Content uploaded by Erin Ash
Author content
All content in this area was uploaded by Erin Ash on Aug 25, 2015
Content may be subject to copyright.
Priming or Proteus
Effect? Examining the
Effects of Avatar Race
on In-Game Behavior
and Post-Play Aggressive
Cognition and Affect in Video Games
Erin Ash
1
Abstract
This research uses a 2 2 factorial experiment to further investigate the Proteus effect
for avatar race observed in previous research by measuring in-game behavior, the use of
stereotypes to describe the avatar, and perceived embodiment of the avatar. Partici-
pants played a boxing video game as White or Black avatar against a White or Black
avatar. Results revealed no main effects for avatar race, but embodiment was found to
moderate the relationship between avatar race and in-game behavior. Subsequent
probing of this interaction revealed that the Proteus effect was demonstrated for those
who experienced greater embodiment. Avatar race was unrelated to the use of
stereotypes, but main effects for aggressive cognition and affect were found. Also,
same-race conditions demonstrated increased aggressive behavior compared to
opposite-race conditions. Theoretical implications of these findings are discussed.
Keywords
Proteus effect, avatars, race, stereotypes, priming, video games
1
Department of Communication Studies, Clemson University, Clemson, SC, USA
Corresponding Author:
Erin Ash, Department of Communication Studies, Clemson University, 415 Strode Tower, Clemson,
SC 29634, USA.
Email: ash3@clemson.edu
Games and Culture
1-19
ªThe Author(s) 2015
Reprints and permission:
sagepub.com/journalsPermissions.nav
DOI: 10.1177/1555412014568870
gac.sagepub.com
at CLEMSON UNIV on April 13, 2015gac.sagepub.comDownloaded from
Video games offer users a significant amount of interactivity and customization
options compared to more traditional forms of media. While immersed in video
game worlds, players take on the role of a character, or avatar, and play the game
through the eyes of this avatar. Previous research has demonstrated that cues derived
from a player’s avatar can lead to attitudinal and behavioral effects based on
expected qualities inferred from those cues, a phenomenon referred to as the Proteus
effect (Yee & Bailenson, 2007). To illustrate, a researcher studying the Proteus
effect would likely predict that a player using a female avatar would behave less
aggressively due to popular views about femininity. The effect could also be applied
to assumed generalizations about other social groups, also known as stereotypes.
The purpose of this study was to test the Proteus effect as it applies to the relation-
ship between behavior, stereotyping, and avatar race. Specifically, this research
sought to determine whether players ascribe stereotypes associated with aggression
to Black avatars when playing a video game and how that impacts aggressive play.
The goals of this research were 2-fold. First, this research investigates whether racial
stereotypes that associate African Americans with aggression might be attributed to
Black avatars, leading to increased levels of aggressive in-game behavior and post-
game feelings and cognitions—a direct test of the Proteus effect and its application
to the context of sports games. Second, this research also sought to explore video
games as a medium for assessing implicit racial attitudes by considering whether the
Proteus effect can serve as a framework for examining negative racial attitudes that
are expressed behaviorally in a video game world. That is, individuals’ experience of
embodiment of the avatar during game play, in which they ‘‘take on’’ the character-
istics they ascribe to the avatar, is investigated as a state in which stereotypes they
associate with the avatar’s racial group may be revealed.
Literature Review
Racial Stereotypes About Aggression
Black males have been stereotyped as criminal, aggressive, and violence-prone
throughout American history (Collins, 2005; Ferber, 2007). Research has demon-
strated that individuals in the United States are more fearful of African Americans
and are more likely to associate Blacks than Whites with crime (Oliver, Jackson,
Moses, & Dangerfield, 2004) and, further, that darker skin tones are positively asso-
ciated with perceived criminality and aggressiveness (Maddox & Gray, 2002). In
fact, ‘‘aggressive’’ was listed by White Americans as one of the five most common
characteristics of Blacks in one study and was one of the traits most frequently used
to describe Blacks in another (Madon et al., 2001). Further, the results of a U.S.
national television survey published in 1990 concluded that African Americans are
identified as violent more often than any other racial group (Hoberman, 1997).
Evidence of the stereotype that links African Americans to aggression can also be
found in American media. A significant body of research has supported the idea that
2Games and Culture
at CLEMSON UNIV on April 13, 2015gac.sagepub.comDownloaded from
Blacks are represented in the media as criminal and as associated with drugs, gang
violence, and other negative issues (see R. Entman & Rojecki, 2000). In addition,
numerous studies have demonstrated that African Americans dominate the popula-
tion of criminal suspects presented in media (Dixon & Linz, 2000; R. M. Entman,
1990; Oliver, 1994; Romer, Jamieson, & de Couteau, 1998). As a result, knowledge
structures shaped by media may induce an automatic categorization of all African
Americans that associates them with negative traits, such as aggression and hostility
(Devine, 1989; Gaertner & Dovidio, 1986).
Racial stereotyping is not limited to traditional media, however, yet little empiri-
cal research has been conducted to examine the effects of race portrayals in more
interactive forms of media, particularly in games. Children Now (2001), a
community-based organization in Oakland, California, conducted a study in 2002
to quantify the proliferation of stereotypes in video games and found that when pre-
sented in sports video games, Black characters were more likely than White charac-
ters to display aggressive behavior, such as trash talking and pushing, 79%compared
to 57%, respectively. Another problematic finding from the study was that 61%of
Black characters were unaffected by violence. In fact, only 15%displayed physical
pain or bodily harm when the victims of violence. As stereotypical portrayals are
being enacted rather than passively consumed, video games may have an even
greater effect on racial attitudes (Leonard, 2003).
Video Game Avatar Identification and Embodiment
Identification, a concept studied in the context of interactive and traditional media, is
important to understanding how racial attitudes and stereotypes can manifest them-
selves in virtual environments. Identification is especially important in the context of
video games, because users act ‘‘as,’’ as opposed to ‘‘with,’’ a media character
(Cohen, 2001). Klimmt, Hefner, and Vorderer (2007, p. 2) argue that traditional the-
ories of identification are not applicable to identification with video game charac-
ters; rather than capturing the degree to which an individual perceives the avatar
is like him or her, identification in video games is better described as a ‘‘temporary
alteration of self-perception by inclusion of properties of the target media charac-
ter.’’ Recent research from Van Looy, Courtois, and De Vocht (2010) further expli-
cates the concept of player identification in role-playing games and offers the
following three main dimensions of identification: avatar identification, group iden-
tification, and game identification, with avatar identification consisting of perceived
similarity, wishful identification, and embodied presence. The latter, embodied pres-
ence or embodiment of the avatar, is the identification component stressed by
Klimmt et al. (2007), and the one that is most useful for understanding how individ-
uals’ perceptions of the characteristics of the avatar would lead them to behave dif-
ferently within a game.
Because identification is an important factor of media enjoyment, individuals
may use shortcuts when attempting to identify with or in the context of interactive
Ash 3
at CLEMSON UNIV on April 13, 2015gac.sagepub.comDownloaded from
media embody a character dissimilar to them. Thus, the social identity of the avatar
may lead individuals unfamiliar with the avatar and the environment to infer appro-
priate behavior through the use of stereotypes about the avatar’s identity and social
group. However, contrary to the idea of deliberate uncertainty reduction accom-
plished through the use of stereotypes, a study investigating the effect of racial
embodiment on racial bias found that the embodiment of a Black avatar led to
increase in implicit racial bias. The authors conclude that racial embodiment acti-
vated individuals’ race-based schema and influenced their implicit negative racial
attitudes (Groom, Bailenson, & Nass, 2009). The implicit nature of the negative bias
produced suggests that the social judgments used to embody an avatar may not be a
conscious process by which individuals deliberately make inferences about expected
behavior based on visual cues, but that embodying an avatar of a different social
identity may unconsciously activate stereotypes that influence behavior in ways the
individual is not aware of.
Proteus Effect
One effect of identification with characters through the use of stereotypes is
explained by the Proteus effect. Conceptualized in computer-mediated communica-
tion research by Yee and his colleagues (Yee & Bailenson, 2007), the effect
‘‘expects users to make inferences about their expected dispositions from their ava-
tar’s appearance and then conform to the expected attitudes and behavior’’ (Yee,
Bailenson, & Ducheneaut, 2009, p. 294). Thus, it predicts individuals will embody
the expected characteristics of an avatar with a salient social identity, which ulti-
mately influences their behavior while acting as the avatar.
The Proteus effect has been examined in several direct empirical investigations
conducted in virtual settings. For example, one early study determined that individ-
uals using an attractive avatar walked closer and disclose more personal information
to a confederate. Similarly, individuals who were delivered cues in a virtual setting
that they were embodying a tall avatar behaved more confidently, as would be pre-
dicted in a real-world setting (Yee & Bailenson, 2007). Other research has moved
outside of the virtual reality lab to investigate whether Proteus effects can occur dur-
ing video game play, and research suggests that they may. Pena, Hancock, and Mer-
ola (2009) found that participants whose video game avatar was Black-cloaked
displayed more aggressive intentions and attitudes compared to those using
White-cloaked avatars in the game. In a second experiment, participants using a
Ku Klux Klan (KKK)
1
-affiliated avatar demonstrated less affiliative behavior than
participants using avatars dressed as doctors. In each of these studies, participants’
game play was consistent with behaviors and play styles that could be inferred about
the avatar based on expectations or stereotypes.
Research by Eastin, Appiah, and Cicchirllo (2009) extended that research to
explore how inferences about a Black avatar would influence video game experi-
ences. Their research demonstrated effects of the avatar race on postgame hostile
4Games and Culture
at CLEMSON UNIV on April 13, 2015gac.sagepub.comDownloaded from
thoughts, or aggressive cognition, and that White participants who played a violent
game as a Black avatar showed higher levels of hostile thoughts after game play
compared to White participants playing as a White avatar. The authors discussed this
finding as support for the Proteus effect, concluding that the increase in aggressive
thoughts for Whites playing as a Black avatar is explained by those participants’
embodiment of the avatar that lead, in a manifestation of negative stereotypes about
African Americans that associate the group with aggression-related concepts, to an
increase in aggressive behavior within the game. They cite the General Aggression
Model (GAM) conceived of by Bushman and Anderson (2002), which suggests that
priming of aggression via exposure to violence or aggressive acts causes increases in
aggressive thoughts, feelings, and behavior, and the model’s association with the
cognitive neoassociation model (Berkowitz, 1984) as their framework for conclud-
ing that the aggressive thoughts measured were a product of the level of aggression
enacted in game play. Essentially, they are arguing that aggressive game play leads
to the activation of aggression-related ideas, such that any differences in thoughts are
a direct result of differences in in-game behavior. This led them to conclude that
players using Black avatars showed greater levels of hostile thoughts because, in
support of the Proteus effect, they had associated African Americans with hostility
and, thus, behaved more aggressively in the game.
Aggression is defined in the GAM and most research on video games and aggres-
sion as an outcome that results from the violence present in games; the GAM pre-
dicts both short-term and eventual long-term effects caused by learning processes
and the development of violence-associated scripts over time. In short, the violence
in game content leads to increases in aggression both in and subsequent to game play
(Bushman & Anderson, 2002). The Proteus effect, however, involves predicting in-
game behavior and attitudes based on identity cues derived from the avatar. Thus, if
the in-game behavior under examination is aggression, a Proteus effect would be
supported if individuals playing with an avatar whose group identity is associated
with greater levels of aggression behaved more aggressively in the game. That is,
a Proteus effect is only observed if individuals embodying the avatar that would
be stereotyped as aggressive show an increase in aggressive in-game actions (Yee
et al., 2009).
Although it may be the case that Eastin et al.’s argument that aggressive cognition
is derived from aggressive behavior is valid, it is also possible that playing as a Black
avatar primed cognitions that did not affect behavior. Aggressive cognition was
measured in the Eastin et al. study using a word-completion test developed by
Anderson (1999) in which participants fill in letter blanks to form words that are
later coded as aggressive and nonaggressive. Although this is a widely used and reli-
able scale, its use is potentially problematic in this line of research because many of
the words in the task are associated with crime that, as previously discussed, is
strongly associated with Black individuals (e.g., murder, mugger, prison, rapist, and
shoot). Thus, it is possible the finding that Black avatars were associated with
aggression, operationalized as hostile thoughts, was an artifact of the priming effect
Ash 5
at CLEMSON UNIV on April 13, 2015gac.sagepub.comDownloaded from
of Black avatars on the activation of crime-related ideas. In that case, in-game beha-
vior may not have been the determinant of aggression and no Proteus effect
observed.
This Research
Therefore, this research seeks to replicate and build on the research of Eastin et al.
(2009) and others in several ways. Primarily, this study investigates whether there
exists a distinction between the Proteus effect and associative priming as explana-
tions for the effects of avatar race on aggression. If increases in aggressive in-
game actions are found for individuals playing as a Black avatar, the Proteus effect
and priming remain plausible explanations of avatar race effects. However, if no
differences for in-game behavior are revealed, the Proteus effect can essentially
be ruled out as the mechanism through which avatar race affects aggression in this
study. Also, research on avatar identification and the Proteus effect offer stereotyp-
ing as a potential mechanism for understanding avatar effects. Thus, this study
attempts to gauge the extent to which individuals understood their avatar in terms
of racial stereotypes. Finally, a third-person video game rather than a first-person
game, which has been more common in this line of research, is used in this study.
Perceived embodiment is measured to collect evidence as to whether Proteus effect
research can be applied to non-first-person games. Specifically, five hypotheses
were proposed and are as follows.
As a first step in investigating the Proteus effect in this study, the use of stereo-
types to describe the avatar, an important tenant of the effect, will be examined. In
order for a Proteus effect to occur, a player must assign some traits to the avatar
that would translate to behavioral expectations. In this study, specifically, it is pre-
dicted that aggressive behavior on the part of those playing with a Black avatar
would stem from stereotypes about African Americans. The documented existence
of race-based schemas that associate Black individuals with certain traits, such as
aggression and violence, leads to the prediction that individuals will describe
Black avatars in these terms.
Hypothesis 1: Players in Black avatar conditions will use more stereotypes
about African Americans in describing their avatar than will players in White
avatar conditions.
It is also expected that playing the game as a Black avatar will lead to differ-
ences in game play through the embodiment of aggression-related characteristics
associated with Black identity in support of the Proteus effect of avatar race. Addi-
tionally, assumptions of the GAM using the cognitive neoassociation model used
by Eastin et al. (2009) to interpret their hostile thoughts finding as a Proteus effect
and the potential for avatar race to prime aggression-related concepts will be
tested by measuring individuals’ aggressive cognition and aggressive affect fol-
lowing game play.
6Games and Culture
at CLEMSON UNIV on April 13, 2015gac.sagepub.comDownloaded from
Hypothesis 2: Playing with a Black avatar will lead to greater (a) aggressive
in-game behavior, (b) aggressive cognition, and (c) aggressive affect than will
playing with a White avatar.
A potential limitation of some studies examining the Proteus effect is that the
embodiment of assumed avatar characteristics is assumed, rather than measured.
Therefore, the study considers whether the extent to which a person embodies the
avatar plays a role in the effects of avatar race on aggression. Specifically, partici-
pants indicated the extent to which they felt that they were themselves the avatar dur-
ing game play. It is predicted that participants playing with a Black avatar that have
inferred appropriate behavior through the use of stereotypes about African Ameri-
cans will demonstrate greater effects for avatar race on aggression.
Hypothesis 3: The positive relationship between playing with a Black avatar
and (a) aggressive in-game behavior, (b) aggressive cognition, and (c) aggres-
sive affect will be stronger for individuals who use Black stereotypes to
describe their avatar.
One of the key assumptions of the Proteus effect is that individuals behave as
their avatar when immersed in virtual worlds. Thus, it is expected that individuals
who felt they had embodied the avatar would be more likely to behave in ways that
match the perceived traits of the avatar, resulting in a stronger effect.
Hypothesis 4: The positive relationship between playing with a Black avatar
and (a) aggressive in-game behavior, (b) aggressive cognition, and (c) aggres-
sive affect will be stronger for individuals who experience greater levels of
embodiment.
Social identities, such as race, often have implications for perceptions of others, as
well. Social identities can be cued by social comparisons between groups to which we
belong and other (out) groups. Differing perceptions of self and other may emerge
depending on which identity is most salient in that particular context or at any specific
point in time (Ellemers et al., 2002). Thus, when self and others’ identities are differ-
ent, socialidentities, in this case race, may be more salient and race-based schema may
be more likely to be activated as a means to infer appropriate behavior for the avatar.
In that case, playing the game as a Black avatar against a White opponent would lead
to the highest levels of in-game and postgame aggression.
Hypothesis 5: (a) Aggressive behavior, (b) aggressive cognition, and (c)
aggressive affect will be greater for participants playing as a Black avatar
against a White opponent, controlling for gender.
Method
Participants were students (N¼84) from undergraduate communication courses at a
large northeastern university in the United States and were awarded a small amount
Ash 7
at CLEMSON UNIV on April 13, 2015gac.sagepub.comDownloaded from
of extra credit for their participation. Participants were majority female (60.7%) and
ranged in age from 18 to 25 (M¼20.93, SD ¼0.49). A large majority was White
(85.7%), followed by Asian (7.1%), Hispanic or Latino, and African American (each
3.6%).
Procedure and Manipulation
Stimulus. A video game produced by Electronic Arts Sports, Fight Night 4, was used
for this study. The game simulates boxing matches between two opponents and has
several modes that allow players to compete in a single match or develop the career
of a chosen boxer. Participants played in the game’s single match mode, called
‘‘Fight Now.’’ Participants competed in three 3-minute rounds. The first round
served as a practice round, and data from two subsequent matches were collected.
Fight Night 4 allows users to select a venue for their match. Game play was con-
ducted through an Xbox 360 console and wireless controller and was displayed on
a 42-in high definition television.
Manipulation. The game allows users to customize boxers, so the race of customized
boxers was able to be easily manipulated. Four avatars (boxers) were designed by the
researcher. The first two boxers that were designed were used as the participant’s
avatar. The two avatars, both named ‘‘Jimmy Bennett,’’ were identical in ability,
skill level, and features, except for skin color, hair color and style, and eye color.
One version of Jimmy Bennett was White, the other is Black. The other two boxers,
named ‘‘Donny Davidson,’’ were used as the participant’s opponent, and, like the
Jimmy Bennett avatars, were identical in every feature, except for skin color, hair
color and style, and eye color, such that one avatar was White and the other was
Black. The ability and skill levels of Bennett and Davidson also matched and were
set to a medium ability with the expectation that this would prevent participants from
being easily beaten or easily winning.
Study Design
The study employed a 2 (Avatar Race)2 (Opponent Race) experimental design.
Each participant was randomly assigned to play as a White or Black boxer against a
White or Black boxer, creating the following four conditions: Black avatar, White
opponent; Black avatar, Black opponent; White avatar, White opponent; and White
avatar, Black opponent (see Table 1). Twenty-one participants were assigned to each
condition.
Procedure
Before entering the lab, participants completed a pencil and paper questionnaire to
determine their video game play experience and demographic characteristics. When
8Games and Culture
at CLEMSON UNIV on April 13, 2015gac.sagepub.comDownloaded from
completed, they entered the lab and were seated in front of the game; the avatar and
opponent were selected prior to participant entry. The researcher first gave the par-
ticipant a tutorial on how to use the game’s controls and allowed him or her to com-
plete one 3-minute round of play as a practice round. The participant then played two
3-minute rounds to complete a match and his or her performance was recorded by
the researcher. After completing game play, a second questionnaire was adminis-
tered to assess participants’ aggressive affect and cognition, use of stereotypes to
describe the avatar, and perceived embodiment. After completion of Part 2, he or she
was thanked and debriefed.
Study Measures
Aggression. Aggressive cognition was measured using the word completion task
developed and used by Anderson and his colleagues (Anderson, Carnagey, &
Eubanks, 2003; Anderson et al., 2004) and Eastin et al. (2009). For this study, 20
words from the original 98-item word-completion task were chosen at random; 11
of the 20 could be completed as aggressive words. Participants were asked to fill out
each word immediately following game play. Each word was then coded as aggres-
sive or nonaggressive (neutral, ambiguous, or nonwords), using the coding key pro-
vided by Anderson and colleagues. Aggressive cognition was then computed as the
sum of aggressive words, with a possible range of 0 to 11, for each participant (M¼
5.46, SD ¼1.85).
Aggressive affect was measured using Anderson, Deuser, and DeNeve’s (1995)
state hostility scale on a 5-point scale. The 35-item scale (a¼.93, M¼2.23,
SD ¼0.50) determines participants’ current mood by asking them how much they
agree with feeling statements, such as Ifeelfurious,I feel good-natured,andI fee l vexed.
Aggressive behavior was operationalized as making contact with the opponent
through punching (M¼235.15, SD ¼36.12). During game instruction, participants
were given specific directions regarding how to make aggressive and defensive
plays. Aggressive moves were performed using the left side of the game controller,
whereas defensive moves are performed using the right side of the controller. This
was explained with the expectation that it would make each type of move more dis-
tinct for participants. Information was displayed on-screen after each match that
indicated how many punches were thrown by each participant.
Table 1. Study Conditions.
Avatar Opponent
Condition 1 Black boxer White boxer
Condition 2 Black boxer Black boxer
Condition 3 White boxer White boxer
Condition 4 White boxer Black boxer
Ash 9
at CLEMSON UNIV on April 13, 2015gac.sagepub.comDownloaded from
Describing the avatar. Participants’ use of stereotypes in perceiving their avatar were
measured using a scale developed by Kawakami and Dovidio (2001) to examine the
reliability of latent response times in measuring implicit stereotyping. Although this
study does not measure participant response time to assess implicit attitudes, the
instrument may be an effective measurement tool for capturing to what extent par-
ticipants responded to their avatar in stereotypical ways.
The scale consists of 48 adjectives, 8 for each category, which can be used to
describe a stereotypically positive or negative quality of Whites or Blacks. For
example, Black stereotypes include ‘‘loud,’’ ‘‘hostile,’’ ‘‘intimidating,’’ ‘‘strong,’’
‘‘humorous,’’ and ‘‘rhythmic.’’ The scale also includes 16 filter items. Participants
were asked to indicate how well each adjective described their avatar on a scale
ranging from 1 (describes very poorly)to5(describes very well). A scale represent-
ing Black stereotype endorsement (a¼.74, M¼3.00, SD ¼0.44) was created by
computing the mean for all of the Black stereotypical characteristics.
Avatar embodiment. Embodiment with avatar was measured using a 4-item identifi-
cation scale (a¼.84, M¼2.70, SD ¼1.00) developed by C. Klimmt (personal cor-
respondence, February 23, 2010). The scale consists of the following items, such as I
literally had the feeling I was in the character’s skin and I almost had the feeling of
actually being the character, measured on a 5-point scale ranging from 1 (strongly
disagree)to5(strongly agree).
Control measures. Participants were also asked how often they play video games of
different genres using various game consoles. They were also asked how often they
play the stimulus game used for the study, Fight Night 4. Participants indicated their
responses using 7-point scale with Didn’t play last year as 1 and More than 4 times a
week as 7. Video game experience (M¼2.13, SD ¼1.19) was computed as the mean
for participants’ reported time spent playing first-person shooter,role-playing,real-
time strategy, and sports games (a¼.75). Demographic information, including gen-
der, age, and ethnicity, was also collected. Video game experience and gender (see
Lucas & Sherry, 2004) were controlled for statistically in all analyses.
2
Results
A series of regression tests were conducted to determine whether support was found
for each hypothesis. Table 2 presents the means of the dependent variables for each
avatar race condition.
Avatar Race and Stereotype Use
Hypothesis 1 predicted that players in Black avatar conditions would describe the
avatar using stereotypes associated with African Americans and those with White
avatars would describe the avatar using stereotypes about Whites. A regression
10 Games and Culture
at CLEMSON UNIV on April 13, 2015gac.sagepub.comDownloaded from
model examining avatar race as a predictor of Black stereotype use was tested. This
analysis revealed avatar race was not associated with Black stereotypes. A second
regression model was conducted to determine whether avatar race predicts White
stereotype use and was nonsignificant.
A subsequent paired sample ttest showed that Black stereotype use (M¼3.00,
SD ¼0.44) was significantly higher than was White stereotype use (M¼2.39,
SD ¼0.52), t(82) ¼12.96, p< .001, suggesting that, contrary to the prediction in
Hypothesis 1, the boxing avatars, regardless of race, were described by participants
as having more Black stereotypes than White stereotypes.
Proteus Effect of Avatar Race on Aggression
Hypotheses 2a through 2c were tested using hierarchical regression models testing
avatar race as a predictor of aggressive behavior, cognition, and affect.
Hypotheses 2a was a direct test of the Proteus effect, predicting playing with a
Black avatar would lead to greater levels of aggressive in-game behavior. Analysis
of the model predicting aggressive behavior revealed no significant effect of avatar
race.
Hypotheses 2b and 2c were included to test whether avatar race was associated
with postgame aggressive cognition and affect. Analysis of the models predicting
cognition and affect revealed playing with a Black avatar is not associated with
increased levels of either form of aggression.
Stereotype Use and Avatar Race on Aggression
Hypotheses 3a through 3c examined Black stereotype use as a moderator of the
effects of avatar race on aggressive in-game behavior, cognition, and affect such
that the positive relationship between using a Black avatar and aggression would
be stronger for players describing the avatar using stereotypes. Hierarchical regres-
sion was employed to examine the Black Avatar Stereotype Use interaction as a
predictor of aggressive behavior, cognition, and affect. The model consisted of the
following two blocks: the first including avatar race, Black stereotype use, and the
Table 2. Means of Aggression and Stereotype Use.
Black avatar M(SD) White avatar M(SD)
Black stereotypes 2.98 (0.44) 3.02 (0.45)
White stereotypes 2.38 (0.47) 2.40 (0.56)
Aggressive behavior 227.88 (34.80) 242.43 (36.35)
Aggressive cognition 5.60 (1.82) 5.33 (1.90)
Aggressive affect 2.22 (0.49) 2.24 (0.53)
Embodiment 2.62 (0.96) 2.78 (1.05)
Ash 11
at CLEMSON UNIV on April 13, 2015gac.sagepub.comDownloaded from
control variables, and the second including the interaction term between avatar race
and stereotype use.
Analysis of the model predicting aggressive behavior indicated that the interac-
tion between avatar race and stereotyping does not significantly predict behavior.
The overall model is also nonsignificant.
A model examining the avatar race and stereotyping interaction as a predictor of
aggressive cognition was then tested. The interaction is not significant. However,
use of Black stereotypes to describe ones’ avatar does emerge as a significant pre-
dictor of aggressive thoughts, b¼.28, p< .05. The negative standardized coeffi-
cient indicates that participants who played as a Black avatar experienced decreased
levels of aggressive cognition after playing. However, the overall model containing
all three blocks was not significant, and Black stereotype use emerged the only sig-
nificant individual predictor in the model. The statistics associated with this model
are reported in Table 3.
The interaction between avatar race and stereotype use was then examined as a
predictor of aggressive affect. Once again, the interaction term is not significant. A
main effect for Black stereotype use is again revealed, b¼.23, p¼.05; however, the
standardized coefficient is positive, revealing that, contrary to the negative relation-
ship between using Black stereotypes to describe one’s avatar and aggressive cogni-
tion, stereotype use was associated with increased levels of aggressive affect. The
overall model with inclusion of the interaction term is not significant. The statistics
from this analysis are presented in Table 4.
Embodiment and Avatar Race on Aggression
Hypotheses 4a through 4c predicted that the relationships between avatar race and
aggressive in-game behavior, cognition, and affect would be moderated by per-
ceived embodiment of the avatar while playing. To test these hypotheses, a series
of hierarchical regression models were tested, with lower order terms entered in the
first step, and the interaction term entered on the second step of the analysis.
Table 3. Avatar Race and Black Stereotype Use Predicting Cognition.
bR
2
DR
2
F (df)
Step 1
Gender .12 .06 — 2.26 (4, 78)
Game experience .08
Avatar race .05
Black stereotype use .28*
Step 2
Race Stereotype Use 1.28 .08 .03 2.34 (5, 77)
Note. Gender was coded as females ¼1, males ¼0. Avatar race was coded as Black ¼1, White ¼2.
*p< .05.
12 Games and Culture
at CLEMSON UNIV on April 13, 2015gac.sagepub.comDownloaded from
In the model for Hypothesis 4a, examining an effect for aggressive behavior, the
results of the Avatar Race Embodiment interaction, as well as the model overall
with the interaction term added, while not significant at the .05 level, approached
statistical significance, b¼.89, p< .10 (see Table 5) and thus the interaction was
further probed. Simple slopes analysis revealed a positive relationship between
aggressive behavior and embodiment for players using a Black avatar (b¼.19) and
a negative relationship between behavior and embodiment for those with a White
avatar (b¼.22); however, neither simple slope is significant. Thus, the interaction
was further probed using a script for SPSS developed by Hayes and Matthes (2009).
The results of this analysis revealed avatar race is a significant predictor of aggres-
sive in-game behavior when individuals experienced levels of embodiment greater
than 1 SD above the mean of embodiment, b¼.55, p< .05, whereas race did not
affect behavior at lower levels of embodiment.
Hypothesis 4b predicted that embodiment would moderate the relationship
between avatar race and aggressive cognition. No significant effect was revealed.
A model testing Hypothesis 4c, which predicted a relationship between avatar race
and aggressive affect would be moderated by embodiment revealed no significant
Table 4. Avatar Race and Black Stereotype Use Predicting Affect.
bR
2
DR
2
F (df)
Step 1
Gender .12 .00 — 1.03 (4, 78)
Game experience .03
Avatar race .01
Black stereotype use .23*
Step 2
Race Stereotype Use .43 .01 .00 .87 (5, 77)
Note. Gender was coded as females ¼1, males ¼0. Avatar race was coded as Black ¼1, White ¼2.
*p¼.05.
Table 5. Avatar Race and Embodiment Predicting In-Game Behavior.
bR
2
DR
2
F (df)
Step 1
Gender .04 .03 — 1.59 (4, 78)
Game experience .17
Avatar race .14
Embodiment .01
Step 2
Race Embodiment .89* .06 .04* 2.01* (5, 77)
Note. Gender was coded as females ¼1, males ¼0. Avatar race was coded as Black ¼1, White ¼0.
*p< .10.
Ash 13
at CLEMSON UNIV on April 13, 2015gac.sagepub.comDownloaded from
interaction. Embodiment did emerge as a significant predictor of aggressive affect,
b¼.50, p< .001, indicating that those participants who experienced greater levels of
embodiment reported more hostile feelings subsequent to game play, as presented in
Table 6.
Opponent Race and Aggression
Hypotheses 5a through 5c investigated the effects of playing the game against a
same-race or opposite-race opponent with the expectation that levels of aggression
would be greatest for those with Black avatars playing against a White opponent. In
order to test this, the avatar and opponent race conditions were dummy coded to
determine whether each condition differed significantly from the Black avatar,
White opponent condition. Subsequently, a series of multiple regression models
were employed to examine differences between Black avatar, White opponent game
play, and all other conditions for aggressive behavior, cognition, and affect. Results
revealed all three models were not significant; playing as a Black avatar against a
White opponent was not associated with higher levels of any form of aggression.
Discussion
To summarize, participants playing the video game with a Black avatar did not char-
acterize the avatar using Black stereotypes to a greater degree than those with a
White avatar or display greater levels of aggressive behavior, cognition, or affect.
Characterizing Black avatars using Black stereotypes did not produce higher levels
of aggression or moderate any effects of avatar race on aggression. Embodiment
was revealed as a marginally significant moderator of the positive relationship
between playing with a Black avatar and aggressive within-game behavior and also
led to greater postgame aggressive affect.
Ultimately, the results of this study demonstrate the need to measure specific
mechanisms underlying the Proteus effect when testing the effect for third-person
Table 6. Avatar Race and Embodiment Predicting Affect.
bR
2
DR
2
F (df)
Step 1
Gender .05 .20 — 6.05*** (4, 78)
Game experience .11
Avatar race .00
Embodiment .50***
Step 2
Race Embodiment .03 .19 .00 4.78*** (5, 77)
Note. Gender was coded as females ¼1, males ¼0. Avatar race was coded as Black ¼1, White ¼0.
***p< .001.
14 Games and Culture
at CLEMSON UNIV on April 13, 2015gac.sagepub.comDownloaded from
games and suggest that testing these mechanisms might be a useful tool for future
research. Had embodiment been assumed rather than measured, the conclusion
would have been that no Proteus effect was found for avatar race. Instead, beha-
vioral differences were only revealed for those who perceived embodiment at a high
level. Furthermore, the use of stereotypes associated with the race of the avatar was
measured but revealed no difference based on race and no effect for behavior or
postgame cognition and affect. Post hoc regression tests were employed to determine
whether, as suggested in the literature, cues from the avatar via heuristic processing
activated stereotypes that influenced in-game behavior (Nowak, 2004; Nowak &
Rauh, 2005). This analysis found no effect of stereotype use on embodiment sug-
gesting, as has also been offered as an explanation for avatar identification and
embodiment of the Proteus effect, that the process of identification based on avatar
characteristics is implicit and automatic rather than a deliberate process of orienting
oneself to the virtual or game world (Groom et al., 2009).
Stereotyping Blacks or Boxers?
The lack of effects for stereotyping could also be interpreted as a function of the sti-
mulus chosen, however, because boxing is often regarded as a raced sport. The stron-
gest evidence for this alternate explanation is the finding that participants were more
likely to use Black stereotypes to describe their avatars—both White and Black.
Furthermore, many of the adjectives from the list Black stereotypes participants used
to evaluate their boxer are words that could be used to describe any boxer or athlete
(e.g., athletic, strong, muscular, tough, intimidating). Therefore, the concept of
Black stereotyping may have been confounded with the concepts of being a good
athlete and boxer. In that case, the Proteus effect of avatar race on behavior could
be interpreted as the embodiment of the athletic superiority stereotype attributed to
Black men, rather than a manifestation of stereotypes associating Black men with
aggression or violence. Research has repeatedly demonstrated the prevalence of this
stereotype in society (Ferber, 2007; Hoberman, 1997) and the existence of a schema
associating Blackness with athleticism (Buffington & Fraley, 2008).
Other findings may also be considered further evidence that effects were driven
by stereotypes about athleticism, rather than aggression. Specifically, the use of
Black stereotypes to describe one’s avatar led to a decrease in aggressive cognition
but to an increase in aggressive affect. On the surface, this appears to be a peculiar
finding, as one would expect aggressive cognition and affect to be related, but it may
be the case that game play and subsequent stereotyping actually had no effect on
aggression. Instead, participants who described their avatar in ways consistent with
being a good boxer may have had fewer aggressive thoughts because they under-
stood the match as a sporting event rather than a series of aggressive or violent acts.
This would also explain the positive relationship between Black stereotype use and
aggressive affect. If participants conceived of their avatars as boxers and perceived
the context in terms of competition rather than violence, they might feel more
Ash 15
at CLEMSON UNIV on April 13, 2015gac.sagepub.comDownloaded from
‘‘hostile’’ because of the emotional intensity that can be associated with competitive
contexts and not as a result of enacting violence.
Limitations and Future Research
The task of choosing an appropriate game to use as the stimulus is always challen-
ging, as video games are not easily customized for manipulation as would be the case
with a news story, website, or video clip. As evidenced in the results, the choice of a
boxing game was not a perfect one. The researcher wanted to examine racial stereo-
typing in a video game more reflective of a real-world environment and thus a sports
game was chosen. However, the abundance of other potential priming devices, such
as racial stereotypes associated with boxing, the setting of the match, or the simple
fact that the game features a physically violent sport, certainly had the potential to
override or dilute effects of the race of the avatar on aggression and stereotype use.
Another significant difference between this research and previous studies of the
Proteus effect is the type of game chosen for study. Previous research on the effect
has typically used first-person fantasy game worlds as stimuli. It might be consid-
ered both a strength and a weakness of this study that a more realistic setting was
chosen for game play—the strength being that any results found are more applicable
to real-world environments, and the weakness being that more realistic games, such
as sports games, bring with them considerable sets of cultural baggage. To elaborate,
in a completely virtual fantasy game environment, the only cues to infer appropriate
behavior would be avatar identity characteristics, causing traits such as race to be
extremely salient. The game environment in Fight Night was filled with cues other
than the color of the avatar’s skin, which may have become salient to a player and
influence play. The question then is whether the Proteus effect can occur in a more
realistic context with competing cues or is a phenomenon only observed in truly
‘‘virtual’’ settings. Furthermore, if it is the latter, what is the application, if any, for
understanding identity, embodiment, and stereotyping in the real world?
Eastin et al. (2009) previously was able to demonstrate a Proteus effect of race on
cognition—participants in their study who were assigned Black avatars showed an
increase in hostile thoughts in comparison to players with White avatars. No effect
for avatar race on aggressive thoughts was revealed in this study. However, it is pos-
sible that in that study, aggression-related traits associated with African Americans,
in the absence of other prominent cues such as those present in the game chosen for
this study and previously discussed, made hostile words more accessible. Ulti-
mately, future research examining the effects of avatar race is both warranted and
necessary. Tests of the Proteus effect have typically relied on assumptions about
what characteristics of an avatar will be most salient and how those characteristics
influence behavior within a virtual world, for example, that individuals with tall or
attractive avatars will display greater confidence than those with short avatars, or
that concepts measured post-play are indicative of in-game behavior. However, to
better test this effect and improve our understanding of the mechanisms that explain
16 Games and Culture
at CLEMSON UNIV on April 13, 2015gac.sagepub.comDownloaded from
it, we should continue to explore novel research designs that will determine more
accurately: what stereotypes individuals use when making decisions about how to
behave in a game world; how to apply those findings to the operationalization and
measurement of in-game behavior as they relate to the avatar’s identity; and how this
phenomenon might be useful for our understanding of racial stereotyping in general.
Finally, these findings should be considered in light of the specific cultural con-
text in which the research was conducted. The participants who took part in the
research were mostly White, likely middle to upper-middle class, and American, all
of which play a significant role in their beliefs regarding both race and sport and, in
turn, the results of this research. More research is necessary to examine how the
Proteus effect and embodied stereotyping function in cross-cultural contexts.
Declaration of Conflicting Interests
The author(s) declared no potential conflicts of interest with respect to the research, authorship,
and/or publication of this article.
Funding
The author(s) received no financial support for the research, authorship, and/or publication of
this article.
Notes
1. The KKK is a White supremacy hate group that has been active in the United States since
1860s.
2. Although participant ethnicity was considered as another control, so few participants were
non-White (n ¼12), the decision was made not to do so in an effort to preserve statistical
power.
References
Anderson, C. A. (1999). Attributional style, depression, and loneliness: A cross-cultural com-
parison of American and Chinese students. Personality and Social Psychology Bulletin,
25, 482–499.
Anderson, C. A., Carnagey, N. L., & Eubanks, J. (2003). Exposure to violent media: The
effects of songs with violent lyrics on aggressive thoughts and feelings. Journal of
Personality and Social Psychology,84, 960–971.
Anderson, C. A., Carnagey, N. L., Flanagan, M., Benjamin, A. J., Eubanks, J., & Valentine, J.
C. (2004). Violent video games: Specific effects of violent content on aggressive thoughts
and behavior. Advances in Experimental Social Psychology,36, 199–249.
Anderson, C. A., Deuser, W. E., & DeNeve, K. (1995). Hot temperatures, hostile affect, hos-
tile cognition, and arousal: Tests of a general model of affective aggression. Personality
and Social Psychology Bulletin,21, 434–448.
Berkowitz, L. (1984). Some effects of thoughts on anti- and prosocial influences of media
events: A cognitive neoassociation analysis. Psychological Bulletin,95, 410–427.
Ash 17
at CLEMSON UNIV on April 13, 2015gac.sagepub.comDownloaded from
Buffington, D., & Fraley, T. (2008). Skill in Black and White: Negotiating media images of
race in a sporting context. Journal of Communication Inquiry,32, 292–310.
Bushman, B. J., & Anderson, C. A. (2002). Violent video games and hostile expectations: A
test of the general aggression model. Personality and Social Psychology Bulletin,28,
1679–1686.
Children, Now. (2001) Fair play? Violence, gender and race in video games. Retrieved
August 1, 2009, from www.childrennow.org/media/video-games/2001
Cohen, J. (2001). Defining identification: A theoretical look at the identification of audiences
with media characters. Mass Communication & Society,4, 245–264.
Collins, P. H. (2005). Black sexual politics: African Americans, gender, and the new racism.
New York, NY: Routledge.
Devine, P. G. (1989). Stereotypes and prejudice: Their automatic and controlled components.
Journal of Personality and Social Psychology,56, 5–18.
Dixon, T. L., & Linz, D. (2000). Race and the misrepresentation of victimization on local tele-
vision news. Communication Research,27, 547–573.
Eastin, M. S., Appiah, O., & Cicchirllo, V. (2009). Identification and the influence of cultural
stereotyping on post video game play hostility. Human Communication Research,35,
337–356.
Ellemers, N., Spears, R., & Doosje, B. (2002). Self and social identity. Annual Review of
Psychology,53, 161–186.
Entman, R., & Rojecki, A. (2000). Black image in the white mind. London, England:
University of Chicago Press.
Entman, R. M. (1990). Modern racism and the images of Blacks in local television news.
Critical Studies in Mass Communication,7, 332–345.
Ferber, A. L. (2007). The construction of Black masculinity: White supremacy now and then.
Journal of Sport and Social Issues,31, 11–24.
Gaertner, S. L., & Dovidio, J. F. (1986). The aversive form of racism. In S. L. Gaertner & J. F.
Dovidio (Eds.), Prejudice, discrimination, and racism (pp. 61–89). San Diego, CA:
Academic Press.
Groom, V., Bailenson, J., & Nass, C. (2009). The influence of racial embodiment on racial
bias in immersive virtual environments. Social Influence,4, 231–248.
Hayes, A. F., & Matthes, J. (2009). Computational procedures for probing interactions in OLS
and logistic regression: SPSS and SAS implementations. Behavior Research Methods,41,
924–936.
Hoberman, J. (1997). Darwin’s Athletes: How sport has damaged Black America and
preserved the myth of race. New York, NY: Houghton Mifflin Company.
Klimmt, C., Hefner, D., & Vorderer, P. (2007). ‘‘Identification with Media Characters as
Temporary Alteration of Media Users’ Self-Concept’’ Paper presented at the annual meet-
ing of the International Communication Association, 5-23-07, San Francisco, CA Online.
Kawakami, K., & Dovidio, J. F. (2001). The reliability of implicit stereotyping. Personality
and Social Psychology Bulletin,27, 212–225.
Leonard, D. (2003). ‘‘Live in your world, play in ours’’: Race, video games, and consuming
the other. Studies in Media & Information Literacy Education,3, 1–9.
18 Games and Culture
at CLEMSON UNIV on April 13, 2015gac.sagepub.comDownloaded from
Lucas, K., & Sherry, J. L. (2004). Sex differences in video game play: A communication-
based explanation. Communication Research,31, 499–523.
Maddox, K. B., & Gray, S. A. (2002). Cognitive representations of Black Americans:
Reexploring the role of skin tone. Personality and Social Psychology Bulletin,28,
250–259.
Madon, S., Guyll, M., Aboufadel, K., Montiel, E., Smith, A., Polumbo, P., & Jussim, L.
(2001). Ethnic and national stereotypes: The Princeton trilogy revisited and revised.
Personality and Social Psychology Bulletin,27, 996–1010.
Nowak, K. L. (2004). The influence of anthropomorphism and agency on social judgment in
virtual environments. Journal of Computer-Mediated Communication,9, 00.
Nowak, K. L., & Rauh, C. (2005). The influence of the avatar on online perceptions of
anthropomorphism, androgyny, credibility, homophily, and attraction. Journal of
Computer-Mediated Communication,11, 153–178.
Oliver, M. B. (1994). Portrayals of crime, race, and aggression in ‘‘reality-based’’ police
shows: A content analysis. Journal of Broadcasting & Electronic Media,38, 179–192.
Oliver, M. B., Jackson, R. L., Moses, N. N., & Dangerfield, C. L. (2004). The face of crime:
Viewers’ memory of race-related facial features of individuals pictured in the news.
Journal of Communication,54, 88–104.
Pena, J., Hancock, J. T., & Merola, N. (2009). The priming effects of avatars in virtual
settings. Communication Research,36, 838–856.
Romer, D., Jamieson, K. H., & de Coteau, N. J. (1998). The treatment of persons of color in
local television news: Ethnic blame discourse or realistic group conflict? Communication
Research,25, 286–305.
Van Looy, J., Courtois, C., & De Vocht, M. (2010). Player identification in online games:
Validation of a scale for measuring identification in MMORPGs. Fun and Games,
126–134.
Yee, N., & Bailenson, J. N. (2007). The Proteus effect: The effect of transformed
self-representation on behavior. Human Communication Research,33, 271–290.
Yee, N., Bailenson, J. N., & Ducheneaut, N. (2009). The Proteus effect: Implications of
transformed digital self-representation on online and offline behavior. Communication
Research,36, 285–312.
Author Biography
Erin Ash (PhD, Penn State University, 2013) is an assistant professor in the Department of
Communication Studies at Clemson University. Her research focuses on representations of
social groups in media, with an emphasis on the psychological processes that explain
media-based stereotyping.
Ash 19
at CLEMSON UNIV on April 13, 2015gac.sagepub.comDownloaded from
