"This SuXX0rz; I quit!!11!": An Examination of Player Exit Motivations in Massively Multiplayer Online Games (MMOGs)
by: UnSub
Overview
Date: 15 March 2004
Version: 0.991
Email: research@thebeholder.org
Massively multiplayer online games (MMOGs) are a rapidly growing form of personal entertainment. Investigation into why players exit the persistent or semi-persistent world these games offer indicates that players will most commonly leave due to boredom, in-game advancement issues or changes in social networks. It was found the Bartle player type and time spent in game had minimal impact on exit motivation. Players who rejoin MMOGs after exit tend to do so for social reasons, trial of new content or desire to play a MMOG. Special offers appear to somewhat encourage players to rejoin MMOGs they have previously exited from.
Introduction
Massively multiplayer online games (MMOGs) are a computer game genre that is rapidly attracting the attention both players and development companies around the world. Able to simultaneously link thousands of players together in a single play environment, MMOG growth in terms of both subscribers / players and number of available products has shown rapid growth across the past few years (Woodcock, 2003).
There are several reasons behind this growth. On the product developer side, the computer game industry typically has a short-term product focus; MMOGs allow companies the opportunity to produce a game with a longer shelf-life that can be improved (through the release of patches / updates) over time. Ongoing subscription fees paid by players offer potential long-term profitability that offsets the high initial cost of MMOG development.
On the player side, increasing adoption of the internet has allowed more computer game players access to this type of game. MMOGs allow players the thrill of competing and socialising with other players from around the world. There are generally no set end-points within MMOGs, meaning a player can experience the game world for as long as they want without reaching a point where they ‘win’ and the game experience ends. Further investigation into online world player motivation has been conducted by Yee (2003) and Bartle (1996).
Despite this rapid growth, to date little information is available on player exits from MMOGs. As with most industries in their birth / growth stages, it is likely the focus is more on attracting new participants / players than developing retention strategies or determining why players are leaving. Regardless, plenty of anecdotal evidence exists of players leaving MMOGs in large numbers for a variety of reasons. This research shall attempt to determine some of the reasons behind players leaving MMOGs as well as providing a starting point for further research into this area.
Two existing models will be used to investigate player exit motivations. The first is the voice, exit and loyalty model proposed by Hirschman (1972). This model shall be used to discuss the concept of exiting a MMOG and what effects this would have to both the player and MMOG in question. The second model is the player type model as proposed by Bartle (1996). Although developed for MUDs, the Bartle player type model is often used as a benchmark in existing MMOG research for player motivation assessment. It shall be used in an attempt to classify the playing style of respondents in order to gain further insight into their exit motivations. Some brief investigation will also be made into MMOG player behaviour in rejoining a MMOG they have previously exited from.
Definition of a MMOG
For the purpose of this study, a MMOG was defined as "any game that is played in an online graphical environment with greater than 64 other players, where every player has a persistent avatar that they can use across multiple sessions". An avatar was defined as a "representation of the player that is capable of development and improvement over time". (It should be noted that this definition has been changed slightly from the one that originally appeared on the online survey – this change was made after discussion with several respondents about weaknesses in the original definition.)
"Leaving" a MMOG is defined as ceasing active involvement with the MMOG with no intention of returning to play at a later date. This does not exclude the possibility that the player did return to play the game; what is important is that at the time of ending play the player did not intend to start again in the short-term. The cancelling of accounts would be considered "leaving" a MMOG, but is not essential – anecdotal reports exist of players "leaving" a game yet keeping their game account active in case they feel the urge to return and don't want to start from scratch.
These definitions were determined because while online games exist that allow mutiple players to participate within a single game, not all keep track of a player’s history or achievements with their avatar. For example, a player involved in CounterStrike starts every play session in exactly the same basic condition regardless of how well they played their last game but an Everquest player starts every play session in the condition they last left their avatar (assuming they are using the same avatar). It was decided that for a game to be ‘massively’ multiplayer, it needed to be able to support a large group of players (hundreds or even thousands) at the same time. Currently first person shooter games are being released that can support up to 64 players at once. It was thought that instead of the vagueness involved with ‘hundreds’ of players, a game that could support greater than 64 players on a near permanent basis would meet current perceptions of MMOG player capabilities. Finally, MMOGs are very closely associated with a highly graphical environment, usually one that appears to be or is three-dimensional. Online games that permanently track a player’s accomplishments but do so with a minimal graphical interface are more likely to be considered multi-user dungeons (MUDs) and are perceived to be an older genre of game that MMOGs have superseded.
As within any genre, MMOGs come in a variety of types. Below is a brief overview of 1) different game categories within the MMOG genre, 2) similar world settings within the MMOG genre and 3) some discussion on the confusion between MMOGs and MOGs (multiplayer online games).
MMOG Sub-genres
Within the MMOG genre there are several different sub-genres. For those unaware of the differences between these genres a brief overview is provided below.
MMORPGs (massively multiplayer online role playing games) see players take control of a single avatar, improving that avatar’s abilities by gaining experience through activity to spend on skills, attributes and / or levels. In-game death can have a moderate to severe impact on an avatar, with the player often losing experience, items and the time required to return to their pre-death state. This type of game sees player skill as having less impact on game play than the abilities of the avatar – in the majority of MMORPGs an incompetent player with a highly experienced character will defeat a skilled player with a low-experience character. The majority of MMOGs available today are MMORPGs with Everquest and Earth and Beyond being examples of this sub-genre.
MMOFPSs (massively multiplayer online first person shooters) see players control a single avatar in a first-person perspective within a world representation of three-dimensional space (Bell, 2003). Typically these games have a highly dynamic play style that sees player fight directly against player in a ‘kill-or-be-killed’ environment. In-game death typically has little long-term consequence. Player skill is very important to MMOFPSs – although some benefit may be given to avatars that have been in-game for a long time (eg higher rank, access to different weapons / armour / equipment) it is the player’s skill that sees them survive against competition. Examples of this fairly small sub-genre include Planetside and World War 2 Online.
MMORTSs (massively multiplayer real-time strategy games) see players control multiple specialised units, typically from a top-down or isometric viewpoint. This sub-genre sees unit management as the key to success, with resource management (ie the collection of certain resources for activities such as unit production) also potentially being very important. Game play usually sees players needing to use their units tactically to defeat other players, meaning that player skill and planning is important to achieving in-game goals. Unit death often means little overall, but a player that loses all their units can face in-game penalties. This sub-genre is very small but has a devoted audience; Shattered Galaxy and the as-yet unreleased Ballerium being examples of MMORTSs.
MMOTBGs (massively multiplayer turn-based games) see players involved in something similar to a MMORPG or MMORTS, but with inputs limited to orders that are executed in ‘turns’. A player must input all the commands to their avatar(s) prior to a set point of time that can be weekly, monthly or even hourly (depending on the game rules). Once that point of time is reached, a turn is executed, where all avatars follow the instructions set for them by their players. If a player has entered no inputs, the avatar(s) will remain idle for that turn or go to a default action that has been previously set for such an event. Players are unable to influence their avatars in between turns. MMOTBGs are arguably a throwback to play-by-mail games or to the time before high-speed internet access was available, often having minimal graphics and an inflexible interface. However, some players prefer this type of game since MMOTBGs often have victory conditions (meaning the game is ‘winnable’) and can be less time-intensive to play. Examples of MMOTBGs include Evernight.
The sub-genres above are reasonably defined, but some MMOGs see their game play cross sub-genre boundaries. Other sub-genres may develop in time, but the above four are believed to cover the majority of the current market.
World Settings
In developing MMOGs, game designers have typically drawn from broad generic world settings to provide the backdrop for game play. The most popular MMOG setting is fantasy, with science fiction settings coming in a distant second and realistic / historical settings occasionally raising their head. Non-fantasy MMOGs have not performed as well financially as fantasy MMOGs (such as Everquest and Dark Ages of Camelot in North America or Lineage in Korea; all are MMORPGs), meaning that some development companies may have determined that non-fantasy games are less profitable. Woodcock (2003) has suggested that the MMOG fantasy market may have peaked in actuality and may just be consuming itself, but it will take time to see if the fantasy setting’s dominance is actually being eroded.
Massively Multiplayer Online Game Versus Multiplayer Online Game
Based on some of the responses received during the data collection of this study, it is apparent that confusion exists even amongst players over the difference between MMOGs and MOGs. MOGs such as Diablo II and Counter-Strike do give players the opportunity to compete against each other and / or the game world in an online setting against multiple players. The difference between MOGs and MMOGs is in the areas of scale, permanence and player interaction.
MOGs typically have a small limit on the number of players who can be involved in a single game world at a time. This limit is usually a multiple of four (eg four, eight, thirty-two) but usually doesn’t exceed sixty-four on the one game world. A MMOG world, on the other hand, may see hundreds or thousands of players engaged in the same environment at the same time. MMOGs often also contain a greater scale of things for the player to do or see, whereas MOGs often limit the player to a smaller number of opportunities or actions.
Conversely, MOGs are often more stable and specialised than MMOGs. MOG in-game systems, being more limited than those found in MMOG environments, are more capable of being tested and therefore keeping coding bugs to a minimum. Also, since MOGs are more specialised, they are often able to do one or two things very well, whereas some MMOGs have on occasion been accused of having little depth in any of the numerous systems they offer.
MOGs may see player avatars gain experience and / or items over time, but typically rely on player skill for achievement. Such achievement is generally not recorded over the longer term by the MOG in question; scores for Counter-Strike or Unreal Tournament 2K3 may only be measured until the end of the round and are then reset. MMOGs require that achievements are recorded over a long period of time and thus have a permanence that is lacking in MMOGs.
MMOG permanence isn’t unavoidable; players can delete or create avatars as they see fit. In doing so, they reset their in-game abilities and only have the added advantage of their game knowledge to help them gain experience with their new avatar – the advantage of player skill is minimised in this state. MOGs generally depend on all players starting in the same condition at the beginning of a new game and then succeeding based on player ability.
Player involvement levels differ between MOGs and MMOGs. In general, MOGs determine achievement as a function of player skill, whereas MMOGs determine achievement as a function of time in game. Although player skill can be developed through practise and thus increasing future player achievement, this aspect is solely player-based. MMOGs require avatar achievements to be constantly recorded meaning that future achievement of an avatar requires both player-side and game-side involvement.
As an example, in order to kill a powerful creature in a MOG, a player is generally required to actively direct their avatar away from the creature’s attacks while aiming their attacks manually. Killing a powerful creature in a MMOG requires a player to have raised their avatar’s experience to a point where the automatically calculated defence and attack ratings the avatar has will see it emerge victorious. Some player interaction is required in MMOG combat, but for the most part the success of a skill is determined by an abstract calculation of ability versus the difficulty of the task.
Exit, Voice and Loyalty & Bartle’s Player Types
Two theories were used to develop hypotheses and assess the findings of this study. These theories were the exit, voice and loyalty model proposed by Hirschman (1972) and the player type model proposed by Bartle (1996).
Exit, Voice and Loyalty
Hirschman (1972) suggests that when a consumer perceives a drop in the level of quality in a product or service they are purchasing, they have two options. The first is to exit the consumption of the product or service while the second is to voice their complaints to the organisation about the drop in quality. In deciding to exit, a consumer is making an economic decision that, if it occurs in large enough numbers, can hurt the financial position of an organisation. A decision to engage the organisation in discussion through the use of voice is a social or political decision that can lead to the organisation fixing the perceived decline in quality. It is easier for a consumer to exit than exercise their voice options since (in competitive industries) they have full control over exiting; using voice takes time and may still result in something that is unsatisfactory to the consumer.
Loyalty adds another level of complexity for the consumer. In some cases a consumer will feel personally involved with the organisation and product / service that is experiencing the quality decline. Rather than leaving immediately, loyalty will cause a consumer to continue their involvement with the organisation and see them more likely to explore voice options than exit options. Loyalty is not an indefinite behaviour; it is calculated to benefit the consumer but over a longer period of time than would otherwise be accepted. If the organisation does not show improvement or does not appear to be responding to the consumer voice, even most loyal of consumers will eventually exit.
One area that the exit, voice and loyalty model has overlooked according to Zhang (2000) is that the voice option may not always be helpful to the organisation in question. Consumers can be very self-centred or complain about things that would not be worth changing; an organisation may have to go through a lot of voice "noise" in order to find a "signal" that contains a message worth listening to.
This theory arises from the author attempting to provide organisations with some framework on which to assess and rectify their decline, but it is equally applicable to consumer behaviour studies. It appears that research into consumer complaint behaviour has seen the main application of Hirschman’s model. Maute and Forrester (1993) have provided support for the exit, voice and loyalty behaviour classifications in their investigation of complaint behaviour. Bolton and Bronkhurst (1995) have linked customer satisfaction with loyalty, suggesting that consumers with low levels of satisfaction are more likely to exit, particularly if they have tried using voice options through complaining. It was also found that the types of complaint can impact on the chance of consumer exit, particularly if the complaint was centred around the failure of the organisation’s core business.
In examining MMOGs under Hirschman’s framework, it appears they would fall under the title of a "Competitive business enterprise in relation to customers". The MMOG industry is highly competitive, with an increasing number of MMOGs released each year and a larger number of proposed MMOGs in production that are all out to attract customers’ eyes, wallets and hopefully hearts to them. If MMOGs as a whole are competitive business enterprises, Hirschman suggests that decline in quality would result in exit and that the MMOG organisations would be most responsive to the exiting of members. An implication of this is that MMOG organisations potentially will not react as actively to members who vocalise their dissatisfaction as they will to members who cancel their subscriptions. Anecdotal evidence suggests that this behaviour is generally true.
Bartle’s Player Types
Bartle (1996) is generally credited with giving names to four general player types that populate MUDs – Achievers, Killers, Socialisers and Explorers. These player types / motivations were developed based on discussions and observations within one MUD Bartle was involved in, but are seen to be reliable across a large number of MUDs. It should be noted that MUDs and MMOGs are potentially very different creatures, but this will be discussed in more depth later on in this section.
Bartle describes the player types briefly as such:
The difference between acting and interacting is cooperation. Socialisers seek permission to involve themselves with other players while Killers simply force themselves on others. Explorers seek to understand the game world in one way or another while Achievers seek to dominate the game mechanics.
Of all the player types, Killers are the most worrying for game designers. Due to their ability (and in some cases, desire) to ruin the game experience for other players, this type of player can actively drive players away from a MMOG / MUD. Killers are generally despised as a player type since they often exploit the game mechanic / people’s naivety / whatever they can to ‘win’ over another player. Almost all MMOGs seek to take steps against Killers to curb their effects, but these attempts are not always successful. Other player types do less ‘damage’ to the game world and are generally not as problematic.
It should be recognised that a player can be a combination of player types / motivations in their behaviour. Some players will seek to search out all of the game world (Explorer behaviour) in order to find more powerful weapons / equipment (Achiever behaviour). Other players enjoy forming groups (Socialiser behaviour) before entering a new game area (Explorer behaviour) and maybe killing a few other player characters who are weaker than them (Killer behaviour).
Such behaviours are also influenced by the nature of the MMOG they are involved in. Shadowbane allows direct player-versus-player conflict, meaning that Killers (or Killer behaviour) is permitted. Everquest is perceived as most rewarding to those who are predominantly Achievers, with little Killer behaviour permitted.
A potential weakness of the Bartle player types is that it was developed for MUDs, so may not be as applicable to MMOGs. MUDs are text-based environments, often free to enter and maintained by volunteers. MMOGs are graphical environments that are developed by professional game design organisations, requiring subscription fees to be maintained. It is very possible that the player motivations for those involved in a free text MUD is different to those involved in a costly, graphically intense MMOG.
The nature of the players themselves may also have changed. MUDs originated and were most popular in the days before the internet was widely adopted as a social and commercial tool – the distribution of computers and modems was less advanced while actual connection to computer networks was more complicated, requiring some specialist knowledge. MMOGs are arising in popularity as high-speed bandwidth options are becoming more available, with little to no specialist computer knowledge required to connect to the game world. It is very possible that MUD players, as early adopters of online gaming worlds, have very different characteristics and motivations to those players connecting today – the rise of the ‘casual MMOG player’ (players who play MMOGs for only a few hours per week) being a possible example of such a change.
With that said, it is recognised that Bartle (1996) set a benchmark in regards to online player motivation research that has arguably yet to be improved upon. Yee (2002) has added Leadership as a potential motivational factor, but this was based on research of three arguably similar games and has not been tested across other MMOGs. It is yet to be seen if Bartle’s or Yee’s findings will be as applicable to the players who sign up to the MMOFPSs or MMORTSs that have been / will be released as they are to MMORPGs.
Hypotheses
A number of hypotheses were developed for testing in this study.
H1: Players of similar Bartle types will quit for similar reasons.
Bartle scores prepose to indicate a players motivation for involvement in a MMOG. It is believed that players with similar Bartle scores (particularly those with one or two dominant player type scores) will exit a MMOG for similar reasons. For example, it is expected that if players with Explorer motivations are unable to find areas to explore, they are likely to leave a MMOG rather than change their playing motivation.
H2: Players who have played a MMOG for similar periods of time will exit for similar reasons.
It is expected that players who involve themselves in a MMOG for similar periods of time will exit the game citing similar reasons, regardless of the MMOG being played. Although MMOGs promise a permanent game world where avatars can exist indefinitely, player burnout should be recognised as a natural event. This may mean that despite player claims that faulty game design is the cause of their exit, the actual reason will be one arising from normal human behaviour.
H3: Special offers encouraging players to return to a MMOG after they have left it are effective attracting players back.
Having given up involvement in a time-intensive MMOG a player will be unlikely to return if no encouragement exists to do so. It is expected that if a special offer exists (such as free play time) players are more likely to return than to a MMOG where no special offer has been made.
Methodology
Early in 2003 an online questionnaire was developed and placed online. Information about the survey and requests for participants were placed on several official MMOG (ie Fung Wan Online, City of Heroes) and MMOG-interest (ie Gamefaqs MMOG forums, Waterthread.org, Stratics) online forums. The official MMOG forums that were used had not released games at the point the survey was linked to the forum. This was done in an attempt to avoid exit reasons from that MMOG from dominating responses from respondents of that forum.
The online questionnaire included a link to an online Bartle Test that defined a player’s motivation type based on their responses to paired questions. Respondents were asked to take that test and to enter their scores into the relevant section of the questionnaire.
After a period of several months 434 responses had been collected. These responses were then examined to ensure that respondents had completed the questionnaire fully and correctly. A number of responses were removed due to 1) a lack of inclusion of a Bartle player motivation score or 2) an non-MMOG being detailed by the respondent such as CounterStrike, The Sims or Diablo2 – these games are muliplayer online games (MOGs) not MMOGs. An attempt has been made to ensure that all the MMOGs listed by players fit the MMOG definition set by this study, but difficulties (such as lack of English-language information) made complete verification impossible. However, the majority of responses to the survey came from those involved in a recognised MMOG.
After the data had been cleaned, a total of 392 responses remained for analysis.
Analysis of Data
DEMOGRAPHICS AND GENERAL MMOG EXPERIENCE
Demographic variables indicated that the majority of respondents were male (96%) aged 25 and under (63%). Only a small number (n=15) of respondents were female. Given claims that between 12% and 16% of all MMOG players are female (Yee, 2002b), this possibly suggests that the sources used to collect respondents from may be male dominated or that female MMOG players were less likely to respond to the exit survey.
The majority of respondents came from North America (71%). European respondents made up 17% of the sample, with a small number coming from Asia / Middle East, Australia and South America. Despite the popularity of MMOGs in some Asian countries, these results are to be expected given the sources used to attract respondents and the language (ie English) used in the survey.
The results of the Bartle scores returned by respondents sees Explorer return the highest mean (57.18), followed by Achiever (51.61), Socialiser (47.19) and Killer (42.83). Killer scores had the highest standard deviation of all the Bartle scores, suggesting that a wide range of values were returned for this score. Socialiser scores had the second highest standard deviation. Mean and median values for these scores were not significantly different.
Examining the frequency distribution of these Bartle scores shows that the Achiever and Explorer scores have an approximate bell curve (although the Achiever scores skew to the right of the scale), while Killer and Socialiser distributions are flatter.
The distribution of these Bartle scores suggests that the sample tended to have participants who were most likely to be Achievers than any other Bartle type.
Reported involvement in MMOGs is quite high from participants, with 92% having played two or more MMOGs in their lifetime and 82% having played two or more MMOGs in the last twelve months. Over 30% of participants have played in more than 5 MMOGs, with nearly 11% of participants have been involved in 5 MMOGs in the last 12 months.
These results suggest that there are numerous players who look a large number of MMOGs over a short period of time – given that a key attraction of MMOGs is their ability to offer a lasting or persistent game play experience, this type of player appears to consume them at an incredible rate. It is also possible that these results are biased by the sources used to obtain the participants; these forums are likely to contain the more "hard core" MMOG player. However, it should be recognised that such players exist and that just because a persistent game experience is offered does not mean players will hang around waiting to experience everything.
These players may not be paying for these game experiences either. Many MMOG developers hold beta tests prior to game launch. In beta, players are allowed to enter a MMOG for free in exchange for testing out game features and reporting back to the developers any system bugs or problems they find; in reality, for many players this is a free trial period to assess if they would pay for their involvement after the game launches (also known as "pay to play" or "p2p"). Betas often only last for fixed periods, meaning players may be involved in several across the course of a year.
Given the high number of MMOGs that players are reporting involvement in, it would be expected that exit from MMOGs would also be reasonably high. MMOGs require a solid time commitment to achieve in-game results, so it would appear difficult to be involved in several MMOGs at once; hence, reported exit should be high.
This appears to be mostly true, but MMOGs do appear to execute some "stickiness" that keeps players involved. Over 94% of participants reported exiting a MMOG in the last twelve months, with over 62% reporting leaving two or more.
MMOG PLAY BEHAVIOUR
Participants were asked to consider one MMOG they had exited from in when providing information about leaving a MMOG. The names of the MMOGs considered when completing this survey shows a wide variety. Everquest (22%), Dark Ages of Camelot (19%) and Ragnarok Online (7%) are the three most frequently mentioned MMOGs that players have exited from. The most often exited MMOGs reflected their relative size in the market – Everquest and Dark Ages of Camelot both have wide recognition in North American MMOG players and are seen to be the two largest english-speaking MMOGs.
Participants appear willing to put long hours and several months into a MMOG before exit. Only 10% of participants played a MMOG for less than a month and at the other end of the scale nearly 14% spent more than two years involved in one MMOG. The mean score indicates that most participants spend about four to six months involved in a MMOG before exit.
Participants also indicate they play MMOGs frequently and for long periods of time. Over 55% indicated they played 7 or more times per week, with this figure increasing to 81% for those playing 5 or more times per week. When asked about the average length of these sessions, almost 67% of participants indicated they would play more than three hours, with 6% indicating that average play sessions that lasted longer than 18 hours. These statistics would seem to indicate that for some participants MMOGs are a lifestyle, consuming the majority of their time.
The vast majority (80%) of participants use high-speed internet access to play MMOGs, which would provide them with a better play experience. It also perhaps indicates that these MMOG players are on the leading edge of internet technology, but whether they have purchased high-speed access to improve their MMOG experience or already had high-speed access and found it useful to have when playing MMOGs is impossible to say.
MMOG EXIT MOTIVATION
Participants were invited to offer a primary and secondary reason for exiting their specified MMOG. In many cases this lead to several reasons being stated which led to such responses being examined to try and find the two core issues that caused their exit. Once these issues had been examined, a total of 24 exit motivation categories for were developed. These issues were:
Advancement issues (13%) appears as the second highest reason for exit, with other reasons for exit being spread over the other categories.
These are just topline results, but seem to indicate that boredom, advancement issues and social reasons are the main reported motivators of player exit, with payment issues having some impact. Player boredom is perhaps past the MMOGs control – it is perhaps too generic a term to be directly actionable, but anyone is likely to get bored in any task given long enough – but advancement issues should be of concern. Large numbers of players appear to dislike an advancement system that involves large numbers of repetitious actions. It is also possible that boredom and advancement issues are two terms for the same thing, worded differently in participant responses.
Social reasons appear to offset boredom and advancement issues to some extent, showing up more strongly in the secondary reasons than the primary reasons for exit. It is very possible that players stick out tedious advancement systems past the point of boredom if they have a social reason for continuing eg friends in game or they are part of a guild. Once this support fails (eg friends leave the game or the guild splits up) players are no longer bound to the game and exit.
A table of the combined MMOG exit reasons is below:
Group $REASON Reason for MMOG exit
Pct of Pct of
Category label Code Count Responses Cases
Poor management / customer service 1 29 4.3 7.4
Pay-to-play or financial issues / end of 2 57 8.5 14.5
Boredom 3 85 12.7 21.7
Advancement issues 4 83 12.4 21.2
Endgame issues 5 35 5.2 8.9
External time pressures 6 17 2.5 4.3
Internal time pressures 7 18 2.7 4.6
Social reasons 8 70 10.4 17.9
Dissatisfaction with changes 9 26 3.9 6.6
Technical issues / bugs 10 36 5.4 9.2
Lack of content / updates 11 31 4.6 7.9
Lack of player impact on game world 12 11 1.6 2.8
Lack of roleplaying / story 13 12 1.8 3.1
Didn't meet expectations / no fun 14 27 4.0 6.9
Administration error 15 1 .1 .3
Player technical limitations 16 5 .7 1.3
Dissatisfaction with PVP 17 22 3.3 5.6
Game balance / design issues 18 37 5.5 9.4
Poor user interface / game appearance 19 15 2.2 3.8
Grief / exploits 20 14 2.1 3.6
Another MMOG / game 21 24 3.6 6.1
Addiction / burnout 22 9 1.3 2.3
Completed own goals / lost interest 23 4 .6 1.0
Sold account 24 3 .4 .8
------- ----- -----
Total responses 671 100.0 171.2
0 missing cases; 392 valid cases
REJOINING BEHAVIOUR
Participants were also asked if they had every rejoined a MMOG after exit. Out of the sample of 392, 199 or nearly 51% had engaged in such behaviour.
As for exit, reasons participants gave for rejoining were classified into categories, with 13 categories being created in this case. These categories are:
Of these motivations, it appears that social reasons (29%), trying new content / retrying MMOG (22%) and wanting a MMOG to play / bored without a MMOG (17%) were the main reasons behind resubscribing to a previously-exited MMOG.
In addition to this, over 23% of rejoining participants indicated that a special offer (eg free play time or bonus items) encouraged them to rejoin. Given that only 4% of participants indicated they rejoined primarily due to the special offer, this would suggest that players may rejoin for another main reason, but having access to a special offer reduces the barriers to MMOG re-entry. That said, a special offer on its own may not be enough to encourage players to re-join – something else must be there to help pull them in.
Most participants only reported rejoining once. Since no questions were asked regarding how long they stayed after re-joining, it is impossible to investigate the behaviour of re-joining players. A small number (2%) of players indicated they had rejoined the same MMOG more than five times – it is suggested that this behaviour would be more likely to take place over a long period of time.
Results
The above tables of stated MMOG exit motivations provides some indication behind why players exit from MMOGs. However, in order to gain a deeper insight into how player behaviours / motivations impact on these exit reasons, data collected from during this research will be tested against the hypotheses stated previously in this research.
For hypothesis testing, exit motivations that had received less than 10 responses were removed, leaving 351 of the original 392 data points and 14 of the original 24 exit motivations in the analysis. Instead of primary reason for exit being used in the analysis, a new variable called reasons for MMOG exit by rank will be used as the dependent variable. This has been done in an attempt to reduce the impact of exit motivations with a small number of responses from distorting results.
H1: Players of similar Bartle types will quit for similar reasons.
Due to the large number of categorical variables used in the research, discriminant analysis was used to examine the structure of the potential relationships between the Bartle scores and primary reason for exit.
In examining the data descriptives, it can be seen that the Explorer Score has a univariate mean that is significantly different from the others. The Box’s M score is not significant, indicating that the covariance matrix is homogenous and the discriminant analysis can go ahead.
The Socialiser Score failed the Tolerance test, excluding it from further analysis.
The standardised discriminant function coefficients indicate the relative importance of independent variables in predicting the dependent. For each of these scores, one Bartle score can be seen as the main determinant of that function eg function 1 is dominated by the Explorer Score.
Due to the Bartle scores dominating a function each, it is not surprising to see them dominate a structure coefficient either, since structure coefficients simply show a correlation between functions and independent variables. If one were to label the functions in the structure matrix, it would be simple to do so down existing Bartle score lines.
When the discriminant function is used to classify cases about their exit motivations based on Bartle score, it shows a large tendency to misclassify cases. Assuming equal probability of an exit motivation (in this case this is 7.1% based on 14 exit motivations), the correct classification rate for participants is only 7.7% - hardly greater than chance. When the leaving-one-out method is used to correct some of the classification bias, this successful classification rate further drops to 4.3%.
| Reasons for MMOG exit by rank | Predicted Group Membership | Total | |||||||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Boredom | Advancement issues | Pay-to-play or financial issues / end of beta | Social reasons | Endgame issues | Technical issues / bugs | Didn't meet expectations / no fun | Game balance / design issues | Dissatisfaction with changes | Lack of content / updates | External time pressures | Internal time pressures | Poor management / customer service | Another MMOG / game | ||||
| Original | Count | Boredom | 5 | 0 | 6 | 0 | 1 | 3 | 0 | 4 | 2 | 7 | 6 | 0 | 16 | 18 | 68 |
| Advancement issues | 0 | 0 | 5 | 0 | 3 | 2 | 0 | 6 | 1 | 9 | 4 | 0 | 16 | 6 | 52 | ||
| Pay-to-play or financial issues / end of beta | 4 | 0 | 4 | 0 | 0 | 3 | 0 | 2 | 3 | 6 | 1 | 1 | 3 | 12 | 39 | ||
| Social reasons | 2 | 0 | 2 | 0 | 1 | 3 | 0 | 2 | 4 | 9 | 2 | 1 | 5 | 4 | 35 | ||
| Endgame issues | 0 | 0 | 1 | 0 | 0 | 2 | 0 | 3 | 1 | 3 | 0 | 0 | 8 | 6 | 24 | ||
| Technical issues / bugs | 1 | 0 | 0 | 1 | 0 | 1 | 1 | 1 | 0 | 4 | 2 | 0 | 4 | 4 | 19 | ||
| Didn't meet expectations / no fun | 2 | 0 | 1 | 0 | 0 | 1 | 0 | 3 | 0 | 4 | 0 | 0 | 5 | 2 | 18 | ||
| Game balance / design issues | 0 | 0 | 3 | 0 | 0 | 0 | 0 | 4 | 2 | 2 | 0 | 0 | 6 | 1 | 18 | ||
| Dissatisfaction with changes | 0 | 0 | 3 | 0 | 0 | 2 | 1 | 1 | 3 | 1 | 1 | 0 | 3 | 2 | 17 | ||
| Lack of content / updates | 1 | 0 | 0 | 0 | 0 | 0 | 0 | 1 | 2 | 4 | 1 | 1 | 4 | 1 | 15 | ||
| External time pressures | 0 | 0 | 2 | 0 | 0 | 0 | 0 | 0 | 1 | 4 | 0 | 0 | 3 | 2 | 12 | ||
| Internal time pressures | 0 | 0 | 0 | 0 | 0 | 0 | 0 | 3 | 0 | 3 | 0 | 0 | 3 | 3 | 12 | ||
| Poor management / customer service | 1 | 0 | 0 | 0 | 0 | 0 | 1 | 0 | 0 | 1 | 1 | 1 | 4 | 2 | 11 | ||
| Another MMOG / game | 0 | 0 | 2 | 0 | 0 | 0 | 1 | 1 | 3 | 0 | 1 | 0 | 1 | 2 | 11 | ||
| Ungrouped cases | 1 | 0 | 0 | 0 | 0 | 1 | 1 | 4 | 5 | 12 | 0 | 0 | 8 | 9 | 41 | ||
| % | Boredom | 7.4 | .0 | 8.8 | .0 | 1.5 | 4.4 | .0 | 5.9 | 2.9 | 10.3 | 8.8 | .0 | 23.5 | 26.5 | 100.0 | |
| Advancement issues | .0 | .0 | 9.6 | .0 | 5.8 | 3.8 | .0 | 11.5 | 1.9 | 17.3 | 7.7 | .0 | 30.8 | 11.5 | 100.0 | ||
| Pay-to-play or financial issues / end of beta | 10.3 | .0 | 10.3 | .0 | .0 | 7.7 | .0 | 5.1 | 7.7 | 15.4 | 2.6 | 2.6 | 7.7 | 30.8 | 100.0 | ||
| Social reasons | 5.7 | .0 | 5.7 | .0 | 2.9 | 8.6 | .0 | 5.7 | 11.4 | 25.7 | 5.7 | 2.9 | 14.3 | 11.4 | 100.0 | ||
| Endgame issues | .0 | .0 | 4.2 | .0 | .0 | 8.3 | .0 | 12.5 | 4.2 | 12.5 | .0 | .0 | 33.3 | 25.0 | 100.0 | ||
| Technical issues / bugs | 5.3 | .0 | .0 | 5.3 | .0 | 5.3 | 5.3 | 5.3 | .0 | 21.1 | 10.5 | .0 | 21.1 | 21.1 | 100.0 | ||
| Didn't meet expectations / no fun | 11.1 | .0 | 5.6 | .0 | .0 | 5.6 | .0 | 16.7 | .0 | 22.2 | .0 | .0 | 27.8 | 11.1 | 100.0 | ||
| Game balance / design issues | .0 | .0 | 16.7 | .0 | .0 | .0 | .0 | 22.2 | 11.1 | 11.1 | .0 | .0 | 33.3 | 5.6 | 100.0 | ||
| Dissatisfaction with changes | .0 | .0 | 17.6 | .0 | .0 | 11.8 | 5.9 | 5.9 | 17.6 | 5.9 | 5.9 | .0 | 17.6 | 11.8 | 100.0 | ||
| Lack of content / updates | 6.7 | .0 | .0 | .0 | .0 | .0 | .0 | 6.7 | 13.3 | 26.7 | 6.7 | 6.7 | 26.7 | 6.7 | 100.0 | ||
| External time pressures | .0 | .0 | 16.7 | .0 | .0 | .0 | .0 | .0 | 8.3 | 33.3 | .0 | .0 | 25.0 | 16.7 | 100.0 | ||
| Internal time pressures | .0 | .0 | .0 | .0 | .0 | .0 | .0 | 25.0 | .0 | 25.0 | .0 | .0 | 25.0 | 25.0 | 100.0 | ||
| Poor management / customer service | 9.1 | .0 | .0 | .0 | .0 | .0 | 9.1 | .0 | .0 | 9.1 | 9.1 | 9.1 | 36.4 | 18.2 | 100.0 | ||
| Another MMOG / game | .0 | .0 | 18.2 | .0 | .0 | .0 | 9.1 | 9.1 | 27.3 | .0 | 9.1 | .0 | 9.1 | 18.2 | 100.0 | ||
| Ungrouped cases | 2.4 | .0 | .0 | .0 | .0 | 2.4 | 2.4 | 9.8 | 12.2 | 29.3 | .0 | .0 | 19.5 | 22.0 | 100.0 | ||
| Cross-validated(a) | Count | Boredom | 4 | 0 | 6 | 0 | 1 | 3 | 0 | 4 | 3 | 7 | 6 | 0 | 16 | 18 | 68 |
| Advancement issues | 0 | 0 | 5 | 0 | 3 | 2 | 0 | 6 | 1 | 9 | 4 | 0 | 16 | 6 | 52 | ||
| Pay-to-play or financial issues / end of beta | 4 | 0 | 3 | 0 | 0 | 4 | 0 | 2 | 3 | 6 | 1 | 1 | 3 | 12 | 39 | ||
| Social reasons | 2 | 0 | 2 | 0 | 1 | 3 | 0 | 2 | 4 | 9 | 2 | 1 | 5 | 4 | 35 | ||
| Endgame issues | 0 | 0 | 1 | 0 | 0 | 2 | 0 | 3 | 1 | 3 | 0 | 0 | 8 | 6 | 24 | ||
| Technical issues / bugs | 1 | 0 | 0 | 2 | 0 | 0 | 1 | 1 | 0 | 4 | 2 | 0 | 4 | 4 | 19 | ||
| Didn't meet expectations / no fun | 2 | 0 | 1 | 0 | 0 | 1 | 0 | 3 | 0 | 4 | 0 | 0 | 5 | 2 | 18 | ||
| Game balance / design issues | 0 | 0 | 3 | 0 | 0 | 0 | 0 | 2 | 2 | 2 | 0 | 2 | 6 | 1 | 18 | ||
| Dissatisfaction with changes | 0 | 0 | 3 | 0 | 0 | 2 | 1 | 3 | 1 | 1 | 1 | 0 | 3 | 2 | 17 | ||
| Lack of content / updates | 1 | 0 | 0 | 1 | 0 | 1 | 0 | 1 | 2 | 2 | 1 | 1 | 4 | 1 | 15 | ||
| External time pressures | 0 | 0 | 2 | 0 | 0 | 0 | 0 | 0 | 1 | 4 | 0 | 0 | 3 | 2 | 12 | ||
| Internal time pressures | 0 | 0 | 0 | 0 | 0 | 0 | 0 | 3 | 0 | 3 | 0 | 0 | 3 | 3 | 12 | ||
| Poor management / customer service | 1 | 0 | 0 | 0 | 0 | 0 | 1 | 1 | 0 | 1 | 1 | 1 | 3 | 2 | 11 | ||
| Another MMOG / game | 0 | 0 | 4 | 0 | 0 | 0 | 1 | 1 | 3 | 0 | 1 | 0 | 1 | 0 | 11 | ||
| Ungrouped cases | |||||||||||||||||
| % | Boredom | 5.9 | .0 | 8.8 | .0 | 1.5 | 4.4 | .0 | 5.9 | 4.4 | 10.3 | 8.8 | .0 | 23.5 | 26.5 | 100.0 | |
| Advancement issues | .0 | .0 | 9.6 | .0 | 5.8 | 3.8 | .0 | 11.5 | 1.9 | 17.3 | 7.7 | .0 | 30.8 | 11.5 | 100.0 | ||
| Pay-to-play or financial issues / end of beta | 10.3 | .0 | 7.7 | .0 | .0 | 10.3 | .0 | 5.1 | 7.7 | 15.4 | 2.6 | 2.6 | 7.7 | 30.8 | 100.0 | ||
| Social reasons | 5.7 | .0 | 5.7 | .0 | 2.9 | 8.6 | .0 | 5.7 | 11.4 | 25.7 | 5.7 | 2.9 | 14.3 | 11.4 | 100.0 | ||
| Endgame issues | .0 | .0 | 4.2 | .0 | .0 | 8.3 | .0 | 12.5 | 4.2 | 12.5 | .0 | .0 | 33.3 | 25.0 | 100.0 | ||
| Technical issues / bugs | 5.3 | .0 | .0 | 10.5 | .0 | .0 | 5.3 | 5.3 | .0 | 21.1 | 10.5 | .0 | 21.1 | 21.1 | 100.0 | ||
| Didn't meet expectations / no fun | 11.1 | .0 | 5.6 | .0 | .0 | 5.6 | .0 | 16.7 | .0 | 22.2 | .0 | .0 | 27.8 | 11.1 | 100.0 | ||
| Game balance / design issues | .0 | .0 | 16.7 | .0 | .0 | .0 | .0 | 11.1 | 11.1 | 11.1 | .0 | 11.1 | 33.3 | 5.6 | 100.0 | ||
| Dissatisfaction with changes | .0 | .0 | 17.6 | .0 | .0 | 11.8 | 5.9 | 17.6 | 5.9 | 5.9 | 5.9 | .0 | 17.6 | 11.8 | 100.0 | ||
| Lack of content / updates | 6.7 | .0 | .0 | 6.7 | .0 | 6.7 | .0 | 6.7 | 13.3 | 13.3 | 6.7 | 6.7 | 26.7 | 6.7 | 100.0 | ||
| External time pressures | .0 | .0 | 16.7 | .0 | .0 | .0 | .0 | .0 | 8.3 | 33.3 | .0 | .0 | 25.0 | 16.7 | 100.0 | ||
| Internal time pressures | .0 | .0 | .0 | .0 | .0 | .0 | .0 | 25.0 | .0 | 25.0 | .0 | .0 | 25.0 | 25.0 | 100.0 | ||
| Poor management / customer service | 9.1 | .0 | .0 | .0 | .0 | .0 | 9.1 | 9.1 | .0 | 9.1 | 9.1 | 9.1 | 27.3 | 18.2 | 100.0 | ||
| Another MMOG / game | .0 | .0 | 36.4 | .0 | .0 | .0 | 9.1 | 9.1 | 27.3 | .0 | 9.1 | .0 | 9.1 | .0 | 100.0 | ||
| Ungrouped cases | |||||||||||||||||
| a Cross validation is done only for those cases in the analysis. In cross validation, each case is classified by the functions derived from all cases other than that case. | |||||||||||||||||
| b 7.7% of original grouped cases correctly classified. | |||||||||||||||||
| c 4.3% of cross-validated grouped cases correctly classified. | |||||||||||||||||
If it is assumed that the frequency of exit reasons collected in this research is reflective of the overall MMOG player population behaviour, then classifying cases can be based on the prior probabilities of the given exit reasons being used in this analysis (ie boredom was the most often given exit motivation, so it is the most likely reason for player exit). In this case up to 19% of cases are correctly classified using the leaving-one-out method – a definite improvement over classification assuming equal probability, but still not particularly reliable.
This analysis has failed to prove the hypothesis, indicating that players of similar Bartle types will not necessarily exit for the same reasons.
H2: Players who have played a MMOG for similar periods of time will exit for similar reasons.
Hierarchical regression analysis was used to test this hypothesis. Reasons for MMOG exit by rank was used as the dependent variable while number of months played before exit, number of sessions played per week and length of average play session (in hours) were selected as the predictors.
The model summary of the regression analysis indicates that only a small percentage of the variance can be explained using these three predictors – only around 3% of exit motivation is determined by number of months played before exit, number of sessions played per week and length of average player session. Number of months played before exit appears to be the most important predictor, providing a significant unique contribution of 2% when included in the analysis.
The beta values of the coefficients indicate that number of months played before exit offers the best predictor of exit motivation of the variables examined – the other variables do not contribute significantly to the regression.
The scatterplot of regression residuals against predicted values appears to show a slightly negative relationship existing between them. This relationship could violate the assumption of linearity and call into question the reliability of any modelling developed from the regression equation.
Given that the hierarchical multiple regression model developed may violate an assumption of regression analysis, it would normally be the case to examine the relationship between the variables using a different statistical technique. However, since the regression analysis has indicated a very small degree of information can be gained about exit motivation based on the amount of time spent in-game, it is questionable what any further analysis would reveal. It is interesting to note that the number of months spent in a MMOG is more important to exit than other timing variables examined. Perhaps it is easier to conceptualise time spent in game (and goals achieved / creatures killed / items collected etc) in terms of months rather than hours or sessions.
Based on the results above, the initial hypothesis has not been proven, indicating that players who have been in a MMOG for a similar period of time will not necessarily exit for similar reasons.
H3: Special offers encouraging players to return to a MMOG after they have left it are effective in attracting players back.
A chi-square and crosstabulation of if a participant has ever rejoined a MMOG against if the MMOG had a special offer indicates that a significant relationship exists between them. Of those participants who did rejoin a MMOG, 46 responses (23%) indicated that a special offer encouraged them to rejoin. This is significantly larger than the expected value of 25 responses if the two variables were independent.
Based off these results, the hypothesis that special offers are effective in attracting players back is considered to be accepted. Although not an absolute way of attracting players back, offering free time or special equipment appears to have some influence on a player's decision to rejoin a MMOG they had previously exited.
Conclusions
Player MMOG exit motivations appear to be consistent with the theories of Hirschman (1972). The majority of exit motivations can be seen to tie in decline in MMOG quality, at least in the eyes of that individual player. Due to the reliance of player enjoyment on in-game advancement, any wane in advancement will lead to players becoming dissatisfied with the quality of the MMOG and considering exit. Arguably player in-game advancement is the "core business" of any MMOG (from the players’ point of view) – flaws in this area are expected to cause the greatest number of exits.
The three largest total exit motivations were boredom (22% of cases), advancement issues (21% of cases) and social reasons (18% of cases). Boredom can be seen as a decline in satisfaction as players feel they have precious little novelty to find in the MMOG. Advancement issues can be linked to a decline in satisfaction with their in-game ability to meet goals important to players. Social reasons, often tied to the removal or change in a social network, can be seen as perhaps associated to loyalty in a MMOG – players would continue involvement in a game they no longer fully enjoyed because they were part of something larger than themselves. Once this net was removed, players had little reason to stay and leave soon after.
This research has not given any consideration to voice, so it is impossible to determine how the treatment of voice impacts on a player’s decision to exit. Future research in this area would uncover a potential treasure trove of information about service issues in MMOGs and how players would prefer them handled.
Bartle’s player types do not appear to be closely tied to exit motivations. Given the potential link that exists between why a player will play a MMOG and why they will exit one, it is surprising to find that the accepted model of MMOG player type did not provide much insight into exit motivations. It is recognised errors in the analysis of the sample may have impacted on results, but it is also possible that the accepted Bartle scores are less applicable to MMOGs than they are believed to be.
Some evidence was provided by this research that indicates that time spent involved in a MMOG (notably in months) has some relationship to exit motivation, but not to large extent. Further research would be required to determine what better predictors of MMOG exit motivation exist and how player type and time in-game impact on these motivations.
Special offers appear to somewhat help reduce the barriers of re-entry for MMOG players. Although it is very likely that a player will reconsider a MMOG for a reason other than an available special offer, the offer may be enough to help a hesitant player over concerns they may have. Their behaviour once they rejoin was not looked at in this research – it is possible that rejoiners only involve themselves in a MMOG for a short time or for only as long as the special offer lasts.
When it comes to rejoining, social reasons (29% of rejoining cases) were the main stated motivation – players had friends entering or still in the MMOG and wished to join them. Trying out new content / retrying MMOG (22%) and wanted or bored without a MMOG to play (17%) were the other two main reasons. Both appear to have some link to a player’s loyalty to a MMOG; despite prior exit, these players are willing to come back and give the MMOG a second- (or greater) chance for different reasons. Players potentially test out new content or retry the MMOG to see if changes made to the game make it more suitable for them. Other players potentially enjoy playing MMOGs and the persistent world it offers them; with no true end to the game, they can play indefinitely, even though they may not be completely satisfied with the current condition of the MMOG.
Limitations of the Research
A number of issues raised themselves over the course of this research that may impact on the validity and applicability of this study. The author is aware of the problems listed below and has noted them as a warning for others trying to conduct research in this area.
It is recognised that little attention was paid to the role of voice prior to or during exit from a MMOG, despite the use of Hirschman’s model as a theoretical base for this study. This was done in an attempt to focus the research only on motivations for exiting a MMOG rather than also including questions about behaviour leading up to an exit decision being made.
The online Bartle test used to measure a respondent’s Bartle score may not be reliable across multiple uses since it randomises questions each time the test is taken. The possibility exists that a person taking this test will not be exposed to the same types of questions each time the test is taken, meaning their Bartle score will vary even though their underlying motivation for playing MUDs / MMOGs will not.
An error in the data collection form saw respondents unable to enter triple digits (ie ‘100’) into the text boxes used to collect the Bartle scores. This error came to light after a reasonable number of responses had been collected. It was decided that the survey would continue as it was rather than fixing the error and discarding the responses collected to that point. It is believed that the difference of one point in these Bartle scores (with respondents entering ‘99’ instead of ‘100’) will have little impact on the result of the study, with only about 3% of respondents for each category running into this problem.
The majority of respondents came from MMORPGs. It is possible that players involved in other MMOG genres (eg MMOFPSs) have different motivations for playing and leaving a MMOG than those involved in MMORPGs. Further research would be required to determine if the player motivation for playing / exiting is consistent across all genres of MMOGs.
Some respondents gave more than two reasons for exit. Future analysis of exit motivations should allow for multiple exit reasons to be considered in analysis as players may potentially leave for a number of equally important reasons rather than one single exit motivation.
Further Research
A number of areas exist in which further research can be conducted regarding MMOG and player behaviour. Some of these areas include:
The role of voice in exit from a MMOG. Do players that attempt to contact MMOG developers / customer service about their problems before exiting? If they successfully have their problem dealt with are they more likely to stay involved in the MMOG? If they are unable to achieve satisfactory results through voice, how likely are players to exit the MMOG?
Do player motivations for playing MMOGs remain consistent across different MMOG genres? At the moment MMORPGs dominate the MMOG market, but other genres of MMOGs exist. Do players involved in MMORTSs or MMOFPSs play for the same reasons as those involved in MMORPGs?
Are Bartle’s player motivation types applicable to MMOGs? Although widely treated as though they are, little research appears to have been conducted into this area. Is Yee’s player motivation model a more accurate replacement for Bartle’s? Does a better model than the two mentioned above exist for examining player motivation?
References:
Bartle R. (1996). "Hearts, Diamonds, Clubs, Spades: Players Who Suit MUDS". Retrieved October 16 2003 from:
http://www.mud.co.uk/richard/hcds.htm
Bell P. (2003). "Realism and Subjectivity in First Person Shooter Video Games". Page 2. Retrieved October 16 2003 from:
http://gnovis.georgetown.edu/articles/pr009_Realism&SubjectivityF.pdf
Bolton R.N. Bronkhurst T.M. (1995). "The Relationship Between Customer Complaints to the Firm and Subsequent Exit Behaviour". Advances in Consumer Research. Volume 22. Pages 94-100. Retrieved October 17 from:
http://oxygen.vancouver.wsu.edu/otcgi/llscgi60.exe?DB=0&ACTION=ViewNative&QUERY=%65%78%69%74&RGN=%61%63%72%61%62%73%74%72%61%63%74&OP=and&QUERY=&RGN=%61%63%72%63%69%74%65&OP=and&QUERY=&RGN=%61%63%72%63%69%74%65&OP=and&QUERY=&RGN=%61%63%72%63%69%74%65&OP=and&QUERY=&RGN=%61%63%72%63%69%74%65&OP=and&QUERY=&RGN=%61%63%72%63%69%74%65&OP=and&QUERY=&RGN=%61%63%72%63%69%74%65&OP=and&SUBSET=SUBSET&FROM=1&SIZE=10&ITEM=1
Hirschman A.O. (1972). Exit, Voice and Loyalty: Responses to Decline in Firms, Organisations, and States. Harvard University Press. Cambridge.
Maute M.F. Forrester W.R. Jr. (1993). "The Structure and Determinants of Consumer Complaints Intentions and Behaviour." Journal of Economic Psychology. Volume 14. Issue 2. Pages 219-247.
Norusis M.J. (1993). SPSS for Windows: Professional Statistics, Release 6.0. SPSS Inc. USA.
Woodcock B.S. (2003). "An Analysis of MMOG Subscription Growth – Version 7.0". Retrieved October 16 2003 from:
http://pw1.netcom.com/~sirbruce/Subscriptions.html
Yee N. (2002). "Facets: 5 Motivation Factors for Why People Play MMORPGS". Retrieved October 16 2003 from:
http://www.nickyee.com/facets/home.html
Yee N. (2002b). "Codename Blue: An Ongoing Study of MMORPG Players". Retrieved March 1 2004 from:
http://www.nickyee.com/codeblue/demographics.html
Zhang. Y. (2000). "Book Review of Exit, Voice and Loyalty of Hirschman". Retrieved October 17 2003 from:
http://www.som.yale.edu/faculty/Sunder/PhdAccountingControl/HirschmanReviewClement.doc