Friday, January 28, 2005

Bartle's Player Types and Keirsey's Temperaments


Some time after discovering the four types of virtual world game players as described in "Hearts, Clubs, Diamonds, Spades: Players Who Suit MUDs" by Richard Bartle, it seemed to me that I might be able to relate those four types to the four more general temperaments as described by David Keirsey in Please Understand Me.

Over the years since then, I've worked out some of the correspondences I think I can see. As a result I've come to the conclusion that the four Bartle Types are indeed aspects of the four Keirsey Temperaments expressed in an online game context. Obviously there are a lot of assumptions built into that statement, so it seems to me that if I'm going to use this correspondence of models as a basis for subsequent ideas about player-centered game design, I need to spell out why I buy into these ideas -- not so much to persuade others as to document my observations and let others decide for themselves if they think there's anything of value here.

Before I get into the details, however, I'd like to note that any and all references to the work of both Richard Bartle and David Keirsey that I make here that aren't clearly identified as quotes are purely my own interpretations, and are not to be considered official explications of these authors' ideas. If I fail to correctly or completely capture some aspect of meaning intended by either author, the error is mine and should not be attributed to either Bartle or Keirsey. Basically, if you think any of my ideas here are nuts, don't blame Keirsey or Bartle -- blame me. (And this includes the point that the four-type model I use here isn't the most current model preferred by Richard as described in his book Designing Virtual Worlds. I hope I can be forgiven for using the Web-available four-type version, which I believe still has considerable explanatory power.)


I start with a number of supporting observations:

  • Type classification systems aren't the only way to to study human personality -- they're just one useful way.
  • An example that uses multiple codes (like the Bartle Types) is the Holland Codes assessment for job placement.
  • No system for classifying human personality will ever be perfect. People are complex; no one is purely one type only, nor can any theoretical system or model fully describe an actual person.
  • A classification system doesn't have to be perfect to be useful; it just has to have some reliable explanatory and predictive power.
  • "Temperament" is a form of typing that considers the fundamental, distinct, and possibly innate motivations behind behavior.
  • Many different theories of temperament have been suggested over thousands of years.
  • Many of these temperament theories (by Hippocrates, Plato, Aristotle, Paracelsus, Spranger, etc.) perceive four temperaments.
  • These models all tend to have certain expressions of human nature in common.
  • Jung's theory of archetypes and the Myers-Briggs type system based on Jung's work rationalized these common traits.
  • David Keirsey's temperament model is a modern synthesis of past models organized around Myers-Briggs empirical data.
After I encountered Richard Bartle's original system of four types of players of virtual-world games (Killers, Achievers, Explorers, and Socializers), it immediately appeared to me that these four player types are strongly analogous to the four general temperaments that Keirsey describes (Artisans, Guardians, Rationals, and Idealists).

To see how I got there, let's look at the foundations of Keirsey's temperament theory.

The Myers-Briggs model posits 16 types, based on four dimensions of personality. The four dimensions (and the two poles of each dimension, along with the letter abbreviation) are:

  • Need for social interaction: Introversion vs. Extraversion (I/E)
  • Source of truth: iNtuition vs. Sensing (N/S)
  • Decision-making style: Thinking vs. Feeling (T/F)
  • Time focus: Judging vs. Perceiving (J/P)
Combining all possibilities yields sixteen "types":


These are the sixteen personality types as recognized by Myers-Briggs type theory.

Through years of observation and years of psychological practice, David Keirsey concluded that there were certain strong relationships among some of these sixteen types, and that these relationships clustered into four groups. Eventually he developed a theoretical model that groups the sixteen types into four groups of four temperaments, and for convenience he gave each temperament a descriptive name:


These four temperaments -- Artisan, Guardian, Rational, and Idealist -- are the four temperaments I believe correspond to the original four player types -- Killer, Achiever, Explorer, and Socializer -- as described by Bartle.


Excessively brief portraits of Keirsey's temperaments are:

  • Artisan: realistic, tactical, manipulative (of things or people), pragmatic, impulsive, action-focused, experience-seeking
  • Guardian: practical, logistical, hierarchical, organized, detail-oriented, possessive, process-focused, security-seeking
  • Rational: innovative, strategic, logical, scientific/technological, future-oriented, result-focused, knowledge-seeking
  • Idealist: imaginative, diplomatic, emotional, relationship-oriented, dramatic, person-focused, identity-seeking
In the second edition of his primary work, Please Understand Me II, Keirsey groups his four temperaments according to their use of tools ("cooperative" or "utilitarian") and words ("abstract" or "concrete"):












There's nothing wrong with that system, but by the time Keirsey proposed it I had already worked out an arrangement I thought (and still think) better reflects how people interact with the world. Rather than two dimensions of tool-use and word-use, I think of the two most important dimensions of behavior as being "internals vs. externals" and "change vs. structure."










This structure has an additional feature: Each temperament is most unlike (and usually misunderstands or even opposes) the temperament diagonal to it in ways that are instinctively familiar to most people.

Thus, Artisans (who seek External Change) tend to perceive Rationals (who seek Internal Structure) as ineffective creators of imaginary and useless ideas, while Rationals see Artisans as ignorant, energy-wasting jocks and gamblers.

Similarly, Guardians (who seek External Structure) often regard Idealists as crusading liberal artistes, while Idealists (who seek Internal Change) see Guardians as boring, bourgeois reactionaries.

(Note that none of these perceptions is really accurate -- they're all the most negative, stereotyped view of those whose ways of looking at the world are different from our own. But they're also very common, given that most of us are strongly inclined to believe that our particular way of looking at the world is the "right" way, and therefore all others are, to various degrees, wrong.)


Similar to the table showing Keirsey's four temperaments, here is the table showing the four Bartle types (rotated 90 degrees clockwise from the way they're presented in "Players Who Suit MUDs").















I see these two graphs -- the four Keirsey temperaments and the four Bartle types -- as isomorphic. I believe they both describe the same thing, the same four fundamental temperaments, just in different contexts. The four general temperaments (Artisan, Guardian, Rational, Idealist) are expressed as the four Bartle Types (Killer [Manipulator], Achiever, Explorer, Socializer) when someone participates in a massively multiplayer gameworld.

Here's a chart that shows those correspondences.









Acting [on]








Acting [on]








[with] World







[with] Players


Considering Bartle's descriptions of Achievers, Explorers, Socializers and Killers, and comparing these descriptions against Keirsey's descriptions of the four temperaments in his model, and further noting the congruence between the 2x2 structure I saw in Keirsey's model versus Bartle's diagram, the four Bartle types give every appearance of being context-specific analogues to Keirsey's four temperaments.

First, let's consider the dimensions of behavior that determine the 2x2 organizations of player types and temperaments.

In "Players Who Suit MUDs", I interpret Bartle's use of the terms ACTING and INTERACTING as "doing" and "learning about," respectively. (Bartle uses the terms "doing-to" and "doing-with" to describe ACTING and INTERACTING, but he also seems to suggest that INTERACTING is more about understanding the properties of things than actually using those things.) INTERACTING corresponds to my notion of INTERNAL in that interacting with a thing is something you do to discover the internal nature of that thing. This contrasts with the concept of ACTING with or on a thing, which is an EXTERNAL usage of a thing.

The case for PLAYERS and WORLD as analogous to CHANGE and STRUCTURE is less clear, but the key can be found in remembering that PLAYERS/WORLD are concepts appropriate for a game context, while a CHANGE/STRUCTURE model is intended to apply to a larger set of human behaviors. In this case, the concept of PLAYERS is a special case of CHANGE, while WORLD is a special case of STRUCTURE. CHANGE happens to be the word I used for the pole of the axis that describes how much or how little order/control a person needs or wants, but "freedom" and "opportunity" are also words that could convey this meaning... and these are all attributes that are unique to human players in a virtual world. Conversely, STRUCTURE/ORDER/BOUNDARIES are attributes of WORLD objects, which must all predefined by the game's creators.

As Artisans (External Change) demand to be free to manipulate the people in their environment as they will, Killers (ACTING on PLAYERS) won't play if they are denied power over other players. Socializers as game-specific cases of Idealists and Achievers as game-specific cases of Guardians follow the same interpretations.


Let's look at each of the types and temperaments, starting with the Rationals/Explorers.

Rationals play the same way they do everything else -- they look for patterns behind the raw data. These can be patterns in space (as in geography) or patterns in time (as in morphology). Or they can be cause-and-effect patterns (entailment) or relationship patterns (connections). Ultimately, it's all about achieving a strategic understanding of the system as a whole thing.

Compare that to Bartle's description of Explorers: "The real fun comes only from discovery, and making the most complete set of maps in existence."

Socializers, meanwhile, are described as "... interested in people, and what they have to say. ... Inter-player relationships are important ... seeing [people] grow as individuals, maturing over time. ... The only ultimately fulfilling thing is ... getting to know people, to understand them, and to form beautiful, lasting relationships."

This sounds very similar to the Keirseian description of Idealists, who interact with other people as part of their lifelong journey of self-discovery. In a way, the highly imaginative Idealists are always roleplaying; they are constantly creating images of themselves (or others) that they feel they should model through their own actions.

Although Guardians can be very interested in forming relationships, they do so for very different reasons than the Idealist. For the Guardian, the world is an insecure place, so it's necessary to protect oneself by accumulating material possessions... just in case. Thus the Guardian focuses on earning money, on buying nice things and maintaining them, on forming stable and formal group relationships, and generally on working hard to acquire possessions.

I see strong echoes of this motivation in Achievers. As Bartle puts it: "Achievers regard points-gathering and rising in levels as their main goal" and "Achievers are proud of their formal status in the game's built-in level hierarchy, and of how short a time they took to reach it."

Finally, there are the Killers. These can be difficult to understand because most virtual worlds have encoded rules that marginalize their playstyle. However, they can be understood as distinct from the other player types through Bartle's description of them as ACTING on PLAYERS -- in short, they enjoy manipulating the participants in the game. "Killers get their kicks from imposing themselves on others." He also points out that Killers "wish only to demonstrate their superiority over fellow humans, preferably in a world which serves to legitimise actions that could mean imprisonment in real life."

This aligns nicely with the Keirseian description of Artisans, who (as their temperament name suggests) delight in skillful manipulation of their environment. These are the tool-users, the combat vehicle pilots, and the negotiators par excellence. They instinctively find and exploit advantages in any tactical situation, and they express this need for dominance of their world in order to retain the greatest amount of personal freedom possible.

(I should mention here that I prefer the term "Manipulator" over "Killer" as I think "Manipulator" captures more of the motivation behind that sensation-seeking gameplay style. Although player-killing as a griefing tactic is undesirable, there are aspects of the Killer personality that could add to a game world rather than detract from it if properly recognized and enabled through appropriate gameplay features. But we're not going to get there if the pejorative term Killer is used to describe those who excel at the tactical manipulation of their environment. So I'd prefer to use the term Manipulators, but I realize I'm fighting established convention here.)


One of the steps I had to take in evaluating this theory of a correspondence between player types and temperaments was whether I was merely seeing what I wanted to see. Was temperament theory my hammer, and Bartle's four player types a handy nail?

The obvious test is to think about a particular Bartle type and consider whether it might be equally or better applied to any other temperament. Take Explorers, for example. Although they can be people-oriented, they're not as interested in social interaction for its own sake as the Idealists. Nor are they so concerned with operations and maintenance or with position in a formal hierarchy as the Guardians are. And even though they will occasionally demonstrate great skill in manipulating people or things, unlike the Artisan they'd much rather discover knowledge than apply it.

A similar analysis can be conducted for the Socializers, Killers and Achievers. Although there are aspects of behavior and motivation common to many kinds of people, each of these three types has many more things in common with one of Idealist, Artisan, and Guardian temperaments respectively. Everyone socializes, but Idealists cannot not socialize; likewise, Artisans are the most natural Killers while Guardians can't help but Achieve.

Finally, a word must be said about the idea of "temperament" and "personality type" overall. Nick Yee has consistently expressed strong doubt about the value of personality models that identify clusters of behavior patterns. Instead, he defends models that offer numerous behavioral preferences, any of which can be expressed simultaneously by anyone. While I believe there may be expressive value in the multiple-mode models Nick supports, he has so far not been willing to similarly endorse models that identify two axes of behavior and derive four "types" from the four quadrants... such as Bartle's player types or Keirsey's temperaments.

His technical doubts notwithstanding, I've personally found both the player types and temperament models to have expressive value. Each in its context has successfully helped me understand why people behave the way they do. Neither model perfectly represents the totality of human behavior, but what model ever can? I don't need perfection, I just need good enough -- and both Bartle's player types and Keirsey's temperaments have demonstrated repeatedly that they are good enough in explaining behavior to be useful, both analytically and predictively. If statistical rigor doesn't back up this empirical finding, then there's something wrong with the design of the statistical test.

But that's my reaction. You should decide for yourself whether you think there's anything to either Bartle's types or Keirsey's temperaments, and if so whether I'm seeing something real when I perceive a congruence between the two models of personality.


  1. You might be interested in some stuff I've been noodling on for awhile. I think that people who have misgivings about these kinds of models are concerned that they're too confining to express the diversity of human affective/conative cognition. I think they are right on one level, but still, *something* is going on that a *lot* of 4-quadrant models seem to be touching upon in a lot of different fields. I have many more examples of such models than currently exist on my wiki, but it's a start. There are a lot of blind men arguing about this elephant.

    1. Thanks, Neil -- as it is, that's quite a list! I can see I have some additional reading to do. :)

      Since writing this piece I've since expanded it considerably. There's a full version I hope to have on this blog soon; in the meantime, you may find the condensed (but still pretty complete) version published by the Gamasutra computer game website to be useful:

      Also, when reading a list of any kind it's nigh onto impossible not to immediately see the things that aren't there that might be. There are several books that would, I think, be very good fits for the various categories on your PAEI list, such as _Living Systems_ by James Grier Miller and _The Evolution of Cooperation_ by Robert Axelrod.

      But sticking strictly to works that describe or at least reference fourfold models, I'm a little surprised that Charles Handy's _Understanding Organizations_ is not listed. I'm even more surprised not to see David Keirsey's _Please Understand Me_ given that the Myers-Briggs Type Indicator is referenced along with Linda Beren's _The Four Temperament Patterns_ which (like my model) depends greatly on Keirsey's book.

      That said, these are quibbles. Your effort to identify works that reference the remarkably popular occurence of fourfold models is impressive. Thanks for taking the time to comment!

  2. Excellent article!!! Very useful, I was trying to find this a long time ago.