Abstract: We use survey data—interpreted through ethnographic interviews and our own game-playing experiences—to model the way culture impacts the therapeutic dynamics of play in the popular online game World of Warcraft (WoW). To do so, we utilize cognitive anthropological understandings of ‘cultural consonance’ (Dressler and Bindon 2000)—that is, the extent to which individuals embody or fail to embody socially shared and sanctioned models of success. We find that players who report more individual ‘consonance’ with culturally shared models of ‘real-life’ or offline success are more likely to play in healthier ways as assessed through players’ self-reports of the impact of WoW on their life happiness, stress relief, and patterns of problematic play. We uncover both direct relationships between an individual’s relative degree of cultural consonance and these wellness outcomes and also indirect ones mediated by ‘absorption-immersion’ (defined as the extent that players feel like they are in a virtual world and in some cases actually their character). Overall, we suggest that WoW—and more generally multiplayer online role-playing games (‘MMORPGs’ or ‘MMOs’ for short) of which WoW is one example—can be thought of as cultural-cognitive technologies promoting a partitioned or ‘dissociated’ consciousness (Lynn 2005) in which players can attribute dimensions of self to in-game characters for potential psychological benefit or harm.
We examine how experiences in the World of Warcraft—a popular online videogame that facilitates deeply ‘absorptive,’ ‘immersive,’ and even ‘dissociative’ experiences (Bartle 2003; Lynn 2005; Yee 2006, 2007)—alternately promotes and compromises players’ subjective experience of wellness. We hypothesize that players who report greater success in actual-world contexts—that is, players who see themselves to be more culturally ‘consonant’ with shared models of offline or so-called ‘real-life’ success (Dressler and Bindon 2000)—are more likely to play WoW in a healthy rather than problematic manner. By contrast, those with low real-life consonance search in WoW for the success they feel they lack in real-life. Such individuals are more vulnerable to lose themselves too deeply in WoW—experiencing what one scholar has called ‘toxic immersion’ (Castronova 2005, 2007)—as they pursue WoW virtual success to compensate for their perceived real-life failings. We test these ideas through an online survey, interpreting the results through ethnographic interviews and our own game-playing experiences.
1.2. Setting: The Persistent and Immersive World of Warcraft In 2008, there were approximately 11.5 million monthly subscribers to WoW, making it the largest subscription-based massively multiplayer online role-playing game (‘MMORPG’ or ‘MMO’ for short) and virtual community in the West, approximately 62% of this part of the world’s MMO population at this time (Blizzard Entertainment 2008).
Central to new online realities like WoW is their quality of persistence: these worlds provide virtual places where thousands of users interact in a world that persists independently of any particular player. Typically, they persist 24 hours a day, with only brief interruptions for maintenance. Any individual player may have logged off his or her computer and thus disappeared from the game-world. Nevertheless, events continue to happen in the world, and players continue to compete and interact in ways that advance and change the contours of the game-space. Of equal importance, these massively multiple play spaces are highly immersive. In part, sophisticated software and computers with powerful 3D graphical processing create the absorptive spaces that feel virtually real. The way one’s avatar or visual representation of a player’s character-self responds to commands adds to this sensation, as do the mentally and emotionally absorbing quests and plot-lines. It is easy to become immersed, to feel as if one actually inhabits another space, no longer aware of the so-called ‘real’ world that surrounds one’s physical body and computer play station. Indeed, in such circumstances, one’s mind or consciousness, if not one’s real-world body, can be said to reside in that other virtual space (Castronova 2005).
World of Warcraft forms a fantastical virtual world that stretches over numerous continents and planets. WoW’s designers have taken great care to render beautifully this virtual reality’s landscapes—its craggy mountains and grassy plains, lakes and oceans, and wastelands. Equal attention has been given to WoW’s social settings, for example, its castles, fortresses, dungeons, battlegrounds, villages, and cities. In both natural and social landscapes, players encounter a range of denizens, other player-avatars animated by real-life persons as well as myriad computer-programmed entities, both menacing like monsters and demons (referred to as ‘mobs’ for ‘mobiles’) and also friendly and helpful such as innkeepers, auctioneers, and quest givers. Perhaps the most important and engaging aspects of WoW relate to character creation and advancement. Each player creates characters, who themselves are represented as avatars, their visual representations in the game-world. Avatars are manipulated through a variety of keyboard and mouse commands, and they can communicate through gestures (like dancing, waving, pointing, jumping, and flirting) that complement chatting through in-game text channels or headsets and voice programs (either built into the game or through third-party modifications like Ventrilo).
WoW offers gamers a seemingly endless choice of tasks, which can increase in complexity, time necessary to complete, and rewards as one advances in the game. Many of these are offered in the forms of quests with specific goals offered by computer-controlled non-player characters (‘NPC’s’). Players can accept these quests, travel to where the tasks can be accomplished, complete them, and then return to the quest NPC for rewards of treasure and experience points. In completing quests, exploring the world, and defeating mobs, players advance in experience and actual levels and amass treasure. Each level acquired, like won gear in the form of swords, armor, and jewelry, bestows additional power and ability on a given character, allowing them to complete more difficult game challenges, which in turn allow them to advance even further in the game. Even after completing the game’s highest level, currently level 85, many players decide to compete in highly challenging in-game content such as multiplayer instances like dungeons or raids requiring hours of cooperation between 5-40 players with groups balanced between different character classes.
1.3. Literature Review: The Cultural Therapeutics of ‘Absorption-Immersion’
Following anthropological scholars, we characterize mental ‘absorption’ in these gaming contexts as a profound narrowing or concentration of attention, cognitive resources, and sensory experience (Luhrmann 2005; Luhrmann et al. 2010). ‘Immersion,’ in our terms, emerges from an initial focused ‘absorptive’ narrowing and subsequent broadening of such attention, cognition, and sensory experience, which allows gamers to be so concentrated in rich game-world that they imaginatively ‘lose themselves’ (Bartle 2003; Yee 2006, 2007). Certain players immerse so fully that they feel like they really are in the game and sometimes actually their avatar-character. For these players, in-game events and identities can feel as vivid and real as the so-called ‘real’ or ‘actual’ world. We refer to such extreme states of imaginative ‘absorption-immersion,’ in which actual-world physics and psychological processes like perception, memory, and identity feel greatly distorted, as ‘dissociative’ (Krippner 1997; Luhrmann 2005; Luhrmann et al. 2010; Lynn 2005; Seligman and Kirmayer 2008). As one scholar puts it: “‘Dissociative’ is an English-language adjective that attempts to describe reported experiences and observed behaviors that seem to exist apart from, or appear to have been disconnected from, the mainstream, or flow of one’s conscious awareness, behavioral repertoire, and/or self-identity. ‘Dissociation’ is a noun used to describe a person’s involvement in these reported dissociative experiences or observed dissociative behaviors (Krippner 1997: 8).”
Further, based on previous research, we anticipate that the extent to which gamers imaginatively absorb, immerse, and even dissociate into WoW determines in part both this game-world’s therapeutic and distressful dimensions (Snodgrass et al. 2011a). Positively, scholars suggest that many everyday ‘absorptive’ activities—losing oneself in a good book or film, communing with nature, daydreaming and reverie, fantasy play, running and also team sports, driving, yoga and meditation—can lead to lightly altered, ‘dissociative’ experiences in which the dominant stream of one’s consciousness feels distorted and fragmented (Butler 2006; Luhrmann 2005; Luhrmann et al. 2010). These activities contribute to the textures and pleasures of daily life in Western and non-Western contexts alike—in some cases tied to deeply pleasurable ‘flow’ states of consciousness described by Csikszentmihalyi (2008 )—and the dissociative experiences associated with them are often judged normal both by cultural insiders and scholars alike (Butler 2006; Seligman and Kirmayer 2008). Indeed, Lynn (2005) argues that such states of consciousness are so important that we should also theorize pathological under-dissociation, as when one does not dissociate enough, and thus misses out on the positive and even therapeutic experiences associated with such states. Negatively, and by contrast, the pathological character of some forms of deep dissociation is well recognized by clinicians and scholars. For example, Dissociative Identity Disorder (DID) is associated with extreme and dysfunctional discontinuities of experience. Such seeming over-detachment from the real-world, alongside distressful accompanying symptoms like amnesia, depersonalization, derealization, and also disruptions to interpersonal relations, has led clinicians and researchers to frame this condition as a mental disorder (APA 2000; Kihlstrom 2005; Ross 1996, 1997). This prior research suggests that gamers’ relative degrees of imaginative dissociation might help explain why some players find MMOs such as WoW so terribly fun and even therapeutic, while others, immersing too deeply and thus ‘toxically’ in this alternative virtual world, compromise their offline lives and even become ‘addicted’ to their play (Castronova 2005, 2007).
As social scientists, we also examine how the positive or negative dimensions of such gaming experiences and states of consciousness will vary according to sociocultural contexts. In the context of this article, we draw in particular on cognitive frames or models theory to understand such cultural contexts, rather than on the concept of offline social networks as we do in previous research (Snodgrass et al. 2011b). Psychologists, linguists, and philosophers have devoted considerable attention to the study of schemas (or schemata): simple cognitive elements or prototypes which help individuals organize and process information in relationship to their social and natural environments. These schemas, and also models or frames, which are understood to be more complex concatenations of schemas, help individuals understand the world around them and attribute meaning and significance to events and experiences (foundational studies of schemas and models include Johnson-Laird 1983; Lakoff 1987; Mandler 1984; Minsky 1975; Rosch 1975; Rumelhart 1980; Schank and Abelson 1977).
We are particularly influenced by cognitive anthropological studies of cultural models and frames, which are opposed to idiosyncratic or personal models. Cultural models are abstract and simplified mental representations of the world that are socially learned and widely shared within a group (see for example D’Andrade 1995; Holland and Quinn 1987; Hutchins 1995; Ross 2006; Shore 1996; Strauss and Quinn 1997; Shweder 1991, 2003). Of special interest for our research is a growing body of cognitive anthropological literature that uses cultural models or frames theory as a starting point to understand health-related processes. William Dressler and his collaborators (e.g., Dressler and Bindon 2000) explore how being in or out of sync with culturally normative models of success and idealized life-style can produce stress. Such stress can manifest itself in negative health outcomes such as high blood pressure or depression. Dressler’s research forms the touchstone of our own examination of WoW play. Specifically, we hypothesize that WoW players’ successful or failed attempts to incarnate cultural ideals related to success and achievement channel gamers’ behavior and experience in patterned ways, steering players alternately into healthy or unhealthy forms of play and immersion.
2. 1. Measures
The data and analysis presented in this article are based largely on responses to a web survey, which we used to assess players’ level of involvement with WoW. The survey included several psychocultural scales, three of which are of particular interest to our arguments in this paper (see Appendix A for actual scale items). In the first, we adapted two scales commonly used in psychological anthropology to assess absorptive and dissociative experiences—the Tellegen Absorption Scale and the Dissociative Experiences Scale (DES) (Tellegen and Atkinson 1974; Bernstein and Putnam 1986)—to create our own WoW-specific measure of ‘absorptive-immersion-dissociation.’ Here, gamers were asked to respond to WoW-specific questions related to the extent to which their play elicited distortions of perception, memory, and identity typical of absorption and dissociation in other contexts. For example, players were asked about their levels of imaginative identification with their characters, as well as the extent to which WoW play led them to become unaware of events happening around them in the real-world, leading them to, for example, ignore the demands and discomforts of their real-world bodies, lose track of actual-world time, and so forth.
In the second scale, we assessed the extent to which survey respondents were ‘consonant’ with culturally shared models of real-life success, that is, how they saw themselves as having achieved relative to that model. In constructing this scale, we followed a four-step process, which is standard within cognitive anthropology. First, we asked respondents to ‘free-list’ different terms associated with offline or ‘real-world’ success. Second, we asked individuals to rate in importance on a 7-point Likert scale the most salient or commonly recurring free-listed success items. Specifically, individuals were asked to rate the importance of these items according to the expectations and values of mainstream U.S. culture (rather than according to their own personal views).1 Third, to determine the level of cultural sharing of these normative models, we subjected responses to cultural consensus analysis (Romney et al. 1986, 1987; Ross 2006; Weller 2007). This method and suite of statistical routines allows one to quantify the extent to which knowledge across a series of statements and a set of respondents is shared. Sets of statements with high consensus are presumed to be, under the tenets of this theoretical model, potentially socially learned and thus cultural in nature.2 Fourth, and finally, individual success items were placed on our web survey, and we asked respondents to assess how much they themselves had accomplished or embodied each of them. Following routines outlined by Dressler et al (2005b), responses across the success items were summed, thus allowing us to gauge each respondent’s level of individual ‘consonance’ with the potentially culturally shared model of success.3
The third and final scale was a WoW Problematic Use measure based on Kimberly Young’s commonly used Internet Addiction Test (IAT) (Young 1998a, b, 2009). Items in this scale measure the extent that individuals reported playing compulsively in ways that negatively affected other dimensions of their lives, such as jobs and relationships. In addition, questions asked about experiences of compulsive play, cognitive preoccupation with the game, maladaptive use of the game to regulate mood, symptoms of withdrawal when unable to play, preferences of for game world over actual-world interactions, and play of excessive duration.
Further, the web survey also asked players to rate their perceived positive emotional experience relative to WoW play, such as how much they believed WoW play added to their happiness and alleviated stress in their lives. Basic demographic data (e.g., gender, education, employment and relationship status), degree of WoW usage and accomplishment within the game, motivation and styles of game-play, social interactions in the game, and numerous other topics were also included in the survey.
We asked 30 respondents to free-list items related to real-life success. These individuals were drawn from local gaming communities and centers as well as through our own WoW player guilds and play networks, followed by snowball sampling from these initial respondents. The real-life success rating task was presented to undergraduate students in an upper division anthropology course taught by this article’s first author. Justifying this second sample, we reasoned that these students were familiar with a general U.S. achievement model that might exert normative pressures on all U.S. inhabitants (including gamers), whether or not they themselves personally believed in each item’s importance or not. We thought eliciting responses from a non-gamer sample would also help us understand this model’s more general distribution in U.S. culture. Twenty-eight students performed this rating task.4
The larger web survey was posted on WoW blogs and gamer sites; it was also circulated opportunistically through our own guilds and networks, with instructions for respondents to pass the survey on to other players. An ongoing version of our web survey, similar to that on which our analysis is based, can be found at http://tinyurl.com/WoWwellness. The analysis presented in this paper is based on the 258 responses received at the time we began our analysis in Fall 2009.
2.3. Analysis and Interpretation
Cultural consensus analysis of our real-life success model was conducted with the software UCINET (Analytic Technologies). The core of our analysis and arguments rests on linear and ordinal regression models, which treat our cultural consonance and dissociation variables as major predictors and positive and negative MMO play experiences as outcomes, while adjusting for several demographic control variables. The controls presented in our regressions resulted from exploring various models with different demographic controls and picking for inclusion in further analysis only the three strongest predictors (as determined by hypothesis tests on the overall model), thus helping to avoid an excessive number of predictor variables in relation to the modest (n = 258) sample size. Finally, to facilitate interpreting results involving the consonance and dissociation scales, we standardized for all analyses, i.e, scores are in standard deviation units. Substantive predictor variables in our models were chosen on a theoretical basis, reflecting our ideas about how consonance and immersion should affect personal experience, with the control variables chosen on an empirical basis as described above.
Both the content of our survey and the interpretation of its results were guided by our qualitative phases of research and anthropological perspectives. We treated this game-world as a culture characterized by shared and socially inherited patterns of thought and practice—for example, group norms that shaped players’ goals, styles of play, speech, social networks of play, and normative assessments of whether they had played the game well or not. To grasp this culture, we engaged in seeing and experiencing this world as co-inhabitants of this online virtual or synthetic reality. Participants at times, we created characters and immersed ourselves fully in the game, losing ourselves in its tasks and its relationships, in its pleasures and absorptive fun. Observers at other moments, we stepped outside of our playing selves and self-reflexively noted and analyzed why we enjoyed the game, as well as how certain game-styles and dynamics contributed more or less to our enjoyment and overall positive and negative experiences in this online fantasy world.
We also conducted interviews of various forms, some face-to-face in the real world, others virtually through in-game interactions and chat programs, some relatively unstructured and open-ended, others with standardized protocols. These interviews were ‘semistructured’ (Bernard 2006) in the sense that we asked interviewees to expand and clarify their answers and we were free to pursue interesting leads not listed on our protocols. Interviewees were also non-randomly chosen through our own play networks. Thirty of these interviews were digitally recorded, transcribed, and coded for common themes by the same two individuals to ensure accuracy and consistency.
Both our participant-observation and interviews structure our discussion of our survey results—and extracts from these phases of inquiry feature in our ‘Discussion’ section—though readers will need to consult other work of ours for a more formal analysis of this qualitative data (see especially Snodgrass et al. 2011c, but also 2011a and 2011b).
3. Survey Results
There were 258 complete responses to the survey. The mean respondent was 26.6 years old (9.0 SD), male (78%), in a relationship at the time of the survey (54%), and a student (53%). In addition a little less than a third of respondents were unemployed (30%) and 36 (14%) of the participants did not classify themselves as white or Caucasian. There were 30 (12%) participants who played WoW forty or more hours per week.
Results from our web survey show that certain gamer respondents report detaching from their real-world bodies and environments and losing themselves in the virtual realities established by the game. For example, two-thirds of our sample reported at least sometimes losing themselves to the point of becoming unaware of events happening around them in the real-world and also losing track of the passage of time while playing. Twenty-three percent said they at least sometimes become so involved in WoW that game events feel actually real, as opposed to merely virtually so. Twenty-seven percent said that they sometimes, often, or always experience events in WoW as more vivid or memorable than those of their real lives. Twenty-two percent have felt themselves (sometimes or more often) to actually be their character, and similar percentages reported having other dissociative experiences contained in this scale, such as ignoring bodily needs while playing, vividly reliving past game events, and confusing one’s actual-world name with that of one’s character.
Survey results also reveal that WoW players report that the game promotes both positive and negative emotional experience. Positively, 45% of our respondents agreed or strongly agreed that WoW increases their happiness. Nearly two-thirds of the sample agreed or strongly agreed that WoW helps them to relax and combat stress. Negatively, high numbers (40-50%) of respondents reported that they experience at least sometimes each of the 19 items included in our multi-item WoW Problematic Use Scale (see Appendix A for these items and also for each scale’s Cronbach’s alpha, each of which was over 0.90, thus demonstrating excellent internal consistency).
In fact, results from our online survey show that these two processes—absorption-immersion-dissociation, on the one hand, and mental wellness, on the other—are associated. The first column in Table 1 shows the results of a linear regression analysis with the WoW Problematic Use scale as the response, and the dissociation scale (standardized) as a predictor. In that analysis, dissociation shows a substantial and significant relation to problematic use, with a one standard deviation change in the dissociation scale being associated with an 11.05 point (2/3 SD) increase in problematic use. The only control variable with a statistically reliable relation to problematic use was relationship status, with players in a relationship showing a 3.08 point lower score on problematic use.
A similar set of analyses using two positive self-rated emotional outcomes appears in Table 2. The models with just dissociation and the control variables as predictors (first and fourth columns) show that dissociation is relatively strongly and positively related to both outcomes. A one standard deviation increase in level of dissociation has an odds ratio of 1.88 in relation to the player’s rating of the extent to which s/he reports WoW as increasing happiness and an odds ratio of 1.56 with respect to feeling that WoW helped them to relax/combat stress. So, while dissociative/absorptive play is linked to more problematic use, it also was related to experiences of better emotional affect. (For more detail on dissociation in relation to such outcomes, see Snodgrass et al. 2011a.)
In a further set of analyses of these same outcomes, we examined the role of ‘cultural consonance’ in these processes—that is, the extent to which players assessed themselves as successful in the so-called ‘real’ world and thus ‘consonant’ with offline culture’s shared and socially transmitted model of achievement. Before proceeding to these regression models, we should note that our cultural consensus analyses from UCINET did indicate an adequate degree of cultural sharing amongst our respondents in regards to how they conceptualized success. This consensus analysis of the real-life success model generated a 1st to 2nd eigenvalue ratio of 3.50, an average competence score of 0.56, and no negative values. In the consensus framework, this showed substantial cultural sharing of knowledge and values related to this domain (see footnote 2). The fact that this model was culturally shared gave us confidence to include these items on our web survey as a potential way to measure the way respondents conceived themselves to be ‘consonant’ with this shared model.This real-life success scale also had high reliability with Cronbach’s α = 0.91.
The second and third columns of Table 1 show the results of considering cultural consonance as a predictor of Problematic Use. The model in the second column contains cultural consonance along with the control variables but excludes dissociation. The slope shows a predicted decrease of 6.05 points in Problematic Use (1/3 standard deviation) for a one standard deviation increase in cultural consonance: that is, persons who see themselves as successful in conventional society are less likely to report adverse consequences from their WoW play. We hypothesized that dissociation might function as a mediator variable between consonance and problematic online play and consequently estimated a model (column 3 of Table 1) with both dissociation and cultural consonance as predictors along with the control variables. In that model, the coefficient for cultural consonance is smaller in absolute size than in the previous model without dissociation as a covariate (-2.64 vs. -6.05). This is precisely as would be expected if there were an indirect or mediated effect, with consonance affecting dissociation, which in turn affects problematic use.
Further, we conducted a formal test of mediation, using a publicly available Stata add-on program. The program is called –khb- and is documented in Karlson and Anders 2011. These authors offer a new method that enables measuring mediation in both linear (as is the case here) as well as logistic and other nonlinear probability regression models (our subsequent analysis). To do this, they have addressed the difficulties that are caused by the fact that the coefficients in full and reduced models are not comparable when using logistic and similar methods, due to the non-independence of coefficients and error variability in these contexts. Their method solves this problem by re-scaling coefficients so that direct and indirect effects can be measured on the same scale. They are able to provide a valid forma test of mediation and also a meaningful numerical assessment of the relative sizes of direct and indirect effects. Using this approach, we find that the mediated or indirect effect of consonance is negative (-3.40) and 56% of the magnitude of the total effect, with p < 0.01.
Parallel analyses for the self-rated positive outcomes (increases happiness, helps relax/combat stress) appear in (respectively) the second and third and also the fifth and sixth columns of Table 2. A model with cultural consonance by itself (column two) indicates that consonance had a weak negative relation to reporting that WoW increases happiness (nonsignificant odds ratio of 0.90 for a one standard deviation increase in consonance). Controlling for dissociation (the model in the third column) resulted in a positive but again small (odds ratio 1.07) coefficient. However, the estimated mediation effect of consonance affecting perceived happiness through dissociation was negative and relatively strong (140% of the total effect, p < 0.01).
For ‘helps relax/combat stress,’ the pattern is different. In the model with cultural consonance as a predictor but without dissociation (column five), consonance shows a weak and nonsignificant positive relation (odds ratio of 1.17) to reporting stress reduction. But when dissociation is introduced as a control (column six), the relationship of consonance to stress reduction becomes stronger and statistically reliable (odds ratio 1.34, p < 0.05). The estimated mediated effect is negative and is 150% of the size of the total effect (p < 0.01).