From the discussion in the previous section, it is possible to summarize the interactions
between player types as follows:

To increase the number of achievers:

reduce the number of killers, but not by too much.
if killer numbers are high, increase the number of

To decrease the number of achievers:

increase the number of killers.
if killer numbers are low, reduce the number of

To increase the number of explorers:

increase the number of explorers.

To decrease the number of explorers:

massively increase the number of killers.

To increase the number of socialisers:

slightly decrease the number of killers.
increase the number of socialisers.

To decrease the number of socialisers:

slightly increase the number of killers.
massively increase the number of achievers.
massively decrease the number of achievers.
decrease the number of socialisers.

To increase the number of killers:

increase the number of achievers.
massively decrease the number of explorers.
increase the number of socialisers.

To decrease the number of killers

decrease the number of achievers.
massively increase the number of explorers.
decrease the number of socialisers.

What are the dynamics of this model? In other words, if players of each type were to
trickle into a system, how would it affect the overall make-up of the player population?

The following diagram illustrates the flow of influence. Each arrow shows a relationship,
from the blunt end to the pointed end. Ends are marked with a plus or minus to show an
increase or decrease respectively; the symbols are doubled up to indicate a massive
increase or decrease. Example: the line

killers + ------------> - achievers

means that increasing the number of killers will decrease the number of achievers.

+ <------------ +
- <------------ -
killers + ------------> - achievers
- + + - - ------------> +
^ ^ | | - + ++ ++ --
| | | | ^ ^ \ / /
| | | | | \ \ / /
| | | | \ \ X /
| | | | \ \/ X
| | | | \ / \/ \
| | | | / \ / \ \
| | | | / / \ \ \
| | | | / / \ \ \
| | | | | / \ \ |
| | v v v v \ | v
- + --++ - - ++ -- -
socialisers explorers
+ - - + + +
^ ^ | | ^ |
| | | | | |
\ \___/ / \___/

A graphical version of the figure appears at the end of the paper. [2]

From this, it can be seen that the numbers of killers and achievers is basically an
equilibrium: increasing the number of achievers will increase the number of killers, which
will in turn dampen down the increase in the number of achievers and thereby reduce
the number of excess killers.

The explorer population is almost inert: only huge numbers of killers will reduce it. It
should be noted, however, that massively increasing the number of explorers is the
*only* way to reduce the number of killers without also reducing the player numbers in
other groups. Because increasing the number of explorers in a MUD generally
encourages others to join (and non-explorers to experiment with exploration), this gives
a positive feedback which will eventually reduce the killer population (although recall the
earlier point concerning how few people are, by nature, explorers).

The most volatile group of people is that of the socialisers. Not only is it highly sensitive
to the number of killers, but it has both positive and negative feedback on itself, which
amplifies any changes. An increase in the number of socialisers will lead to yet more
socialisers, but it will also increase the number of killers; this, in turn, will reduce the
number of socialisers drastically, which will feed back into a yet greater reduction. It is
possible for new socialisers to arrive in large enough quantities for a downward spiral in
numbers not to be inevitable, but it is unlikely that such a system could remain viable in
over a long period of time.

This analysis of the dynamics of the relationships between players leads naturally to a
consideration of what configurations could be considered stable. There are four:

1) Killers and achievers in equilibrium. If the number of killers gets too high, then the
achievers will be driven off, which will cause the number of killers to fall also (through
lack of victims). If there aren't enough killers, then achievers feel the MUD isn't a
sufficient challenge (there being no way to "lose" in it), and they will gradually leave; new
killers could appear, attracted by the glut of potential prey, however this happens so
slowly that its impact is less than that of the disaffection among achievers. Socialisers
who venture out of whatever safe rooms are available eventually fall prey to killers, and
leave the game. Those who stay find that there aren't many interesting (to them) people
around with whom to talk, and they too drift off. Explorers potter around, but are not a
sufficient presence to affect the number of killers.

2) A MUD dominated by socialisers. Software changes to the MUD are made which
prevent (or at least seriously discourage) killers from practicing their craft on socialisers;
incoming socialisers are encouraged by those already there, and a chain reaction
starts. There are still achievers and explorers, but they are swamped by the sheer
volume of socialisers. The number of socialisers is limited only by external factors, or
the presence of killers masquerading as socialisers. If the population of socialisers
drops below a certain critical level, then the chain reaction reverses and almost all the
players will leave, however only events outside the MUD would cause that to happen
once the critical mass had been reached.

3) A MUD where all groups have a similar influence (although not necessarily similar
numbers). By nurturing explorers using software means (i.e. giving the game great depth
or "mystique", or encouraging non-explorers to dabble for a while by regularly adding
new areas and features), the overall population of explorers will gradually rise, and the
killer population will be held in check by them. The killers who remain do exert an
influence on the number of socialisers, sufficient to stop them from going into
fast-breeder mode, but insufficient to initiate an exodus. Achievers are set upon by
killers often enough to feel that their achievements in the game have meaning. This is
perhaps the most balanced form of MUD, since players can change their position on the
interest graph far more freely: achievers can become explorers, explorers can become
socialisers, socialisers can become achievers - all without sacrificing stability. However,
actually attaining that stability in the first place is very difficult indeed; it requires not only
a level of game design beyond what most MUDs can draw on, but time and player
management skills that aren't usually available to MUD administrators. Furthermore, the
administrators need to recognize that they are aiming for a player mix of this kind in
advance, because the chances of its occurring accidentally are slim.

4) A MUD with no players. The killers have killed/frightened off everyone else, and left to
find some other MUD in which to ply their trade. Alternatively, a MUD structured
expressly for socialisers never managed to acquire a critical mass of them.

Other types could conceivably exist, but they are very rare if they do. The dynamics
model is, however, imprecise: it takes no account of outside factors which may influence
player types or the relationships between then. It is thus possible that some of the more
regimented MUDs (e.g. role-playing MUDs, educational MUDs, group therapy MUDs)
have an external dynamic (e.g. fandom interest in a subject, instructions from a
teacher/trainer, tolerance of others as a means to advance the self) which adds to their
cohesion, and that this could make an otherwise flaky configuration hold together. So
other stable MUD forms may, therefore, still be out there.

It might be argued that "role-playing" MUDs form a separate category, on a par with
"gamelike" and "social" MUDs. However, I personally favor the view that role-playing is
merely a strong framework within which the four types of player still operate: some
people will role-play to increase their power over the game (achievers); others will do so
to explore the wonder of the game world (explorers); others will do so because they
enjoy interacting and co-operating within the context that the role-playing environment
offers (socialisers); others will do it because it gives them a legitimate excuse to hurt
other players (killers). I have not, however, undertaken a study of role-playing MUDs, and
it could well be that there is a configuration of player types peculiar to many of them
which would be unstable were it not for the order imposed by enforcing role-play. It
certainly seems likely that robust role-playing rules could make it easier for a MUD to
achieve type 3) stability, whatever.

At this point, we return to the social/gamelike MUD debate.

Ignoring the fourth (null) case from the above, it is now much easier to see why there is a
schism. Left to market forces, a MUD will either gravitate towards type 1) ("gamelike")
or type 2) ("social"), depending on its administrators' line on player-killing (more
precisely: how much being "killed" annoys socialisers). However, the existence of type
3) MUDs, albeit in smaller numbers because of the difficulty of reaching the steady
state, does show that it is possible to have both socialisers and achievers co-existing in
significant numbers in the same MUD.

It's very easy to label a MUD as either "hack-and-slash" or "slack-and-hash", depending
on whether or not player-killing is allowed. However, using player-killing as the only
defining factor in any distinction is an over-generalization, as it groups together type 1)
and type 3) MUDs. These two types of MUD should *not* be considered as identical
forms: the socializing which occurs in a type 3) MUD simply isn't possible in a type 1),
and as a result the sense of community in type 3)s is very strong. It is no accident that
type 3) MUDs are the ones preferred commercially, because they can hold onto their
players for far longer than the other two forms. A type 1) MUD is only viable
commercially if there is a sufficiently large well of potential players to draw upon,
because of the much greater churn rate these games have. Type 2)s have a similarly
high turnover; indeed, when TinyMUD first arrived on the scene it was almost
slash-and-burn, with games lasting around six months on university computers before a
combination of management breakdown (brought on by player boredom) and resource
hogging would force them to close down - with no other MUDs permitted on the site for
perhaps years afterwards.

This explains why some MUDs perceived by socialisers to be "gamelike" can actually
be warm, friendly places, while others are nasty and vicious: the former are type 3), and
the latter are type 1). Players who enter the type 3)s, expecting them to be type 1)s, may
be pleasantly surprised (Bruckman, 1993). However, it should be noted that this initial
warm behavior is sometimes the approach used by administrators to ensure a new
player's further participation in their particular MUD, and that, once hooked, a player
may find that attitudes undergo a subtle change (Epperson, 1995).

As mentioned earlier, this paper is not intended to promote any one particular style of
MUD. Whether administrators aim for type 1), 2) or 3) is up to them - they're all MUDs,
and they address different needs. However, the fact that they *are* all MUDs, and not
"MU*s" (or any other abbreviation-of-the-day), really should be emphasized.

To summarize: "gamelike" MUDs are the ones in which the killer-achiever equilibrium
has been reached, i.e. type 1); "social" MUDs are the ones in which the pure-social
stability point has been reached, i.e. type 2), and this is the basis upon which they differ.
There is a type 3) "all round" (my term) MUD, which exhibits both social and gamelike
traits, however such MUDs are scarce because the conditions necessary to reach the
stable point are difficult or time-consuming to arrange.


Earlier, the effect of taking each axis on the interest graph to its extremes was used to
give an indication of what would happen if a MUD was pushed so far that it lost its
MUDness. It was noted, though, that along the axes was not the only way a MUD could
be tilted.

What would happen if, in an effort to appeal to certain types of player, a MUD was
overcompensated in their favor?

Tilting a MUD towards achievers would make it obsessed with gameplay. Players would
spend their time looking for tactics to improve their position, and the presence of other
players would become unnecessary. The result would be effectively a single-player
adventure game (SUD?).

Tilting towards explorers would add depth and interest, but remove much of the activity.
Spectacle would dominate over action, and again there would be no need for other
players. The result of this is basically an online book.

Tilting towards socialisers removes all gameplay, and centers on communication.
Eventually, all sense of the virtual world is lost, and a chatline or IRC-style CB program

Tilting towards killers is more difficult, because this type of player is parasitic on the
other three types. The emphasis on causing grief has to be sacrificed in favor of the
thrill of the chase, and bolstered by the use of quick-thinking and skill to overcome
adversity in clever (but violent) ways. In other words, this becomes an arcade ("shoot
'em up") type of game.

It's a question of balance: if something is added to a MUD to tilt the graph one way,
other mechanisms will need to be in place to counterbalance it (preferably
automatically). Otherwise, what results is a SUD, book, chatline or arcade game. It's the
*combination* that makes MUDs unique - and special. It *is* legitimate to say that
anything which goes too far in any direction is not a MUD; it is *not* legitimate to say
that something which doesn't go far enough in any direction is not a MUD. So long as a
system is a (text-based) multi-user virtual world, that's enough.


To answer the questions posed in the preface:

Are MUDs

games? Like chess, tennis, D&D?
Yes - to achievers.
pastimes? Like reading, gardening, cooking?
Yes - to explorers.
sports? Like huntin', shooting', fishin'?
Yes - to killers.
entertainments? Like nightclubs, TV, concerts?
Yes - to socialisers.


[1] This paper is an April 1996 extension of an earlier article, "Who Plays MUAs" (Bartle,
1990a). As a result of this, and of the fact that I am not a trained psychologist, do not
expect a conventionally rigorous approach to the subject matter.

Permission to redistribute freely for academic purposes is granted provided that no
material changes are made to the text. [2] In the figure below, green indicates increasing
numbers and red indicates decreasing numbers. A red line with a green arrowhead
means that decreasing numbers of the box pointed from lead to increasing numbers of
the box pointed to; a red line with a red arrowhead would mean that a decrease in one
leads to a decrease in the other, and so on. The thickness of the line shows the strength
of the effect: thin lines mean there's only a small effect; medium lines mean there's an
effect involving roughly equal numbers of players from both boxes; thick lines means
there's a great effect, magnifying the influence of the origin box.


Aspnes, J. (1989). TinyMUD [C]

Bartle, R. A. (1985). MUD2 [MUDDLE] MUSE LTD, Colchester, Essex, UK.

Bartle, R. A. (1990a). Who Plays MUAs? Comms Plus!, October/November 1990

Bartle, R. A. (1990b). Interactive Multi-Player Computer Games. MUSE LTD, Colchester,
Essex, UK ftp://parcftp.xerox.COM/pub/MOO/papers/mudreport.txt

Bruckman, A. S. (1992). Identity Workshop: Emergent Social and Psychological
Phenomena in Text-Based Virtual Reality. MIT Media Laboratory, Cambridge,
Massachusetts. ftp://media.mit.edu/pub/asb/papers/identity-workshop.ps

Bruckman, A. S. (1993). Gender Swapping on the Internet Proc. INET-93

Bruckman, A. S. & Resnick, M. (1993). Virtual Professional Community: Results from
the MediaMOO Project. MIT Media Laboratory, Cambridge, Massachusetts.

Bruckman, A. S. (1994a). Workshop: "Serious" Uses of MUDs? Proc. DIAC-94

Bruckman, A. S. (1994b). Approaches to Managing Deviant Behavior in Virtual
Communities. MIT Media Laboratory, Cambridge, Massachusetts.

Burka, L. P. (1995). The MUDline. http://www.ccs.neu.edu/home/lpb/mudline.html

Carton, S. (1995). Internet Virtual Worlds Quick Tour: MUDs, MOOs and MUSHes:
Interactive games, Conferences and Forums Ventana Press, Chapel Hill, North

Caspian-Kaufman, J. (1995). Sid Meier's CivNET: Instruction Manual Microprose, Hunt
Valley, Maryland.

Cherny, L. (1995a). The Modal Complexity of Speech Events in a Social MUD.
Electronic Journal of Communication, Summer 1995.

Cherny, L. (1995b). The Situated Behavior of MUD Back Channels. Dept. Linguistics,
Stanford University, California. ftp://bhasha.stanford.edu/pub/cherny/aaai.ps

Clodius, J. A. (1994). Concepts of Space in a Virtual Community.

Curtis, P. (1992). Mudding: Social Phenomena in Text-Based Virtual Realities. Proc.
DIAC-92 ftp://parcftp.xerox.com/pub/MOO/papers/DIAC92.txt

Curtis, P. & Nichols, D. A. (1993). MUDs Grow Up: Social Virtual Reality in the Real
World. Xerox PARC, Palo Alto, California.

Dibbell, J. (1993). A Rape in Cyberspace. The Village Voice, December 21, 1993.

Emert, H. G. (1993). "X" Marks the Spot. East Stroudsburg University, Pennsylvania.

Eddy, A. (1994). Internet After Hours Prima, Rocklin, California.

Epperson, H. L. (1995). Patterns of Social Behaviour in Computer-Mediated
Communications. Dept. Sociology, Rice University.
http://www.eff.org/pub/Net_culture/Misc_net_culture/ web_social_behaviour.paper

Evard, R. (1993). Collaborative Networked Communication: MUDs as System Tools.
Proc. LISA-93 http://www.ccs.neu.edu/home/remy/documents/cncmast.html

Fanderclai, T. F. (1995). MUDs in Education: New Environments, New Pedagogies.
Computer-Mediated Communication Magazine, 2(1), 8.

Farmer, F. R., Morningstar, C. & Crockford, D. (1994). From Habitat to Global
Cyberspace. Proc. CompCon-94, IEEE

Kort, B. (1991). The MUSE as an Educational Medium BBN Labs, Cambridge,
Massachusetts. ftp://musenet.bbn.com/pub/micromuse

Mauldin, M. L. (1994). Chatterbots, TinyMUDs and the Turing Test: Entering the
Loebner Prize Competition. Proc. AAAI-94 http://fuzine.mt.cs.cmu.edu/mlm/aaai94.html

Moock, C. (1996). Virtual Campus at the University of Waterloo.

Norrish, J. (1995). MU*s. http://www.vuw.ac.nz/~jamie/mud/mud.html

Poirier, J. R. (1994). Interactive Multiuser Realities: MUDs, MOOs, MUCKs, and
MUSHes. The Internet Unleashed, 1192-1127. SAMS Publishing, Indianapolis, Indiana.

Reid, E. (1994). Cultural Formations in Text-Based Virtual Realities. Dept. English,
University of Melbourne, Australia.

Riner, R. D. & Clodius, J. A. (1995). Simulating Future Histories: The NAU Solar
System Simulation and Mars Settlement. Anthropology & Education Quarterly
26(1):95-104. http://tinylondon.ucsd.edu/~jen/solsys.html

Rosenberg, M. S. (1992). Virtual Reality: Reflections of Life, Dreams and Technology.
An Ethnography of a Computer Society.

Roush, W. (1993). The Virtual STS Centre on MediaMOO: Issues and Challenges as
Non-Technical Users Enter Social Virtual Spaces. MIT Media Laboratory, Cambridge,
Massachusetts. ftp://media.mit.edu/pub/MediaMOO/Papers/STS-Centre

Urdang, L. & Manser, M. (1980). The Pan Dictionary of Synonyms and Antonyms Pan
Reference, London, UK.

Whitlock, T. D. (1994). Fuck Art, Let's Kill!: Towards a Post Modern Community.

Whitlock, T. D. (1994b). Technological Hierarchy in MOO: Reflections on Power in