Thompson KM, Haninger K. Violence in E-Rated Video Games. JAMA. 2001;286(5):591-598. doi:10.1001/jama.286.5.591
Author Affiliations: Harvard School of Public Health, Harvard University, Boston, Mass.
Medicine and the Media Section Editor: Annette
Flanagin, RN, MA, Managing Senior Editor.
Context Children's exposure to violence, alcohol, tobacco and other substances,
and sexual messages in the media are a source of public health concern; however,
content in video games commonly played by children has not been quantified.
Objectives To quantify and characterize the depiction of violence, alcohol, tobacco
and other substances, and sex in video games rated E (for "Everyone"), analogous
to the G rating of films, which suggests suitability for all audiences.
Design We created a database of all existing E-rated video games available
for rent or sale in the United States by April 1, 2001, to identify the distribution
of games by genre and to characterize the distribution of content descriptors
associated with these games. We played and assessed the content of a convenience
sample of 55 E-rated video games released for major home video game consoles
between 1985 and 2000.
Main Outcome Measures Game genre; duration of violence; number of fatalities; types of weapons
used; whether injuring characters or destroying objects is rewarded or is
required to advance in the game; depiction of alcohol, tobacco and other substances;
and sexual content.
Results Based on analysis of the 672 current E-rated video games played on home
consoles, 77% were in sports, racing, or action genres and 57% did not receive
any content descriptors. We found that 35 of the 55 games we played (64%)
involved intentional violence for an average of 30.7% of game play (range,
1.5%-91.2%), and we noted significant differences in the amount of violence
among game genres. Injuring characters was rewarded or required for advancement
in 33 games (60%). The presence of any content descriptor for violence (n
= 23 games) was significantly correlated with the presence of intentional
violence in the game (at a 5% significance level based on a 2-sided Wilcoxon
rank-sum test, t53 = 2.59). Notably, 14
of 32 games (44%) that did not receive a content descriptor for violence contained
acts of violence. Action and shooting games led to the largest numbers of
deaths from violent acts, and we found a significant correlation between the
proportion of violent game play and the number of deaths per minute of play.
We noted potentially objectionable sexual content in 2 games and the presence
of alcohol in 1 game.
Conclusions Content analysis suggests a significant amount of violence in some E-rated
video games. The content descriptors provide some information to parents and
should be used along with the rating, but the game's genre also appears to
play a role in the amount of violent play. Physicians and parents should understand
that popular E-rated video games may be a source of exposure to violence and
other unexpected content for children and that games may reward the players
for violent actions.
Created in 1994, the Entertainment Software Rating Board (ESRB) rates
video games according to the categories listed in the
Box and using content
descriptors, which game manufacturers display on the game box to inform consumer
choices.1 Analogous to the G rating of films,2 the E rating (for "Everyone") of video games suggests
suitability for all audiences, but the E rating does not mean violence-free.
1. Early Childhood (EC)
Titles rated EC have content suitable for children ages 3 and older
and do not contain any material that parents would find inappropriate.
2. Everyone (E)*
Titles rated E have content suitable for persons ages 6 and older. These
titles will appeal to people of many ages and tastes. They may contain minimal
violence, some comic mischief (for example, slapstick comedy), or some crude
3. Teen (T)
Titles rated T have content suitable for persons ages 13 and older.
Titles in this category may contain violent content, mild or strong language,
and/or suggestive themes.
4. Mature (M)
Titles rated M have content suitable for persons ages 17 and older.
These products may include more intense violence or language than products
in the Teen category. In addition, these titles may also include mature sexual
5. Adults Only (AO)
Titles rated AO have content suitable only for adults. These products
may include graphic depictions of sex and/or violence. Adults Only products
are not intended to be sold or rented to persons under the age of 18.
6. Rating Pending (RP)
Product has been submitted to the ESRB and is awaiting final rating.
*Prior to 1998, the "Everyone (E)" rating was named "Kids to Adults
Studies on children's use of various media document the popularity of
video games as a major source of entertainment. A recent study found that
70% of children (age, 2-18 years) live in homes that have at least 1 video
game console, 33% of children have video game consoles in their bedrooms,
and 30% of children in the study played video games the previous day.3 Children in the study reported playing video games
for 20 min/d on average, although older children (age, 8-18 years) accounted
for most of this use (average, 27 min/d), with boys spending significantly
more time playing video games than girls and white children playing video
games for significantly less time than black or Hispanic children.3 Unfortunately, little information exists about the
ratings and genres of the games that children play as a function of their
age, sex, family income level, and ethnicity, although some differences in
preferences exist.3 Overall, children appear
to play relatively more games in the action, adventure, and sports genres,3 but this may simply reflect the types of games available.
The health implications of exposure to video games and other media with
violent content remain uncertain, but considerable concern about the potential
impacts of children experiencing media violence exists within the broad medical
community.4,5 Although several
recent studies repeat concerns about the content of video games6- 10
and the marketing of violent entertainment to children,11
more research on the impact of violent interactive entertainment, including
video games, is needed.4,5,9
Remarkably, no quantitative analysis exists on content in E-rated video games
or on the relationship between game content and the ESRB content descriptors.
This study focuses on providing quantitative information to physicians and
parents about the content of E-rated video games.
Video game console systems continue to evolve12,13
with 3 manufacturers presently dominating the market: Nintendo, maker of Nintendo
64 (N64); Sony, maker of PlayStation (PS) and PlayStation 2 (PS2); and Sega,
maker of Dreamcast (DC). Popular arcade games featuring different types of
game play (eg, Space Invaders, Pole Position, Donkey Kong) served as the
first home video games and gave rise to the modern video game market with
its wide variety of games of different genres.14
We created a database of information about the universe of E-rated games available
for rent or sale in the United States by April 1, 2001 (accessible at http://www.kidsrisk.harvard.edu) because we expected that the level
of violence in video games might depend on genre. The process involved using
data from the ESRB1 and several Internet sites15- 18 to
identify all 672 E-rated console games, verify that each game was released
in the United States, determine each game's content descriptor(s), and classify
each game by 1 of 11 primary genres: action, adventure, casino, fighting,
puzzle, racing, role-playing, shooting, simulation, sports, and strategy.
A small number of games (n = 9) could not be classified by these genres and
were labeled as "other." Unfortunately, the subjective process of characterizing
game play, as well as the complexity introduced by the growing presence of
games of hybrid genres, have prevented a universal system for classifying
video games by genre. For example, one of the most popular games in our sample, The Legend of Zelda: Ocarina of Time, is classified as
action, adventure, or role-playing on different Internet sites.15,17,18
In such cases, we selected the genre that was most commonly used to describe
the game. Using the database, we then performed statistical analyses to summarize
the distribution of games by genre and content descriptors.
To quantitatively assess the content of games, we selected 55 E-rated
video games that represented the distribution of content descriptors and genres
and that were available for play on one of the current major home video game
consoles in the United States (DC, N64, PS, or PS2). We designed the study
to include several games on each console and to play a mixture of both the
highly popular games as well as ones that did not receive widespread consumer
To explore the possibility of trends in series of video games, we also
selected 2 of the most popular series by sales for study: The Legend of Zelda series in the adventure genre and the Super Mario Bros. series in the action genre. We played all of the
games in these series, including games released for older consoles like Nintendo
Entertainment System and Super Nintendo Entertainment System. Since the 2
oldest games in The Legend of Zelda series were released
prior to the creation of the ESRB and have not been rated, we did not include
them in our analysis of E-rated games even though we are confident that these
games would receive E ratings. Finally, for historical comparison, we assessed
the content of 8 classic arcade games that have been rereleased as E-rated
compilations or paired with E-rated remakes of the original games. Overall,
we played a total of 65 games.
For consistency, an undergraduate student with considerable video gaming
experience played all of the games and recorded all game play directly onto
videocassettes for later coding. The student played each game to its conclusion
or for at least 90 minutes, whichever occurred first. Some action and adventure
games that allow the player to save game progress are designed for very long
play times; consequently, not playing these games to their conclusion means
that some content is missed. In particular, some games may become more difficult
as the player advances and they may offer additional weapons or other more
mature content. However, in our effort to strike a reasonable balance between
playing more games and playing individual games for longer times, we determined
that playing the game to its conclusion or for at least 90 minutes allowed
us to obtain a reasonably good sample of game play for any single game. Video
games often start with an introduction and setup, which the player may elect
to bypass. Consequently, we did not include introductions and game setup in
our coding or calculations of the duration of game play, although we did generally
observe them. For consistency, we defined the beginning of game play as the
first scene where autonomous movement occurred.
With the game play recorded on videocassettes for consistency, one author
(K.H.), who also has considerable video gaming experience, reviewed and coded
all of the games using a standard coding instrument (available on request)
and entered the data into a database constructed with Microsoft Access (version
2000, Microsoft Corp, Redmond, Wash). The first author and the game player
each independently coded a subset of 10 games to assess intercoder reliability.
We discussed all instances of games that presented difficulty in coding with
verification of game details from the undergraduate student who played the
games. We performed descriptive and statistical analyses using Microsoft Access
and Excel (version 2000, Microsoft Corp) to compare our sample to the universe
of E-rated games. We calculated intercoder reliability using the κ statistic
comparing the duration of violence coded by both the game player and the primary
coder (K.H.) in each 10-second interval for 30 minutes of game play (or until
the end of the game) for 10 games.19
We defined violence as acts in which the aggressor
causes or attempts to cause physical injury or death to another character.
We did not include accidental actions that led to unintentional physical harm,
the effects of natural disasters, or the presence of dangerous obstacles that
could not be attributed to the actions of a particular character. A violent incident was defined as an uninterrupted display
of violence by a character or a group of characters. We defined characters broadly, including personified objects that attacked either
the player or other characters. We did not code as violence any intentional
acts of physical force that represent normal play in a sports game (eg, tackling
in football or checking in hockey), because the intention of the player is
technically to stop the other player without causing injury. We did code excessive
physical contact in sports games, such as punching or otherwise attacking
another player (eg, after the football play was over). To quantify the amount
of violence, we manually recorded the starting and ending times of each incident
of violence toward other characters (hours, minutes, and seconds from the
beginning of the tape).
In video games, characters often engage in a series of violent acts
that are punctuated by brief periods of time spent running toward the next
encounter. For consistency, we established a rule that a series of violent
acts would be coded as 1 violent incident only if individual acts of violence
were separated by fewer than 10 seconds of nonviolent behavior. For each violent
incident, we recorded the type of weapons used for violence, whether the violent
incident resulted in injury, and the number of character deaths attributable
to the violent incident. In addition, for each game, we noted whether injuring
characters or destroying objects is rewarded or is necessary to advance in
the game, whether the player could select weapons, and whether any of the
following content was present: alcohol, tobacco, illegal drugs, profanity,
and sex. Finally, we also looked for the presence of music from explicit-content–labeled
recordings, because a recent Federal Trade Commission report found that 2
music companies had approved the use of music with parental advisory labels
in E-rated video games.10
Our analysis of the universe of E-rated video games led to a database
of 672 games with 99% of these games available for play on at least 1 of the
major home consoles in our study (DC, N64, PS, or PS2). Of these 672 games,
our sample included 55 games (8.2%). Half (28 of 55 games) appeared on the
monthly list that ranks the 25 best-selling games in the United States by
units sold (regardless of ratings).20 In the
context of coding, we found good agreement between the author who coded all
of the games and the student who played them (κ = 0.90).
Table 1 shows the comparison
of the 58 content descriptors for the 55 nonarcade games played in our sample
vs the 710 content descriptors for the universe of 672 E-rated games. Overall,
the sample has a similar distribution to the universe of content descriptors.
Our sample does not include any of the 20 games that received content ratings
for animated blood, realistic violence, violence, mild violence, suggestive
themes, informational, strong language, suitable for all users, or edutainment.
Based on analysis of the 672 E-rated video games released for home consoles,
384 games (57%) did not receive any content descriptors.
Table 2 shows the comparison
of genres for the 55 games in our sample and the universe of E-rated games.
Based on analysis of the 672 E-rated video games released for home consoles,
games in the sports (28%), racing (26%), and action (23%) genres account for
most of the games.3 Again, our sample has a
similar distribution to the universe of E-rated games, although our effort
to explore trends in 2 series of video games contributed to our oversampling
of games from the action and adventure genres.
Table 3a summarizes the content
of violence in our sample of video games. We report the percentage of game
play depicting violence because comparisons of the absolute durations of violence
for the games would be meaningless given the differences in game play durations
that resulted from the student completing some games relatively quickly and
requiring at least 90 minutes (mean game play, 57 minutes; range, 6.6-136
minutes) for other games. In our sample of 55 games played, we found 20 games
that did not include violent game play, and 35 games (64%) that involved intentional
violence, with an average of 30.7% of the game duration representing violent
game play for these games (range, 1.5%-91.2%). We found that injuring characters
was rewarded or required for advancement in 33 (60%) games. Separating the
55 games into 2 groups, 1 group containing 23 games that received a content
descriptor for violence and 1 group containing 32 games that did not receive
a content descriptor for violence, we found that the games with a violence
descriptor contained significantly more violence (at a 5% significance level
based on a 2-sided Wilcoxon rank-sum test, t53= 2.59). Remarkably, we also found that 14 of the 32 games (44%) that
received no content descriptors contained acts of violence an average of 37%
of the game duration (range, 3%-88%). All of the games we played in the action
(n = 22), adventure (n = 3), fighting (n = 2), shooting (n = 1), strategy
(n = 1), and simulation (n = 1) genres included violence, while only 2 of
the sports games (17%) included violence not associated with normal game play.
Given the relatively small sample size, however, we caution against overgeneralization
of these particular results.
In the sample of 55 games played, 27 games (49%) depicted deaths from
violence. Not surprisingly, the shooting game showed the highest numbers of
deaths per minute (23.8). In all 22 of the action games, we found that injuring
characters was rewarded or was required to advance in the game. Nearly all
of the action games (21/22 [95%]) depicted deaths from violence, with an average
(arithmetic mean) of 2.3 deaths per minute (range, 0 deaths per minute for Paperboy to 8.4 deaths per minute for Rat Attack).
We observed that each successive game within The Legend
of Zelda series had progressively less violence (Figure 1) and fewer deaths per minute; a less clear trend was demonstrated
for the Super Mario Bros. series. One explanation
that is consistent with our experience is that successive games in series
may tend to involve more complexity in character development and engage the
player in more exploration and discovery activities that will help him/her
achieve a goal. However, this trend of less violence may be offset by the
tendency for successive games to portray violence more graphically and more
realistically as technology advances. The limited evidence of these 2 series
should not be overgeneralized.
Although damage to objects was not coded as violence in our analysis,
we found that games rewarded characters for destroying objects or required
object destruction for advancement in 29 of 55 games (53%). Table 3 indicates that 32 of 55 games (58%) in our sample depicted
weapons other than the body and that the player could select weapons in 16
(50%) of these 32 games. A total of 30 of 55 games (55%) used the body as
a weapon, 27 games (49%) used projectiles, 16 games (29%) used magic, 13 games
(24%) used guns, 6 games (11%) used a knife or sword, 2 games (4%) used toxic
substances (poisons), 17 games (33%) used explosives, and 26 games (47%) used
other weapons (eg, fire, hammers, snowboard). This is not an exhaustive list
of the weapons that might be encountered in the games because of the limited
amount of time that each game was played; consequently, it should be viewed
as a subset of the weapons depicted in these games.
In addition to coding for violence, we also noted other content in the
games that might have led the ESRB to assign content descriptors to the game.
For example, Goemon's Great Adventure and NFL Blitz 2000 received ESRB content descriptors for "mild language."
We found the word "damn" printed on the screen in Goemon's
Great Adventure and noted that the players taunt each other in NFL Blitz 2000. Although none of the games received a content
descriptor for "suggestive themes," we noted the provocative leather outfit
worn by Ai Fukami in Ridge Racer V, the screen shot
between her thighs, and the phrases "curb your desire" and "push it to the
limit" in the introduction.21 We also noted
sexual innuendo in Gex 3: Deep Cover Gecko. Finally,
in Harvest Moon 64, which received a content rating
"use of tobacco and alcohol," the player can choose to purchase and consume
beer, wine, or liquor resulting in a red face and a fall to the floor. We
did not find any depiction of tobacco in our play of that game or find any
music with parental advisories in any of the games played in our sample.
The first public outcry over violence in video games occurred in 1976,
when Exidy Games withdrew from the market Death Race 2000, a game that awarded players points for running over stick figures.12,13 In the 1980s, the US government began
using video games for military training purposes, and recently Grossman and
DeGaetano22 publicized the use of off-the-shelf
video games in military battle training. Controversy and concern about the
effects of video games on children continue,4- 10
although much remains to be learned.
With all of the questions about the impact of violence in video games
on children, this is the first study to our knowledge to quantify the amount
of violence in E-rated video games and to show that many E-rated games do
involve violence, killing, and the use of weapons in the course of normal
play. No games provide messages about not using violence, and some games reward
or require violence and the destruction of objects.
The video game genre, ESRB rating, and ESRB content descriptors provide
important information about the content of the game, and overall illustrate
the considerable variability that exists in the universe of E-rated games.
One implication of this finding is that studies that assess video game content
with a mix of games of different ratings and genres might produce very different
results than studies that focus on a single rating and genre; future researchers
will need to carefully consider the process of selecting the games for their
samples. Clearly, efforts to standardize definitions for genres would be both
challenging and helpful.
The content descriptors appear to provide limited information about
violence. We found that receiving any content descriptor for violence (animated
violence, mild animated violence, etc) provided a good indication of violence
in the game, but the absence of a descriptor did not mean violence-free. The
definition for the E rating states that the game "may contain minimal violence,"
yet our experience shows that many E-games contain a significant amount of
violence and demonstrates ambiguity in what constitutes "minimal violence."
We did not see how the ESRB distinguishes between different content descriptors
for violence and we believe that efforts to standardize the definitions of
content descriptors would be helpful. Another approach to consider would be
to have content descriptors that provide information about the amount of violence
using a scale instead of noting simply its presence. This might help consumers
distinguish among games that receive the same descriptors but contain very
different amounts of violence (eg, Nuclear Strike 64
vs 40 Winks or Rat Attack
vs The Smurfs). We also noted some inconsistencies
between games that received a content descriptor and games that did not.
Currently, the ESRB rates games based on information and excerpts submitted
by the game manufacturer, but does not play the game before assigning the
rating. While giving the same materials to raters may promote consistency,
our experience playing the games leads us to believe that the ESRB raters
should play the finished game, including the introduction, before assigning
Remarkably, we found some nearly identical games that received different
ratings on different consoles (eg, Nuclear Strike 64
and Gex 3: Deep Cover Gecko are E-rated on N64 but
rated T for "Teen" on PS), which may make game selection more confusing. We
believe that the ESRB should avoid assigning different ratings of the same
game on different platforms and should assign the highest rating to all of
the games of the same title to avoid inadvertently misleading consumers who
may not appreciate the differences between platforms.
A few important limitations exist in this study. First, the sample of
games represents only a small subset of the available E-rated games. Second,
the results depend on the actual game play that we recorded and the methods
we used for coding information, which include subjective judgment in the definitions
and their implementation. While our approach of having a single person play
all of the games and another independent person code all of the games reduces
variability in both the game play and the coding, it may have limited the
representativeness of the game play for different types and skill levels of
players and accuracy in the coding. However, our method of recording the game
play on videocassette and recording actual screen times for incidents provided
opportunity for validation of the coding, which showed excellent reliability
between coders. Third, our use of a broad definition of violence focused on
the intention of the character may differ from other similarly legitimate
definitions. For example, the sports and racing games, essentially all of
which receive an E rating, provided the greatest challenges to coding because
they contain intended acts of physical contact like checking in hockey and
tackling in football that are not intended to cause injury, although other
studies or coders might deem these acts to be "violent."
Despite these limitations, this study provides important and useful
information to physicians and parents about the content of E-rated games.
Parents should be aware of games' ratings, content descriptors, and genres,
and parents whose children play games should actively participate in game
selection and engage their children in discussion of the game content. Several
Internet sites also provide helpful information for parents who want to better
understand the content of video games. In addition to the ESRB Web site, the
National Institute on Media and the Family's KidScore media evaluation system
offers information to parents about many types of content in video games.23 We compared the 24 games in our sample that also
appear in the KidScore database and noted a few differences. In particular,
2 games (Crash Team Racing and Kirby 64: The Crystal Shards) that received "green lights" in the KidScore
database contained significant amounts of violence in our sample. Also, we
noted that the KidScore database did not note any sexual content or nudity
in Ridge Racer V. Collaborative efforts among industry,
parents, physicians, advocacy groups, and the research community to standardize
media rating systems and to develop a universal rating system should be pursued.24
Our content analysis suggests that many E-rated video games contain
a significant amount of violence and that an "E" rating does not automatically
signify a level of violence acceptable for very young game players. Physicians
and parents should understand that popular E-rated video games may be a source
of exposure to violence for children that rewards them for violent actions
and that they may contain other content that is not expected given the E rating.
We believe that physicians, particularly pediatricians, should consider asking
patients about their experience with video games and the medical and public
health communities should play an active role in informing parents about the
content in video games.