Sponsored By

What causes addiction in massively-multiplayer online games? Is it team play? Is it socialization? Perhaps it's the so-called "grinding" towards an end goal? Today's Gamasutra feature explores this by polling real MMO players, with fascinating results.

Neils Clark

August 22, 2006

13 Min Read


I recently completed a study at the University of Hawai‘i at Manoa finding that gamers’ activities and preferences within games could be linked to addiction. The study looked at players within massively multiplayer online games, for instance Blizzard’s World of Warcraft or Square Enix’s Final Fantasy XI.

The study, which examined in-game behaviors on a number of levels, found that playing with real life friends, side activities like exploration or taking pictures, and membership to social guilds may be related to less harmful play. On the other hand, stealing from or otherwise manipulating players, along with membership to more goal-oriented “hardcore” raid guilds may be related to addiction. Player versus player activity was related to both less damaging and addictive behaviors on different levels of data analysis. While these relationships are present, it is not known whether games are actually the cause of this behavior, or if these are simply behaviors that already addicted players seek out.

Based on these relationships, the study suggested that the process of becoming addicted is likely very complex. It is suggested that more research connecting in-game activities and addiction is needed, and that determining whether games are to blame will most likely require a longitudinal study, looking at a group of gamers over a length of time. It is also noted that studying this link between gameplay and addiction could open doors for addicts generally. From the study:

“There are a million little pieces working together in these games. Understanding this process not only holds the potential for helping the people with real gaming problems. Research within prototypical game worlds may have real implications for helping people with other kinds of non-game addictions.”

There were also a number of suggestions on improving the ways in which future studies of videogames collect their data. After making the attempt at using one new method for getting data on gamers, the study made a number of observations on the technical and ethical limitations for data collection in videogame studies.

It is also noted that despite a lack of good information on what “addiction” could actually mean, gamers and advertisers use the term far too much. Popularizing the word “addiction” as a mark of quality hurts not only players, but also sales. Players already widely confuse good games with harmful overuse. From chapter one of the study:

“The sustained misuse of perceptions and stereotypes on the part of marketers will likely have an increasingly devastating impact on game players that do have problems. Understanding the intersect between addiction and videogames is a necessary precursor toward first, understanding what to regard as addiction, and second, search for clues as to how we might begin to help the people whose play is having a clearly negative affect on their lives.”

What follows are select extracts from this thesis. At the end of this article, you may download the thesis in its entirety, in PDF format, which also includes details of the works cited in the following pages.

Player Populations and Structural Characteristics of MMO games

While we may not know a great deal about addiction as it relates to these games,
there has nonetheless been a great deal of work exploring these worlds. The methods of these studies may not be perfect, yet they nonetheless give strong hints as to who is playing and why. Even if many demographic variables have yet to be sampled scientifically, what exists can be compared against information on addiction in order to garner information regarding these worlds.

Figure 2. Blizzard Entertainment’s World of Warcraft

Each MMO, for instance EverQuest or World of Warcraft, is a different game, with
different mechanics. Within games there can be large differences as well, as each game will, on average run one, 20, or even 100+ servers, depending on the popularity. Blizzard’s World of Warcraft, currently the real world’s largest MMO, runs at least 108 servers, which they call “realms.” This number increases as World of Warcraft gains popularity, yet may collapse if newer games grow in favor. A server is a single copy of the game world, with its own internal mechanics; likely variations between servers are community, economy, or nationality.

To be more specific, after purchasing an MMO and creating a billing account, one will have to choose which particular “server” of this game world they would like to play in. The bulk of player population studies do not explicitly explain how they treat sampling with regard to servers or games, though this is not the major downfall to such studies so far. A major sampling downfall is that the bulk of studies looking at player populations have been elicited through outside websites, particularly “community websites” for particular games. Two major concerns with this are (1) the sheer volume of such popular community websites, and (2) the low likelihood that those seeking community outside the game will be representative of those playing within (especially those pathologically addicted).

Structural Characteristics as a Theoretical Concept

Though literature pertaining to videogames is growing rapidly, few have stopped to
actually define videogames or analyze their structure (Wood, Griffiths, et al., 2004). In a
study of more traditional single-player games, Wood et al. set out to assess video game
structure by way of asking which structural characteristics were most salient among game players (Wood et al., 2004). Many of his structural characteristics were not entirely applicable to MMO play, for instance ‘duration of game’ and ‘mapping’ (creating custom levels for a single player game). Other characteristics apply to MMO games, such as interface options, use of humor, and brand assurance, yet were omitted from this study in order to restrict its size.

Figure 3. Structural characteristics map the structures available for interaction within
these games, and include exploring new areas.

Yee’s MMO Motivations

Nicholas Yee, one of the more recognizable names in the study of MMO effects, in a
3 year period surveyed over 30,000 players from the MMO games Ultima Online, EverQuest, Dark Age of Camelot, and Star Wars Galaxies. In a paper set to be published in 2006, he explores positive and negative effects of MMO playing. Qualitatively, Yee identified five factors of MMO motivation: achievement, relationship, immersion, escapism, and manipulation.

It is important to differentiate motivations for play from structural characteristics, the
actual structure within the game that is being manipulated. This distinction notwithstanding, the criteria used by Yee can be modified and clarified so that these motivations may be measured as if they were structural characteristics. For instance, Yee’s concept of manipulation includes the questions, “I scam other people out of their money or equipment”, and, “I like to taunt or annoy other players.” His questions do not have to be changed drastically for a successful query of MMO structures.

The wording of some measures taken from Yee remain unchanged. In “immersion”,
for example, creating a back-story for one’s character can be entirely in the mind of the game player, or it can be an integral part of the game. Eve Online and City of Heroes are two games where players are encouraged to enter information about their character’s history, in a structure that allows other players to then read that information and/or back-story. World of Warcraft has no structure for this, but it is not uncommon to hear a player talking about their created back-story on an RP, or role-playing server. The same structure is here operating in radically different ways, sometimes even within a single game.

Demographics of MMO game players

What we know about gaming comes primarily from Yee’s studies, particularly his
Daedalus Project. While nearly all of the information collected by Yee came by way of selfselected respondents seeking out his surveys, the many thousands of people attracted for his studies have made his data likely the most widely cited in academic work on game player demographics.

In terms of employment, 50.0% of respondents were shown to work full time, 22.2% were full-time students, and 13% of female players referred to themselves as “homemaker.” Additionally, the number of female MMO players seems to increase with age, surpassing the number of males in the 23-28 age range, and in each subsequent age range (Yee, 2006). Yee argues that this data dispels the notion that all gamer players are unemployed, male, and young; rather games have a universal appeal.

Other findings from Yee are that 60.9% of respondents had played for at least 10
contiguous hours, this effect being roughly equivalent along age groups. 15.8% of men and 59.8% of women play MMOs with a romantic partner, while 25.5% of men and 39.5% of women play with a family member, suggesting that women are primarily being introduced to MMO games by a spouse or family member. Most importantly, Yee points out, “…the data demonstrate that MMORPGs appeal to a very wide demographic and that this appeal is strong and elicits high time investment from users.” (Yee, 2006).

Comparing the Social Networks within MMOs with those of the Mafia

“…the mafia initially grew out of an ancient honor system where elders were entrusted to negotiate in conflicts and pass judgments that the others were obliged to adhere to. The fact that Sicily historically has been targeted by outside interests such as the Spanish and fascists has also contributed to a need for organized resistance against outside oppression. The transition into a criminal organization came later, possibly more or less because the mafia realized that they could use their powerful organization to achieve fortune for themselves. This pattern is repeated in EQ [EverQuest]. The strong emphasis on reputation in the creation of social networks grows out of a need from the players to self-govern their gaming environment in order to secure a positive experience in the presence of potential disturbances and a simultaneous absence of an effective and reliable governing system. But ultimately these networks are also used to take shortcuts through, or trick, the formal rules of the system.” (Jakobsson & Taylor, 2003)

Comparing EverQuest’s prominent framework for social structure, the guild, to the
social structure of the mafia should seem laughable at first. The above quote however
illustrates and contextualizes ways in which play style shifts strongly within MMO games, favoring these social networks as players approach the highest levels of in-game
achievement. Put another way, where a player’s guild, online friends, and real-life
connections at early stages of play provided the support required to succeed, at the end-game they become the connections that allow a player to dominate.

Figure 4. The members of both social and goal-oriented guilds will occasionally line
up for group photographs.

Here it begins to become apparent that Yee’s (2006) conceptions that deal with
interaction (how much a person talks, shares feelings, etc.) and perhaps also individualism, a person’s preference toward playing on their own, may need to be expanded. In the mafia, family provides a strong foundation for commitment. You stick with your family, and they stick with you. Jakobsson and Taylor are here arguing that MMO games work similarly, where people who know each other outside of the game have a much higher commitment to each other than to friends that they know strictly through the game.

The idea here was then to split up communication between these two types of connections; real-life friend, and strictly online friend (individuals or guildmates), and then to rework these criteria in order to better reflect levels of interaction. Individualism was expanded from Yee’s group/solo criteria in part due to this emphasis on interaction. If some players prefer playing with reallife friends, and others with online friends, then perhaps those that prefer no interaction at all differ in equally significant ways.

The Study

As part of Clark's thesis, a pool of 291 MMO players were given a 91-question survey. The following extract explains the tools used, and an accompanying image shows the survey's results.

The current research provides data which suggests that a player’s perceived use of varied structural characteristics within a game is related to differing addiction and engagement levels. In particular, the constructs of negative valence, side activities, and interaction with real life friends within MMO games had distinct relationships with each addiction and engagement levels.

When structural characteristics were controlled for, PvP advancement and guild preference appeared to be significant predictors of addiction or engagement. While the primary aim was to offer preliminary data comparing structural characteristics and addiction, this work also discusses the benefits and limitations of sampling respondents within MMO game worlds.

Table 1 - Game Element Reliability (Cronbach’s E) and Correlation (Pearson’s r) with Addiction and Engagement


Addiction Composite. Level of addiction was measured by adapting the Asheron’s
Call Addiction (ACAddiction) measure from Charlton & Danforth, 2004. This scale contains
14 items and showed good reliability in the current study with an alpha of .790. In this scale, participants respond to questions on a 5-point Likert-type scale on which 1 = very strongly disagree, and 5 = very strongly agree. A high score on the Addiction Composite represents high addiction, and a considerable inability to control play.

Engagement Composite. Level of engagement was measured by adapting the
Asheron’s Call Engagement (ACEngagement) measure from Charlton & Danforth, 2004.
This scale contains 15 items and shows good reliability in this study with an alpha of .845. In this scale, participants respond to questions on a 5-point Likert-type scale on which 1 = very strongly disagree, and 5 = very strongly agree. A high score on the Addiction Composite represents high addiction, and a considerable inability to control play.

Videogame Structural Characteristics. These characteristics came from a variety of
diverse sources, as covered in the concepts section. These scales vary in size, and generally showed acceptable reliability, with alpha scores ranging from .723 to .906. The exact reliabilities for each measure are provided as part of Table 1. In these scales, which are part of the questionnaire provided in Appendix B, participants respond to questions on 5-point Likert-type scales which are discussed at length in the concepts section.

Abstract of the Thesis

This work also provides a comprehensive review of videogame-related
literature in fields as diverse as communications, psychology, sociology, and neuroscience; this reading suggested that behaviors noted as detrimental in psychology and neuroscience were being observed in studies of Massively Multiplayer Online (MMO) gamers.

For further reading, you may download the entirety of "Addiction and the Structural Characteristics of Massively Multiplayer Online Games" (*.pdf format, 660kb).

Read more about:


About the Author(s)

Neils Clark


Neils Clark is co-author of Game Addiction: The Experience and the Effects. He is a professor at the DigiPen Institute of Technology, in Redmond.

Daily news, dev blogs, and stories from Game Developer straight to your inbox

You May Also Like