 

 

 

.2 of? LIBRARY

Michigan State
lnivereity

r“-

This is to certiw that the
thesis entitled

 

SHIFTS IN ACCEPTANCE OF TYPE-SPECIFIC VIOLENCE AS AN
OUTCOME OF CHARACTER-ROLE IN VIDEO GAMES

presented by
Matthew N. Grizzard

has been accepted towards fulﬁllment
of the requirements for the

MA. degree in Communication

ﬂ {LAM

Major Professor’s Signature
Aﬁd Z 1/1 JOIO

Date

 

 

 

 

MSU is an Afﬁrmative Action/Equal Opportunity Employer

 

—;ZZ_‘—:

 

 

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

OCT 2 3 28:3

 

 

1-4
(a
C"!
1—.—
(‘3

,-_A

 

 

 

 

 

 

 

 

 

 

 

 

 

 

5/08 K:lProj/Acc&Pres/ClRC/Dat90ue.indd

 

SHIFTS IN ACCEPTANCE OF TYPE-SPECIFIC VIOLENCE AS AN OUTCOME OF
CHARACTER-ROLE IN VIDEO GAMES

By

Matthew N. Grizzard

A THESIS

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

MASTER OF ARTS
Communication

2010

ABSTRACT

SHIFTS IN ACCEPTANCE OF TYPE-SPECIFIC VIOLENCE AS AN OUTCOME OF
CHARACTER-ROLE IN VIDEO GAMES

By

Matthew N. Grizzard
Research on violent video games has largely focused on differences in outcomes as a
function of violent vs. non-violent games. The current experiment examines the effects of
narrative role (hero/villain) of a video game character within a violent game on
acceptance of violent behaviors after game play. It was predicted in accordance with
social cognitive theory that acceptance of type-speciﬁc violence (i.e., lawful or criminal)
would interact with the character-role assumed in the game such that acceptance of
lawful violence would be higher and acceptance of criminal violence would be lower for
participants who played the heroic character than participants who played the villainous
character. Results indicate an interaction between the role of a game character and the
acceptance of speciﬁc types of violence (lawful/criminal) in the opposite direction of
what was predicted. This interaction is discussed in terms of a lawfulness effect, whereby
lawful violence became more acceptable and criminal violence became less acceptable

after playing a villainous character in a video game.

ACKNOWLEDGMENTS

I am extremely grateful to my advisor, Ron Tamborini, whose guidance and
tireless dedication helped me conceptualize and complete the following project.

I would also like to thank Wei Peng and John Sherry for serving on my committee
and providing excellent insights throughout the process.

I am also grateful to my entire family. Your constant help and support made it
possible for me to pursue my Master’s degree. Thank you Mom; Dad; Michael; Grandma
and Papa Grizzard; Grandma Rodgester; Pam and Pete; Dan and Erin; Pat and Jay;
Heather; Karen and Hardee; Kim, Calvin, and Dylan; Tina, Jerry, Chase, and Stephanie.

Thanks to my friends, Dan, Ashley, Robert, Rikki, and Jack, who provided me
with much needed distractions from work.

Last, but not least, I would like to thank my wife, Alison. Thanks for leaving the
warm summers of North Carolina for the cold winters of Michigan. I could not have done

it without you, and I love you very much.

iii

TABLE OF CONTENT

LIST OF TABLES ............................................................................................................. vi
LIST OF FIGURES .......................................................................................................... vii
Introduction ......................................................................................................................... 1
Media Violence and Acceptance of Violence ......................................................... 2

Social Cognitive Theory ......................................................................................... 4
Character-Roles in Video Games and Identiﬁcation .............................................. 6
Method .............................................................................................................................. 10
Participants ............................................................................................................ 10
Stimulus Material .................................................................................................. 11
Design and Procedure ........................................................................................... 11
Character-role conditions. ......................................................................... 12

Control condition. ..................................................................................... 13

Measures and Data Scoring .................................................................................. 14
Acceptance of violence. ............................................................................ 14

Identiﬁcation. ............................................................................................ 1 6

Control variables. ...................................................................................... 16

Results ............................................................................................................................... 17
Measurement Model Tests .................................................................................... 17
Acceptance of violence. ............................................................................ 18

Identiﬁcation scale. ................................................................................... 18

Enjoyment scale. ....................................................................................... 20

Game familiarity. ...................................................................................... 20

Trait empathy. ........................................................................................... 20

Character liking ......................................................................................... 21

Descriptive Statistics ............................................................................................. 21
Hypothesis Testing ................................................................................................ 23
Discussion ......................................................................................................................... 26
Limitations and Future Research .......................................................................... 29
Conclusion ............................................................................................................ 30
Endnotes ............................................................................................................................ 31
Appendix A ....................................................................................................................... 32
Acceptance of Violence Measure ......................................................................... 32
Appendix B ....................................................................................................................... 34
Identiﬁcation Measure .......................................................................................... 34

iv

Appendix C ....................................................................................................................... 35

Control Measures .................................................................................................. 35
Enjoyment Measure .................................................................................. 35

Character Liking Measure ......................................................................... 35

Empathy Measure ..................................................................................... 35

Video Game Playing Measure .................................................................. 36

Newspaper Reading Measure ................................................................... 37

Appendix D ....................................................................................................................... 38
Correlation Matrices for 8-Item Identiﬁcation Scale ............................................ 38
Appendix E ....................................................................................................................... 39
Correlation Matrices for 7-Item Identiﬁcation Scale ............................................ 39
Appendix F ........................................................................................................................ 4O
Correlation Matrices for Enjoyment Scale ........................................................... 40
References ......................................................................................................................... 41

LIST OF TABLES

Table 1 Descriptive Statistics for Key Variables in Study ............................................... 21
Table 2 Zero-order Correlations among Key Variables in Study ..................................... 23
Table 3 Acceptance of Violence Means by Condition ..................................................... 24

vi

LIST OF FIGURES

Figure 1 Hypothesis 1 presented in graphical format. ........................................................ 9

Figure 2 Weighted means from AN COVA including character-role and violence type as
independent factors, game familiarity as a covariate, and acceptance of violence
as the dependent measure ...................................................................................... 26

Figure 3 A graphical depiction of the weighted means from two separate analyses. The
means from the hero and villain conditions come from an ANCOVA which
includes game familiarity as a covariate. The means from the control condition
comes from an ANCOVA which included game play as a covariate. .................. 28

vii

Shifts in acceptance of type-speciﬁc violence as an outcome of character-roles in video
games
Introduction

Increases in acceptance of violence are posited to be the result of a social learning
process where viewers come to adopt the beliefs, attitudes, and behaviors of violent
media characters (Anderson & Bushman, 2001). Notably, these changes can affect
narrow aspects of attitudinal domains. This is apparent in research showing that
increases in acceptance of speciﬁc types of violence can occur withbut regard to other
types of violence changing. For example, after September 11, 2001 , researchers found
that acceptance of war violence increased but acceptance of intimate violence did not
(Carnagey & Anderson, 2007). The current study examines the relationship between
type-speciﬁc changes in acceptance of violence and video game play. Speciﬁcally, the
current study seeks to examine how the video game character—role assumed by the player
might affect learned aggressive outcomes.

Past research on media violence (particularly television violence) suggests that
exposure to unjustiﬁed or criminal violence should lead to less acceptance of violence
(National Television Violence Study, 1996; NTVS). One reason given to explain this
outcome is that viewers are motivated to identify with moral protagonists and
disassociate themselves from the acts of immoral antagonists. Since disliked antagonists
in television and ﬁlm habitually engage in criminal violence, viewers exposed to these
behaviors should form negative attitudes toward these behaviors and become less
accepting of criminally violent acts. However, there is reason to believe this outcome

will differ with video games. Unlike television, ﬁlm, and other passive forms of media,

video games present players with exposure experiences (e.g., assuming the role of one
speciﬁc character and vicariously enacting the behaviors of that character) that might be
expected to promote association with characters engaging in unjustiﬁed violence. By
playing a character that engages in criminal violence, the user might be expected to
identify with that character and form positive attitudes toward the character and his/her
behavior.

Currently, there is a growing body of evidence demonstrating that violent video
games increase aggression, desensitization, and acceptance of violence (Anderson &
Bushman, 2001; Sherry, 2001). Whereas playing the role of anti—social characters has
been a paramount concern in the video game violence debate (Gentile & Anderson,
2003), little research has examined the effect of character-role on acceptance of violence.
This study examines the inﬂuence of character-roles in video games that might
distinguish the effects of violence in video games from the effects of violence in
traditional media. We argue in accordance with the logic of social cognitive theory that
video games present media experiences where users identify with villainous characters
and subsequently model those characters’ attitudes. Our study begins with a discussion
of media violence and attitudinal changes, speciﬁcally with regard to acceptance of
violence. Next, social cognitive theory (SCT) is presented followed by a discussion of
character-roles. In our discussion of character-roles, we argue that video games present
users with unique opportunities where villainous characters act as the primary model for
media users. Finally, hypotheses are presented in accordance with SCT logic and our
discussions of character-role.

Media Violence and Acceptance of Violence

Questions about the potential antisocial inﬂuence of violent mass media have
been a concern for scholars and media critics dating back at least to the Payne Fund
Studies in the 19305 (Sparks & Sparks, 2002). Most recently, video games have garnered
the most concern. Media violence literature suggests that exposure to violent video
games results in generalized aggressive outcomes, such as increased aggressive attitudes,
cognitions, and behaviors. Of particular concern for this study are increases in aggressive
attitudes, speciﬁcally attitudes regarding the acceptance of violence.

Acceptance of violence is conceptually deﬁned for this paper as a collection of
attitudes held by an individual regarding different types of violent behavior. Drawing
from persuasion research, here attitude is deﬁned as a person’s “general predispositions
to evaluate other people, objects, and issues favorably or unfavorably” (Petty, Priester, &
Briflol, 2001); violence is deﬁned as “any behavior directed toward another individual...
with extreme harm as its goal (e.g., death),” (Anderson & Bushman, 2002, pp. 28-29).
Taken together, we deﬁne acceptance of violence as predispositions to evaluate favorably
or unfavorably those behaviors directed toward another individual with extreme harm as
their goal. Consistent with recent media effects research, this deﬁnition treats acceptance
of violence as a multi-dimensional concept, encompassing attitudes regarding multiple
types of violence, including war violence, intimate violence, interpersonal violence, penal
code violence, and corporal punishment of children violence (Anderson, Benjamin,
Wood, & Bonacci, 2006). Past research has demonstrated a relationship between media
exposure and changes in attitudes and acceptance of violence related to the content.

In a naturalistic experiment, Malamuth and Check (1981) found that acceptance

of violence against women increased for males after they viewed a ﬁlm in which violence

committed against women was portrayed as having positive consequences. In a
longitudinal experiment, Zillmann and Bryant (1982) found that exposure to pornography
resulted in changes in a wide array of beliefs and attitudes related to women. Subjects in
the treatment conditions became less accepting of the women’s rights movement and
became more callous toward rape.

Researchers have had similar ﬁndings with regard to reward and punishment in
video games. Individuals who played a video game where violence was rewarded
showed increased aggressive cognitions and behaviors compared to those who played a
non-violent version of the same game, whereas those who played a version of the game
where violence was punished did not show this increased aggressive response (Carnagey
& Anderson, 2005). Notably, recent research has demonstrated that changes can be very
speciﬁc with regard to the type of violence. Research conducted during and after
September 11, 2001, determined that environmental events can affect acceptance for
some types of violence without affecting others. In a longitudinal study, subjects’

lth attacks, whereas

acceptance of war and penal violence increased after the September 1
their acceptance of interpersonal violence did not change (Carnagey & Anderson, 2007).
These studies demonstrate that changes in acceptance of violence can occur through
media exposure. These changes are thought to be related to a process of social learning
where behaviors are observed and modeled by viewers. This process is explicated in
social cognitive theory (SCT).

Social Cognitive Theory

Social cognitive theory (SCT) is a broad theory that combines cognition,

behavior, and environment to explain human agency (Bandura, 2002). A large portion of

SCT focuses on observational learning (Bandura, 1986, 2002). Rather than learning only
by individual experience, SCT argues that humans possess certain unique capacities that
allow them to learn through the observation of models. Due to the ubiquity of mass
media, media characters are considered by SCT to be some of the most prominent and
important models for observational learning (Bandura, 2002). Whereas social learning
theory, the predecessor of SCT, focused primarily on learned behaviors, SCT applies to
both simple behavioral mimicry and more abstract learning, such as shifts in attitudes,
language use, and personal standards for behavior (Bandura, 1986, 2002). Central for the
present study is observational learning’s impact on attitude change.

According to SCT, individuals form attitudes by observing a model’s behavior
and the rewards and punishments received by the model for enacting said behavior
(Bandura, 1986, 2002). Witnessing models being rewarded should cause attitudinal shifts
regarding the behavior from negative to positive due to a process called disinhibition;
witnessing models being punished should cause shifts regarding the behavior performed
from positive to negative due to a process called inhibition (Bandura, Ross, & Ross,
1963). In addition to reward and punishment inﬂuencing attitude change, observational
learning theories argue that individuals are likely to adopt the attitudes of models with
whom they associate closely (Anderson et al., 2003). In television and ﬁlm, viewers are
motivated to identify with moral protagonists and disassociate themselves from the acts
of immoral antagonists (Bandura, 1986; NTVS, 1996). Since disliked antagonists in
television and ﬁlm habitually engage in criminal violence, viewers exposed to these
behaviors should form negative attitudes toward these behaviors and become less

accepting of criminally violent acts. By contrast, since liked protagonists in television

and ﬁlm habitually engage in lawful violence, viewers exposed to these behaviors should
form positive attitudes toward these behaviors and become more accepting of lawfully
violent acts.

There is reason to believe, however, that the manner in which people process
violent media and the outcomes that result from exposure might differ with video games.
Unlike television, ﬁlm, and other more passive forms of media, video games present
players with exposure experiences (e.g., assuming the role of a speciﬁc character and
vicariously enacting their behaviors) that might be expected to promote identiﬁcation
with characters engaging in violence of all types, be it the type of lawful violence
typically enacted by liked protagonists or the type of unlawful violence performed by
antagonists. Research has shown that players tend to identify with the character they
control in a video game (i.e., the player-character; Hefner, Klimmt, & Vorderer, 2007).
Following this logic, the user might be expected to form positive attitudes toward the
player-character and his/her behaviors. In video games, the type of behaviors in which a
player engages is determined by his/her character’s role within the game’s narrative.
Character-Roles in Video Games and Identification

The actions of the player-character in a video game, while controlled by the
player, are guided by the character’s role in the game narrative. Character-role, deﬁned
as the motivational goals and attributes of a character that determine his/her function
within the media context, can vary on any number of dimensions; however, character-role
in video games usually varies along the hero/villain dichotomy. Whereas this dichotomy
may seem simplistic when considering forms of media such as novels and ﬁlms, video

games have traditionally employed a more basic narrative focused on a battle between a

heroic character and a villainous character. This is prevalent in many video games, such
as the Legend of Zelda series, the Super Mario Brothers series, and various other popular
games. Whereas both heroes and villains tend to be violent, each role is guided by very
disparate motives (Sanders, 2004). Heroes are guided by prosocial tenets; they protect
innocents and ﬁght criminals through lawful acts. Villains, however, tend to be guided
by selﬁsh goals; they exploit the innocent and enact criminal behavior (Sanders, 2004).

Unlike television and ﬁlm, where the narrative structure of storytelling habitually
encourages viewers to adopt the point of View of the moral protagonist, some play
conditions in video games encourage the player to adopt the point of view of the
antagonist. By playing a villainous character in a violent video game, the player will be
vicariously engaging in criminal violence and might be expected to form positive
attitudes toward that type of violence. By playing a heroic character, the player will be
engaging in lawful violence and might be expected to form the same type of positive
attitudes toward lawful violence that we would expect from watching television. When
we consider the outcomes expected from playing a heroic character, the predictions are
fairly straight forward. Playing the role of a hero tends to place the character in
opposition to criminal violence due to the fact that heroes traditionally ﬁght villains. As
a result, we expect that players assuming the role of a hero will become more accepting
of lawful violence and less accepting of criminal violence. This is due to the fact that the
player is encouraged by the game experience to identify with a lawfully violent character
and to disassociate from a criminally violent character.

By contrast, villains are not always pitted against heroes. For example, assuming

the role of a robber might put the player in conﬂict with innocents rather than heroic

characters. At the same time, there are occasions when villains are pitted against heroic
characters. Whereas playing the role of a villain pitted against a heroic character should
result in the player identifying with a criminally violent character and disassociating from
a lawfully violent character (thus increasing acceptance of criminal violence and
decreasing acceptance of lawful violence), playing the role of a villain in conﬂict with
innocents should cause the player to become more accepting of criminal violence, and
lawful violence should be unaffected. The logic here argues, essentially, that playing a
hero role would increase acceptance of lawful behavior (and decrease acceptance of
criminal behavior), whereas playing a criminal would increase acceptance of criminal
behavior (and decrease acceptance of lawful behavior). This logic is represented as the
interaction depicted in Figure 1 and is formally stated as the hypothesis below:
H1: The heroic/villainous nature of the character-role adopted when playing a
video game will interact with the lawfulness of behaviors evaluated after game
play to inﬂuence the perceived acceptability of those behaviors such that
acceptance of criminal violence will be highest for individuals playing a criminal
character, second highest for individuals not playing a game, and lowest for
individuals playing a heroic character, whereas acceptance of lawful violence will
be highest for individuals playing a heroic character, second highest for
individuals not playing a game, and lowest for individuals playing a villainous

character pitted against a heroic character.

 

 

 

 

 

 

 

I /
,r/ ‘ 4 -0- Hero
‘ x +Villain
\ -t- Control

 

 

 

 

 

 

Acceptance of Violence
Ir

 

 

 

Criminal Lawful

Type of Violence

Figure 1. Hypothesis 1 presented in graphical format.

SCT argues that identiﬁcation serves as a moderator of the learning process
(Bandura, 1986). Within media research, identiﬁcation is an oﬁen used but rarely well-
deﬁned construct (Zillmann, 1995). Zillmann (1995, 2006) discussed identiﬁcation (in
the context of Freud’s psychoanalytic literature) as a process of ego confusion, in which a
person experiences the actions of another as if s/he were the person in question. Zillmann
(1995, 2006) argues against using this concept of identiﬁcation with regard to the
dramatic experience. Rather than the viewer experiencing the action from within,
Zillmann advocates the viewer is a third party who merely witnesses the actions of the
dramatic characters. Yet this position seems at odds with the experience of video game
play. Recent research on identiﬁcation argues that while passive forms of media, such as
ﬁlm and television, are more closely connected to a witness perspective, video games can
be considered to possess varying levels of identiﬁcation due to their interactive nature

(Heﬁter, Klimmt, & Vorderer, 2007). Past research has argued that higher identiﬁcation

leads to stronger learning effects (Anderson et al., 2003; Konijn, Bivnak, & Bushman,
2007). Following this logic and evidence, we present the following hypothesis.

H2: Identiﬁcation will moderate the inﬂuence of character-role on the acceptance

of criminal violence such that the interaction will be greater for subjects high on

identiﬁcation than those low on identiﬁcation.

Method

To examine the proposed hypotheses, respondents took part in a 2 x 2 experiment
varying the character-role assumed in a video game (hero vs. villain) and the type of
violence (lawful vs. criminal) evaluated after game play, with the addition of an offset
control. Prior to game play, subjects completed video game and newspaper exposure
questionnaires. After playing the game, respondents completed a measure of acceptance
of violence as well as multi-item scales measuring identiﬁcation, empathic concern,
character liking, and enjoyment. To measure acceptance of lawful and criminal violence,
subjects read and evaluated scenarios in which violence was used either lawfully or
unlawfully.
Participants

One-hundred fourteen students from a large Midwestern university took part in an
experiment. Two subjects were dropped from the study: one due to a computer error
during game play and the other due to missing data; this left a ﬁnal sample of 112. A
convenience sample was recruited from the communication department’s subject pool.
The use of college students is well-suited for the purposes of this study due to the fact
that the college population represents a large portion of the gaming community (Jones,

2003). The sample included both men (n = 56) and women (n = 56), and though sex is

10

not a key variable of interest in this study, stratiﬁed random sampling was used to
distribute men and women equally across conditions to prevent possible inﬂuence
resulting from the potential that gender would get confounded with condition. Subjects
received course credit and were entered into a rafﬂe for a $25 dollar Best Buy gift card.
Stimulus Material

The video game in this experiment as the stimulus material was Fable (Lionhead
Studios, 2004). Fable is a violent, action/role-playing game in which players control a
character that can vary between being heroic and villainous based on the version of
quests selected within the game. Based on the subject’s condition, s/he either completed
a heroic or a villainous version of a quest. The heroic version was entitled “Protect
Orchard Farm,” and the villainous version of the quest was entitled “Attack Orchard
Farm.” In the heroic quest, the player must “help guards defend Orchard Farm from a
Bandit assault” (Lionhead Studios, 2004). The player must kill bandits in order to
prevent them from killing the guards and stealing the property on Orchard Farm. In the
villain quest, the player’s charge is to “lead the Bandits to the crates and cover their backs
while they steal them” (Ibid., 2004). The player accomplishes this task by killing the
guards of the property and assisting in the bandits’ getaway.
Design and Procedure

An experiment varied the character-role adopted by participants assigned to play
the violent video game (hero vs. villain) and the type of violence evaluated after game
play (lawful vs. criminal). In addition, an offset control was also collected where
subjects did not play the game at all. The character-role conditions varied the lawfulness

of the violence perpetrated by the player’s character. In the hero condition, the player

11

completed the Protect Orchard Farm quest, where they engaged in normatively lawful
violence (i.e., protecting property and innocents); in the villain condition, the player
completed the Attack Orchard Farm quest, where they engaged in nonnatively unlawful
violence (i.e., stealing property and harming law ofﬁcers). Both character-role conditions
were collected within a laboratory during the same time period. The offset control was
collected separately in a classroom.

Character-role conditions. Participants were randomly assigned to one of the
two character-role conditions (nucro = 38, nvmam = 40). Upon entering the lab,
participants were greeted by the experimenter and informed that the research being
undertaken consisted of two separate studies; study one examined the effects of varying
game controllers on the evaluation of video games, whereas study two examined the
relationship between newspaper readership and perceptions of two social issues (i.e., jury
sentencing and political knowledge). The subject was told that that the two data
collections were occurring together due to the fact that the video game study required a
short break to set-up some additional equipment, which provided enough time to collect
data for the other study. This deception was implemented to ensure that the subjects were
blind to the hypotheses. The experimenter then presented a consent form to the subject
outlining the amount of experimental credit s/he would receive (1 hr) and that s/he would
be entered into a rafﬂe for a $25 gift certiﬁcate to Best Buy. Participants then ﬁlled out
measures regarding their past video game and newspaper exposure. Both measures were
ﬁlled out in the beginning to avoid any break between game play and the measure of the
true dependent variable, acceptance of violence. In order to maintain the illusion of

separate studies, these two measures were administered separately. The participants were

12

then taken to the room containing the game. Once in the room, the experimenter told the
subject about the game and played a video made up of cutscenes from the game to
familiarize the subject with the game’s backstory. After this video was over, the subject
received game play instructions on playing the game and was given an opportunity to
become familiar with the controls. After the participant became comfortable with the
controls, s/he was shown a second video based on his/her condition. This video consisted
of a cutscene from the game that informed the participant of his/her character-role in the
game as either a hero or villain. The experimenter then left the room allowing the subject
to complete the level. Upon leaving the room, the experimenter began video recording
the game play of the subject. Once the level was completed or ten minutes elapsed, the
experimenter then reentered the room and informed the subject that the ﬁrst game session
was over and that time was needed to set up the second controller. The subject was then
escorted to another room, where the experimenter gave the subject the survey containing
the acceptability of violence measures (described as a jury sentencing measure) and a
political knowledge survey for the purported newspaper study. Once these measures
were completed, the experimenter reentered the room and told the subject that the second
control device was not functioning. The experimenter then asked the subject to ﬁll out
the remaining video game survey items which included measures of identiﬁcation, game
familiarity, character liking, enjoyment, and empathic concern. After these scales, the
subject was fully debriefed and asked not to talk about the experiment with other
students.

Control condition. The data from the control condition was collected in a

classroom. Subjects (n = 34) were recruited from the same population as the character-

13

role conditions. Subjects in the control condition completed the same measures as the
character-role conditions, except the identiﬁcation, game familiarity, character liking, and
enjoyment measures.

Measures and Data Scoring

Acceptance of violence. The acceptability of lawful and criminal violence
measure consisted of six scenarios in which violence was varied to be lawful or criminal.
The US Department of Justice considers ﬁve acts to be violent criminal acts: simple
assault, aggravated assault, robbery, rape, and homicide (US Department of Justice,
2006). Two of the ﬁve violent criminal acts (aggravated assault and homicide) were the
subject of the six scenarios (three for assault and three for homicide). These crimes were
selected because both aggravated assault and homicide ﬁt the deﬁnition of violence
advanced earlier in the paper as being acts aimed at causing extreme harm (unlike
robbery and simple assault). F Luthermore, both aggravated assault and homicide allow
for the motive of the act to vary between lawfulness and criminality (unlike rape and
robbery, for which there are no lawful equivalents).

In order to avoid potential confounds resulting from response biasing content
found in a scenario used only in the scale measuring the acceptability of either lawful or
unlawful violence, two different versions (one lawful and one criminal) were created for
each of the six scenarios (see Appendix A). In the lawful version, the violence is
described in a manner that would make the act legally justiﬁable (i.e., committed in self-
defense or committed by law enforcement); in the criminal version, the violence is
described in a manner indicating that it is an illegal act (i.e., performed out of malice or

as part of a larger criminal act). Two different forms of the survey used to measure

14

acceptance of violence were created from the twelve scenarios and the form was entered
as a between subjects factor. Both forms featured three acts of lawful violence and three
acts of unlawful violence. In the ﬁrst form, the lawful acts were comprised by using two
of the three assault scenarios and one of the two homicide scenarios; the criminal acts
included the one other assault scenario and two other homicide scenarios. The second
form was a mirror image of the ﬁrst, with the lawful version of each scenario replacing
each criminal scenario used in the ﬁrst version, and the criminal version of each scenario
replacing each lawful scenario used in the ﬁrst version. For example, if we refer to the
two versions of each scenario as version A and version B, the ﬁrst form contained
scenarios for assault l-A, assault 3-A and homicide 2-A, plus assault 2-B, homicide l-B,
and homicide 3-B. The second form contained scenarios for assault l-B, assault 3-B and
homicide 2-B, plus assault 2-A, homicide l-A, and homicide 3-A. The two forms were
distributed evenly among the conditions by random assignment.

After each scenario, the subjects were presented with four questions modiﬁed from
Hill and Zillmann (1999). The ﬁrst question asked, “If you were to serve on the jury for
the above case, would you ﬁnd (the perpetrator of violence) guilty or not guilty?” Should
the subject answer “guilty” to this question, a follow up question asked what a fair
sentence (in years and months) for the crime would be. The sentence length (in years and
months) is considered to be an inverse measure of acceptance. The ﬁnal two questions
were included to form a measure of respondent attitudes regarding the acceptability of
violence in each scenario. The two items ask the subject to rate how justiﬁed and how
acceptable the perpetrator’s actions were on a 7-point Likert-type scale with anchors 1

(Not at all) and 7 (Completely).

15

Identification. To measure identiﬁcation, an 8-item 5-point Likert-type scale
modiﬁed from Hefner, Klimmt, and Vorderer (2007) was used (see Appendix B). The
scale included items such as, “I felt like I was the game character,” and “The goals of the
character became my own goals.” The scale had to be modiﬁed due to the fact that the
items were originally in German. In its previous unmodiﬁed use, the scale demonstrated
a high reliability (or = .84).

Control variables. The following variables were measured for use as possible
control variables: subject’s sex, game enjoyment, familiarity with the game in question,
video game exposure, number of kills during game play, level success, trait empathy, and
character liking. Sex was included to account for the fact that the player-character in the
game was male, and similarity is known to lead to identiﬁcation (Eastin, 2006).
Enjoyment was measured because identiﬁcation has been shown to contribute to the
enjoyment of a video game (Hefner, Klimmt, and Vorderer, 2007), and it is unclear
whether enjoyment contributes to identiﬁcation; a 4-item scale taken from a larger ﬂow
and enjoyment scale developed by Sherry, Lucas, Greenberg, and Lachlan (2006) was
employed to measure enjoyment. In addition to the four items from the original scale, a
ﬁfth reverse coded item was added. Game familiarity was measured by the following
(yes/no) questions: “Had you ever played the game used in the study before?” and “Had
you ever heard of the game used in the study before?,” and a 7-point Likert-type question
asking how familiar the subject was with the game before exposure. Video game
exposure in general was measured by asking how often a subject plays video games on an
average weekday, Friday, Saturday, and Sunday. The video game exposure questionnaire

was developed by Lachlan, Tamborini, Bowman, and Eden (2006). Newspaper exposure

16

was measured by a similar item to video game exposure. Number of kills committed by
the player was measured by recording the game play and counting number of kills
committed by each subject. Success of quest was measured by identifying whether or not
the subject completed the level. Whether the level was completed successfully was used
as an indirect measure of player-skill and quality of exposure with players who
completed the level being considered more skilled. Trait empathy was measured to
account for concerns based on debate within the communication discipline regarding
whether individuals experience all media interactions empathically or whether some
might be experienced through identiﬁcation. This measure was included for comparison
with the identiﬁcation measure to examine whether identiﬁcation or empathy was a better
predictor of changes in acceptance of violence. The trait empathy scale was taken from
Tamborini, Salomonson, and Bahk (1993). Character liking was measured to account for
differences in liking between the heroic character and the villainous character. The
measure of character liking consisted of seven statements with Likert-style responses
measuring agreement with the items. See Appendix C for all of the control measures.
Results

Before testing the hypotheses, the data were examined, and the measurement
model for each of the scales used in the study was tested through conﬁrmatory factor
analysis (CF A). Composites were then created for each of the scales based on the results
of the factor analysis. The hypotheses were then tested using these composites through
ANOVA and ANCOVA.

Measurement Model Tests

17

Acceptance of violence. Before testing the measurement model of the
acceptance of violence measure, the distribution of responses to the questions examining
the acceptability and morality of the violence in each scenario were examined for each
form of the survey. Examination of the distribution revealed a severe lack of variance for
the criminal violence in all scenarios except the third scenario. Because of the severe
lack of variance in the acceptance of criminal violence for all scenarios except the third, it
was determined that that a test of the measurement model would be uninformative. In
place of testing the measurement model, reliability was calculated for the items of the
third scenario to determine whether responses to that scenario could be averaged to create
a composite score for the acceptance of violence measure. The two questions measuring
acceptability and morality of the violence for the third scenario were highly correlated, r
(110) = .76, p < .001 and reliable, or = .86. Because of the high correlation between the
two questions and the high reliability, the two questions were averaged to create a
composite called acceptance of violence. For analysis purposes, acceptance of violence
will be used as one dependent measure across the two character-role conditions and the
two survey forms, which will be used as independent factors in hypothesis testing.

Identification scale. A CF A was conducted to test the ﬁt of a one-factor solution
for the identiﬁcation scale. Internal consistency was examined for the identiﬁcation scale
by using Hamilton and Hunter’s (1988) Conﬁrmatory Factor Analysis program. The
CFA program generates predicted correlations between items using the internal
consistency theorem and then examines deviations of the observed correlations from the

predicted correlations. If the deviations are found to be small based on signiﬁcance tests

18

(i.e., not signiﬁcantly differ from zero), then the measurement model being tested is not
rejected.

The correlations between the items were entered into the CF A program. Based on
the observed correlations, factor loadings were extracted for each of the eight items and a
predicted correlation matrix was created. The factor loadings ranged from .36 to .77.
Errors between the observed correlation matrix and the predicted correlation matrix were
signiﬁcantly different from zero based on a signiﬁcance test for deviation from a
unidimensional solution that allows for a gradient, x2 (27, N = 78) = 40.83, p < .05,
indicating error in the measurement model. Because the one-factor solution that
contained all eight items contained errors, the residual correlation matrix was examined
to determine whether one of the eight items was particularly responsible for the errors
(see Appendix D). Item 1 was responsible for the largest deviations from the predicted
correlation matrix (-.17 and .15). Due to this fact and the fact that Item 1 had the smallest
factor loading, it was dropped from the scale and the remaining seven items were
examined to determine if they ﬁt a one-factor solution. The factor loadings for the new
seven-item scale ranged from .55 to .82. Errors between the observed correlation matrix
and the predicted correlation matrix did not differ signiﬁcantly from zero based on a
signiﬁcance test for deviation from a unidimensional solution that allows for a gradient,
x2 (20, N = 78) = 19.27, us (see Appendix E). Thus based on the results of the factor
analysis, the hypothesis that the remaining seven items of the identiﬁcation scale ﬁt a
one-factor solution could not be rejected. These items were then averaged to create a

composite variable labeled identiﬁcation (SI or = .87).

19

Enjoyment scale. A conﬁrmatory factor analysis was conducted on the ﬁve
enjoyment items. The factor loadings ranged from .65 to .90. The signiﬁcance test for
deviation from a unidiminensional model that allows for a gradient yielded a signiﬁcant
result, x2 (9, N = 78) = 53.92, p < .001, indicating error in the measurement model. The
errors in the residual correlation matrix were not speciﬁc to one particular item (see
Appendix F), so ﬁve different four-item versions of the scale (each with a different item
dropped) were tested. The tests for deviation from a unidimensional model conducted on
each of the potential four item scales also yielded signiﬁcant results indicating problems
with these solutions. No internal consistency tests were conducted for a possible three-
item scale as the measurement model would be just-identiﬁed and would ﬁt a one-factor
solution by deﬁnition. However, because the standard item alpha for the ﬁve-item
version of the scale was very high (SI or = .90), the items were averaged to create a
composite variable called enjoyment.

Game familiarity. Since there were only three indicators of game familiarity, the
measurement model is just-identiﬁed and ﬁts a one-factor solution by deﬁnition. After
standardizing the scores to the responses to the three questions regarding game
familiarity, reliability was calculated for the three items. Because reliability was high, or
= .80, the three standardized scores were averaged to create a composite variable called
game familiarity.

Trait empathy. The measurement model was examined using CF A for the trait
empathy scale. The measurement model failed with large errors throughout the multiple
dimensions of the empathy scale, and as such, trait empathy was not utilized in further

analysis.

20

Character liking. CFA was conducted on the character liking items. The lowest
factor loading was .45 and the highest was .87. The signiﬁcance test for deviation from a
unidimensional model that allows for a gradient in factor loadings was not signiﬁcant, x2
(20, N = 77) = 22.22, ns. Because the predicted correlations did not signiﬁcantly deviate
from what was observed, the factors were considered to be alternate indicators of the
same construct. The seven items were averaged to create a composite score called
character liking, or = .90.
Descriptive Statistics

Table 1 Descriptive Statistics for Key Variables in Study

 

 

Scale Range Mean SD Minimum Maximum
Acceptance of 7-point 3.04 1.69 1.00 7.00
Violence
-Lawful 3.89“ 1.65 1.00 7.00
Violence
—Criminal 2.15** 1.22 1.00 6.50
Violence
Identiﬁcation 5-point 2.49 0.80 1 .00 4.33
—Hero 2.75 0.68 1.17 4.00
—Villain 2.51 0.90 1.00 4.33
Enjoyment 5-point 3 .48 0.75 1.00 4.80
—Hero 3.55 0.73 2.20 4.80
—Villain 3 .42 0.78 1.00 4.80
Game Familiarity standardized 0.00 0.84 -2.29 .56
Character Liking 5-point 3.47 0.67 1.00 4.86
—Hero 3.69* 0.46 2.57 4.86
-Villain 3.27* 0.77 1.00 4.57

 

* indicates p < .01 , ** indicates p < .001

Table 1 shows descriptive statistics for the key variables in the current study.
Participants were slightly lower than the midpoint (4) of the scale on the acceptance of
violence measure across both forms of the survey (M = 3.13, SD = 1.69). For criminal
violence, participants’ scores centered below the midpoint (M = 2.15, SD = 1.22), and for

lawful violence, participants’ scores centered at the midpoint (M = 3.89, SD = 1.65). An

21

independent samples t-test indicated that this difference was signiﬁcant, t (110) = -6.33, p
< .001. Identiﬁcation was slightly less than the midpoint (3) of the scale, (M = 2.63, SD
= 0.76) with participants in the hero condition (M = 2.61, SD = 0.68) scoring slightly
higher than participants in the villain condition (M = 2.38, SD = 0.90). However, an
independent samples t-test indicated that this difference was not signiﬁcant, 1 (76) = -
1.30, p = .20. Overall, subjects tended to enjoy the game as indicated by their Enjoyment
scores (M = 3.48, SD = 0.75). A one-sample t-test indicates that Enjoyment was
signiﬁcantly higher than the midpoint (3) of the scale, t (77) = 5.64, p < .001. However,
results from an independent samples t-test indicated that participants’ enjoyment did not
differ, t (76) = -0.78, p = .44, based on whether or not they were assigned to play the hero
role (M = 3.55, SD = 0.73) or the villain role (M = 3.42, SD = 0.78). Participants overall
tended to like their game character, (M = 3.47, SD = 0.67), but an independent samples t-
test showed that subjects in the hero condition (M = 3.69, SD = 0.46) liked their character
signiﬁcantly more than subjects in the villain condition (M = 3.27, SD = .77), t (76) =
2.92, p < .01.

Zero order correlations for subjects in the two game play conditions (i.e., hero and
villain) were examined to determine if variables included in the study as potential control
measures were associated with acceptance of violence (see Table 2). Character liking,
identiﬁcation, and enjoyment were unrelated to acceptance of violence. Yet game
familiarity was negatively correlated with acceptance of violence (r = .22, p =.056),

suggesting its use as a control in hypothesis testing.

22

Table 2

Zero-order Correlations among Key Variables in Study

 

 

1 2 3 4
1. Identiﬁcation
2. Character Liking .47**
3. Enjoyment .46** .48**
4. Game Familiarity .10 -.07 -.22*
5. Acceptance of Violence -.01 .19 .19 -.22*

 

*indicates p < .06, two-tailed. “indicates p < .01, two-tailed.
Note. Character Role was coded such that 0 = villain and l = hero.

Hypothesis Testing

The ﬁrst hypothesis predicted that the heroic/villainous nature of the character-
role assumed by the participant would interact with the lawfulness of violence evaluated
after game play. The hypothesis predicted a disordinal symmetrical interaction in which
playing a hero role would increase acceptance of lawful behavior (and decrease
acceptance of criminal behavior), whereas playing a criminal would increase acceptance
of criminal behavior (and decrease acceptance of lawﬁrl behavior).

To test this hypothesis a 3 x 2 ANOVA was conducted with character-role (hero
vs. control vs. villain) as one between-subject’s factor and violence type (lawful vs.
criminal) as the other. The results of the ANOVA revealed a signiﬁcant main effect for
violence type, F (1, 111) = 40.03, p < .001, n2 = .26, such that acceptance of violence was
signiﬁcantly higher for the lawful violence (M = 3.89, SD = 1.65) than the criminal
violence (M = 2.15, SD = 1.22). The main effect for character-role was not signiﬁcant, F
(2, 111) =.74, ns, n2 < .01. The interaction term was also not signiﬁcant, F (2, 111) =
1.80, p = .17, n2 = .02. Although the interaction was not signiﬁcant, examination of the

means for the interaction term indicated a trend in the opposite direction of what was

23

predicted. Acceptance of criminal violence was higher for the hero condition (M = 2.63,
SD = 1.53) than the villain condition (M = 1.88, SD = 0.94) and acceptance of lawful
violence was higher for the villain condition (M = 4.25, SD = 1.65) than the hero
condition (M = 3.76, SD = 1.48). For the control condition, acceptance of criminal
violence fell between the hero and villain condition (M = 1.94, SD = 1.00), but
acceptance of lawful violence was lower than both the hero and villain conditions (see
Table 3).

Table 3

Acceptance of Violence Means by Condition

 

 

Lawful Violence Unlawful Violence
Hero M = 3.76, SD = 1.48 M = 2.63, SD = 1.53
Control M = 3.64, SD = 1.83 M = 1.94, SD = 1.00
Villain M = 4.25, SD = 1.65 M = 1.88, SD = 0.94

 

Based on the indication from our zero order correlations showing a signiﬁcant
correlation between acceptance of violence and game familiarity, hypothesis one was
further examined by including game familiarity as a covariate. Because game familiarity
was not measured with control subjects (as these subjects did not play the game), the
control condition was not included in this analysis. As such, a 2 x 2 ANCOVA was
conducted with character-role (hero vs. villain) and violence type (lawful vs. criminal) as
between-subjects’ factors and game familiarity as a covariate. The results of the
ANCOVA yielded no effect for character-role, F (1, 77) < 1, but a signiﬁcant main effect
for violence type, F (1, 77) = 22.74, p < .001, n2 = .22, and a signiﬁcant interaction

between character-role and violence type, F (1, 77) = 4.47, p < .05, n2 = .04. No effect

was found for game familiarity, F (1, 77) = 1.01, ns, 112 < .01.

24

Although the main effect for violence type showed that acceptance of violence
overall was greater for lawful violence than for criminal violence, interpretation of this
effect is qualiﬁed by the symmetrical interaction between character-role and violence
type. Most notably here, the pattern of means associated with the interaction was not in
the direction predicted. Although playing a hero role did, as expected, lead to greater
acceptance of lawful behavior (estimated marginal M = 3.67, SE = .34) than criminal
behavior (estimated marginal M = 2.73, SE = 0.34), the pattern of means for the villain
condition was in the opposite direction of predictions (see Figure 2). Whereas we
predicted that playing the villain would increase acceptance of criminal behavior (and
decrease acceptance of lawful behavior), we found that playing the villain lead to greater
acceptance of lawful behavior (estimated marginal M = 4.23, SE = 0.32) than criminal
behavior (estimated marginal M = 1.89, SE = 0.32). Instead of its effect being the reverse
of playing a hero, the effect of playing a villain was in the same direction of playing a
hero, and even stronger.

Hypothesis two predicted that identiﬁcation would moderate the inﬂuence of
character-role on the acceptance of violence such that the differences between character-
roles would be more pronounced for acceptance of criminal and lawful violence. To test
this hypothesis, identiﬁcation was divided by a median split into a high/low identiﬁcation
factor and added to the previous ANCOVA. Of note here, the main effect of
identiﬁcation, as well as all two and three way interactions of identiﬁcation with
character-role and violence type were not signiﬁcant, F < l in the ANCOVA. Thus,

hypothesis 2 is not supported.

25

 

 

 

 

 

 

 

 

 

 

 

 

 

 

4. ,..._ ._ -. --.. ..-..- .......
g i f 4.22 l
E 3 5 /) 3 67
> /’
n5 3 , ’ —0- Hero
4, 2.73 0’ / _ _
g 2 5 / +Vlllam
*3 2 i
a, 1.89
o 1.5
O
< 1 .

Criminal Lawful

Type of Violence

Figure 2. Weighted means from ANCOVA including character—role and violence type as
independent factors, game familiarity as a covariate, and acceptance of violence as the
dependent measure.

Discussion

The central hypothesis in our study argued that changes in acceptance of violence
through video game play would be affected by combinations of the character-role
assumed by players and the type of violence considered, with acceptance of lawful
violence being higher after playing heroic characters versus villainous characters and
acceptance of criminal violence being higher aﬁer playing villainous characters versus
heroic characters. Despite the fact that this hypothesis was not supported, the ﬁndings in
the paper are noteworthy. Perhaps the most interesting ﬁnding is that while we did
observe a signiﬁcant interaction between character-role and violence type, the interaction
was not in the predicted direction.

Our hypothesis was based on the expectation that playing a hero role would

promote acceptance of lawful behavior, whereas playing a criminal would promote

26

acceptance of criminal behavior. Instead, when controlling for game familiarity,
character-role and violence type interacted such that acceptance of criminal violence was
lower and acceptance of lawful violence was higher for subjects in the villainous
condition than subjects in the heroic condition. This pattern of means may indicate that
playing the role of the villain had a “lawfulness” effect on players. In other words,
playing the role of the villain may have stimulated a drift toward “lawfulness” (i.e.,
becoming more accepting of lawful behaviors and less accepting of criminal behaviors).
Although we can only offer post hoc speculation on reason for this outcome, we might
explain this as a function of the uncommon attentional frame likely created by playing a
villainous character. Player characters in video games have traditionally been heroic
characters. As such, playing the role of a heroic character is somewhat normative,
whereas playing the role of a villainous character is novel. Because playing the role of a
heroic character is a normative experience for video game play, it is unlikely that playing
a heroic character would make lawfulness of violence more salient than usual. If the
novelty of playing a villainous character made lawﬁrlness and unlawfulness more salient,
this increased salience may have strengthened the player’s instinctive bias towards

lawfulness. This interpretation seems consistent with the pattern of means seen in Figure

3.1

From Figure 3, it appears that playing the heroic character may lead to an increase
overall in the acceptance of violence. The heroic condition is almost parallel, but slightly
higher, than the control condition. The villain condition, on the other hand, looks quite

different. It appears that playing the villain condition led to a sharp increase in

27

acceptance of lawful violence and a decrease in the acceptance of criminal violence,

when compared to the other two conditions.

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

o 4.5 ~ ~ \ -~
g 4 [4.22
{5’ 35 / 3.67
s ' 74—. . ‘
H5 3 f/é 3.47 _.— Hero
or 2.73 V ' +Villain
o 25 /
C . 2.06 —" COHthI
g 2
0 1.89
E 1.5

1 .

Criminal Lawful

Type of Violence

Figure 3. A graphical depiction of the weighted means from two separate analyses. The
means from the hero and villain conditions come from an AN COVA which includes
game familiarity as a covariate. The means from the control condition comes from an
ANCOVA which included game play as a covariate.

This ﬁnding could be particularly important for video game research, as it may
indicate a similarity between the outcomes from video game play and other forms of
narrative media (i.e., ﬁlm and television), but for different reasons. Within television and
ﬁlm, it is likely the case that viewers will dislike and not identify with villainous
characters. This is due to the fact that these characters are typically portrayed as
antagonists and are not the main characters within the narrative. When playing a villain
in a video game, the villain becomes the central character.

The idea that television viewers would not like villainous characters led the
NTVS (1996) to conclude that violence committed by a villain was not as likely to lead to

negative outcomes as violence committed by heroic characters. The ﬁndings in the

28

current study may indicate that video game play leads to similar outcomes through
different mechanisms, perhaps through an increase in the salience of issues related to
lawfulness that results form playing a villainous character. Notably, the explanation
offered by NTVS for why witnessing narrative violence committed by villainous
characters was less likely to produce negative outcomes than violence committed by
heroes (i.e., that viewers will dislike villainous characters) is not consistent with the
ﬁndings in the present study. Respondents who played villains did in fact like their
characters (M = 3.27 on a 5-point scale), but playing the villainous character produced the
same type of positive outcomes reported by NTVS. As the data here are unable to test
competing explanations for the positive outcomes resulting from exposure to villainous
characters here and in research on TV exposure, future research should examine the
extent to which these competing rationales are speciﬁc to different media.
Limitations and Future Research

The ﬁrst limitation, and perhaps the most important limitation in this study, has to
do with the measurement of the dependent variable. The lack of variance in the
dependent variable created a restriction in range which attenuated the relationship
between the independent variables and the dependent variable. The current research
project should be replicated with more sensitive measures of the dependent variable. The
dependent measure in this study was scenarios that featured clearly lawful of clearly
criminal behavior. By doing so, ceiling and basement effects occurred which reduced the
variance. Scenarios that featured more questionable examples of lawful and criminal
violence could overcome these basement and ceiling effects and could provide a better

test of the hypotheses presented in the paper. Another option that may provide more

29

sensitive measures would be creating acceptance of criminal and lawful violence scales,
which if valid, could reduce measurement error and provide a more consistent measure.

The second limitation is the low number of participants in the current study.
Because of not being able to collapse across the two forms of the dependent measure to
create two variables (one measuring acceptance of lawful violence and one measuring
acceptance of unlawful violence) the form of the survey had to be used as a between-
subjects factor in the analysis. This effectively cut the power of the study in half by
reducing the number of the participants in each cell to half of the number of subjects in
each condition. Future research should create separate measures for the acceptance of
lawful violence and the acceptance of unlawful violence, which would increase the
number of subjects in each cell and consequently increase the power.
Conclusion

The current study sought to examine a possible relationship between character-p
roles in video games and their inﬂuence on acceptance of criminal and lawful violence.
The study found an interaction inconsistent with the one predicted. This is the ﬁrst study
to examine the effect of playing a hero or villain character on violence-related outcomes
from video game play. The ﬁndings reveal a surprising inﬂuence pattern suggesting a
drift toward lawfulness. Compared to playing a hero character, playing a villainous
character produced an increase in the acceptance of lawful violence and in decrease in the
acceptance of criminal violence. This study represents an important ﬁrst step in
examining how the character-role assumed by a game player may moderate the effects of

game play.

30

Endnotes

In addition to showrng the estimated marginal means for acceptance of Violence in the

heroic and villainous conditions weighted by game familiarity, Figure 3 presents the
estimated acceptance of violence means of the control condition weighted by video game
exposure in general. As previously stated, the design of the experiment did not allow us
to extract means for the control condition weighted by game familiarity. To address this
limitation, we extracted acceptance of violence means for the control condition weighted
by general video game play, the measure most comparable to game familiarity. This was
done to provide our best estimate of how acceptance of violence scores might look in the
control condition when controlling for game familiarity. As such, the chart provides only

our best estimate and does represent the outcome of statistical analysis.

31

Appendix A

Acceptance of Violence Measure

1.

On September 28, 2007, a young man was walking through a neighborhood in Grand
Rapids, MI. Two men approached the young man and began to threaten him for
walking through their neighborhood. Then, the two men began to attack the man with
bats.
a. [Criminal Violence] The two perpetrators beat the young man badly, but left
him alive.

b. [Lawful Violence] The young man defended himself and beat the perpetrators
off with a stick injuring one of them very badly.

On December 17, 2007, two business associates began arguing at their annual
Christmas party in Lansing, MI. One of the business associates pulled out a knife
during the argument, and the two began to struggle.
a. [Criminal Violence] During the struggle, the man who pulled out the knife
stabbed the other associate. The injured man lived.

b. [Lawful Violence] During the struggle, the man who pulled out the knife was
stabbed by the other associate. The injured man lived.

On May 19, 2007, at a house party in Ann Arbor, MI, a young man began yelling at
one of the hosts. A few days earlier, the enraged man and his girlfriend broke up, and
he blamed the host of the party for the break up. The man then threw a punch at the
host of the party.

a. [Criminal Violence] The punch connected and broke the jaw of the host.

b. [Lawful Violence] The punch missed, and the host counterpunched breaking
the man’s jaw.

On November 19, 2007, a fender bender occurred in Charlotte, MI. The accident
occurred when a man couldn’t stop his car at a stoplight and ran into the back of the
driver who was in front of him. The man who was rear-ended jumped out of his car
and started yelling. The driver of the other car stayed inside his car. The man who
was rear-ended then reached inside his car and pulled out a hunting riﬂe and pointed

it at the other man.
a. [Criminal Violence] The man then shot and killed the man who had rear-

ended him.

b. [Lawful Violence] The man in the car then hit the gas running over and killing
the man he rear-ended.

32

5. On June 15, 2007, a man broke into a home in West Bloomﬁeld, MI, by breaking a

window. A resident of the house heard the glass break and went down stairs with a
ﬂashlight and gun to investigate the sound. When he got to the bottom of the stairs,
the robber pointed a gun at him, and the man shined his ﬂashlight at the robber.

a. [Criminal Violence] Then the robber shot and killed the homeowner.

b. [Lawful Violence] Then the homeowner shot and killed the robber.

. Police responded to a bank’s silent alarm on June 30, 2007, outside of Saginaw, MI.
The robber was still inside the bank when the police arrived, and held the bank’s
employees and customers as hostages. After several hours of negotiations, the robber
had released all but one hostage, the banks security guard. After the robber continued
to threaten the guard’s life, the police then positioned a sharp shooter in a position to
kill the robber.

a. [Criminal Violence] The robber then shot and killed the guard.

b. [Lawful Violence] The police then shot and killed the robber.

If you were to serve on the jury for the above case, would you ﬁnd

guilty or not

guilty? What, in your judgment, is a fair sentence for this crime as committed under the
indicated circumstances? Give your response in years and months. year(s)

month(s)

How justiﬁed was the perpetrator in his actions?

Not at all justiﬁed 1 2 3 4 Completely
justiﬁed

How acceptable were the perpetrator’s actions?

Not at all acceptable 1 2 3 4 Completely
acceptable

33

Appendix B

Identification Measure

1. While I was playing the game, I forgot everything around me.

2. I had the feeling I was literally in the character’s skin.

3. I forgot myself because I was focusing so much on the game character’s actions.

4. I had the feeling I was the game character more so than myself.

5. The game character’s goals became my goals.

6. While playing the game, the game world was more real to me than my “actual
reality.”

7. I felt as if I was really participating in the shown/depicted happenings.

8. I almost had the feeling of actually being the character.

34

Appendix C

Control Measures

Enjoyment Measure

P'PP’NT‘

I enjoyed playing this game.

I would play this game longer if I had the opportunity.
This game was not very interesting to me. (R)

I would recommend this game to a friend.

This game was fun.

Character Liking Measure

NP‘MPPJN?’

The character was likeable.

The character was fascinating.

The character was appealing.

The character was admirable.

The character was heroic.

The character was virtuous.

I think the character could be a friend of mine.

Empathy Measure

NP‘S‘PP’NT‘

9°

10.

11.
12.
13.
14.
15.

16.

I cannot continue to feel okay if others around me are feeling depressed.

Before criticizing someone, I try to imagine how I would feel in their place.

I am the type of person who is concerned when other people are unhappy.

I am the type of person who can say the right thing at the right time.

I really get involved with the feelings and characters in a novel or ﬁlm.

I don’t become upset just because a friend is acting upset.

I sometimes try to understand my friends better by imaging things from their
perspective.

When I see someone being taken advantage of, I feel kind of protective toward them.
Even though I often try to console someone who is feeling bad, I never seem to be
able to say the right thing.

When I am reading an interesting story or novel, I imagine how I would feel if the
events were to happen to me.

I become nervous if others around me seem nervous.

I sometimes ﬁnd it difﬁcult to see things from another’s perspective.

I often have tender, concerned feelings for people less fortunate than myself.

I usually respond appropriately to the feelings of others.

After acting in a play or seeing a play or a movie, I have felt partly as though I were
one of the characters.

The people around me have a great inﬂuence on my moods.

35

17. I try to look at everyone’s side of a disagreement before I make a decision.

18. I would describe myself as a pretty soft-hearted person.

19. Others think of me as an empathetic person.

20. When I watch a good movie, I can easily put myself I the place of the lead character.
21. When I am upset, I usually try to put myself in his or her shoes for awhile.

22. I sometimes don’t feel very sorry for people when they are having problems.

23. My friends come to me with their problems because I am a good listener.

24. I become very involved when I watch a movie.

25. Other people’s misfortunes do not usually disturb me a great deal.

26. I am often touched by the things that I see happen.

 

Scale Item Scoring (‘R’ indicates reverse scored-items): Emotional Contagion: 1, 6R, 11,
16; Perspective Taking: 2, 7, 12R, 17, 21; Empathetic Concern: 3, 8, 13, 18, 22R, 25R,
26; Communicative Response: 4, 9R, 14, 19, 23; Fictional Involvement: 5, 10, 15, 20, 24.

Video Game Playing Measure

1. On an average Weekday (Monday, Tuesday, Wednesday, Thursday), how many hours
do you usually spend playing video games during these time periods?

2. On an average Friday, how many hours do you spend playing video games during
these time periods?

3. On an average Saturday, how many hours do you spend playing video games during
these time periods?

4. On an average Sunday, how many hours do you spend playing video games during
these time periods?

 

 

 

(I) in the (2) in the (3) in the (4) at night
morning afternoon evening (11 p.m. to 6
(6 a.m. to (noon to 7 (7p.m. to 11 a.m.)
noon) p.m.) p.m.)

0 never E] El El E]

0 up to 1 hr. 13 D D D

o 1 hr. to 2 hrs. D 1:] D D

o 2 hrs. to 3 hrs. D D E] El

0 3 hrs. to 4 hrs. [:1 D E] [:1

0 4 hrs. to 5 hrs. D D El E1

0 5 hrs. to 6 hrs. E] D D D

o more than 6 D D D D

hrs

 

36

Newspaper Reading Measure

1. On an average Weekday (Monday, Tuesday, Wednesday, Thursday), how many hours
do you usually spend reading newspapers during these time periods?

2. On an average Friday, how many hours do you spend reading newspapers during these
time periods?

3. On an average Saturday, how many hours do you spend reading newspapers during
these time periods?

4. On an average Sunday, how many hours do you spend reading newspapers during
these time periods?

 

 

 

 

 

(1) in the (2) in the (3) in the (4) at night
morning afternoon evening (11 p.m. to 6
(6 a.m. to (noon to 7 (7p.m. to II a.m.)
noon) p.m.) p.m.)
0 never El E] Cl C]
0 up to 1 hr. D D D D
o 1 hr. to 2 hrs. El 1:] E] D
g02hrs.to3hrs. D D D [:1
o 3 hrs. to 4 hrs. D D D E]
0 4 hrs. to 5 hrs. D D [:1 1:]
o 5 hrs. to 6 hrs. D [:1 E] D
.- 0 more than 6 D D D D
i hrs

 

37

Appendix D

Correlation Matrices for 8-Item Identification Scale

Observed Correlation Matrix

 

 

 

 

 

 

 

 

Item 1 Item 2 Item 3 Item 4 Item 5 Item 6 Item7 Factor
Loadings
Item 1 .36
Item 2 .24 .70
Item 3 .41 .47 .71
Item 4 .11 .64 .49 .77
Item 5 .15 .32 .43 .49 .54
Item 6 .26 .43 .45 .54 .38 .69
Item 7 .39 .47 .52 .51 .33 .59 .74
Item 8 .19 .63 .45 .69 .45 .50 .54 .77
Predicted Correlation Matrix
Item 1 Item 2 Item 3 Item 4 Item 5 Item 6 Item7 Factor
Loadings
Item 1 .36
Item 2 .25 .70
Item 3 .26 .50 .71
Item 4 .28 .54 .55 .77
Item 5 .19 .38 .38 .42 .54
Item 6 .25 .48 .49 .53 .37 .69
Item 7 .27 .52 .53 .57 .40 .51 .74
Item 8 .28 .54 .57 .59 .42 .53 .57 .77
Residual Correlation Matrix
Item 1 Item 2 Item 3 Item 4 Item 5 Item 6 Item7 Factor
Loadings
Item 1 .36
Item 2 -.01 .70
Item 3 .15 -.03 .71
Item 4 -.17 .10 -.06 .77
Item 5 -.04 -.06 .05 .07 .54
Item 6 .01 -.05 -.04 .01 .01 .69
Item 7 .12 -.05 -.01 -.06 -.07 .08 .74
Item 8 -.09 .09 -.10 .10 .03 -.03 -.03 .77

 

38

Appendix E

Correlation Matrices for 7-Item Identification Scale

Observed Correlation Matrix

 

 

 

 

 

 

 

 

Item 2 Item 3 Item 4 Item 5 Item 6 Item7 Factor
Loadings
Item 2 .70
Item 3 .47 .66
Item 4 .64 .49 .82
Item 5 .32 .43 .49 .55
Item 6 .43 .45 .54 .38 .68
Item 7 .47 .52 .51 .33 .59 .70
Item 8 .63 .45 .69 .45 .50 .54 .79
Predicted Correlation Matrix
Item 2 Item 3 Item 4 Item 5 Item 6 Item7 Factor
Loadings
Item 2 .70
Item 3 .46 .66
Item 4 .57 .54 .82
Item 5 .39 .36 .45 .55
Item 6 .48 .45 .56 .37 .68
Item 7 .49 .46 .57 .39 .48 .70
Item 8 .55 .52 .65 .43 .54 .55 .79
Residual Correlation Matrix
Item 2 Item 3 Item 4 Item 5 Item 6 Item7 Factor
Loadings
Item 2 .70
Item 3 .01 .66
Item 4 .07 -.05 .82
Item 5 -.07 .07 .04 .55
Item 6 -.05 .00 -.02 .01 .68
Item 7 -.02 .06 -.06 -.06 .11 .70
Item 8 .08 -.07 .04 .02 -.04 -.01 .79

 

39

Appendix F

Correlation Matrices for Enjoyment Scale

Observed Correlation Matrix

 

 

 

 

 

 

 

 

Item 1 Item 2 Item 3 Item 4 Item 5 Factor
Loadings
Item 1 .79
Item 2 .59 .81
Item 3 .71 .72 .90
Item 4 .39 .70 .60 .65
Item 5 .86 .59 .77 .51 .87
Predicted Correlation Matrix
Item 1 Item 2 Item 3 Item 4 Item 5 Factor
Loadings
Item 1 .79
Item 2 .64 .81
Item 3 .71 .73 .90
Item 4 .51 .53 .59 .65
Item 5 .69 .70 .78 .57 .87
Residual Correlation Matrix
Item 1 Item 2 Item 3 Item 4 Item 5 Factor
Loadings
Item 1 .79
Item 2 -.05 .81
Item 3 .00 -.01 .90
Item 4 -.12 .17 .01 .65
Item 5 .17 -.ll -.01 -.06 .87

 

40

References

Anderson, C. A., & Bushman, B. J. (2001). Effects of violent video games on aggressive
behavior, aggressive cognition, aggressive affect, physiological arousal, and
prosocial behavior: A meta-analytic review of the scientiﬁc literature.
Psychological Science, 12, 353-359. doi:10.1111/1467-9280.00366

Anderson, C. A., & Bushman, B. J. (2002). Human aggression. Annual Review of
Psychology, 53, 27-51. doi:10.1146/annurev.psych.53.100901.135231

Anderson, C. A., Benjamin, A. J ., Wood, P. K., & Bonacci, A. M. (2006). Development
and testing of the velicer attitudes toward violence scale: evidence for a four-
factor model. Aggressive Behavior, 32, 122-136. doi:10.1002/ab.20112

Anderson, C. A., Berkowitz, L.,Donnerstein, E., Huesmann, L., Johnson, J., Linz, D.,
Malamuth, N., & Wartella, E. (2003). The inﬂuence of media violence on youth.
Psychological Science in the Public Interest, 4, 81-110. doi:10.1111/j.1529-
1006.2003.pspi_1433.x

Bandura, A. (1986). Social foundations of thought and action: A social cognitive theory.
Englewood Cliffs, NJ: Prentice-Hall.

Bandura, A. (2002). Social cognitive theory of mass communication. In J. Bryant & D.
Zillmann (Eds) 2nd Edition, Media effects: Advances in theory and research (pp.
121-153). Mahwah, NJ: Lawrence Erlbaum Associates.

Bandura, A., Ross, D., & Ross, S. A. (1963). Imitation of ﬁlm-mediated aggressive
models. Journal of Abnormal Social Psychology, 66, 3-11. doi:10.1037/h0048687

Carnagey, N. L. & Anderson, C. A. (2005). The effects of reward and punishment in
violent video games on aggressive affect, cognition, and behavior. Psychological
Science, 16, 882-889.

Carnagey, N. L. & Anderson, C. A. (2007). Changes in attitudes towards war and

violence after September 11, 2001. Aggressive Behavior, 33, 118-129.
doi:10.1111/j.l467-9280.2005.01632.x

Eastin, M. (2006). Video game violence and the female game player: Self— and opponent
gender effects on presence and aggressive thoughts. Human Communication
Research, 32, 351-372. doi:10.111l/j.1468-2958.2006.00279.x

Gentile, D. A., & Anderson, C. A. (2003). Violent video games: The newest media

violence hazard. In D. Gentile (Ed.) Media Violence and Children (pp. 131-152),
Westport, CT: Praeger.

41

Hamitlon, M. A., & Hunter, J. E. (1988). Conﬁrmatory Factor Analysis [Computer
Software].

Hefner, D., Klimmt, C., & Vorderer, P. (2007). Identiﬁcation with the player character as

determinant of video game enjoyment. Lecture Notes in Computer Science, 4 740,
39-48. doi:10.1007/978-3-540-74873-1_6

Hill, J. R., & Zillmann, D. (1999). The oprahization of America: Sympathetic crime talk
and leniency. Journal of Broadcasting & Electronic Media, 43, 67-68.
doi:10.1080/08838159909364475

Jones, S. (2003). Let the games begin: Gaming technology and entertainment
among college students. Pew Internet and American Life Project. Retrieved
from http://www.pewinternet.org/pdfs/PIP_College_Gaming_Reporta.pdf

Konijn, E. A., Bijvank, M. N., & Bushman, B. J. (2007). I wish I were a warrior: The role
of wishful identiﬁcation in the effects of violent video games on aggression in
adolescent boys. Developmental Psychology, 43, 1038-1044. doi:10.1037/0012-
l649.43.4.1038

Lachlan, K., Tamborini, R., Bowman, N. and Eden, A. (2006) Aﬂective disposition theory
and accepting violent reprisal: The moral monitoring of motives and behavior.
Paper presented at the 92nd annual meeting of the National Communication
Association, San Antonio, TX.

Lionhead Studios. Fable. (2004). [Xbox Software]. Redmond, WA: Microsoft Game
Studios.

Malamuth, N. M., & Check, J. V. P. (1981). The effects of mass media exposure on
acceptance of violence against women: A ﬁeld experiment. Journal of Research
in Personality, 15, 436-446. doi: 1 0.1016/0092-6566(81)90040-4

National Television Violence Study. (1996). Scientiﬁc Papers, 1994-95. Studio City, CA:
Mediascope.

Petty, R. E., Priester, J. R., & Brifiol, P. (2001). Mass media attitude change: Implications
of the elaboration likelihood model of persuasion. In J. Bryant & D. Zillmann
(Eds) 2nd Edition, Media effects: Advances in theory and research (pp. 155-198).
Mahwah, NJ: Lawrence Erlbaum Associates.

Sanders, M. S. (2004). Is it a male of female thing?.' Identification and enjoyment of
media characters. Paper presented at the 54th Annual Meeting of the International

Communication Association, New Orleans, LA.

Sherry, J. (2001). The effects of violent video games on aggression. Human
Communication Research, 27, 409-431. doi: 10.111 1/j.1468-2958.2001.tb00787.x

42

Sherry, J. L., Lucas, K., Greenberg, B. S., & Lachlan, B. (2006) Video game uses and
gratiﬁcations as predictors of use and game preference. In P. Vorderer & J. Bryant
(Eds), Playing video games: Motives, responses, and consequences (pp. 213—

224). New York: Routledge.

Sparks, G. G. & Sparks, S. W. (2002). Effects of media violence. In J. Bryant & D.
Zillmann (Eds) 2nd Edition, Media effects: Advances in theory and research (pp.
269-285). Mahwah, NJ: Lawrence Erlbaum Associates.

Tamborini, R., Salomonson, K., & Bahk, C. (1993). The relationship of empathy to
comforting behavior following ﬁlm exposure. Communication Research, 20, 723-
738. doi:10.1177/009365093020005005

US Department of Justice. (2006). Bureau of Justice Statistics Violent Crime Rate
Trends. Retrieved from http://www.ojp.usdoj.gov/bj s/glance/viort.htm.

Zillmann, D. (1995). Mechanisms of emotional involvement with drama. Poetics, 23, 33-
51 . doi:10.1016/0304-422X(94)00020-7

Zillmann, D. (2006). Dramang for emotions from ﬁctional narration. In J. Bryant and
P. Vorderer (Eds) Psychology of entertainment (pp. 215-238). Mahwah, NJ:
Lawrence Earlbaum Associates.

Zillmann, D., & Bryant, J. (1982). Pornography, sexual callousness, and the trivialization

of rape. Journal of Communication, 32, 10-21. doi:10.1111/j.1460-
2466.1982.tb02514.x

43

   

"11111111111111[illiljijlllllilijisliillﬂ