Sexualization and aggression against women: A focus on

UNIVERSITE DE LIEGE

Faculté de Psychologie, Logopédie et Sciences de l’Education

Unité de Recherche en Psychologie et Neuroscience Cognitives

Sexualization and aggression against

women: A focus on sexualized characters in

video games

Jonathan Burnay

Thèse présentée en vue de l’obtention du titre de Docteur en Sciences Psychologiques

Sous la direction du Pr. Frank Larøi

Thèse co-dirigée par le Pr. Brad Bushman

Membres du jury : Pr. Stéphanie Demoulin, Pr. Joël Billieux, Pr. Björn-Olav Dozo, Pr.

Benoît Dardenne, Pr. Brad Bushman, Pr. Frank Larøi

2018-2019

Acknowlegments

‚It’s dangerous to go alone! Take this.‛

When I started my Ph.D., I did not meet an old sage in a cavern and I did not

receive a sword to face the difficulties along the road. Unarmed, I was sure to lose all

my hearts before succeeding in my quest. But it was without counting on all the

fairies I would meet on the way!

First, I want to thank Frank Larøi, my advisor, for the confidence he placed in me.

Thank you for your time, patience, thoroughness, and words of encouragements.

You taught me to be obsessive and I sure hope that I took a few pages out of your

book!

I also want to thank Brad Bushman, my co-advisor. Beyond your precious

methodological and theoretical help, you offered me one of the greatest experiences

of my life by inviting me for a research stay in your lab. During that research stay,

you did not treat me as a Ph.D. student, but as a friend.

Thank to my thesis committee, Professor Joël Billieux and Professor Benoît

Dardenne who helped me during the development of this thesis through their

advices and guidance. I also thank Professor Stephanie Demoulin and Professor

Björn-Olav Dozo giving of their time and being part of my jury. I sincerely hope you

will enjoy the reading.

Thanks to all my colleagues for their help, support, and all the good moments

spent during lunchtime: Audrey, Aurélie, Bénédicte, Benjamin, Céline, Christina,

Clara, Coline, Fabienne, Fanny, Hedwige, Julien, Line, Lucie, Maëlle, Marie, Marion,

Murielle, Sylvie, Sylvie, Vincent, and Thierry. There are a few that I would like to

thank particularly. Thank you to Marie Geurten for the many (some might say too

many) geek discussions we had. You reminded me that beyond the PhD, I still have

other passions. I do not have any excuse left not to finish that story! Thank you as

well for your precious statistical assistance. Thank you to Julien Laloyaux, the Great

Methodology Destroyer, without whom my studies would have ever reached half

their current quality. Thank you to Aurélie Wagener for your problem solving skills

and the coffee breaks, in Kawa we trust! Thank you to Clara Della Libera, some

If you do not know the reference, the quotation comes from a 1986 video game called The

Legend of Zelda. In that game, your health is represented by hearts you lose every time you

get hit by a monster. Fairies are the best way to recover all your hearts and they can even

resurrect you if you die!

would say you are too loud, I would say you bring more joy to the office. Your

capacity for hard work is a model for me! Thank you to Audrey Krings for being one

of the most smiling and enthusiastic persons I know. No matter the situation, you

never seem down. Thank you to Vincent Marinelli for always finding the most

uncomfortable theoretical questions. You always seem to calculate the best next

moves ahead in everything. Thank you to Bénédicte Thonon for the friendly and

productive atmosphere you create in the office. I want to express my gratitude to

Sylvie Blairy who trusted me in the past and to Sylvie Willems who trusts me for the

future.

I have another large group of people I need to thank: my friends, the old as well as

the new ones. Many thanks to Adriano Tosoni, Amour D’Haene, Damien Schwanen,

Damien Vannespenne, Dimitri Fagnoul, Emilie Casagrande, Gaëlle Cox, Kevin

Debois, Laurence Lecrompe, Loïc Diricken, Mégane Likin, Milo Schoonjans, Robin

Schoonjans, Sarah Denis, Sarah Schroyen, Tangui Horel, and Vanessa Verest. I

particularly want to thank Adriano Tosoni – I know, you do not like to be called

Adriano, but I have to make it official – for being my initiator to the scientific

method. You taught me that any affirmation should survive multiple sources

verifications before being considered as facts. You punched so many of my beliefs in

the face I hardly recognize who I was before I met you. Thank you to Laurence

Lecrompe, my personal physician, for patching me up every time I needed it. I know

I will remain a faithful customer thanks to your cocktails and barbecues. Thank you

Tangui Horel for all these hours of co-working and for never giving me any excuse

for not going out and taking a break. Thank you to Damien Vannespenne for being

the insomniac confidant everyone needs. Thank you to Sarah Denis, your stalking

skills made you the first close friend I made in college and since then, your subtle

sense of humor remains my favorite. Thank you to Robin Schoonjans for proving that

no one should be afraid of changing career track; you have always been a model of

resilience and valor. Thank you to Kevin Debois for your perpetual good mood, your

jokes always brighten even the worst situations. Thank you to Gaëlle Cox for being

the first who explicitly expressed feminist ideas around me. The idea caught on.

Thank you to Amour D’Haene for the joy you communicate around you with your

everlasting smile. And because no one is too young to appear in the

acknowledgment of a thesis, I want to address a special thank you to the newest

member of our little band, Milo Schoonjans.

I do not really have the words to express how much Charlotte Delleur helped me

during these past four years. Thank you for being such a supportive wife. You stood

by my side during the best and the worst moments. In the darkest time, you were the

one who turned on the light for me and sharing the best moments with you is what

made them real. Every day spent with you makes me a better man.

The final ‚thank you‛ goes to my parents, Murielle Claes and Olivier Burnay. You

have built the foundations that have led to this moment.

Table of Content

PROLOGUE ....................................................................................................................................................... 11

THEORETICAL SECTION ............................................................................................................................. 13

CHAPTER 1: AGGRESSION AGAINST WOMEN .................................................................................... 15

DEFINITION ...................................................................................................................................................... 15

AGGRESSIVE COGNITION AND AFFECTS ......................................................................................................... 16

AGGRESSIVE BEHAVIOR AGAINST WOMEN ..................................................................................................... 16

AGGRESSIVE ATTITUDES TOWARD WOMEN .................................................................................................... 18

SUMMARY ......................................................................................................................................................... 20

CHAPTER 2: AN INTEGRATED MODEL TO EXPLAIN AGGRESSION AGAINST WOMEN: THE

CONFLUENCE MODEL INTEGRATED WITH THE GENERAL AGGRESSION MODEL.............. 21

ONE MODEL TO INTEGRATE THEM ALL .......................................................................................................... 21

Cognitive-Neoassociationistic Theory .......................................................................................................... 21

Social Learning Theory ................................................................................................................................ 22

Script Theory ................................................................................................................................................ 22

Excitation Transfer Theory .......................................................................................................................... 23

Social Interaction Theory ............................................................................................................................. 23

THE GENERAL AGGRESSION MODEL .............................................................................................................. 23

Proximate Causes and Processes .................................................................................................................. 24

Distal Causes and Processes......................................................................................................................... 26

THE CONFLUENCE MODEL .............................................................................................................................. 27

Hostile Masculinity...................................................................................................................................... 27

Promiscuous-Impersonal Sex ....................................................................................................................... 27

THE CONFLUENCE MODEL INTEGRATED INTO THE GENERAL AGGRESSION MODEL ................................... 28

SUMMARY ......................................................................................................................................................... 29

CHAPTER 3: REPRESENTATION OF WOMEN AND MEN IN VIDEO GAMES .............................. 31

DEFININING SEXUALIZATION .......................................................................................................................... 32

THE VIDEO GAME INDUSTRY, THE VIDEO GAME COMMUNITY, AND VIDEO GAME CONTENT AS A HOSTILE

ENVIRONMENT FOR FEMALE PLAYERS ............................................................................................................ 32

Video Game Content Analysis ..................................................................................................................... 33

THE NEED FOR AN OBJECTIVE INSTRUMENT TO EVALUATE VIDEO GAME CONTENT REGARDING

SEXUALIZATION AND STEREOTYPED ROLES .................................................................................................... 40

SUMMARY ......................................................................................................................................................... 42

CHAPTER 4: IMPACT OF FEMALE SEXUALIZED CONTENT ON AGGRESSION AGAINST

WOMEN ............................................................................................................................................................. 45

A BRIEF DESCRIPTION OF PREVIOUS VIDEO GAME CONTENT RESEARCH ..................................................... 45

THEORETICAL ARGUMENTS TOWARD AN IMPACT OF SEXUALIZED FEMALE CONTENT IN VIDEO GAME ON

AGGRESSION AND AGGRESSIVE ATTITUDES TOWARD WOMEN ..................................................................... 47

STATE OF THE ACTUAL EMPIRICAL RESEARCH ABOUT THE IMPACT OF FEMALE SEXUALIZED CONTENT ON

AGGRESSION AGAINST WOMEN ...................................................................................................................... 48

SUMMARY ......................................................................................................................................................... 54

AIMS OF THE THESIS .................................................................................................................................... 55

EXPERIMENTAL SECTION ........................................................................................................................... 57

THE DEVELOPMENT AND VALIDATION OF AN OBJECTIVE MEASURE OF THE

SEXUALIZED CONTENT OF VIDEO GAMES: THE VIDEO GAME SEXUALIZATION

PROTOCOL ....................................................................................................................................................... 59

ABSTRACT ......................................................................................................................................................... 61

INTRODUCTION ................................................................................................................................................ 63

STUDY 1 ............................................................................................................................................................ 64

METHOD ........................................................................................................................................................... 64

Sampling ...................................................................................................................................................... 64

Coding .......................................................................................................................................................... 65

Coder Training ............................................................................................................................................. 66

Statistical Analyses ...................................................................................................................................... 66

RESULTS ............................................................................................................................................................ 66

Intercoder Reliability .................................................................................................................................... 66

Exploratory Factor Analysis (EFA) ............................................................................................................. 67

Descriptive Statistics, Internal Consistency, and Inter-Correlation of the Female and Male Two-factor

Model............................................................................................................................................................ 68

DISCUSSION ...................................................................................................................................................... 69

STUDY 2 ............................................................................................................................................................ 69

METHOD ........................................................................................................................................................... 71

Participants .................................................................................................................................................. 71

Questionnaires ............................................................................................................................................. 72

Sampling ...................................................................................................................................................... 73

Coding .......................................................................................................................................................... 73

Coder Training ............................................................................................................................................. 75

Statistical Analyses ...................................................................................................................................... 75

RESULTS ............................................................................................................................................................ 75

Intercoder Reliability .................................................................................................................................... 75

Exploratory Factor Analysis (EFA) ............................................................................................................. 76

Descriptive Statistics, Internal Consistency, and Inter-Correlation of the Female and Male Three-Factor

Models .......................................................................................................................................................... 77

Convergent Validity ..................................................................................................................................... 78

Regression Analyses ..................................................................................................................................... 78

DISCUSSION ...................................................................................................................................................... 88

EFFECTS OF SEXUALIZED VIDEO GAMES ON ONLINE SEXUAL HARASSMENT ..................... 93

ABSTRACT ......................................................................................................................................................... 95

INTRODUCTION ................................................................................................................................................ 97

Theoretical Foundations of the Present Research ......................................................................................... 98

Objective of the Present Research............................................................................................................... 101

METHOD ......................................................................................................................................................... 102

Participants ................................................................................................................................................ 102

Materials .................................................................................................................................................... 102

Questionnaires ........................................................................................................................................... 104

Procedure ................................................................................................................................................... 104

RESULTS .......................................................................................................................................................... 106

DISCUSSION .................................................................................................................................................... 108

Theoretical and Practical Implications ....................................................................................................... 109

Limitations and Future Research ............................................................................................................... 111

Conclusion.................................................................................................................................................. 112

AN EXAMINATION OF THE POSSIBLE IMPACT OF THE SEXUALIZED CONTENT OF VIDEO

GAMES AND COGNITIVE LOAD ON IMPLICIT EVALUATIONS OF WOMEN ......................... 113

ABSTRACT ....................................................................................................................................................... 115

INTRODUCTION .............................................................................................................................................. 117

METHOD ......................................................................................................................................................... 120

Participants ................................................................................................................................................ 120

Materials .................................................................................................................................................... 120

Questionnaires ........................................................................................................................................... 123

Procedure ................................................................................................................................................... 124

RESULTS .......................................................................................................................................................... 126

Individual Differences ................................................................................................................................ 126

Implicit Responses in Experimental Conditions ........................................................................................ 126

Bayesian Statistics...................................................................................................................................... 129

DISCUSSION .................................................................................................................................................... 130

Theoretical and Practical Implications ....................................................................................................... 130

Limitations and Future Research ............................................................................................................... 132

Conclusion.................................................................................................................................................. 133

IMPACT OF SEXUALIZED VIDEO GAME AND COGNITIVE LOAD ON RAPE MYTH

ACCEPTANCE AND DEHUMANIZATION OF THE PERPETRATOR.............................................. 135

ABSTRACT ....................................................................................................................................................... 137

INTRODUCTION .............................................................................................................................................. 139

METHOD ......................................................................................................................................................... 143

Participants ................................................................................................................................................ 143

Materials .................................................................................................................................................... 143

Questionnaires ........................................................................................................................................... 145

Procedure ................................................................................................................................................... 146

Statistical Analyses .................................................................................................................................... 147

RESULTS .......................................................................................................................................................... 148

DISCUSSION .................................................................................................................................................... 152

EFFECTS OF VIOLENT AND NONVIOLENT SEXUALIZED MEDIA ON AGGRESSION-

RELATED THOUGHTS, FEELINGS, ATTITUDES, AND BEHAVIORS: A META-ANALYTIC

REVIEW ............................................................................................................................................................ 159

ABSTRACT ....................................................................................................................................................... 161

INTRODUCTION .............................................................................................................................................. 163

Nonviolent and Violent Sex in the Mass Media ........................................................................................ 163

Definitions .................................................................................................................................................. 164

Types of Research Designs ......................................................................................................................... 165

Why Do Sexualized and Violent Sexualized Media Increase Aggression? ................................................ 169

Previous Reviews ....................................................................................................................................... 171

Present Review ........................................................................................................................................... 172

METHOD ......................................................................................................................................................... 173

Literature Search Procedures ..................................................................................................................... 173

Inclusion Criteria ....................................................................................................................................... 173

Moderator Variables Coded ........................................................................................................................ 174

Intercoder Reliability .................................................................................................................................. 176

Meta-Analytic Procedures ......................................................................................................................... 177

RESULTS .......................................................................................................................................................... 178

Global Effect of Violent and Nonviolent Media on Aggression ................................................................. 178

Outliers ...................................................................................................................................................... 179

Moderators ................................................................................................................................................. 179

Publication Bias Analyses .......................................................................................................................... 181

DISCUSSION .................................................................................................................................................... 182

Main Findings............................................................................................................................................ 182

Magnitude of Average Effect Sizes ............................................................................................................ 183

Limitations and Future Research ............................................................................................................... 184

Conclusions ................................................................................................................................................ 185

DISCUSSION .................................................................................................................................................. 209

AN OBJECTIVE INSTRUMENT OF EVALUATION OF SEXUALIZED AND ATTITUDE CONTENT OF VIDEO GAME

........................................................................................................................................................................ 212

IMPACT OF SEXUALIZED CONTENT IN VIDEO GAMES ON AGGRESSIVE BEHAVIOR AND ATTITUDES AGAINST

WOMEN .......................................................................................................................................................... 216

A SYSTEMATIC REVIEW ABOUT THE IMPACT OF SEXUALIZED CONTENT ON AGGRESSION ......................... 221

GENERAL DISCUSSION ................................................................................................................................... 223

Theoretical Implications ............................................................................................................................. 223

Directions for Future Studies..................................................................................................................... 228

Practical Implications................................................................................................................................. 230

CONCLUSION.................................................................................................................................................. 232

REFERENCES .................................................................................................................................................. 235

ANNEX .............................................................................................................................................................. 261

ANNEX 1: SUPPLEMENTARY MATERIAL FROM ‚THE DEVELOPMENT AND VALIDATION OF AN OBJECTIVE

MEASURE OF THE SEXUALIZED CONTENT OF VIDEO GAMES: THE VIDEO GAME SEXUALIZATION

PROTOCOL‛ .................................................................................................................................................... 262

ANNEX 2: JOKES USED IN ‚EFFECTS OF SEXUALIZED VIDEO GAMES ON ONLINE SEXUAL HARASSMENT‛ 278

ANNEX 3: SUPPLEMENTARY MATERIALS FROM ‚EFFECTS OF SEXUALIZED VIDEO GAMES ON ONLINE

SEXUAL HARASSMENT‛ ................................................................................................................................. 283

ANNEX 4: SUPPLEMENTARY MATERIALS FROM ‚IMPACT OF SEXUALIZED VIDEO GAME AND COGNITIVE

LOAD ON RAPE MYTH ACCEPTANCE AND DEHUMANIZATION OF THE PERPETRATOR‛ ............................ 285

Prologue

If Pac-Man had affected us as kids, we’d all be running around in dark rooms,

munching pills and listening to repetitive electronic music.

– Marcus Brigstocke, English comedian, actor and satirist

Whether you find them stupid and useless or fun and useful, no one can deny that

video games are part of our everyday life and have an impact on our culture. Video

games have created icons known by everybody such as Mario, Pac-Man, Lara Croft,

and Pikachu. Even if you never touched a video game console in your life, you are

probably familiar with these characters. Video games started on computers and

consoles, but they soon invaded movie theaters, TV screens, literature, and the other

forms of media. Currently, Grand Theft Auto V is considered the most profitable

cultural product of all time (Cherney, 2018). In short, video games are everywhere

and even if you do not like them, you have to deal with them.

Video game players are always able to tell you about the positive impact that

video games have had for them. Some may just say that video games are fun and

help them relax. Others may tell you that the story of a certain video game almost

made them cry – they definitely cried, but they might never admit it. Video game

players sometimes even explain that video games have taught them important things

(e.g., patience, perseverance, friendship, history). However, if they can admit that

video games had a positive influence on them, they rarely admit that they can have a

negative influence.

Potential negative consequences of video games have always been a worry. More

precisely, the content of video games and its potential impact have raised a lot of

concerns. Recently, awareness has been raised among the general public about the

presence of sexualized and sexist content in video games. Both Anita Sarkeesian (in

the United State) and Mar_Lard (in France) have pointed out that women are poorly

represented in video games, are often represented as damsels in distress or as sex

objects. They also denounce the harassment and discrimination faced by female

players in the video game community.

Despite the many specific examples given by Anita Sarkeesian and Mar_Lard, the

real consequences of sexualized video game content still remain unclear. Scientific

research is needed to determine these consequences. The present thesis focuses on

one main question: can sexualized content from video games impact aggression

against women?

Prologue

Five studies will be presented. In the first study, an objective instrument of

evaluation of sexualized content was designed and validated. Three experiments

were then carried to examine the possible impact of sexualized content on sexual

harassment (Study 2), general negative attitudes toward women (Study 3), and

dehumanization and rape myth acceptance (Study 4). Study 4 also evaluated the

impact of sexualized media on aggression in video games but also in other forms of

media.

Theoretical Section

Chapter 1: Aggression against Women

The concept of aggression is paradoxically both simple and complex. Aggression

is simple because most acts of aggression are easily identified, such as by witnesses.

Yet, aggression is complex for two reasons. First, the term aggression is used

differently by lay people and researchers. For example, during a heated argument,

one might be qualified as aggressive because he used a stronger voice. Second,

aggression includes a large variety of behaviors and involves several moderators.

Definition

Among researchers, aggression is defined as any behavior intended to harm

another person who does not want to be harmed (Baron & Richardson, 1994). This

definition includes four important features (Bushman, 2017). First, aggression is an

external and visible behavior (e.g., swearing, hitting, slapping and so on). Even if

aggressive thoughts and emotions exist (see below regarding ‚aggressive cognition

and affects‛), they cannot be qualified as aggression. Second, aggression involves at

least two people, which means that acts of self-harm do not qualify as aggression.

Third, aggression is intentional rather than accidental. For example, painful actions

that help rather than hurt the person (e.g., emergency stitches) are not considered

aggressive behaviors. Fourth, the victim wants to avoid being harmed. Sado-

masochistic practices are therefore excluded from aggression. Finally, it is important

to note that failing to harm another while intending to, is still considered as an act of

aggression. For example, throwing a stone at someone is aggressive, regardless of

whether the stone hits the victim. Based on this definition, the previous example is

not aggressive. The person that raised his voice is only trying to give more weight to

his point of view without intending to harm anybody.

According to this definition, aggression includes a wide variety of behavior that

can range from mild (e.g., dirty look) to extreme violence (e.g., mass shooting).

Aggression must be distinguished from violence. Violence can be defined as any

behavior intended to cause extreme physical harm, such as injury or death, to

another person who does not want to be harmed (Bushman, 2017). For example,

pushing someone to get him out of the way will be considered as aggression, but not

violence, whereas intentionally pushing someone down the stairs would be an act of

violence. All acts of violence are aggression, but not all acts of aggression are

violence.

Chapter 1: Aggression against Women

Aggressive Cognition and Affects

Aggressive cognition and aggressive affect are important precursors to aggression

(J. J. Allen & Anderson, 2017). Aggressive cognition is defined as thoughts,

memories, and ideas that are associated with aggression and violence. Aggressive

cognition includes aggressive beliefs and attitudes (e.g., believing that it is acceptable

for a husband to hit his wife), aggressive perceptual schemata (e.g., the tendency to

perceive ambiguous situations in a hostile manner, such as believing that your wife is

angry because she is quiet), aggressive expectation schemata (e.g., expecting

aggressive comments from a feminist friend), aggressive behavioral scripts (e.g.,

thinking that it is necessary to play rough to sexually turn a woman on), and hostile

attribution bias (e.g., interpreting the intention of others as aggressive, Anderson &

Bushman, 2002). Aggressive affect includes feelings of anger, hostility, and irritability

(Anderson & Bushman, 2002).

Both aggressive cognition and aggressive affect cannot be considered as

aggression because they are not external and visible behaviors. However, their

presence increases the probability of emitting aggressive behavior (J. J. Allen &

Anderson, 2017). Indeed, those that have easily accessible aggressive cognition tend

to perceive their surrounding as more hostile and to find aggressive solutions for

their interpersonal conflicts. Similarly, aggressive affect can be caused by various

situations (e.g., provocation), and will facilitate aggressive behavior. In other words,

aggressive cognition and aggressive affect are mediators of aggressive behavior

(Anderson & Bushman, 2002). However, neither aggressive cognition, nor aggressive

affect are necessary or sufficient conditions to cause aggressive behavior. For

example, someone can leave a provocative situation even if they have aggressive

thoughts and feel angry inside. Similarly, aggressive behavior can occur without the

presence of aggressive affect or cognition (e.g., acts of torture). Such kind of

aggressive behavior is called instrumental aggression.

Aggressive Behavior against Women

In western society, several aggressive and violent acts mostly target women

compared to men. At the top of the list come all types of sexual aggression. In the

United States, women are four times more likely to be sexually assaulted and twice

more likely to be sexually harassed than men (Stop Street Harassment, 2018).

Prevalence numbers show that around 63% of women have been sexually harassed

once in their life (FRA, 2014; Stop Street Harassment, 2018). Violence and sexual

violence by an intimate partner (22%) accounts for two thirds of the global violence

Theoretical Section

and sexual violence committed against women (31.8%; FRA, 2014). During their

lifetime, around 18.3% of women have been raped compared to 1.4% of men (CDC,

2010). European reports are more encouraging, but are still discouraging — 1 out of

20 women have been raped during their lifetime (FRA, 2014). Among all victims of

human trafficking in Europe and North America, 76.5% are women or girls and

mostly for sexual exploitation (UNODC, 2016).

Non-sexual violence against women includes a higher risk of violence in an

intimate environment. For example, homicides tend to affects men four times more

often than women, except for intimate partner/family-related homicides where two

thirds of the victims are women (UNODC, 2013). Around 20% of women have

experienced violence from a partner in Europe and North America, compared to

around 12% of non-partner violence (WHO, 2013). Concerning online harassment,

men and women are globally equally vulnerable to insults, bullying and physical

treats. However, stalking and sexual harassment occurs 2 to 3 times more often to

women than men (Pew Research Center, 2014).

Although it cannot be qualified as aggression (i.e., because we do not know if the

motives are to hurt women), it is important to mention that women also suffer direct

discrimination in various environments. Women are less likely than men to access

senior and middle management positions and, for equivalent job positions, they

usually earn significantly less than men (Blau & Kahn, 2016; United Nations Statistics

Division, 2015). Access to some type of job is more difficult for women because they

are considered as less competent than men for those jobs, for example, less than 30

percent of the world’s researchers are women (Computer Science, 2018) and men are

four times more likely to be hired in technology jobs (Brown, 2018). Further, some of

these discriminations are even institutionalized. Indeed, 155 countries have at least

one law discriminating on basis of gender (World Bank Group, 2016). In most

countries, women have to pay more than men for similar products. This phenomena

is called the pink tax and concern several product that cost more for women (e.g.,

toiletries, toys, clothing; Cone, 2019).

This list of aggressive behaviors against women is not exhaustive, but

demonstrates the importance of studying the phenomenon. Aggression against

women is an important societal problem. Despite obvious consequences such as

injuries and death, violence and sexual violence by a partner causes sexual health

problems (e.g., STDs), mental health difficulties (e.g., higher risk of depression,

anxiety, PTSD, alcohol use disorders), and even suicide (WHO, 2013). Each of these

behaviors have a cost due to increased demand in medical care, health care, police

Chapter 1: Aggression against Women

intervention, diminution of the victim quality of life and productivity. For example,

in the United States, rape alone has an estimated cost for society of $129 billion per

year (National Institute of Justice, 1996).

Aggressive Attitudes toward Women

Attitudes can be defined as general evaluations that people hold about

themselves, other people, objects, and issues (Petty & Cacioppo, 1986). Among the

various aggressive attitudes toward women, four seem particularly important for

predicting aggression against women: sexism, rape myth acceptance,

dehumanization, and sexual objectification.

Sexism is probably one of the most studied aggressive attitudes in the literature.

Sexism can be defined as ‚a multidimensional construct that encompasses two sets of

sexist attitudes: hostile and benevolent sexism‛ (Glick & Fiske, 1996). Hostile sexism

can be defined as all aspects of sexism that correspond to hostility and prejudice

toward women. Hostile sexism is based on a hostile form of sexuality and on the idea

that men are superior and more dominant than women (Dardenne, Delacollette,

Grégoire, & Lecocq, 1996). Examples of hostile sexism include sexual harassment,

sexist comments or jokes, and physical violence. Benevolent sexism is often wrongly

perceived as being a positive form for the perceiver. This set of interrelated attitudes

tend to elicit outwardly pro-social or intimacy-seeking behaviors, but still places

women in an inferior position by viewing them in a stereotypical manner and in

restricted roles. An example of benevolent sexism would be helping a woman to

carry something heavy, which implies that she is too weak to do it herself.

Benevolent sexism can be separated into three main forms that are: protective

paternalism (i.e., women should be loved, protected and cherished because they are

too weak to protect themselves), complementary gender differentiation (i.e., women

complement men because of their purity, sensitivity, culture and moral sense), and

heterosexual intimacy (i.e., men and women are incomplete without each other).

Rape myth acceptance can be defined as any belief that will cause to trivialize and

minimize the act of rape, or to hold the victims partially or fully responsible for being

raped (Burt, 1980; Lonsway & Fitzgerald, 1994; Loughnan, Pina, Vasquez, & Puvia,

2013). Such beliefs can concern the rape, the rape victim, or the rapist. Rape myths

take many forms but can be generally grouped into four main categories (McMahon

& Farmer, 2011): (1) the victim had it coming (e.g., her outfit was immodest and has

caused the sexual desire of the perpetrator), (2) the rape cannot be considered as such

(e.g., the victim did not clearly say ‚no‛ to the perpetrator), (3) the perpetrator

Theoretical Section

cannot be hold responsible (e.g., he was too drunk to understand what he did), (4)

the victim lied about the rape (e.g., she cheated on her boyfriend so she pretended to

be raped). A recent example of rape myth acceptance occurred during a trial in

Ireland (BBC, 2018). A perpetrator was acquitted because his victim – a 17-year-old

adolescent – was wearing a thong at the time of the rape. The thong was used as a

proof for the victim consent.

Dehumanization is a process through which a person is treated as an animal, an

object, or not completely human (Gervais, Bernard, Klein, & Allen, 2013; Haslam,

2006). The dehumanized person is delegitimized as a human being and morally

excluded (i.e., moral values do not apply to that person). By dehumanizing others,

one disengages himself/herself from moral self-sanction (Bandura, 2002). For

example, a rapist might dehumanize his victim, which means that he will not feel

guilt or remorse because usual moral values do not apply to the victim. Two main

forms of humanness can be denied the person (Haslam, 2006): human uniqueness

and human nature. Human uniqueness includes characteristics that distinguish

human and other related categories of animals. Among these characteristics are

social learning, refinement, civility, moral sensibility, rationality, logic, and maturity.

Human nature includes characteristics that can be found among other species, but

that are still fundamental and shared by all humans. For example, creativity might

not be unique to human, but is an essential attribute of humanity. Such

characteristics are emotional responsiveness, interpersonal warmth, cognitive

openness, agency, individuality, and depth. The denial of these two forms of

humanness corresponds to two forms of dehumanization. Denial of human

uniqueness is called animalistic dehumanization and denial of human nature is

called mechanistic dehumanization. More concretely, those that are animalistically

dehumanized are perceived as being more animal than human. In other words, they

are perceived as being immoral and immature, and their behavior appears to be

driven by instincts, motives, and appetites, instead of mediated by cognition. Those

that are mechanistically dehumanized are perceived as objects or automaton. They

are seen as lacking in depth, as being cold, rigid, superficial, incapable of agency,

interchangeable, and passive. Their behaviors are interpreted as the consequence of

causal events instead of their own will. Even if dehumanization is not specifically an

aggressive attitude toward women, dehumanizing women causes aggressive

behavior toward them, such as sexual aggression, sexual harassment, and

objectifying gazes (Gervais et al., 2013; Rudman & Mescher, 2012).

Chapter 1: Aggression against Women

Objectification can be considered as a specific form of dehumanization. When

applied to women, one of the most common forms of objectification is sexual

objectification. Sexual objectification occurs when a woman become instrumental and

is not seen as an entire human being because one focuses on someone’s physical

attributes or sexual function, instead of focusing on his/her face and other non-

observable attributes, such as thoughts, feelings, and desires (Loughnan et al., 2010;

Vaes, Paladino, & Puvia, 2011). In other words, when women are objectified, their

sexual body parts or functions are separated from their person (Gervais et al., 2013).

Sexual objectification provokes a large variety of behaviors that range from

inappropriate compliments and sexual gaze on the one hand, to sexual assault,

exploitation and trafficking on the other hand. Concrete examples of sexual

objectification can be found among the clients of strippers and prostitutes.

In the literature, a large variety of attitudes toward women exists. However, these

attitudes are often more specific forms or mixed forms of sexism, rape myth

acceptance, dehumanization, and sexual objectification. As example of a specific

form of hostile sexism is attitudes toward women as a manager (Peters, Terborg, &

Taynor, 1974), which designates the attitudes about women’s right and abilities in a

business setting. An example of rape myth acceptance and hostile sexism combined

is adversarial sexual beliefs, which refers to the expectation that sexual relationships

are fundamentally exploitative, that each party to them is manipulative, sly, cheating,

opaque to the other's understanding, and not to be trusted (Burt, 1980). All of these

aggressive attitudes toward women interact with each other and can facilitate

aggression.

Summary

Aggression is a concept that includes a wide range of behaviors from minor acts

(e.g., name calling) to severe ones (e.g., murder). Aggression is influenced by several

mediators, including aggressive attitudes, cognition, and affect. Women are the

prime victims of several specific aggressive behaviors. Compared to men, women are

more often the victims of sexual aggression, sexual harassment, violence committed

by a partner, murder by a partner, stalking, and discrimination. These acts of

aggression are particularly influenced by four aggressive attitudes toward women,

namely sexism, rape myth acceptance, dehumanization, and sexual objectification.

Chapter 2: An Integrated Model to Explain Aggression

against Women: The Confluence Model Integrated with the

General Aggression Model

Two models are particularly relevant to explain aggression against women: the

General Aggression Model (GAM; Anderson & Bushman, 2002, 2018) and the

Confluence Model (Malamuth, Linz, Heavey, Barnes, & Acker, 1995). The first model

is a meta-theory that describes the general underlying mechanisms of aggression.

The second model focuses on the various factors that can specifically predict male-

on-female sexual aggression. Both models have been integrated together to predict

aggression against women (Anderson & Anderson, 2008). This chapter will first

briefly describes the five theories that have been integrated together to create the

original GAM (Anderson & Bushman, 2002), then describe the revised GAM

(Anderson & Bushman, 2018), the Confluence Model (Malamuth et al., 1995), and the

Confluence Model integrated into the GAM that was designed by Anderson and

Anderson (2008).

One Model to Integrate Them All

The original version of the GAM (Anderson & Bushman, 2002) was designed to

integrate and synthetize the five most used theory of aggression and learning at the

time: cognitive-neoassociationistic theory, social learning theory, script theory,

excitation transfer theory, and social interaction theory. It also extends these theories

in important ways and has four advantages compare to individual theories. First, the

GAM offers a more parsimonious theoretical framework. Second, it better explains

aggressive behaviors that are based on multiple motives (e.g., instrumental and

affect-based aggression). Third, the GAM aid for a better understanding of

aggression development across time.

Cognitive-Neoassociationistic Theory

According to cognitive-neoassociationistic theory (Berkowitz, 1990), aversive

events (e.g., provocation, frustration, loud noises) can produce negative affect that

will stimulate thoughts, memories, expressive motor reactions (i.e., ‚automatic

reactions that occur in conjunction with specific emotions, largely in the face‛,

Anderson & Bushman, 2002), and physiological responses associated with fight and

flight tendencies. Cognitive-neoassociationistic theory is based on the idea that

memory consists of a network of concepts and of links between these concepts (A. M.

Collins & Loftus, 1975). Each individual concept can be represented as a node, and

Chapter 2: An Integrated Model to Explain Aggression against Women: The

Confluence Model Integrated with the General Aggression Model

each relation between these concepts can be represented as links. Strongly related

nodes become knowledge structures that can include emotions, behavioral

responses, and beliefs. Nodes can be activated through priming, which leads to the

activation of these knowledge structures. For example, a man who holds attitudes of

hostile sexism and rape myth acceptance might associate together concepts such as

‚sexualized women‛ and ‚sexually available‛. If he tries to seduce a woman that

represents such concepts for him and is turned down, his frustration, associated with

the idea that women like to play ‚hard to get‛, might lead to sexual aggression.

Social Learning Theory

According to social learning theory, direct experience and observation are the two

main ways to learn aggression-related behaviors, attitudes, expectations, beliefs, and

perceptual schemata (Bandura, 1983, 2001; Mischel & Shoda, 1995). Such learning can

occur in a number of different contexts such as direct social interactions, watching

television or movies, or playing video games. Further, the learner not only observes

the behavior itself but also whether the behavior is rewarded or punished. Therefore,

a form of aggressive behavior might have been learned through observation, but the

probability of performing the behavior will depend on whether it was rewarded or

punished (Bandura, 1965; Bandura, Ross, & Ross, 1961). For example, one might

observe a peer performing sexual harassment behavior in the street, but avoid

performing it himself due to the presence of the police. However, later, when the risk

of punishment has vanished (e.g., the police have left), he might perform the

aggressive behavior in order to obtain the approval of his peers.

Script Theory

Scripts are well-learned, well-rehearsed, and highly associated concepts in mind,

which often involve an understanding of causation, goals, and plans of action

(Abelson, 1981; Schank & Abelson, 1977). Scripts are an important part of one’s

learning because they are used to organize knowledge, guide behaviors and define

situations. When facing a situation, a person first selects a corresponding script and

then assumes a role in the script. For example, a restaurant script helps one know

what to expect when going to a restaurant (i.e., enter restaurant, wait for a host, host

seats you at a table and gives you a menu, waiter/waitress takes your order, cook

prepares your order, waiter/waitress brings your food to the table, eat food, ask

waiter/waitress for a check, pay for food, exit restaurant). Therefore, scripts can be

considered as a group of strongly linked concepts in memory that become a unitary

concept. The more a script is rehearsed, the more accessible it becomes in memory.

Rehearsal increases the number of paths through which a script can be activated and

Theoretical Section

the strength of the path themselves (e.g., one might find really difficult to travel by

plane the first time, but the ‚airport script‛ will become stronger after going to an

airport several times). Scripts can guide every kind of behavior, including aggressive

behaviors. For example, being exposed to aggression through media can create

scripts that involve aggressive behaviors (Huesmann, 1998).

Excitation Transfer Theory

According to excitation transfer theory (Zillmann, 1988), when two events occur

within a short amount of time, the arousal caused by the first event might be

misattributed to the second event, even if both events do not share the same hedonic

value. Therefore, if a first event causes arousal (e.g., exercising), and a second event

causes joy (e.g., receiving some good news), then the arousal of both events will add

up to cause even more joy. In the context of aggression, the arousal from a first event

(e.g., having a car accident) can add up with the arousal from an event that caused

anger (e.g., being provoked by the other driver). Further, if the arousal is consciously

attributed to anger, anger can last for long periods of time, leaving the person more

prone to aggression even after the dissipation of the initial arousal. The anger can

also be prolonged through rehearsing the event in memory.

Social Interaction Theory

Social interaction theory is particularly appropriate when trying to explain goal-

motivated aggression (Tedeschi & Felson, 1994), also called ‚instrumental

aggression.‛. In this theory, aggression is defined as a social influence behavior (i.e.,

aggression can be used to cause changes in the victim’s behavior). For example,

aggression can be used to obtain valuables (e.g., sex, services, money, information),

to retaliate against a perceived injustice, or to bring about desired social and self-

identities (e.g., appear manly, competent, tough).

The General Aggression Model

The core of the General Aggression Model (GAM) has remained the same since it

was first proposed (Anderson & Bushman, 2002). However, slight modifications

have occur over time (e.g., Anderson & Bushman, 2018).

The General Aggression Model (GAM; Anderson & Bushman, 2002, 2018; Figure

1) is a model that describes each step leading to behavioral aggression. When facing a

situation that can potentially result in aggression, proximate and distal causes and

processes interact together to determine behavior. The core of the model – proximate

causes and processes – describe an episode of social interaction. During that episode,

Chapter 2: An Integrated Model to Explain Aggression against Women: The

Confluence Model Integrated with the General Aggression Model

the interaction between the characteristics of a person and the features of the

situation influence the present internal state. Modification of the present internal

state will influence a process of appraisal of the situation that might result in

aggression. The distal causes and processes described by the model are all the long-

term influences that have modified the personality of the person, which influences

the characteristics that a person brings into a social interaction.

Figure 1. General Aggression Model.

Proximate Causes and Processes

The core of the GAM is the proximate causes and processes consist of three main

parts: (1) person and situation variables, (2) present internal state, and (3) appraisal

and decision processes (Anderson & Bushman, 2018).

Person and situation variables. Aggression can be influenced by two types of

input variables: personal and situational variables. Personal variables are all the

characteristics that the person brings into the situation, such as gender, age, race, IQ,

biological dispositions, personality traits, beliefs, attitudes, values, scripts, long-term

goals, etc. Situational variables are all the features of the environment that can

Theoretical Section

influence aggression, such as presence of a weapon, a frustrating event, exposure to

violent and/or sexualized media, a provocation, drug and alcohol use, aggressive

peers, etc. Together, personal and situational variables can influence the individual’s

present internal state.

Present internal state. The present internal state is composed of three main routes

that interact with each other: cognition, affect, and arousal (Anderson & Bushman,

2002). Input variables (i.e., personal and situational variables) can prime or activate

aggressive thoughts, schemata, and scripts in memory. For example, playing a

violent sexualized video game might activate hostile thoughts against women in

memory. Affect corresponds to the negative or aggressive mood and emotions that

can be directly influenced by input variables, which can thereafter influence

behaviors (e.g., my car is damaged because of an accident, which makes me angry

and more aggressive toward the other driver). The last route is arousal, which can

initiate a person to be more ready to act (high arousal) or to inhibit (low arousal)

behavior. For example, it has been shown that when someone is aroused through

exercise, that person may act more aggressively (Zillmann, Katcher, & Milavsky,

1972). Each of these routes can either influence behavior directly or through their

interactions. For example, affect and cognition can interact with each other through

mood-congruent cognitions and mood-dependent memory (Buckley & Anderson,

2006). Mood-congruent cognition is a phenomenon in which a person’s mood

increases the processing of information that is affectively similar (e.g., depressed

people have more difficulty to disengage their attention from sad faces than other

expressions; Sears, Thomas, Lehuquet, & Johnson, 2010). Mood-dependent memory

is the finding that information learned in a particular mood is best retrieved in that

mood (e.g., depressed people better recall faces with a negative facial expression than

faces with a positive facial expression; Ridout, Astell, Reid, Glen, & O’Carroll, 2003).

In other words, people tend to pay more attention to mood-matching information

and are more likely to store it in memory.

Appraisal and decision processes. The last step before the apparition of

aggressive behavior is the appraisal and decision processes. The modification of the

internal state caused by the inputs will provoke an immediate appraisal of the

situation. Such automatic (i.e., effortless, occurring without awareness)

interpretations of the situation can result in an impulsive behavior. For example,

after being angered by his wife, a man could slap her. However, if the immediate

appraisal is judged to be unsatisfactory and if the person has sufficient time and

cognitive resources, then the situation might be reappraised and may lead to a

Chapter 2: An Integrated Model to Explain Aggression against Women: The

Confluence Model Integrated with the General Aggression Model

thoughtful action (e.g., revising the initial judgment). For example, if the man does

not want to hit his wife – the outcome is unsatisfactory – he might choose to storm

out of the room instead. However, the result of an immediate appraisal is not

automatically aggression (e.g., facing a harasser, the impulsive reaction of a woman

might be to flee). Similarly, the reappraisal process leads to a thoughtful action that

might be aggressive (e.g., the same woman might find it unsatisfactory to flee in

front of a harasser and therefore might choose to retaliate). Whether or not the

immediate appraisal and reappraisal lead to an aggressive behavior is dependent on

all the preceding mechanisms (i.e., person and situation inputs, and changes to the

present internal state).

Distal Causes and Processes

The distal causal factors of aggression refer to stable factors that influence the

development of personality (Anderson & Bushman, 2018). Distal factors include both

biological and environmental modifiers. Biological modifiers include all the

biological elements of a person that can influence aggressiveness such as impulsivity

deficits, Attention Deficit Disorder/Attention Deficit Hyperactivity Disorder, various

genetic risk factors, hormones, etc. Similarly, environmental modifiers include all the

long-term influences caused by one’s environment such as poor parenting, antisocial

peers, growing up in a violent neighborhood, repeated exposure to violent media,

etc. Each of these variables creates and reinforces beliefs, attitudes, perceptual

schema, expectation schema, behavior scripts, and affective components that are part

of one’s personality.

The GAM can be seen as a closed circle in which distal and proximate causal

factors influence each other. The distal aspects of personality will influence the

various elements that one brings into the proximal personal variable, as well as the

interpretation of proximal situational variables. For example, a man that has

developed hostile sexism through repeated exposure to pornography can more easily

activate such beliefs when discussing with a feminist, which can lead to the

activation of hostile sexist thoughts. In return, each social encounter causes learning,

and the repetition of learning can lead to the modification of the personality. For

example, someone living in a nonviolent neighborhood can expect people in the

street to be nice and helpful. However, if the same person moves into a violent

neighborhood, situations of aggression might occur more often and each of these

social encounters might create the expectation schema that the street is a dangerous

place where one can be assaulted at any moment.

Theoretical Section

The Confluence Model

The Confluence Model (Malamuth, 2003; Malamuth et al., 1995; Vega &

Malamuth, 2007) was initially designed to specifically explain sexual aggression. The

confluence model includes two main groups of variables that predict sexual

aggression against women: hostile masculinity and promiscuous-impersonal sex.

Hostile Masculinity

Hostile masculinity can be described as a personality profile that combines (1)

hostility, distrust, insecurity, defensive, and hypersensitive attitudes toward women,

and (2) willingness to control and dominate women. High hostile masculinity is

related to fear of rejection by women, anxiety about romantic relationship, and

feelings of threat, especially by women in positions of power. In this context,

dominance is a way to maintain control over women.

Further, hostile masculinity accounts for two individual differences: masculine

gender role expectations and general hostility. Stress associated with the traditional

masculine role may cause men to believe that women can challenge their role

fulfillment. In other words, in order to be a ‚real man‛, one must embrace masculine

roles defined by society and culture such as seeking power, toughness, dominance,

aggressiveness and competitiveness. In contrast, stereotypically feminine qualities

must be rejected, such as softness, empathy, and sensitivity. Displaying traditionally

feminine qualities for a hostile masculine male might be associated with loss of male

identity and lead to dominating and/or aggressive behavior in order to recover

masculine superiority. The second individual difference that was accounted for in

hostile masculinity is general hostility. General hostility is a predictor of more

general aggression against women (Malamuth et al., 1995). Indeed, general hostility

appears to be a good predictor of nonsexual aggression, whereas hostile masculinity

is a good predictor of sexual aggression.

Promiscuous-Impersonal Sex

Promiscuous-impersonal sex is defined as the tendency to have a noncommittal,

game-playing orientation in sexual relations (Malamuth et al., 1995). The

promiscuous-impersonal sex concept is similar to the sociosexuality concept, which

corresponds to individual differences in willingness to engage in sexual relations

without closeness or commitment (Gangestad & Simpson, 1990; Simpson &

Gangestad, 1991). Individuals that are high in promiscuous-impersonal sex are more

likely to report having sex earlier in their relationships, more than one current sexual

Chapter 2: An Integrated Model to Explain Aggression against Women: The

Confluence Model Integrated with the General Aggression Model

relationship, many different sexual partners in the past, partners with whom they

had sex only once, and expecting to have many different partners in the future. Such

attitudes toward sexuality and sexual partners have been associated with sexual

aggression and distress in marital relationships (i.e., men with a sociosexual

orientation tend to be less monogamous, which creates tensions in their current

relationship).

The Confluence Model Integrated into the General Aggression Model

The Confluence Model was initially independent from the General Aggression

Model. However, both models have benefits and limitations. The General Aggression

Model has been proven to be highly relevant to predict all kinds and forms of

aggression (Anderson & Bushman, 2002, 2018), but might lack specificity, especially

in its ability to predict more specific forms of aggression such as aggression against

women. In contrast, the Confluence Model proposes various factors that can predict

aggression against women but neglects situational factors such as provocation and

the intermediate psychological processes (i.e., cognition, affects, and arousal;

Anderson & Anderson, 2008). Thus, Anderson and Anderson (2008) integrated the

Confluence Model into the GAM, and found that the integrative version of the model

is highly relevant to predict aggression against women.

All the variables that are proposed as predictors of aggression in the Confluence

Model can be considered as person input variables in the GAM (Anderson &

Anderson, 2008) (Figure 2). When designing the model that integrates the Confluence

Model into the GAM, seven variables have been shown to predispose men to aggress

against women: (1) general hostility, (2) general aggressiveness, (3) violent attitudes,

(4) hostility to women, (5) impersonal-dominant sex, (6) aggressiveness to women,

and (7) violent attitudes to women. The integration of both models has also allowed

the identification of three particularly important situational variables: (1)

provocation, (2) gender of the target, and (3) aggression opportunity. Such variables

may influence one’s present internal state. Further, the Confluence Model adds to the

potential choice of aggressive or non-aggressive behaviors by influencing the two

types of appraisal processes described in the General Aggression Model. Indeed, the

various personal variables proposed by the Confluence Model suggest that different

paths might influence the choice of using verbal, physical, and sexual aggression. For

example, if a man holds hostile attitudes toward women (person input), he might feel

more depreciated and angered (present internal state) by a woman who has rejected

his sexual advances. He might choose to follow her (thoughtful action) and later

Theoretical Section

assault her when she is alone (situational variable). Similarly, if the same man has

less control over his reaction (e.g., he is drunk), he might immediately insult her

(impulsive action).

Figure 2. Confluence Model integrated into the General Aggression Model (Anderson

& Anderson, 2008).

Summary

Two models that are particularly relevant to explain aggression against women

are the General Aggression Model (GAM; Anderson & Bushman, 2002, 2018) and the

Confluence Model (Malamuth, 2003; Malamuth et al., 1995; Vega & Malamuth, 2007).

The GAM is a meta-theory that was initially design to regroup and integrate

mechanisms issued from cognitive-neoassociationistic theory, social learning theory,

script theory, excitation transfer theory, and social interaction theory. According to

Chapter 2: An Integrated Model to Explain Aggression against Women: The

Confluence Model Integrated with the General Aggression Model

the GAM, situation and person inputs cause changes in the person’s present internal

state, which, in turn, can lead to aggressive behavior after either an automatic or

controlled appraisal process. The Confluence Model is a more specific model that

predicts aggression against women using variables issued from two main paths:

hostile masculinity and promiscuous-impersonal sex. Recent work has integrated the

Confluence Model into the General Aggression Model (Anderson & Anderson, 2008).

Indeed, the individual factors described by the Confluence Model can be integrated

as person inputs in the social encounter component of the GAM. According to this

integrated model (Anderson & Anderson, 2008), the person inputs that predict the

best aggression against women are: general hostility, masculine gender role stress,

impersonal sex, hostile masculinity, and violent attitudes toward women.

Chapter 3: Representation of Women and Men in Video

Games

Aggression against women is a problem in our society and the roots of this

problem are complex and multi-determined. The causes of aggression against

women can be examined at a societal, dyadic, individual and situational level (Krahé,

2017). For example, aggression against women is more likely in societies that

promote patriarchal values (i.e., societal explanation). Low marital satisfaction is one

of the best predictor of partner violence (i.e., dyadic explanation). Individual level

risk factors of aggression against women are being younger, less educated, and less

affluent. Examples of situational precipitators of violence against women are alcohol

and drug consumption.

A recent situational precipitator might be sexualized and sexist video games.

Indeed, video games are a recent media that started in the early seventies and is

continuously growing since to become a billion dollar industry (WePc, 2018). In

Western society, video games are played by around 64% of the population

(Interactive Software Federation of Europe, 2018; Nielsen Games, 2017; UKIE, 2018).

When picturing the stereotypical video game players, one typically imagines a young

male. However, several surveys reveal that adolescents and young adults (i.e.,

between 16 and 24 years of age) only represent 26% of players, and that almost as

many women play video games as men (Interactive Software Federation of Europe,

2018; UKIE, 2018). Despite this equivalent number of female and male players, video

games are often described as being a hostile environment for female players (Gray,

2012a, 2012b; Kuznekoff & Rose, 2013). By consequence, video game exposure is

regularly suggested as a potential cause of aggression against women (Dill, Brown, &

Collins, 2008; Driesmans, Vandenbosch, & Eggermont, 2015; Fox & Potocki, 2016;

Yao, Mahood, & Linz, 2010).

This chapter provides definitions of sexualization, examines why video games

might be considered a hostile environment for women, and how this representation

might influence aggression against women. First, a description of the way women

are treated by the video game community and industry is provided. Second, current

content analyses will be described to better understand how male and female

characters are represented in video games. Third, it will be argued that content

analyses have significant limits and that a new objective and accurate classification

system is necessary in order to efficiently evaluate sexualization and roles in video

games.

Chapter 3: Representation of Women and Men in Video Games

Definining Sexualization

Two main definitions of sexualization have been proposed in the scientific

literature. Among these, one is a general one, the other is a multi-dimensional one.

The general definition of sexualization states that ‚sexualization occurs when a

person is held to a standard that equates physical attractiveness (narrowly defined)

with being sexy‛ (R. L. Collins, Lamb, Roberts, & Ward, 2010, p. 1). A second

definition of sexualization adds more precision to the first one by defining it as ‚a

number of complex, interacting factors, such as the extent of nudity and revealing

clothing and poses that are suggestive of sexual activity or availability‛ (Pacilli et al.,

2017). In this thesis, we will use the second definition of sexualization as it allows a

more precise operationalization of the concept of sexualization.

The Video game industry, the Video Game Community, and Video Game Content

as a Hostile Environment for Female Players

Video games in general convey negative attitudes toward women (Stermer &

Burkley, 2012). Attitudes such as objectification, dehumanization, sexism, and rape

myth acceptance can be found both in the video game industry, in the video game

community (Brehm, 2013), and in the video games themselves (Stermer & Burkley,

2012).

Few studies have focused on the video game industry. However, some authors

have suggested that sexualization and objectification of women in video game are

used as a selling point (Jansz & Martis, 2007; Near, 2013; Scharrer, 2004). For

example, one study found that that female sexualization in video game covers is

related to the financial success of the game (Near, 2013). In general, the video game

industry is dominated by males, which has caused an overall masculine culture in

the video game industry (Williams, 2006). In other words, video games are created

by men for men and with the stereotypical idea of what men want in a video game.

Many anecdotal examples of the consequences of such a culture range from the

objectification and dehumanization of female characters in advertisements (e.g., an

advertisement from PlayStation presented a woman’s bust with breasts on both sides

with the text: ‚Psvita, two tactile interfaces, twice more sensations‛) to objectification

and even sexual harassment of real women (e.g., video game conventions often hire

‚babes,‛ which are women wearing revealing outfits to staff the booths). All these

elements convey the idea that male players are the dominating population in video

games (Williams, 2006).

Theoretical Section

This overall idea that video games are essentially a male activity designed for men

has led to consider female players as a minority that suffers the consequences

(Brehm, 2013; Gray, 2012a; Kuznekoff & Rose, 2013). One online study (Brehm, 2013)

examined in detail the sexist behavior of players in World of Warcraft, a massively

multiplayer online role-playing game. The study used both multiple choice questions

and open-ended question. Using both kinds of questions, participants were able to

communicate if they had personally experienced sexism or if they had witnessed

another player experiencing sexism, and to describe the sexist event. Based on the

data collected from 294 participants, male video game players expressed both hostile

and benevolent sexist attitudes toward women. Female video game players

experienced both hostile (e.g., verbal aggression, sexual harassment, discrimination)

and benevolent (e.g., some male players offer gifts to obtain favors from female

players, and often treat them as if they are precious and delicate) sexist related

behaviors. Negative attitudes sometimes even involved rape myth acceptance with

the common use of rape jokes by male video game players.

A regularly suggested cause of these aggressive behaviors against women is the

physical and the role representation of male and female video game characters (Dill

et al., 2008; Driesmans et al., 2015; Fox & Potocki, 2016; Yao et al., 2010). Indeed,

female and male video game characters are often depicted in stereotypical ways,

both physically and in terms of their roles within the game.

Video Game Content Analysis

Usually, content analyses are used to examine the representation of male and

female characters in video games. Six content analyses are particularly important.

Two of them described the evolution of the portrayal of female characters over time

(Lynch, Tompkins, van Driel, & Fritz, 2016; Summers & Miller, 2014), two used the

same classification system for both male and female characters, thus giving the

possibility of comparing both studies (Burgess, Stermer, & Burgess, 2007; Downs &

Smith, 2010), and two were the only ones to focus only on male characters (Martins,

Williams, Ratan, & Harrison, 2011; M. K. Miller & Summers, 2011). Each content

analysis is briefly described below.

Of the two content analyses that described the evolution of female characters over

time, one examined the years 1983 to 2014 (Lynch et al., 2016) whereas the other

examined the years 1988 to 2007. Both examined the sexualization of female

characters, but focused on different attitudes. One focused on the capacities of female

characters (Lynch et al., 2016) whereas the other focused on benevolent sexist

Chapter 3: Representation of Women and Men in Video Games

attributes endorsed by female characters (Summers & Miller, 2014). They also used

different sampling methods.

More precisely, the 1983 to 2014 content analysis focused on the evolution of

sexualization and competence of playable female protagonists (Lynch et al., 2016).

They excluded video games that contained non-anthropomorphized characters,

erotic video games, and video games that contained characters from pre-existing

media franchises to only focus on characters exclusively created for video games.

Their sampling method resulted in 571 video games, which were analyzed for a

randomly selected 5-minute of gameplay. Sexualization of female characters was

analyzed by coding if bare skin was exposed for breasts, buttocks, waists, and legs, if

certain areas (i.e., breasts, buttocks, and waist-to-hips ratio) had a disproportionate

size compared to the body, and if certain areas (i.e., breasts and buttocks) were

accentuated by garments or artistic styling (e.g., shadings). They also coded for

sexualized movement, described as ‚unnecessary undulation or jiggling that drew

attention to their body in a sexual manner‛. Finally, they coded for physical capacity

(i.e., ‚if they engaged in feats of physical strength or agility aside from routine

activities such as walking or picking up objects‛) and degree of violent portrayal (i.e.,

‚if they engaged in threats of physical force or use of such force against an animate

being or group of beings‛). The conclusions from this study were that sexualization

first increased (from 1983 to 2006) then started decreasing in 2007. Further, secondary

female characters were more sexualized than primary characters, and that sexualized

female characters were more capable.

The 1988 to 2007 content analysis examined articles from magazines that included

at least one full page of text and one picture that clearly described the video game

character. Using the text or the picture, coders rated (on 8-point scales): how sexy,

helpless, and innocent the character was, how revealing her clothing was, whether

she was a princess, and whether she was being rescued. A total of 223 female

characters from 175 game magazines articles were coded. Results showed that

attributes associated with benevolent sexism (i.e., needing rescue, helplessness,

innocence) tended to diminish over time, but not the overall presence of a princess.

However, attributes associated with sexualization (i.e., sexiness, revealing clothes)

tended to increase over time.

Both studies have similar limits. Indeed, the sampling method of both studies

might have biased their conclusion. The 1983 to 2014 content analysis (Lynch et al.,

2016) only focused on video games featuring a playable female character. By

choosing such a sampling method, the researchers excluded all video games

Theoretical Section

featuring a male character as the only playable character. This sampling method

excluded many video games in which female characters are not playable characters

(e.g., Grand Theft Auto V). The 1988 to 2007 content analysis (Summers & Miller, 2014)

only coded articles from magazines that described a female character. Therefore, the

female game character had to be sufficiently of interest to appear in a magazine

article along with a full page of text. This sampling method excluded several video

games in which female characters are of less interest. Another limit of both studies is

related to the degree of inaccuracy and subjectivity. For example, Summers and

Miller (2014) described innocence as ‚looking or acting innocent or sweet‛. Such a

description is too inaccurate and subject to the interpretation of each coder. Although

it can be concluded that Lynch et al. (2016) are precise concerning their evaluation of

sexualized content, their description of sexualized movement and physical capacity

are too subjective. For instance, concerning physical capacity, which activity can be

considered as a routine activity or not? Because of these limits, no clear conclusion

can be drawn about the evolution of female sexualization. Further, both studies

cannot be compared with each other because they used different classification

systems. Despite their limits, the authors of both content analyses concluded that

female characters are sexualized, and that their role seems to have evolved from

being submissive, helpless, and innocent, to being more capable and strong.

Several content analyses have included both male and female characters (Burgess

et al., 2007; Downs & Smith, 2010; Ivory, 2006; M. K. Miller & Summers, 2007; Near,

2013), which allows for a comparison between genders. Two of them appear to be

particularly relevant because they used similar sampling methods and classification

systems (Burgess et al., 2007; Downs & Smith, 2010). The earlier one used a thorough

classification system to analyze objectification (Burgess et al., 2007). First, a

distinction was made between objectification role and physical objectification.

Objectification role was coded using five variables: (1) being portrayed without

action, (2) being in a submissive role, (3) being a reward to others, (4) appearing

unnecessary to the goal of the game, and (5) requiring rescue without any clear

attempts to help oneself. Physical objectification was analyzed using any

combinations of five variables: (1) being portrayed in a sexy way, (2) having an

exaggerated bust (groin area for male characters), (3) wearing vulgar or tight clothing

that accentuates body parts, (4) wearing a revealing dress, and (5) only being

portrayed as part of a body without showing the face. They also evaluated the breast

size of female characters and the degree of muscularity of male characters. Breast size

was evaluated using three categories: normal, busty (i.e., ‚having breasts larger than

normal, but not unnaturally so‛) and super-busty (i.e., ‚unnaturally large breasts

Chapter 3: Representation of Women and Men in Video Games

that, due to size and/or shape, were not natural‛). Similarly, muscularity was

evaluated using three categories: normal, muscular (‚healthy body that had muscles

clearly evident; the muscles were not emphasized but readily visible‛), and super-

muscular (i.e., ‚specific muscle groups being visible, unnaturally large muscles,

and/or a muscle group larger than another part of the body that should not normally

be comparable‛). The authors analyzed the portrayals of female and male characters

depicted on video game covers. Results showed that male characters were twice

more likely to be represented on video game covers than female characters. Female

characters were significantly more likely to be role objectified (33%) than male

characters (1%). Similarly, female characters were more likely to be physically

objectified (47%) than male characters (14%). Further, female characters were more

likely to be busty or super-busty (49%) than male characters were to be shown

muscular or super-muscular (26%). Limits of this study include inaccuracy,

subjectivity and poor choice of sampling method. First, instead of using a score to

evaluate role and physical objectification, authors only coded the presence or absence

of these variables. Therefore, a character that had multiple physical objectification

variables (e.g., wearing a revealing dress, having an exaggerated bust, and being

portrayed in a sex way) was considered as equally physically objectified compared to

a character with only one of these physical objectification variables. Further, their

categories were highly subjective. For example, breast size was coded using vague

categories such as: ‚normal‛, ‚busty,‛ and ‚super-busty‛. Busty was defined as

‚having breasts larger than normal, but not unnaturally so‛, super-busty was

defined as ‚unnaturally large breasts that, due to size and/or shape, were not

natural‛. All other types of breasts were defined as being ‚normal‛. This

operationalization of breast size is clearly subjective as well as potentially culturally-

based. What is considered as a ‚normal‛ or a ‚large breast size‛ can potentially vary

greatly from person to person, and from country to country. Finally, their sampling

method was not adapted for some variables. Indeed, the sample only contained

stationary images (i.e., video game covers), which is not suitable for analyzing

variables such as role objectification (i.e., being portrayed without action, being in a

submissive role, being a reward to others, appearing unnecessary to the goal of the

game, requiring rescue without any clear attempts to help oneself). That last

limitation was addressed in the second content analysis (Downs & Smith, 2010),

which analyzed the 60 top-rated games of the three main game consoles popular at

the time. The authors used two types of sampling units: (1) images of video game

characters, and (2) video segments from the video game itself. Specifically, images of

video game characters were used to analyze physical aspects of the characters,

Theoretical Section

whereas 20-minute gameplay videos were used to analyze character roles. Eight

variables were coded from videos of gameplay: (1) revealing clothing, (2) nudity, (3)

body proportion, (4) appropriateness of attire, (5) breast size, (6) waist size, (7)

presence of sex talk, and (8) presence of sexual behavior. Results showed that male

characters were overrepresented compared to female characters, and were more

often the primary character. Further, female characters were more likely than male

characters to be sexualized (i.e., wearing revealing clothing, exposing partial or full

nudity, having unrealistic body proportions, and having a small waist). Sex talk and

sexual behavior were not analyzed due to their infrequent occurrence, and breast size

was only evaluated for female characters. Like other content analyses, the

classification system was too inaccurate and subjective. For example, body

proportion ‚was gauged by the propensity of characters to resemble an average male

or female human‛. However, what resemble and average male or female human

would be highly dependent from one coder to another. These two studies underscore

the importance of an accurate and objective system of classification and sampling

method.

The two last content analyses are particularly relevant to this thesis because they

focused on the representation of male characters. One of them included precise

measures of muscularity (Martins et al., 2011), and the other described the evolution

of male video game characters over time (Miller & Summers, 2011). In the first

content analysis (Martins et al., 2011), the 150 top-selling video games were

examined. Of these, the games that included adult male human characters were

analyzed, which included 1074 male characters. Each character was 3D modeled by a

professional computer graphic artist. All the 3D models were scaled to the same

height. Muscularity was determined by measuring head, chest, waist and hips. These

measures were compared to a similar 3D model based on the average

anthropometric data provided by the CAESAR (Civilian American and European

Surface Anthropometry Resource; Harrisson & Robinette, 1998). Results showed that

in general male video game characters have significantly larger head, chest, waist

and hips than the average American man. Further, they have a significantly smaller

V-shape proportion compared to the average American man. In other word, male

video game characters are larger than the average man, but are not especially more

muscular. Further, video game characters were compared based on age rating of the

video game. Male video game characters have a larger V-shape proportion in games

rated E 10+ or younger than in games rated T or older. Based on these results, the

authors concluded that hypermuscularity is mostly present among video games

rated for children than among video games rated for adults. The measurement

Chapter 3: Representation of Women and Men in Video Games

method (i.e., comparing characters that are scaled to the same height) is the only one

that computes a score for muscularity which, if reproduced could help compare

various sample of video games. However, that method might be difficult to replicate

because it requires the involvement of a professional computer graphic artist.

The second content analysis (Miller & Summers, 2011) used magazine articles

from 1988 to 2007 as the sampling material. Masculinity was coded using six

characteristics associated with masculinity that are powerful, danger-seeking, anti-

social, unsympathetic, successful, and muscular. These characteristics were coded on

an 8-point scale. Other coded variables were: sexiness, attractiveness, innocence,

helpless, mad, happy, afraid, carefree, presence or absence of abilities and weapons.

Results showed that, across time, male characters have become more muscular,

attractive, sexy, and mad, but became also less happy, carefree, and innocent. Again,

inaccuracy and subjectivity were the main limitations of this content analysis.

Conclusions from both studies are not similar, with one study concluding male

characters are more muscular (Miller & Summers, 2011) and the second one showing

that male video game character are larger, but not especially more muscular, than the

average male population (Martins et al., 2011). This stresses the necessity for a

classification system that uses an objective measurement system similar to the one

used by Martins et al. (2011). Indeed, such a measurement method create a precise

score that can be used for statistical analyzes and eventual comparisons between

studies.

Physical representations of female and male characters in video games. Based on

the previous content analyses, several conclusions about the representation of male

and female characters can be drawn. First, female characters are globally

underrepresented in video games. Indeed, there are at least twice as many male

characters as female characters on video game covers and reviews (Ivory, 2006; Near,

2013; Stermer & Burkley, 2015), and over seven times more male characters as female

characters in gameplay footage (Downs & Smith, 2010). Further, all content analyses

concluded that female characters are sexualized. Indeed, female characters frequently

have large breasts and buttocks, small waist-to-hips ratios, large amounts of exposed

skin, often dressed in revealing clothing (e.g., revealing cleavage or midriff) that are

inappropriate for the task at hand, and are portrayed as sexy and attractive (Burgess

et al., 2007; Downs & Smith, 2010; Ivory, 2006; Lynch et al., 2016; M. K. Miller &

Summers, 2007; Near, 2013).

Concerning the physical representation of male characters, muscularity is the main

physical feature analyzed by content analyses. With one exception (Martins et al.,

Theoretical Section

2011), all content analyses have found that male video game characters are muscular,

especially compared to female characters (Burgess et al., 2007; Miller & Summers,

2007, 2011).

Role representations of female and male characters in video games. Female and

male video game characters mainly engage in stereotypical roles. Female characters

assume three main roles (Summers & Miller, 2014). The first role is described by

Summers and Miller (2014) as a benevolent sexist role and corresponds to being

portrayed as needing rescue, and being helpless and innocent. In other words, female

characters often adopt the role of a ‚damsel in distress‛. Further, they are often

passive and secondary characters (i.e., they are usually not playable characters,

possess less capacities than male characters, are not central in the game, and are

unnecessary to the goal of the game; Ivory, 2006; Miller & Summers, 2007; Near,

2013). The second role is being a sex object. Female characters are often objectified

(e.g., the cover of a video game showing only parts of the female bodies; Burgess et

al., 2007). They act as a reward for the player (e.g., in Resident Evil 5, when the players

finish the game, they are rewarded by being allowed to change the outfit of the

female character to a more sexualized outfit) or are sex objects (e.g., the strippers and

prostitutes in Grand Theft Auto V, Burgess et al., 2007). The third role combines three

characteristics: sexy, strong, and secondary (Lynch et al., 2016). An example of such a

character would be Mad Moxxi in the video game Borderlands. Mad Moxxi is

described as sadistic and dangerous (i.e., strong), is a non-playable character that is

the hostess of a fighting arena (i.e., secondary), and is wearing a highly sexualized

outfit that reveals her cleavage, thighs, and undergarments.

In video games, male characters assume two main roles. The first role is being the

main protagonists in the game. Most heroes in video games are male, and most

playable characters are male (Burgess et al., 2007; Downs & Smith, 2010; Ivory, 2006;

M. K. Miller & Summers, 2007). The second – that can pair with the first one – is

being hyper-masculine. Hyper-masculinity can be defined as an exaggeration of

‚macho‛ characteristics, where displays of emotion are considered a weakness,

where physical aggressiveness is a fundamental part of what it is to be masculine,

and where belittling women, engaging in romantic relations, and exhibiting risk-

taking behavior are also typical male characteristics (Scharrer, 2004). Indeed, male

characters are more aggressive, ‚mad‛ (i.e., have an angry appearance or try to hurt

someone), display more weapons and abilities, display more competence, and are

more powerful than female characters (Burgess et al., 2007; M. K. Miller & Summers,

2007, 2011). Male characters are also less happy and carefree (i.e., sweet and caring,

Chapter 3: Representation of Women and Men in Video Games

law abiding) and less innocent in most recent video games compared to less recent

video games (Miller & Summers, 2011).

Furthermore, these male and female roles are the expression of negative attitudes

toward women (i.e., sexism, dehumanization and objectification, rape myth

acceptance). Benevolent sexism is represented by the role of the damsel in distress

among females (Summers & Miller, 2014) and male characters often assume the role

of the hero that attempts to save her (e.g., in Super Mario Bros, the whole objective of

the video game is to save the Princess Peach). Hostile sexism is represented by two

roles, the role of the sexy and strong secondary female character (Lynch et al., 2016)

and the hyper-masculine male character (Miller & Summers, 2007, 2011). Indeed, the

sexy, strong and secondary female character is the female that threatens male

masculinity by seeking to control men, often through sex (Matthews, Lynch, &

Martins, 2016), and the hypermasculine male character exhibits violent macho

characteristics (M. K. Miller & Summers, 2007, 2011; Scharrer, 2004). For example, in

God of War 3, Kratos is a hyper-masculine male character that meets Aphrodite, the

goddess of love (a sexy, strong/powerful, and secondary character). During their

interactions, Aphrodite trades sex for information (i.e., controlled Kratos), and Kratos

is aggressive during their sexual intercourse. Female characters are dehumanized

and objectified when they assume the role of a sex object or reward for players

(Burgess et al., 2007; Stermer & Burkley, 2012). Although we could find no content

analyses that examined rape myth acceptance in video games, some anecdotal

example can be found (e.g., in God of War 3, Aphrodite enjoyed the rough sexual

intercourse). In summary, video game content often conveys negative attitudes and

beliefs toward women.

The Need for an Objective Instrument to Evaluate Video Game Content Regarding

Sexualization and Stereotyped Roles

It is practically impossible to compare the results of the different content analyses

in the literature, and therefore impossible to draw clear conclusions about the

evolution of both the stereotypical appearance and the roles of male and female

characters due to the variability of sampling methods and the inadequacy of most

classification systems. Sampling methods vary mostly in terms of four characteristics:

(1) examined material (e.g., video game covers, magazines, gameplay), (2) selection

of video games (e.g., including the 60 top-rated video games that have sold at least

10,000 copies), (3) type of characters analyzed (e.g., only female, only male, both male

and female; playable vs. non-playable characters), and (4) the age rating of the video

Theoretical Section

game. The variability of the sampling method is highly dependent on the main

objective of the authors. For example, in one study (Near, 2013), the objective was to

determine the relation between video game sales and sexualization. Therefore, the

sampling method only included video game covers because it is one of the most

important selling features of a video game in a store. Further, the authors focused on

video games sold mostly to young males, games rated T (for Teens 13+) or M (for

Mature players 17+) according to the ESRB (2018) system. The results of such a

sampling method meets its objective, however, it does not provide a clear picture of

the current state of the sexualized content of a large number of video games rated

appropriate for younger players (i.e., rated eC to E 10+), and also does not examine

the actual content of the video game (e.g., the cover of Batman: Arkham Asylum only

represents male characters and excludes Harley Queen, a sexualized female

antagonist). Other content analyses only focused on video games with female

playable characters (Lynch et al., 2016). Thus, no conclusions can be drawn about the

real evolution of female characters in video games. Indeed, most of the video games

featured a male character as the only playable character. For instance, in Darksiders,

the only playable character is the male character War. However, the video game still

pictured Uriel, a sexualized female character wearing skin-tight armor. Such video

game would be excluded using Lynch et al.’s sampling method and therefore, the

sexualized female character is not coded.

Further, the classification systems used by previous content analyses often contain

a number of limitations such as: high levels of inaccuracy, high degree of subjectivity,

and narrow focus. When carrying out a content analysis, one must first define the

concept that one wishes to operationalize. Some authors lack accuracy in their

description of the concept. For example, one study did not define ‚muscularity‛ – a

concept they wanted to evaluate among male video game characters (M. K. Miller &

Summers, 2007). Most of the content analyses have used highly subjective

evaluations. For instance, one analysis (Near, 2013) coded sexualization if the female

character was represented with disproportionately large breasts, clothes prominently

showing the midriff or cleavage, clothing similar to swimwear, dress revealing any

undergarments, or a suggestive pose. In this coding, the meaning of

‚disproportionately large‛, ‚prominently‛, ‚similar to‛, and ‚suggestive pose‛

depend on each coder’s estimation. In two other content analyses, all evaluations

were made on a culturally-based training — ‚how a typical person in the US would

view the character‛ (M. K. Miller & Summers, 2007; Summers & Miller, 2014). Yet,

what can be considered as a typical person in the US remains unclear. The last limit

of most classification systems is their narrow focus. Usually, only a limited number

Chapter 3: Representation of Women and Men in Video Games

of variables are analyzed. For instance, in one analysis (Martins et al., 2011) only

muscularity was analyzed to determine how male characters are physically

represented in video games. In another analysis (Near, 2013), only sexualization was

coded and only for female characters.

Therefore, there is a clear need for an appropriate and objective instrument to

evaluate the actual sexualized and role content of video games. Such a classification

system should address the limits of previous content analyses by possessing a certain

number of elements, including defining a clear sampling method, specifying accurate

and objective concepts, and keeping a broad focus. First, a clear sampling method

should be defined. Based on previous content analyses, images should be used to

evaluate the physical features of characters, and videos (e.g., trailers, gameplay

recording) should be used to evaluate player roles. Second, the conceptualization of

each coded concepts should be accurate, objective, and free of any risk of cultural

interpretation. For example, body proportion could be evaluated using a direct

measurement instead of a vague estimation. Finally, the focus of the classification

system should be large and include both male and female video game characters. In

addition, similar concepts should be evaluated for male and female characters (e.g.,

body proportion, sexualized clothing, and muscularity). Such an evaluation

instrument should help researchers to create a score of sexualized content and role

for each character in a game. This would allow the direct comparison of studies and

would allow relating sexualization and stereotyped roles to any other variable of

interest, such as aggression and negative attitudes toward women.

Summary

Despite an equivalent number of female and male players, video games represent

a hostile environment for women. The video game industry and the video game

community are aggressive against women. Indeed, female players are often the

victims of sexual and gender harassment (i.e., hostile sexism), including rape jokes

(i.e., rape myth acceptance). They are often treated as if they were delicate, fragile,

and need protection (i.e., benevolent sexism), and they are surrounded by

dehumanization and objectification (e.g., booth babes and advertisements).

These negative representation and attitudes toward women are also directly

present in the content of video games. Representations of female and male video

game characters are often stereotyped. Female characters in video games are often

globally underrepresented, sexualized, and conform to stereotyped roles (i.e., a

damsel in distress, a sex object, or a strong character but sexy and secondary). Male

Theoretical Section

characters are usually muscular, hyper-masculine (i.e., aggressive, dominant,

powerful, mad), playable characters, and game heroes.

Representations of male and female video game characters have been evaluated

by content analyses. However, these content analyses have limits that prevent one

from drawing clear and definitive conclusions about the current state of

sexualization and roles endorsed by video game characters. These limitations are a

high degree of inaccuracy and subjectivity, a limited focus, and an inadequate

sampling method. There is thus a need for an instrument of evaluation that

addresses these limits.

Chapter 4: Impact of Female Sexualized Content on

Aggression against Women

Questions have been raised about the potential impact of video game content on

behaviors and attitudes. Most of the previous work has focused on the impact of

violent video game content on aggressive behavior (Anderson et al., 2010;

Greitemeyer & Mügge, 2014; Prescott, Sargent, & Hull, 2018), and on the impact of

prosocial video game content on prosocial behavior (Greitemeyer & Mügge, 2014).

Recently, attention has been drawn on the potential impact of sexualized content on

aggression and negative attitudes toward women. As it has been shown by previous

content analyses (Lynch et al., 2016; Martins et al., 2011; M. K. Miller & Summers,

2011; Stermer & Burkley, 2012; Summers & Miller, 2014), video game players are

exposed to several types of contents that include female sexualization, female

stereotyped roles, and dominating males. In addition, if they play an online game,

they might be exposed to negative attitudes toward women expressed by members

of the video game community (Brehm, 2013). Presently, it is difficult to affirm

whether video content is the cause of aggressive behavior and attitudes or whether

video games have adapted their content in order to please their community.

However, based on the Confluence Model integrated into the GAM (Anderson &

Anderson, 2008), female sexualized content from video games might theoretically

lead to aggression against women.

A Brief Description of Previous Video Game Content Research

Until recently, research about the impact of video game content focused on two

areas: (1) the impact of violent video game content on aggression and aggression-

related outcomes, as well as feelings of empathy and prosocial behavior, and (2) the

impact of prosocial video games on prosocial behaviors.

The impact of violent video game content has been largely debated among

scholars during the past years. While a consensus appears to be reached about a

positive impact of violent video game content on aggression (Anderson et al., 2010;

Bushman, Gollwitzer, & Cruz, 2015; Bushman & Huesmann, 2014; Greitemeyer &

Mügge, 2014; Krahé, 2014; Warburton, 2014), some researchers argue that the impact

of violent video game on real-world aggressive behavior is overstated (Elson &

Ferguson, 2013; Ferguson, 2007, 2008, 2009, 2010, 2015a, 2015b; Ferguson & Kilburn,

2009). Other researchers note that there is a link between exposure to violent video

Chapter 4: Impact of Female Sexualized Content on Aggression against Women

games and violent criminal behavior, but the magnitude of the effect is smaller

(Bushman & Anderson, 2015).

Two meta-analyses (Anderson et al., 2010; Greitemeyer & Mügge, 2014) have

provided with arguments in favor of a positive impact of violent video game content

on aggression. The first meta-analysis (Anderson et al., 2010) was conducted on 136

research papers with 130,296 participants published prior to 2009. This meta-analysis

showed that violent video game content had a causal impact on increased aggressive

behavior, aggressive cognition, aggressive affect, and arousal. Further, playing

violent video games also had a negative impact on feelings of empathy and on

prosocial behavior. The second meta-analysis (Greitemeyer & Mügge, 2014) was

conducted on 98 studies published from 2009 to 2014 with 36,965 participants

(Greitemeyer & Mügge, 2014). The objective of this second meta-analysis was to

replicate the results of the first one (Anderson et al., 2010) and to analyze the impact

of prosocial video game content on prosocial outcomes. Results showed a positive

impact of violent video game content on aggressive behavior, thoughts and feelings,

and a negative impact of violent video game content on prosocial behavior, thoughts,

and feelings. Both meta-analyses thus present coherent results despite analyzing

studies from different time periods. In both meta-analyses, the impact of video game

content was present for all methodological designs (i.e., cross-sectional,

experimental, and longitudinal), for both males and females of all ages, and no

presence of a publication bias was found.

However, some authors have argued that these effects were invalid and raised

four main criticisms relating to video game content research (Ferguson, 2009, 2010,

2015a, 2015b; Ferguson & Kilburn, 2009). First, video game researches rarely use

measures of overt aggression. Indeed, several studies, especially experimental

studies, used ‚nonserious‛ aggression measures for ethical reasons (e.g., the

competitive reaction time task or the hot sauce paradigm; (Epstein & Taylor, 1967;

Lieberman, Solomon, Greenberg, & McGregor, 1999). Second, studies did not include

important covariates such as gender, trait aggression, family/parenting, and negative

life events. Therefore, the effect could be attributed to a third variable relationship in

transversal and longitudinal studies. Third, when only the studies that have

addressed the two first criticisms are included in a meta-analysis, a publication bias

might exist in favor of studies showing a relation between violent video game play

and aggression. Forth, even if the reported effect sizes are significant, the effects are

usually weak. According to the same authors, when all these criticisms are

addressed, studies showed an absence of effect of violent video game content. For

Theoretical Section

example, three longitudinal studies have shown an absence of effect of violent video

game content on overt aggression (Ferguson, 2011; Ferguson, Garza, Jerabeck,

Ramos, & Galindo, 2013; Ferguson, Miguel, Garza, & Jerabeck, 2012).

A recent meta-analysis (Prescott et al., 2018) addressed these criticisms by only

including studies that have assessed the relation between exposure to violent video

game content and overt physical aggression. More specifically, 24 studies were

included with over 17,000 participants. Studies included were longitudinal with a

time lag ranging from 3 months to 4 years and controlled for trait aggression as well

as other covariates of interest. Results showed that violent video game play increased

overt physical aggression, even when trait aggression was controlled for. Further, the

effect remained even after controlling for all covariates included in each study.

Finally, the meta-analysis used three different publication bias methods that all

showed an absence of publication bias. In other words, this meta-analysis addressed

three out of four of the main criticisms raised against studies about the impact of

violent video game content on aggression.

In summary, there has been proof that video game content can influence

behaviors. Indeed, violent video game content has been shown to increase behavioral

aggression, and prosocial video game content has been shown to increase prosocial

behavior (Anderson et al., 2010; Greitemeyer & Mügge, 2014; Prescott et al., 2018).

However, few studies have examined the potential impact that sexualized video

game content might have on aggression against women.

Theoretical Arguments toward an Impact of Sexualized Female Content in Video

Game on Aggression and Aggressive Attitudes toward Women

Recall from Chapter 2 that the GAM is a model designed to predict aggressive

behavior. According to the GAM, aggressive behavior is the result of several factors

that interact together. When a person is in a situation that can potentially result in

aggressive behavior, the first elements that can influence the behavior are the

features of the environment (i.e., situational variables) and the features of the person

(i.e., personal variables). The interaction between situational variables and personal

variables influence the present internal state that includes affect, cognition, and

arousal. Modification of the present internal state influence appraisal and decision

processes. First an immediate appraisal of the situation occurs which, if not

reappraised, can lead to an impulsive (i.e., automatic) behavior. However, if the

person judges the initial appraisal as unsatisfactory and if the person has sufficient

time and cognitive resources, he might reappraise the situation, which could lead to

Chapter 4: Impact of Female Sexualized Content on Aggression against Women

a more thoughtful behavior. The result of all the interactions between the

components of the GAM might be an aggressive behavior.

The GAM is a useful theoretical framework to analyze the impact of media content

on aggressive behavior (Anderson & Bushman, 2018). Indeed, the GAM has already

been proved useful to predict aggressive behavior by exposure to violent content of

media in general and by exposure to violent content of video games (Bushman,

2017). According to the GAM, media content can be considered as a situational

variable. In this context, sexualized content from video games would act as a

situational variable and interact with personal variables to modify the present

internal state. In other words, sexualized content from video games might increase

aggressive thoughts, feelings, and physiological arousal levels. Further,

modifications of the present internal state are predicted to influence appraisal and

decision processes. Video games might have a stronger impact on the appraisal and

decision processes than other more passive forms of media (e.g., television, film,

videos). Because of their interactive nature, video games consume more cognitive

resources than passive forms of media (such as film and television; Lin, 2013).

According the GAM, such a higher consumption of cognitive resources might

interfere with the reappraisal process and therefore lead to a more impulsive action

(i.e., a more automatic behavior) instead of a thoughtful action. If sexualized content

did influence the present internal state and the appraisal and decision processes, the

result of the GAM cycle might be an aggressive behavior against women.

Further, the Confluence Model integrated into the GAM (Anderson & Anderson,

2008) postulates that the aggressive outcome should be more likely if several specific

personal variables and situational variables are present. The personal variables

identified by the integrated model are general hostility, general aggressiveness,

violent attitudes, hostility toward women, impersonal-dominant sex, aggressiveness

to women, and violent attitudes to women. The situational variables identified are

provocation, a female target, and an aggressive opportunity.

State of the Actual Empirical Research about the Impact of Female Sexualized

Content on Aggression against Women

If sexualized content from video games can theoretically cause aggression against

women, empirically, only a handful of studies have tried to analyze the impact of

sexualized content from video games on aggression against women. To the best of

our knowledge, only four studies have manipulated sexualization alone (Behm-

morawitz & Mastro, 2009; Driesmans et al., 2015; Fox, Ralston, Cooper, & Jones, 2015;

Theoretical Section

Read, Lynch, & Matthews, 2018). In those studies, sexualization was manipulated by

modifying the outfit of the character or the character itself. Further, four other studies

are of interest concerning the potential impact of video games on aggression against

women (Breuer, Kowert, Festl, & Quandt, 2015; Dill et al., 2008; Fox & Potocki, 2016;

Yao et al., 2010), but have limits or too broad a focus concerning the impact of

sexualized content.

Among the four studies that manipulated sexualized content alone, one focused

on hostile sexist attitudes (Behm-morawitz & Mastro, 2009), one on rape myth

acceptance (Fox et al., 2015), one on both rape myth acceptance and tolerance toward

sexual harassment (Driesmans et al., 2015), and one on both hostile sexism and rape

myth acceptance (Read et al., 2018). None focused on actual aggressive behavior.

The first study (Behm-morawitz & Mastro, 2009) examined the impact of a

sexualized video game on gender stereotyping. To evaluate the impact of sexualized

content, participants were separated into three groups. One group was exposed to a

sexualized video game — Tomb Raider: Legend with the classic character that wears

revealing clothing. A second group was exposed to a non-sexualized version of the

same video game with a fully clothed character (i.e., the character wore a snow suit).

The third group was a control group that did not play any video game. Four

different types of gender stereotypes were measured using the Attitudes Toward

Women Scale (Spence & Helmreich, 1972): (1) female cognitive capacity, (2) female

physical capacity, (3) career and domestic labor (i.e., men’s and women’s

responsibilities related to household chores and childrearing, and appropriateness of

their roles in professional and manual labor jobs), and (4) appearance (i.e., gender

expectation about how women should maintain their bodies and appearance for

others). Results showed that playing a sexualized video game compared to no video

games at all significantly increased the belief that women are less cognitively

capable. No significant differences were found between the sexualized video game

and the non-sexualized video game. Further, results showed that women in the

sexualized condition significantly considered that women were less physically

capable than females in the non-sexualized condition. All other comparisons were

found to be non-significant. Therefore, there was no effect of video game condition

on career and domestic labor and on appearance.

The objective of the second study (Fox et al., 2015) was to determine the impact of

a sexualized characters on self-objectification and rape myth among female

participants. Using the video game Second Life, the authors created two sexualized

and two non-sexualized characters (i.e., Second Life gives a lot of freedom in the

Chapter 4: Impact of Female Sexualized Content on Aggression against Women

customization of avatars). The sexualized characters were rated to be significantly

sexier and suggestively dressed than the non-sexualized characters by an

independent sample of participants. Participants played either with a sexualized

character or with a non-sexualized character. State self-objectification was measured

using the Twenty Statements Test (Fredrickson, Roberts, Noll, Quinn, & Twenge,

1998), and rape myth acceptance was measured using the Rape Myth Acceptance

Scale (Burt, 1980). Results showed no direct effect of sexualized condition on rape

myth acceptance. However, self-objectification significantly mediated the relation

between sexualization exposure and rape myth acceptance. Specifically, exposure to

a sexualized avatar increased self-objectification, which, in turn, led to greater rape

myth acceptance.

The third study (Driesmans et al., 2015) recruited adolescent between 12 and 15

years of age. They played the video game The Story of Arado with either a sexualized

female character or with a non-sexualized male character. After gameplay, tolerance

toward sexual harassment was measured using the Tolerance for Sexual Harassment

Inventory (Lott, Reilly, & Howard, 1982), and rape myth acceptance was measured

using the Rape Myth Scale (Lonsway & Fitzgerald, 1995). The wording of both scales

was adapted to the age group of the study. Results showed that sexualized content

from the video game significantly increased rape myth acceptance and tolerance

toward sexual harassment among adolescents.

The last study that focused on sexualized content alone also manipulated

cognitive load (Read et al., 2018). Their objective was to investigate how task demand

and avatar sexualization influence rape myth acceptance and hostile sexism. First,

participants were played a sexualized or a non-sexualized video game. Sexualization

was manipulated using the same video game: The Elder Scrolls V: Skyrim. The

sexualized female character was wearing an outfit that left sexual regions nude and

accentuated. Her body featured exaggerated body proportions with large breasts, a

small waist, and full hips. The non-sexualized female character wore armor that

covered her torso in such a way that her chest size and hip-to-waist ratio was

proportionate to her body. Further, cognitive load was also manipulated during

gameplay. Participants were asked to retain symbols (7 in the high condition, 2 in the

low condition) that would be used at the end of the game to open a door. In this

study, rape myth acceptance was measured using the Short Form of the Illinois Rape

Myth Acceptance Scale (D. L. Payne, Lonsway, & Fitzgerald, 1999), and hostile

sexism was measured using the Ambivalent Sexism Inventory (Glick & Fiske, 1996).

Results showed that sexualization significantly interacted with cognitive load.

Theoretical Section

Specifically, when high sexualized content was combined with high cognitive load,

both rape myth acceptance and hostile sexism were lower than when cognitive load

was low.

Results from these studies showed mixed conclusions about the impact of

sexualized content on aggressive attitudes toward women. First, one study showed

that sexualization might increase tolerance toward sexual harassment (Driesmans et

al., 2015). Further, another study showed that women are considered as less

cognitively capable when exposed to a sexualized video game content compared to

no video game at all (Behm-morawitz & Mastro, 2009). The same study showed that

among female participants, playing a sexualized video game diminished the belief

that women are physically capable compared to playing a non-sexualized video

game or not playing any video game at all. One study (Read et al., 2018) showed that

aggressive attitudes toward women (i.e., hostile sexism and rape myth acceptance)

diminished when exposed to sexualized video game content, but only if participants

experienced cognitive load. However, that study is in opposition with two others

that have shown that sexualized video game can directly increase rape myth

acceptance (Driesmans et al., 2015) or indirectly increase rape myth acceptance

through an increase of self-objectification (Fox et al., 2015). In summary, results tend

to show that sexualized video game increase aggressive attitudes toward women, but

that various variables (e.g., self-objectification and cognitive load) might need to be

accounted for in order to better understand the exact nature of that relation. One

main limitation can be addressed to these four studies. All of them used trait

measures of aggressive attitudes toward women. Specifically, the Attitudes toward

women Scale (Spence & Helmreich, 1972), the Ambivalent Sexism Inventory (Glick &

Fiske, 1996), the Tolerance for Sexual Harassment Inventory (Lott et al., 1982), the

Rape Myth Scale (Lonsway & Fitzgerald, 1995), the Short Form of the Illinois Rape

Myth Acceptance Scale (D. L. Payne et al., 1999), and the Rape Myth Acceptance

Scale (Burt, 1980). Such measures are supposed to be stable across time and

situations. Therefore, it is uncertain whether the difference of results is obtained

because of the different sexualization conditions or by chance.

Other studies, although not directly relevant to this dissertation, reported

interesting results. One study showed that being exposed to sexualized images from

video games compare to images of female politicians did not significantly influence

rape myth acceptance but significantly increased tolerance toward sexual harassment

(Dill et al., 2008). This study also used a trait measure to evaluate rape myth

acceptance, but was the first to use a state evaluation of tolerance toward sexual

Chapter 4: Impact of Female Sexualized Content on Aggression against Women

harassment. Participants read an ambiguous story of sexual harassment perpetrated

by a male college professor against a female student. After reading, participant

answered six questions about the event: (1) whether the incident constituted sexual

harassment, (2) how serious the event was (3) the degree of damage caused by the

event, (4) blame of the victim, (5) empathy for the victim, and (6) choice of

punishment for the perpetrator. Compared to trait measures of tolerance toward

sexual harassment, state measures evaluate an immediate judgment about the

situation. One main limit of this study is its lack of ecological validity. Indeed,

participants were only exposed to sexualized images from video games instead of

being exposed to actual video game content.

Another study (Yao et al., 2010) focused on the impact of a sexually explicit video

game on the occurrence of sexual thoughts, sexual objectification of women, and self-

reported tendency of sexually harassing a woman. Participants played either the

sexually explicit game Leisure Suit Larry: Magna Cum Laude or one of two non-

sexually explicit control games (i.e., The Sims II, PacMan II). Leisure Suit Larry: Magna

cum Laude is a sexually suggestive video game that contains sexualized female

characters, sexist attitudes and sexuality. After gameplay, participants completed a

Lexical Decision Task to measure sexual thoughts. That task measured how quickly

participants could distinguish real words that were either sexual words (e.g., sex,

penis, etc.), neutral words (e.g., door, bank, etc.), or scrambled words. A Lexical

Decision Task was also used to measure sexual objectification of women, the words

used were sexually objectifying of women (e.g., slut, whore, bitch), neutral words (e.g.,

sister, nurturer, niece), or scrambled words. The last measure was the Likelihood to

Sexually Harass Scale, which measures how participants would act in 10 scenarios

depicting situation in which they have the opportunity to sexually harass a woman.

Results showed that playing the sexually explicit video game, compared to the

neutral ones, significantly increased the occurrence of sexual thoughts, increase the

tendency for men to view women as being sex objects, and increased their self-

reported tendency to sexually harass a woman. In this study, one main limit is that

sexualization was confounded with other factors including: the presence of sexism

(i.e., negative remarks about women) and sexuality (i.e., presence of a sexual act).

Finally, two studies (Breuer et al., 2015; Fox & Potocki, 2016) used overall

consumption of video games to predict hostile sexism and rape myth acceptance.

One study (Fox & Potocki, 2016) carried a transversal study and found a positive

indirect effect of time spent playing video games on rape myth acceptance through

an increase of interpersonal aggression and hostile sexism. The second study (Breuer

Theoretical Section

et al., 2015) was longitudinal and did not find any impact of video game

consumption on hostile sexism. Obviously, these two studies are difficult to interpret

because they did not separate consumption of sexualized video games from overall

consumption of video games.

These previous studies have several limitations. First, one study (Dill et al., 2008)

lacked ecological validity by using static images of video games. Video games are

interactive by nature. Future studies should have participants play a video game

instead. A second limitation is the presence of confounding variables. For example,

one study used a sexually explicit video game that contained sexualization, sexism,

and sexuality (Yao et al., 2010). By using such a video game, it is impossible to

determine to which variable the effect can be imputed. Confounding variables need

to be avoided in control conditions too. Another study (Dill et al., 2008) used images

of female politicians as a control. However, pictures of politicians might have primed

various concepts in memory (e.g., competence, authority, justice, etc.). It is unclear

whether the impact is due to an increase of tolerance toward sexual harassment after

exposure to sexualized video game character, or whether it is due to a diminution of

tolerance toward sexual harassment after exposure to politicians. Future studies need

to avoid any manipulation that does not concern only sexualization in order to

clearly conclude about its impact on aggressive behavior and attitude toward

women. A third limit of previous studies is having a too broad a focus. For example,

two studies in particular have tried to predict aggressive attitudes toward women

using time spent playing video games (Breuer et al., 2015; Fox & Potocki, 2016). Such

a method might lead to diluted results because video games are a complex media

with a lot of various contents. One last limitation is the use of trait measure as

dependent variables in experimental studies (Behm-morawitz & Mastro, 2009; Dill et

al., 2008; Driesmans et al., 2015; Fox et al., 2015; Read et al., 2018). All of these limits

need to be addressed in future studies.

The second reason for which no clear conclusion about the impact of sexualized

video game can be reached yet is that studies need to vary the aggressive variables

examined. Based on the confluence model integrated into the GAM (Anderson &

Anderson, 2008), sexualized content might theoretically influence aggressive

behavior, cognition and affect. However, most studies have focused on aggressive

attitudes toward women. Only two studies have evaluated the impact of sexualized

content on aggressive cognition. One study measured the thoughts of participants

about a sexual harassment situation (Dill et al., 2008), whereas the other study

measured the accessibility of sexually objectifying thoughts (Yao et al., 2010). None

Chapter 4: Impact of Female Sexualized Content on Aggression against Women

of the previous studies have analyzed the impact of sexualized content on actual

aggressive behavior against women, or on aggressive affect. There is a clear need for

future studies about the consequences of sexualized content from video game.

Summary

Female sexualized content from video game has identified as a potential cause of

aggression against women. Theoretically, the confluence model integrated into the

General Aggression Model (Anderson & Anderson, 2008) predicts that female

sexualized content might lead to aggressive behavior and attitudes toward women.

Empirically, no conclusion can be drawn yet about the impact of the sexualized

content of video games on attitudes toward women. Indeed, the existing studies are

both scarce and often have major limitations (i.e., lack of ecological validity, using a

sexualized stimulus that is confounded with other factors, having a too broad focus,

using trait measures as dependent variables). Further, there is a lack of studies that

examined the impact of sexualized content from video game on actual aggressive

behavior against women.

There is a real need for further studies that aim to determine the actual effect of

the sexualized content of video games on aggressive behavior and attitudes toward

women. These future studies should try to avoid the limits of previous studies. First,

experimental research on video games should use an ecological means of exposure to

sexualized content by asking participant to actually play the video game. Second,

studies about the impact of sexualized content should try to avoid confounding

variables in order to clearly determine the cause attributed to sexualization. Finally,

in the case of experimental studies, the dependent variable should be adequately

evaluated using a state measure.

Aims of the Thesis

The present thesis has three main aims:

1) To develop an instrument of evaluation of sexualized and stereotyped role

content of video games. Such an instrument of evaluation should be based on a

classification system that is objective, reliable, and valid. Both male and female video

game characters should be evaluated indiscriminately. The instrument of evaluation

should be sufficiently clear and precise to be used by any researcher regardless of

their degree of video game experience. Furthermore, the predictive validity of such a

classification system concerning negative attitudes toward women will also be

investigated.

2) To explore the impact of female sexualized content on aggressive behavior and

attitudes toward women. Studies on this topic are scarce and there is therefore a need

for more studies that examine various forms of aggression against women. Such

studies need to address the limit of previous ones by being ecologically valid,

avoiding the presence of other confounding variables, and using state measures of

aggressive behavior and attitudes.

3) To draw a clearer picture of the actual impact of sexualized content on

aggression, regardless of the media and/or whether or not this is specific to certain

types of media. No recent quantitative review exists about the impact of any visual

sexualized media on aggressive behavior, cognition, affects and attitudes. A meta-

analysis will therefore be carried out with the aim of providing a better

understanding of how sexualized content of media influences aggressive behavior,

cognition, affect, and attitudes.

Experimental Section

The Development and Validation of an Objective Measure of

the Sexualized Content of Video Games: the Video Game

Sexualization Protocol

Jonathan Burnay

, Brad J. Bushman

, Hedwige Dehon

, Aurélien Cornil

, Frank

Larøi

a,d,e

Psychology and Neuroscience of Cognition Research Unit, Faculty of Psychology, Speech

and Language Therapy, and Education, University of Liège

School of Communication and Department of Psychology, The Ohio State University,

Columbus

Laboratory for Experimental Psychopathology (LEP), Psychological Science Research

Institute, Université catholique de Louvain, Louvain-la-Neuve, Belgium

Department of Biological and Medical Psychology, Faculty of Psychology, University of

Bergen

NORMENT – Norwegian Center of Excellence for Mental Disorders Research, University

of Oslo

Abstract

Sexualization is an integral part of many video games. Based on content analyses,

female characters often have large breasts and buttocks, a small waist, and usually

wear revealing clothes. Male characters often have big muscles with less revealing

clothes. However, there is no objective instrument to measure sexualized content in

video games. The objective of this study was to develop and validate such an

instrument, which is called the ‚Video Game Sexualization Protocol‛ (VGSP). In

Study 1, the initial two-factor (male / female) VGSP was created. The VGSP

possessed good inter-coder validity and internal consistency. Study 2 had two

objectives: (1) to make the measures of female and male sexualization more uniform,

and (2) to determine whether the VGSP can predict negative attitudes toward

women. A three-factor structure (‚sexualized body‛ / ‚muscularity‛ / ‚revealing

outfit‛) was observed for both female and male characters. This revised version of

the VGSP also had a good internal consistency and good inter-coder validity. Results

showed that exposure to various features of video game characters predicted

negative attitudes toward women. Benevolent sexism was predicted positively by

exposure to female characters wearing revealing outfits, and negatively by exposure

to female characters with sexualized bodies. Hostile sexism was predicted positively

by exposure to male characters with muscular bodies, and negatively by the

interaction between exposure to female characters wearing revealing outfits and to

female characters with muscular bodies. Rape myth acceptance was predicted by

exposure to male characters with revealing outfits, but positively when interacting

with exposure to male muscularity and negatively when interacting with exposure to

male characters wearing a sexualized outfit. These findings indicate that, in order to

predict negative attitudes toward women using sexualized video games, it is

necessary to analyze various features of that media.

Introduction

Video games are an important type of media in Western society. According to

several surveys, video games are played by around 64% of the Western population

(Interactive Software Federation of Europe, 2018; Nielsen Games, 2017; UKIE, 2018).

Further, almost as many women as men play video game (Interactive Software

Federation of Europe, 2018; UKIE, 2018).

Frequent video game exposure may have a number of consequences, including

reinforcing stereotyped male and female roles in society (Dill et al., 2008; Driesmans

et al., 2015; Fox & Potocki, 2016; Yao et al., 2010). More specifically, sexualized

content from video games has already been shown to increase negative attitudes

toward women. For example, one study (Dill et al., 2008) showed that being exposed

to images of sexualized content from video games increased tolerance toward sexual

harassment. Another study (Driesmans et al., 2015) showed that playing a sexualized

video game (compared to a non-sexualized video game) increased rape myth

acceptance among adolescents.

Sexualization is an integral feature of many video games. Female characters are

often portrayed with large breasts and buttocks, a small waist, a large amount of

exposed skin, and revealing clothes (Downs & Smith, 2010; Lynch et al., 2016;

Summers & Miller, 2014). In contrast, male characters are often portrayed as being

very muscular (e.g., large arm- or chest muscles) and wearing less revealing clothes

(Dill & Thill, 2007; Downs & Smith, 2010; Scharrer, 2004). The degree of sexualization

of video game characters has been evaluated in previous studies, albeit via a number

of different methods. Further, each content analysis has its own classification system,

without a clear consensus between them. Also, these classification systems are

subject to a high degree of inaccuracy, subjectivity, and a narrow focus.

Some classification systems are too inaccurate. For instance, some studies do not

even take into account the content of video games (Breuer et al., 2015; Fox & Potocki,

2015). Instead, they use time spent playing video games as an evaluation of sexism.

Another limitation of some classification systems is that they are too subjective.

Often, the coding categories and the instructions to the coders were not clearly

defined (Burgess et al., 2007; M. K. Miller & Summers, 2007; Summers & Miller,

2014). For instance, one classification system evaluates breast size using vague

categories such as ‚normal‛, ‚busty‛ and ‚super-busty‛ (Burgess et al., 2007).

However, a ‚normal‛ breast size will vary from one person to another. In two studies

The Development and Validation of an Objective Measure of the Sexualized

Content of Video Games: the Video Game Sexualization Protocol

(M. K. Miller & Summers, 2007; Summers & Miller, 2014), coders were ‚trained to

code based on how a typical person in the US would view the character‛. Yet, what

can be considered as a typical person in the US remains unclear. The consequence of

such a high degree of subjectivity makes the results from content analyses difficult to

replicate.

Finally, some classification systems have a limited focus. For example, one

classification system only focused on clothing and breast size (Beasley & Collins

Standley, 2002). More variables need to be coded if one is to evaluate the presence of

sexualization in an adequate manner. As another example, some classification

systems exclude male characters (Summers & Miller, 2014). Even systems that

include both male and female characters placed more emphasis on female characters.

In theory, both male and female video game characters can influence stereotyped

attitudes about males and females in the real world.

A classification system that addresses these limitations is urgently needed. It

should be based on accurate, precise, and objective measurements that can be easily

replicated. Further, a classification system should assess sexualization in video

games for both genders. Such an instrument is essential for researchers who wish to

study the potential effects of video game characters on stereotypes regarding male

and female roles in society.

Study 1

The objectives Study 1 were twofold. First, to create an objective, reliable and valid

classification system that can be used by researchers regardless of their degree of

video game experience. Second, to examine the psychometric properties of this new

instrument.

Method

Sampling

Participants were 99 adolescents (54 male, 42 female) 12 to 18 years old (M = 15.26,

SD = 1.86) from various secondary schools in Belgium. Participants listed their 5 most

recently played video games. They played video games for an average of 9.8 hours a

week (SD = 11.11, range 1-72 hours).

After duplicates were eliminated, the sample contained 195 unique video games.

The final sample included video games from each of the five Pan European Game

Information (PEGI, 2018) age categories (i.e., 3+, 7+, 12+, 16+, 18+) and from each

Experimental Section

genre based on Entertainment Software Rating Board (ESRB, 2018) classification (e.g.,

Action, Role Play Gaming, Puzzle, Simulation, Adventure, Strategy, Racing, Sports).

Appendix 1

contains a list of the 195 video games included in the sample. In order to

code the degree of sexualization of the video game characters, 4 pictures of

humanoid main characters were used for each game (if the game contained

characters). Preferentially, the 4 pictures had to include two males (the main

protagonist and the main antagonist) and two females (the main protagonist and the

main antagonist).

Coding

Sexualization of female characters. Four observation scores were assigned to

female characters: (1) breast, (2) hips, (3) uncovered body parts, and (4) clothing. To

avoid variation due to image size, the measurements for breasts and hips were

proportions.

Breast. Breast size was measured by dividing the width of the bust (largest point)

by the width of the waist (smallest point).

Hips. Hip size was measured by dividing the belt width (smallest point) by the

width of the waist (smallest point).

Uncovered body parts. One point was given for each of five possible naked body

areas: (1) shoulders, (2) belly and/or the back, (3) neckline and/or the breast, (4)

buttocks, and (5) thighs. Thus, points could range from 0 to 5.

Clothing. Points (0 to 5) were also given for sexualized clothes. The following

clothes were coded 1 point: mini-skirt, shorts, top neckline, bottom neckline, slit

dress, high boots, fishnet stockings, and tattoo. The following clothes were coded 2

points: bikini bottom, bikini top, swimsuit, bra, underpants, corset, and garters.

Naked characters or characters where the breasts or buttocks were fully revealed

were coded 5 points. Skin-tight clothes (e.g., leather) that revealed body shapes were

coded according to the body parts concerned. For example, a leather pant would be

coded two because it concerns the buttocks and the thighs.

Sexualization of male characters. Four observation scores were assigned to male

characters: (1) shoulder span, (2) biceps, (3) uncovered body parts, and (4) clothing.

To avoid variation due to image size, the measurements for shoulder span and biceps

were proportions.

Appendix 1 can be found in Annex 1

The Development and Validation of an Objective Measure of the Sexualized

Content of Video Games: the Video Game Sexualization Protocol

Shoulder span. Shoulder span was measured by dividing the width of the

shoulders (largest point) by the width of the head (largest point), using the bottom of

the eyes as a guide for the ruler.

Biceps. Biceps were measured by dividing the width of the biceps (largest point)

by the width of the head (largest point), using the bottom of the eyes as a guide for

the ruler.

Uncovered body parts. One point was given for each of five possible naked body

areas: (1) shoulders, (2) biceps, (3) abdominal muscles and/or the back, (4) pectoral

muscles, and (5) buttock.

Clothing. Points (0 to 5) were also given for sexualized clothes. The following

clothes were coded 1 point: shorts and tattoo. The following clothes were coded 2

points: underpants, sleeveless shirt, open shirt. Naked characters or characters where

the torso or buttocks were fully revealed were coded 5 points. Skin-tight clothes (e.g.,

leather) that revealed body shapes were coded according to the body parts

concerned.

Coder Training

Two coders (undergraduate University students) coded all the images extracted from

the video games. One coder was a man that described himself as a video game

player, the other coder was a woman that described herself as having little

knowledge about video games.

Statistical Analyses

An exploratory factor analysis was used to determine the factor structure of the

‚Video Game Sexualization Protocol‛ (VGSP). Oblimin rotation was used because

we expected our factors to be correlated with each other (Sass & Schmitt, 2010; Yong

& Pearce, 2013). Pearson’s correlations were then conducted to evaluate the

relationship between the various factors of the VGSP. According to Cohen (1988), a

‚small‛ correlation is .10, a ‚medium‛ correlation is .30, and a ‚large‛ correlation is

.5.

Results

Intercoder Reliability

Pearson’s correlations were used for measures that involved proportions (i.e.,

breast, hips, shoulder span, and biceps), and intraclass correlations were used for

categories (i.e., female and male uncovered body parts and clothing). As can be seen

Experimental Section

in Table 1, intercoder reliability was excellent for coded dimensions for both female

and male characters.

Table 1

Intercoder Reliability using Pearson’s correlations and intraclass correlations

Measure

Pearson’s

correlation

Intraclass

correlation

Breast

.968

Hips

.864

Female uncovered body parts

.963

Female clothing

.939

Biceps

.938

Shoulder Span

.943

Male Uncovered Body Parts

.974

Male Clothing

.956

Exploratory Factor Analysis (EFA)

Two exploratory factor analyses were used to determine the best model for

sexualization of female video game characters and sexualization of male video game

characters. For the sexualized female video game characters, the EFA with Oblimin

rotation suggested extracting two factors (Table 2), which accounted for 92.38% of

the variance (eigenvalue = 0.99). Breasts and hips loaded on Factor 1, and uncovered

body parts and clothing loaded on Factor 2.

Table 2

Corrected Item-Total Correlation and Factor Loadings from EFA for measures related to

female characters

Items

Corrected

item-total

correlation

Factor 1

Factor 2

Breasts

.833

.952

Hips

.833

.962

Female Uncovered Body Parts

.862

.960

Female Clothing

.862

.969

The Development and Validation of an Objective Measure of the Sexualized

Content of Video Games: the Video Game Sexualization Protocol

For the sexualized male video game characters, the EFA with Oblimin rotation

suggested extracting two factors (Table 3), which accounted for 94.93% of the

variance (eigenvalue = 1.27). Biceps and shoulder size loaded on Factor 1, and

uncovered body parts and sexualized clothing loaded on Factor 2.

Table 3

Corrected Item-Total Correlation and Factor Loadings from EFA for measures related to male

characters

Items

Corrected

item-total

correlation

Factor 1

Factor 2

Biceps

.859

.964

Shoulder size

.859

.972

Male uncovered body parts

.920

.992

Male clothing

.920

.966

Descriptive Statistics, Internal Consistency, and Inter-Correlation of the Female

and Male Two-factor Model

Descriptive results, internal consistency, and intercorrelations between all factors of

both two-model factors of the VGSP are reported in Table 4. Internal consistency was

evaluated with Cronbach’s α coefficient, which ranged from .91 to .96. Many

methodologists recommend a minimum coefficient of .65. Thus, these coefficients

were very high.

Table 4

Descriptive statistics, internal consistency and inter-correlations of all facets of the VGSP

Factors

M (SD)

Cronbach

1 - Female Sexualized Body

2.30(1.08)

.91

2 - Female Sexualized Outfit

3.19(3.45)

.92

.46***

3 - Male Sexualized Body

4.11(1.56)

.93

.13

.23***

4 - Male Sexualized Outfit

2.83(3.68)

.96

.20**

.51***

.34***

**p < .01; ***p < .001

Experimental Section

Discussion

The first objective of Study 1 was to create and validate a classification system that

evaluates the degree of sexualization of male and female video game characters,

called the Video Game Sexualization Protocol (VGSP). The VGSP was designed to be

as objective, reliable, and valid as possible, and to be used independently from the

coder’s video game experience. The second objective was to examine the inter-coder

reliability, internal structure, and internal validity of the VGSP.

Exploratory factor analysis revealed that a 2-factor structure best explained

variances for both male and female video game characters. These two factors can be

described as ‚Sexualized Body‛ and ‚Sexualized Outfit‛. All facets possessed

excellent internal consistency (Cronbach α’s > .91). Finally, all facets were

intercorrelated except for ‚male sexualized body‛ and ‚female sexualized body‛.

Results also revealed that the VGSP possesses good intercoder reliability,

suggesting that the VGSP is sufficiently precise and objective. Recall that coder 1 was

an experienced male video game player, whereas coder 2 was a female with little

video game experience. Yet agreement between them was very high, which suggests

that codings were not influenced by video game experience.

Although the results are generally promising, a number of improvements are

required in order for the VGSP to be an even better instrument. First, the protocol

could be more uniform when comparing male and female video game characters. For

example, skirt was coded for female video game character, but not for male video

game characters. If a male video game character would have worn a kilt, it would not

have been coded as being sexualized clothing. Second, the structure of the protocol

should be replicated with more video games. Study 1 only contained 195 video

games, with a total of 167 female characters and 192 male characters analyzed.

Finally, future studies should try to determine if the VGSP possesses predictive

validity. For example, can the VGSP predict negative attitudes toward women? A

second study was conducted to overcome these limitations.

Study 2

Study 1 provided evidence that the Video Game Sexualization Protocol (VGSP) is

a valid measure of the sexualized content of video games. Study 2 aims to improve

the VGSP further and to examine if it can be used to predict negative attitudes

toward women.

The Development and Validation of an Objective Measure of the Sexualized

Content of Video Games: the Video Game Sexualization Protocol

Although the results from Study 1 were promising, the VGSP was improved in

Study 2. First, the protocol was made uniform for some variables between male and

female video game characters. For example, although the outfits were different

between male and female characters, the same outfit wore by a man or a woman

should be considered as being equally sexualized. Second, a measure of a sexualized

body had been added for men. Indeed, all sexualized body measures concerned the

muscularity of the male character. However, one measure – the V-shape – is rarely

evaluated in content analyses even though is an important measure of sexualization.

Indeed, the V-shape in men is more associated with a high degree of male sexual

desirability than large biceps and a large shoulder span (Braun & Bryan, 2006).

Therefore, the V-shape could be seen as an even more suited measure of

sexualization of male characters. Similarly, muscularity of female characters was

included. Muscular female video game characters might be perceived as stereotype

inconsistent information and might interact with sexualization to predict attitudes.

Second, the VGSP might be useful to predict negative attitudes toward women.

Theoretically, sexualized content in video games could impact negative attitudes

toward women. Based on the General Aggression Model (GAM; Anderson &

Bushman, 2018; Bushman, 2017; Figure 1), sexualized content in video games could

act as a situational variable that will influence one’s internal state, including negative

attitudes toward women. However, this hypothesis has not been adequately tested

because there is no instrument that examines, in a precise and valid manner, the

sexualized content of video games. Previous studies (Breuer et al., 2015; Fox &

Potocki, 2016; Stermer & Burkley, 2015) that have tried to relate video game content

to negative attitudes toward women, but have used inadequate measures, leading to

inconsistent results. For example, two studies used time spent playing video games

as a predictor of negative attitudes toward women, and found either no results

(Breuer et al., 2015) or an indirect effect on hostile sexism and rape myth acceptance

(Fox & Potocki, 2016). Another study used the subjective evaluation of sexist content

by male players themselves (Stermer & Burkley, 2015), and found a relationship for

benevolent sexism but not for hostile sexism.

The first objective of Study 2 was to further improve upon the VGSP. The second

objective of Study 2 was to determine whether the VGSP could predict negative

attitudes toward women. Study 2 explored how different characteristics of video

game characters, and their interactions, will predict hostile sexism, benevolent

sexism, trait aggressiveness, and rape myth acceptance. Further, rape myth

acceptance has been shown to be influenced by the indirect effects of video game

Experimental Section

consumption through interpersonal aggression and hostile sexism (Fox & Potocki,

2016). Therefore, a mediating effect of these two variables, and benevolent sexism on

rape myth acceptance, will be analyzed. To test more efficiently the effect of our

objective variables, we will control for subjective sexualization exposure (i.e., the

degree of sexualization perceived by the participants). Further, since our dependent

variable are hostile attitudes, we will control for subjective violence exposure (i.e.,

the degree of violence perceived by the participant).

Figure 1. The General Aggression Model. Source. Anderson and Bushman (2002),

Krahé (2013).

Method

Participants

Participants were 574 video game players (63.94% male) between ages 18 and 51

(M = 24.49, SD = 5.57) that were recruited on social media for an online study.

Participants spent between 0 and 120 hours per week playing video games (M =

13.49, SD = 14.27).

The Development and Validation of an Objective Measure of the Sexualized

Content of Video Games: the Video Game Sexualization Protocol

Questionnaires

Demographic information and video game consumption. Participants reported

their gender, age, and the average number of hours they spend playing video games

per week.

Subjective sexualization exposure. An adapted version of the Video Game

Questionnaire (Anderson & Dill, 2000) was used to measure self-reported

sexualization of the video games they played. Participants listed the three games

they played the most the past year, and indicated the extent to which the female and

the male characters in each game were sexualized (1 = Not at all to 7 = Extremely).

Participants also reported how often they played each game (1 = Rarely to 7 = Often).

The sexualized evaluation was multiplied by the frequency of play. All scores were

added together then pondered by the number of cited video games.

Subjective Violence exposure. The Video Game Questionnaire (Anderson & Dill,

2000) was used to measure participant’s exposure to violent video game content.

Participants rated how violent each game they listed was (1 = Not at all to 7 =

Extremely). The violence evaluation was multiplied by the frequency of play. All

scores were added together then pondered by the number of cited video games.

Ambivalent sexism. Participants completed a French version (Dardenne et al.,

1996) of the Ambivalent Sexism Inventory (ASI; Glick & Fiske, 1996), which contains

an 11-item hostile sexism subscale (e.g., ‚Most women interpret innocent remarks as

sexist; Cronbach α = .92), and an 11-item benevolent sexism subscale (e.g., ‚Women

should be cherished and protected by men‛; Cronbach α = .84). All items were scored

using a 6-point scale (0 = Totally disagree to 5 = Totally agree).

Rape myth acceptance. The Updated Illinois Rape Myth Acceptance Scale

(McMahon & Farmer, 2011) was translated and back translated to create a French

version for the purpose of this study

. This scale contains 19 items (e.g., ‚If a girl is

raped while she is drunk, she is at least somewhat responsible for letting things get

out of hand.‛; Cronbach α = .84) that are scored using a 5-point response scale (1 =

The French version of the Illinois Rape Myth Acceptance scale was validated using a

Confirmatory Factor Analysis. Three items were dropped because they lowered the internal

consistency of the sub-factor ‚He Didn’t Mean To‛. The four-factor model of the French

Illinois Rape Myth Acceptance Scale was validated in a sample of 268 participants. The

model possessed an excellent fit, χ

(148) =1331.31, p < .001; RMSEA = .037; CFI = .994; AGFI

= .720; NFI = .979; NNFI = .993.

Experimental Section

Strongly Agree to 5 = Strongly Disagree). For all statistical analyses, the score of the

Updated Illinois Rape Myth Acceptance Scale was reversed so that higher scores

indicate stronger rape myth beliefs.

Trait aggression. Participants completed a French version (Genoud &

Zimmermann, 2009) of the Aggression Questionnaire (AQ; Bryant & Smith, 2001),

which contains 12 items (e.g., ‚I have threatened people I know‛; Cronbach α = .71)

that are scored using a 6-point response scale (1 = Not at all like me to 6 = Completely

like me).

Sampling

After elimination of the duplicates, the sample of video games resulted in 447

video games that contained a total of 610 female characters and 692 male characters.

The final sample included video games from each of the different Pan European

Game Information (PEGI, 2018) age rating categories (i.e., 3+, 7+, 12+, 16+, 18+) and

from each genre based on the Entertainment Software Rating Board (ESRB, 2018)

classification (e.g., Action, Role Play Gaming, Puzzle, Simulation, Adventure,

Strategy, Racing, Sports). Appendix 2

contains a list of 447 video games included in

the sample. In order to code the degree of sexualization of the video game characters,

4 pictures of humanoid main characters were used for each game (if the game

contained characters). Preferentially, the 4 pictures had to include two males (the

main protagonist and the main antagonist) and two females (the main protagonist

and the main antagonist).

Coding

Female character evaluation. Six observation scores were observation scores were

assigned to female characters: (1) breast, (2) hips, (3) shoulder span, (4) Biceps, (5)

uncovered body parts and (6) Clothing. To avoid variation due to image size, the

measurements for breasts, hips, shoulder span and biceps were proportions.

Breast. Breast size was measured by dividing the width of the bust (largest point)

by the width of the waist (smallest point).

Hips. Hip size was measured by dividing the belt width (smallest point) by the

width of the waist (smallest point).

Appendix 2 can be found in Annex 1

The Development and Validation of an Objective Measure of the Sexualized

Content of Video Games: the Video Game Sexualization Protocol

Shoulder span. Shoulder span was measured by dividing the width of the

shoulders (largest point) by the width of the head (largest point), using the bottom of

the eyes as a guide for the ruler.

Biceps. Biceps were measured by dividing the width of the biceps (largest point)

by the width of the head (largest point), using the bottom of the eyes as a guide for

the ruler.

Uncovered body parts. One point was given for each of five possible naked body

areas: (1) shoulders, (2) belly and/or the back, (3) neckline and/or the breast, (4)

buttocks, and (5) thighs. Thus, points could range from 0 to 5.

Clothing. Points (0 to 5) were also given for sexualized clothes. The following

clothes were coded 1 point: mini-skirt, shorts, top neckline, bottom neckline, slit

dress, high boots, fishnet stockings, and tattoo. The following clothes were coded 2

points: bikini bottom, bikini top, swimsuit, bra, underpants, corset, garters, sleeveless

shirt, open shirt. Naked characters or characters where the breasts or buttocks were

fully revealed were coded 5 points. Skin-tight clothes (e.g., leather) that revealed

body shapes were coded according to the body parts concerned. For example, a

leather pant would be coded two because it concerns the buttocks and the thighs.

Male character evaluation. Five observation scores were assigned to male

characters: (1) shoulder span, (2) biceps, (3) V-Shape, (4) uncovered body parts and

(5) clothing. To avoid variation due to image size, the measurements for shoulder

span, biceps and V-shape were proportions.

Shoulder span. Shoulder span was measured by dividing the width of the

shoulders (largest point) by the width of the head (largest point), using the bottom of

the eyes as a guide for the ruler.

Biceps. Biceps were measured by dividing the width of the biceps (largest point)

by the width of the head (largest point), using the bottom of the eyes as a guide for

the ruler.

V-Shape. V-shape was measured by dividing the width of the shoulder (largest

point) by the width of the (waist smallest point).

Uncovered body parts. One point was given for each of six possible naked body

areas: (1) shoulders, (2) biceps, (3) abdominal muscles and/or the back, (4) pectoral

muscles, (5) thighs, and (6) buttock.