more plausible. How these two variables are related causally deserves
more attention.

To date, homicidal thought has received very little empirical attention. Taken as a
whole, evidence is mounting that suggests homicidal thinkers are not hard to find, at least
among undergraduate students. That said, however, this small body of research has relied
on non-probability samples from a population that may well be more “psychologically
healthy” at least in terms of aggressive cognition and behavior. The findings presented
here suggest that many people may be predisposed to homicidal thought but cannot
remember having them well enough to write them down in open-ended form. Drawing
conclusions about the normality of homicidal thought should be tempered. Homicidal
ideation could either be less or more common in broader populations, depending on the
psychological function it serves.

Several questions stemming from the idea of normality are relevant to this

literature. What distinguishes the homicidal ideation of psychiatrically normal and non-

47

normal patients? Empirical analyses of clinically diagnosed patients seem to suggest that
homicidal thinking is just as common as it is for undiagnosed people. The inventory
presented here could be used to distinguish those who fixate on one or a set of thoughts
about homicide and to explore other factors that may play a role in this fixation, such as
media-related thoughts about homicide.

Perhaps the primary question yet to be answered is whether or not homicidal
thoughts facilitate homicidal behavior. Fitting homicidal thinking into process models of
aggressive behavior could help answer this question. These models currently suggest that
aggressive thoughts mediate aggressive behavior. This is a somewhat troubling
possibility, given the frequency with which people seem to be having homicidal thoughts.
Correlates and consequences of homicidal thinking may yet reveal protective and risk
factors associated with this type of thought, including whether or not media violence

exposure enhances, attenuates, or simply alters the nature of homicidal thinking.

FOOTNOTES

1 In other words, no evidence is expected showing that each dimension is tapping the same broader

concept. At present, the inventory is not expected to examine a unitary latent construct. While there is
some evidence guiding the formation of each dimension there is little reason to expect them to relate with

one another in a second-ordered factor structure.
2 For example, a person listing 3 favorite TV shows would have their violence rating averaged for each

show listed. To take into account missing data, the average violence score for TV would be multiplied by
three. This procedure was done for each medium. To get an overall media violence exposure score, the

medium scores were also averaged. This procedure was followed for both subjective (participant) and

48

objective (coder) ratings with one exception. Because two independent coders were used, the average of
their violence ratings was used to assign a value to each media selection listed. For example, if coder l’s
TV ratings were 2, 3, 4, l and coder 2’s TV ratings were 2, 2, 2, l, the score would be computed by

(2+3+4+l/4) + (2+2+2+l/4)/2.
3 This definition comes from the National Television Violence Study (see, Smith et al., 1998). That

definition focuses on a credible threat of physical harm by one animate being to another.

49

REFERENCES

Anderson, CA. (1997). Effects of violent movies and trait irritability on hostile feelings
and aggressive thoughts. Aggressive Behavior, 23, 161-178.

Anderson, K. B., Anderson, C. A., Dill, K. E. & Deuser, W. E. (1998). The interactive

relations between trait hostility, pain, and aggressive thoughts. Aggressive
Behavior. 24 (3), 161-171.

Anderson, CA. & Huesmann, R. (2003). Human aggression: A social cognitive view.

In M.A. Hogg & J. Cooper (Eds.) Handbook of Social Psychology. (pp. 296-323)
London: Sage Publications

Altman, H, Sletten, I. W., Eaton, M. E., & Ulett, G. A. (1971). Demographic and mental
status profiles: Patients with homicidal, assaultive, suicidal, persecutory and
homosexual ideation. Psychiatric Quarterly, 45(1), 57-64.

Bandura, A. (2001 ). Social cognitive theory of mass communication. Media Psychology,
3(3), 265-299.

Bandura A., Caprara, G. V., Barbaranelli, C., Pastorelli, C., Regalia, C. (2001).
Sociocognitive self-regulatory mechanisms governing transgressive behavior.
Journal of Personality and Social Psychology, 80(1), 125- 1 35.

Baumeister, R. F., Smart, L., & Boden, J. M. (1996). Relation of threatened egotism to
violence and aggression: The dark side of self-esteem. Psychological Review,
103(1), 5-33.

Berkowitz, L. (1984). Some effects of thoughts on anti- and prosocial influences of media
events: A cognitive-neoassociation analysis. Journal of Psychological Bulletin,

95(3), 410-427.

Berkowitz, L. (1993). Aggression: Its Causes, Consequences, and Control. New York:
McGraw-Hill

Berkowitz, L. (2000). Causes and Consequences of Feelings. New York, NY: Cambridge
University Press

Boyson, A.R., Smith, S.L., Rosaen, S. (2004). Thinking about killing: Is exposure to
media violence related to homicidal fantasizing? Manuscript submitted for
publication.

Bower, G. H. (1981). Mood and memory. American Psychologist, 36(2), 129-148.

Brown, R., & Kulik, K.(1977). F lashbulb memories. Cognition, 5, 73-99.

50

Bushman, B. J., (1995). Moderating role of trait aggressiveness in the effects of violent
media on aggression. Journal of Personality and Social Psychology, 69, 950-960. '

Buss, D. M., & Perry, M. (1992). The aggression questionnaire. Journal of Personality
and Social Psychology, 63(3), 452-459.

Caprara G. V., & Pastorelli, C. (1989). Toward a reorientation of research on aggression.
European Journal of Personality, 3, 121-138.

Cohen, D., & Strayer J. (1996). Empathy in conduct-disordered and comparison youth.
Developmental Psychology, 32(6), 988-998.

Condry, J .C. (1989). The Psychology of Television. Hillsdale, NJ: Erlbaum

Crabb, P. B., (2000). The material culture of homicidal fantasies. Aggressive Behavior,
26, 225-234.

Crick, N.R. & Dodge, K.A. (1994). A review and reformulation of social information
processing mechanisms in children’s social adjustment. Psychological Bulletin,
115(1), 74-101.

Davis, M. H. (1983). Measuring individual differences in empathy: Evidence for a
multidimensional approach. Journal of Personality and Social Psychology, 44(1),
113-126.

Diener, E. & De Four, D. (1978). Does television violence enhance programme
popularity? Journal of Personality and Social Psychology, 36, 333-341.

Diener, E. & Woody, L.W. (1981). Television violence, conflict, realism, action: A study
of viewer liking. Communication Research, 8, 281-306.

Dodge, K. A. (1985). Attributional bias in aggressive children. In P. C. Kendall (Vol.
Ed.), Advances in Cognitive-Behavioral Research and Therapy: Vol. 4 (pp.73-
110). London: Academic Press

Dodge, K. A. & Coie, J. D. (1987). Social-information-processing factors in reactive and
proactive aggression in children’s peer groups. Journal of Personality and Social
Psychology, 53 (6), 11469-1158.

Eisenberg, N., Fabes, R.A., Burphy, B., Karbon, M., Maszak, P., Smith, M., O’Boyle, C.,
& Suh, K. (1994). The relations of emotionality and regulation to dispositional

and situational empathy-related responding. Journal of Personality and Social
Psychology, 66 (4), 776-779.

51

Eron, L.D., Huesmann, LR, Lefltowitz, M.M., & Walder, L0. (1972). Does television
viewing cause aggression? American Psychologist, 27, 253-263.

F enigstein, A. (1979). Does aggression cause a preference for viewing media violence?
Journal of Personality and Social Psychology, 3 7, 2307-2317.

Greco, C. M. & Cornell, D. G. (1992). Rorschach object relations of adolescents who
committed homicide. Journal of Personality Assessment, 59(3), 574-583.
Greenberg, BS. (1988). Some uncommon television images and the drench
hypothesis. In S. Oskarnp (Ed.), Applied Social Psychology annual: Television as
a Social Issue (Vol. 8, pp. 88-102). Newbury, CA: Sage

Gunter, B. (1985). Dimensions of Television Violence. Aldershot, England: Gower
Publishing Company Limited

Gunter, B. & Furnham, A. (1984). Perceptions of television violence: Effects of
programme genre and type of violence on viewers’ judgments of violent
portrayals. British Journal of Social Psychology, 23, 155-164.

Honeycutt, J. M. (2003) Imagined Interactions: Daydreaming about Communication.
Creskill, NJ: Hampton Press

Huesmann, L. R. (1998). The role of social information processing and cognitive schema
in the acquisition and maintenance of habitual aggressive behavior. In R. G. Geen
& E. Donnerstein (Eds.), Human Aggression (pp. 73-110). San Diego, CA:
Academic Press

Huston, A. C., Donnerstein, E., Fairchild, H., F eshbach, N. D., Katz, P. A, Murray, J. P.,
Rubinstein, E. A.,Wilcox, B. L., & Zuckerman, D. (1992). Big World, Small
Screen: The Role of Television in American Society. Lincoln, NE: University of
Nebraska Press

Kenrick, D. T., & Sheets, V. (1993). Homicidal fantasies. Ethology and Sociobiology,
14, 231-246.

Korn, M. L., Plutchik, R., & Van Praag, H. M. (1997). Panic-associated suicidal and

aggressive ideation and behavior. Journal of Psychiatric Research, 31(4), 481 -
487.

Kroll-Mensing, D. (1992). Differentiating anxiety and depression: An experience
sampling analysis. Unpublished doctoral dissertation, University of Minnesota,

Minneapolis

Miller, P. & Eisenberg, N. (1988). The relation of empathy to aggressive and
extemalizing/antisocial behavior. Psychological Bulletin, 102, 324-344.

52

Mischel, W. (1968). Personality and Assessment. New York: Wiley
Moyer, K. (1976). The Psychobiology of A ggression. New York: Harper & Row

Potter, W.J. & Smith, S.L. (2000). The context of graphic portrayals of television
violence. Journal of Broadcasting & Electronic Media, 44(2), 301-323.

Rachman, S. & de Silva, P. (1978). Abnormal and normal obsessions. Behavioral
Research & Therapy, 16, 233-248.

Rubin, AM. (2002). The uses and gratifications perspective of media effects. In J.
Bryant & D. Zillman (Eds.) Media Effects: Advances in Theory and Research.
Mahwah, NJ: Erlbaum

Schwartz, R. C., Petersen, S., & Skaggs, J. L. (2001). Predictors of homicidal ideation
and intent in schizophrenia: An empirical study. American Journal of
Orthopsychiatry, 71(3), 379-3 84.

Smith, S. L., Wilson, B. J., Kunkel, D., Linz, D., Potter, W. J., Colvin, C. M., &
Donnerstein, E. (1998). Violence in television programming overall: University
of California, Santa Barbara study. In National Television Violence Study (Vol.
3; pp. 5-220). Thousand Oaks, CA: Sage

Steyer, R., Schmitt, M., & Eid, M. (1999). Latent state-trait theory and research in

personality and individual differences. European Journal of Personality 13, 389-
408.

Webster, G. J ., & Wakshlag, J. (1985). Measuring exposure to television. In D. Zillman
& J. Bryant (Eds.), Selective Exposure to Communication (pp. 35-62). Hillsdale,
NJ: Erlbaum

Wisner, K. L., Peindl, K, & Hanusa, B. H. (1994). Symptomatology of affective and
psychotic illness related to childbearing. Journal of A fi’ective Disorders, 30, 77-
87.

Zagacki, K.S., Edwards, R., & Honeycutt, J .M. (1992). The role of mental imagery and
emotion in imagined interaction. Communication Quarterly, 40, 56-68.

53

APPENDIX

54

Figure 1 - The Proposed Structure of a Predisposition to Think about Homicide

   

Proactive
Thought
Cognitive
Domain -—-> Reactive
Thought

Media-
related
Thought

Positive
Affective
Thought

A flective
Domain

Negative
Affective
Thought

 

Repetitive
Thought

Frequency

 

Domain

Independent
Thought

 

55

Figure 2 — An Inventory of Homicidal Fantasizing Capacity

Reactive Thought
1. Given enough aggravation, I may think about killing someone.
2. I can think of no situation that warrants a homicidal thought.
3. lfsomeone threatens me, I may think about killing that person.
4. I would have no trouble thinking about killing a child molester.
5. When I see someone else attacked, I may think about killing the person who hurt them.
6. If someone were to humiliate me in public, I may have a homicidal fantasy.
7. If I were to have a fight with a romantic partner, I might have a homicidal fantasy.
8. If a boss at work were to reprimand me unfairly, I may have a homicidal fantasy.
9. I can imagine having a family argument that might result in a homicidal fantasy.

Proactive Thought

10.
ll.
12.
l3.
I4.
15.
l6.
I7.
18.
I9.
20.
21.
22.
23.
24.

I may think about killing another person just to pass the time.

When I’m alone with my thoughts, I may conjure a homicidal thought.

I would only have homicidal thoughts if someone threatened me.

I admit that it would be interesting to know how it feels to kill someone.

I could fantasize about killing someone for no good reason.

I would have no trouble thinking about killing a stupid person.

I would love to work for the CIA or the FBI, even if it meant I would have to kill someone.
Thinking about killing is cathartic for me.

As I daydream, I may fantasize about killing someone.

I sometimes wonder what it would be like to be a hit man.

Fantasies about killing another person in battle, or war, are interesting to me.
I have imagined killing someone as a way to get what I want.

I could have a homicidal fantasy about killing someone over money.

When I let my imagination run, I could have homicidal fantasy.

If someone paid me enough, I could be an assassin.

Media-related thought

25.
26.
27.

28.
29.
30.
31.
32.

When I’m at the movies, I have fantasized about being a killer.

I am attracted to media characters, like James Bond, who sometimes have to kill for a living.
When I see someone commit murder on television, I sometimes imagine what it might be like to
do that.

I don’t mind listening to music that talks about killing.

I think it would be fun to play video games that involve killing.

1 have a tendency to replay in my mind TV or movie scenes involving death.

My thoughts about killing are informed by things I’ve seen in the mass media.

I have no trouble thinking about movie or TV scenes involving killing.

56

33.
34.
35.

I sometimes find myself imagining what it would be like to kill someone like I see on television.
Of all types of media characters, I identify most with action-adventure heroes.

1 have pretended to kill people like I see on television or in the movies.

Negative affect thought

36.
37.
38.
39.
40.
41.

I have been so mad that I could think about killing someone.

The rage I feel sometimes could cause me to think about killing someone.
I get so stressed out, sometimes, homicide sometimes crosses my mind.
At times, I have been so scared that I had a homicidal thought.

I can imagine becoming angry enough to have a homicidal fantasy.

I have been fearful enough to imagine killing someone.

Positive affect thought

42.
43.
44.
45.
46.
47.

My homicidal fantasies are enjoyable to me.

I don’t find my homicidal fantasies disturbing.

I don’t just have homicidal fantasies when I’m mad or angry.

I rather like having a homicidal fantasy.

The idea of having a homicidal fantasy sounds kind of thrilling to me.

I become aroused when I imagine killing another person.

Repetitive thought

48.
49.
50.
51.
52.
53.

l have had a recurring fantasy about killing the same person.

When I have had a homicidal thought I tend NOT to dwell on it.

I eon remember obsessing about killing a person for days.

Sometimes my mind replays the same homicidal fantasy over and over.
My thoughts about killing have involved just one or two people.

Under some circumstances, I may contemplate killing someone for days.

Independent thought

54.
55.
56.
57.
58.
59.

I am pretty sure that I’ve had homicidal thoughts, but I can’t remember any of them right now.
I sometimes have homicidal fantasies.

I have all kinds ofimaginary thoughts about killing.

1 have homicidal fantasies quite often.

I can’t remember ever having a homicidal thought.

I could have many, different fantasies about killing.

57

Table 1

Intercorrelations Between Indices of Homicidal Thinking

 

Dimension 1 2 3 4 5 6 7 8
(n = 186)

1. Thought Recollections - .43*** .30*** .23“ .39*** 0.11 .31*** .46***
(Mean=.40, SD=.68)

2, Reactive Factor - .67*** 54*“ .74*** 47*“ .56*** .64***
(Mean=l9.26, SD=6.9I) '

3. Proactive Factor - .76*** .58*** .67*** .57*** .68***
(Mean=24.82, SD=9.85)

4. Media Related Factor - .55*** .54*** .46*** .59***
(Mean=24.02, SD=8.46)

5. Negative Affective Factor - .47*** .63*** .66***
(Mean=10.93, SD=5.73)

6. Positive Affective Factor - .53*** .53***
(Mean=9.05, SD=3.95)

7. Repetitive Thought - .61 Ht

(Mean= l .36, SD=.64)
8. Independent Thought

(Mean= l .86, SD=.76)

*p< .10, **p< .05, "*p< .001

58

Table 2

Correlations Between Predictor Variables

 

Scale 1 2 3 4 5 6
(n = 186)

1. General Media Exposure - -0.01 .21 * * -0. 12 0.04 0.11

(Mean = 404.55; SD = 266.57)

2. Empathy - -0.12 .33*** -0.04 -.l9**

(Mean = 4.19; so = .63)

3. Aggression - -.28*** 0.02 .23**

(Mean = 3.25; SD = .99)

4. Sex - 0.03 -.59***

(60% female)

5. Overall Grade Point Average - 0.01

(Mean = 2.97; SD = .46)
6. Media Violence Exposure

(Mean = 2.89; SD = .60)

 

*p< .10, **p< .05, "*p< .001 Sex coded as l = male, 2 = female. In all other cases, higher scores

indicate greater amounts of variables.

59

Table 3

Summary of Non-trivial predictors of Homicidal Thought

 

 

 

 

 

 

 

 

Variable B SE B ,8
Reactive Thought (df = 185)
Aggression 2.61 .47 .3 7* * *
Violence Exposure 2.55 0.78 .22* *
Proactive Thought (df = 191)

Aggression 2.47 0.62 -.25 * * *

Sex -8.34 1.26 -.42***
Media-Related Thought (df = 183)

Agression 1.44 0.54 0.17M

Sex -6.02 1.31 -.35***

Violence Exposure 3.06 1.7 .22“

Negative Affect Thought (df = 184)

Aggression 1.67 i .40 .29"‘ * * .

Violence Exposure 1.9 .67 .20* *
Positive Affect Thought (df =187)

Sex -3.03 0.54 -.38*

Repetitive Thought (df = 187)
Sex -0.35 0.1 -.26***
Aggression .09 .05 .13*
Independent Thought (df = 190)

Sex -0.56 0.1 -.36***

Aggression 0.22 0.05 .30* * *
Thought Recollections (df = 186)

Violence Exposure 0.24 0.08 .22* *

 

‘p < .10, "p < .05, "*p < .001 Sex coded as l = male, 2 = female. In all other cases,

higher scores indicate greater amounts of variable.

60

Table 4

Partial Correlations Between Media Violence & Homicidal Thought Measures, Controlling for Aggression

Reactive Thought

Proactive Thought

Media-Related Thought

Negative Affect Thought

Positive Affect Thought

Independent Thought

Repetitive Thought

Whole Sample

.18; N=l83;p=.02

.09; N=I83; p=.24

.21; N=l83;p=.01

.14; N=l82; p=.05

.06; N=l78; p=.45

.09; N=l82; p=.21

.06; N=l79; p=.40

Media Violence Exposure

Just Men

.08; N=7l;p=.51

.04; N=7l ; p=.74

.12; N=7l ; p=.37

.01; N=7l;p=.93

.12; N=7l;p=.3l

.13; N=7l ; p=.29

.001; N=7l ; p=.99

Just Women

.24;N=110;p=.01

.15; N=110;p=.ll

.31; N=110;p=.001

.24; N=109; p=.Ol

.01; N=105; p=.93

.05; N=109;p=.59

.l3; N=106;p=.l8

 

6l