I recently witnessed a certified MBTI practitioner giving a brief explanation of the major differences, from his point of view, between Keirsey's system and the MBTI. In the course of his presentation he made a very interesting observation about a curious decision that Keirsey made in building his theory, but did not choose to pursue the matter in greater detail. Had he at that point turned his attention to the principles underlying Keirsey's choice, he would have discovered a strategy - a specific move that can be defined with a degree of precision that is almost mathematical - that is very interesting indeed, because it is generalizable, and has wider applications in any field that deals with classification systems. I shall call it the 'Keirsey Stratagem', not because he invented it, but because he has made the most conspicuous and profitable use of it in the field of personality typology.

In this paper I will show -

1. how the 'Keirsey Strategem' can be used to generate a number of other possible personality typologies;
2. how a clear understanding of it can help us to evaluate the strengths and weaknesses of Keirsey's system, and to understand how it drives some of the recent theoretical moves that Keirsey makes in his new book; and
3. how the strategem can be used to understand the relationship between the Enneagram and the MBTI - in particular, the distribution of MBTI types across the Enneagram.

Section One: The Puzzle

skip to section two

Most individuals who are reading this piece will know that when Keirsey and Bates introduced their system in 1978, in a book called Please Understand Me: Character & Temperament Types, they relied heavily on the MBTI classification system. They freely utilized the MBTI four-letter nomenclature (INFJ, INFP, etc) throughout their book. They also presented a fifty-page appendix, with sixteen sections, each devoted to a description of one of the MBTI types. Furthermore, they included in the book a test that could be self-administered, the 'Keirsey Temperament Sorter'. An easily accessible alternative to the Myers-Briggs Type Indicator (MBTI), it purportedly assists the test subject in classifying him/herself as one the 16 MBTI types.

Most readers will also know that in presenting the MBTI characteristically groups their sixteen personality types in sets of four that are sometimes referred to as the four 'core types' - namely, 'NF', 'NT', 'SF', and 'ST':

Many readers will also realize that in their first book Keirsey and Bates departed from the MBTI practice of parsing the 16 types in the above-mentioned manner, utilizing instead a slightly different breakdown into four groups of four types - the 'NF', 'NT', 'SP', and 'SJ'.

So I can't imagine that many readers will be surprised to learn that when the MBTI practitioner that I mentioned in the introduction to this piece compared the two systems, he immediately honed in on this difference. Pointing to the Keirsey groups, he exclaimed, 'It does not make sense!', and proceeded to explain that whereas 'NF' and 'NT' each refer to combinations of functions (ie, the 'N', 'F' and 'T'), Keirsey's 'SP' and 'SJ' do not - they combine a function (S) with an 'attitude' (i.e., 'J' or 'P'). He left it up to the imagination of the listener to fill in the missing steps in the argument, verifying for themselves that Kiersey's move breaks with our expectations about what makes sense.

But lets choose not to move on so quickly, and linger here for a minute or two to consider this matter more thoroughly. Who amongst us, on first seeing the four labels - NF, NT, SP and SJ - has not wondered about the legitimacy of the this way of parsing the groups, even if our doubts are manifest only in a fleeting double-take and a vague feeling of discomfort in the presence of a possible anomaly? Do we stop to grapple further with the question? Most wince, shrug, table the matter, and move on. A closer look at what is going on will prove enlightening.

Section Two: The 'Keirsey Stratagem'

skip to section three

What is it about the MBTI groupings that seem logically or aesthetically satisfying, comparatively speaking? What is it about the Keirsey groupings that gives pause, even if only momentarily? Is there something logically inconsistent about about Keirsey's groups?

The MBTI group (NF, NT, ST, SF) lists four classes or sets that are mutually exclusive (there is no overlap) and exhaustive (there is no additional set, missing from the list). It is thus a clearly demarcated and complete grouping or classification. Does Keirsey's grouping meet these requirements? If he had taken the 16 MBTI groups and parsed them in the following way -

NF, NT, ST, SJ

- his system would have overlapping groups (ST and SJ), groups that are not mutually exclusive. And this would fail to meet the criteria established above. We could say that such a breakdown is thus not a logical one, or perhaps that it 'doesn't make sense'.

Or if he had parsed groups in the following way -

SP, SJ, EN, INF

- his system would not include all 16 types (ie, the INTJs and INTPs would be missing). It would thus not be an 'exhaustive' set and this attempt to parse the groups, because it does not meet the criteria established above.

But, interestingly, Keirsey's four groups - NF, NT, SP, SJ - ARE mutually exclusive and exhaustive groups! The following chart clearly shows this -

NF	NT	SJ	SP
INFJ	INTJ	ISTJ	ISTP
INFP	INTP	ISFJ	ISFP
ENFJ	ENTJ	ESFJ	ESFP
ENFP	ENTP	ESTJ	ESTP

So the Keirsey parsing does 'make sense', at least in a way in which the two alternative groupings described above DON'T. This fact by itself, however, is just a mildly interesting curiosity. But when one stops to realize that Keirsey did more than parse the groups in an interesting fashion - he attributed an important additional 'meaning' to the groups by associating them with the 'four temperaments' of antiquity. Recognizing this, one begins to see that this parsing strategy of Keirsey's was indeed a brilliant maneuver, deserving of the appellation, 'stratagem'. For it permitted him to continue to utilize the 16 MBTI types while clearly DISTINGUISHING his system from the MBTI in a rather significant way! By cleverly parsing the MBTI types in this unique way and throwing emphasis on the larger groupings, as opposed to the sixteen individual 'types', a new typology was born. Never mind, for the moment at least, that some would dampen our enthusiasm for Keirsey's solution by pointing out that Jung and others have understood the four temperaments in a way that resists correlation with the NF, NT, SP, and SJ groups.

Motivated by our newfound appreciation for how Keirsey tickled new truths out of the MBTI by using a logical stratagem, we might wonder what would prevent some ambitious young personality theorist from squeezing similar pearls of wisdom, as yet undiscovered, from the haggard old Jungian/MBTI cow. What other logical possibilities remain for re-grouping the types?

How about separating them into - NP, NJ, SF and ST groups? This grouping meets the criteria (mutually exclusive and exhaustive groups). But what possible 'meaning' could we attribute to this ways of drawing the lines? I leave this conundrum, along with the puzzle of unearthing of the other possibly meaningful and consistent groupings of the 16 types up to the curious and industrious reader.¹

Before basking too long in the vicarious afterglow of Keirsey's apparently brilliant success, lets take a closer look at the groupings that he arrived at. How, specifically, do they differ from the 'core type' groupings? And why do they still smack of 'trickery' of some kind, despite all that we have said above?

Lets makes the situation less abstract by creating an analogy. To each of the letters involved, we will assign another concrete meaning, in the following way -

F= fat
T= skinny
N= female
S= male
J= poor
P= rich

There are four types of people - fat men (SF), skinny men (ST), fat women (NF) and skinny women (NT)

There are four types of people - fat women (NF), skinny women (NT), rich males (SP) and poor males (SJ)

Although the grouping is logically consistent - describing a set with mutually exhaustive and exclusive members - it does in fact seem to be 'mixing apples and oranges'. One is likely to ask, 'What about the rich women (NP) and poor women (NJ), and the fat men (SF) and skinny men (ST)?!' And this question leads us directly into the topic of the next section.

Section Three: The Drawbacks of Keirsey's System

skip to section four

Just as it occured to us to ask, in the context of the analogy presented in the previous section, 'What about the rich women and poor women, and the fat men and skinny men?', we might ask Keirsey, 'But what about the STs and SFs, let alone the NJs and NPs?'

Certain very important categories - the ST and SF in particular - are in effect rendered virtually invisible by Keirsey parsing. When, in following Keirsey, our emphasis shifts from the 16 categories (the 'types') to the four groupings (the Keirsey 'temperaments'), it is like using a different set of lenses. Although we still have the 'NF' lens and the 'NT' lens in the Keirsey system, we no longer use the 'ST' and 'SF' lenses, for which the 'SP' and 'SJ' lenses have been substituted. Is this at least a fair trade?

The ST and SF lenses will be ones that are sorely missed, theoretically speaking - especially when we attempt to eek culturally relevant conclusions out of Type demographics. We will feel the loss particularly strongly for two reasons -

• Because the ST type is the most populated amongst the four MBTI 'core type' and the one that can be taken as 'prototypical' of our culture. The rational-empirical ST type, diametrically opposed to the NF, is rendered 'un-available' as a concept in Keirsey's system, by virtue of his way of parsing the types.

• The ESTJ is the most populated male category and the ESFJ the most populated female category in Western society, and what is culturally considered 'masculine' and 'feminine' generally follow 'ST' and 'SF' lines respectively. The ST-SF dichotomy is obscured by the (NF, NT, SJ, SP) grouping in Keirsey's system, which makes the system less powerful as a tool for use in cultural analyses that seek to arrive at gender-specific conclusions.

Using demographics that employ the MBTI 'core type' categories, for instance, Pat and I found that it is relatively easy to show that 'familial' forms of organization preferred by the 'SF', a form of organization that parallels culturally 'feminine' assumptions about what it is to be 'socially organized', are underrepresented and undervalued in our society. As one travels up the corporate ladder, the concentration of ESFJs drastically reduces, at a rate that is inversely proportional to their presence in the population at large. These are very interesting observations, which it would not have been possible to make using Keirsey's system, which renders 'SF' and 'ST' invisible by a slight-of-hand that is equivalent to typological gerrymandering.

There are numerous additional consequence of Keirsey's way of parsing MBTI types that may be less obvious, but are nonetheless significant. I shall deal with only two of them in the remaining portion of this section.

1) Opposites

First, whereas amongst the four MBTI groups (SF, NF, NT, and ST) we can locate two pairs of diametrically opposite types (the SF and NT are diametrically opposed, by definition, as are the NF and ST), there are no diametrically opposed pairs amongst Keirsey's group of four - the SP, SJ, NT and NF. This is because the diametrical opposite to each of the members of Keirsey's set of four is not included in the set. The diametrical opposite to NT, which is SF, is not included in the group. The diametrical opposite to NF, which is ST, is not included. Neither are the diametrical opposites of SJ (NP) or SP (NJ).

If the reader gives this matter a moments thought she will realize that this is a distinct drawback for a typology, one which is related to our criticisms above that the 'ST' and 'SF' have been rendered 'invisibile' in Keirsey's grouping. That Keirsey himself misses this feature (of diametrically opposite relations between the members of the primary set of four) is attested to by the fact that in his new book (Please Understand ME II: Temperament, Character, and Intelligence, page 83), he attempts to show - by a rather clever but spurious argument - that the NT and SJ are in fact 'diametrical opposites', as are the NF and SP!

We might call the strategy that Keirsey uses in order to arrive at this conclusion Keirsey's 'second' stratagem, or 'the Keirsey Stratagem II'. It is tantamount to a theoretical sleight of hand in which Keirsey crosses the line that distinctly separates him from his previous efforts to maintain an approach that is CONSISTENT with the MBTI. As we have seen above, the four groups that he identified in his first book were carefully crafted so that the groups could easily be mapped onto the 16 MBTI types. The argument that he launches in his new book, however, which seeks to establish that NT and SJ (and NF and SP) are 'diametrical opposites', illustrates that he is using these familiar labels in a revisionist way that will only create confusion with individuals habituated to the way they are used in the MBTI.

To see this clearly, consider the analogy we created above. To call the NT and SJ 'diametrically opposite' is like saying the 'skinny female' (=NT) is the diametrical opposite to the 'poor male'(=SJ). One can only wonder how this could possibly qualify as an instance of diametrical opposition. Realizing that there is a problem here, Keirsey invokes a new set of traits which he can point to in order to prove that the SJ and NT are exact opposites. All one has to do in order to see that Keirsey's argument involves a real 'sleight of hand' is to ask the question, "How can NF and ST be 'diametrically opposite' (as the MBTI claims) and the NF and SP ALSO be 'diametrically opposite' (as Keirsey now claims)". The only way this can be so is if Keirsey now means something DIFFERENT by the designations 'NF', and so forth. And this, in fact, is the case - although Keirsey's presentation does not make that clear.

In Keirsey's new book, the 'NF', to use one of the core types as an example, is no longer defined in Keirsey by reference to 'N' and 'F' - ie, to the 'four function' theory that Pat and I have called the Jungian 'infrastructure' of MBTI theory. He has replaced this infrastructure with another, which provides his four 'temperaments' (unfortunately still labeled NF, NT, SP and SJ) with their meaning. The new infrastructure, not unlike the Jungian one in this respect, is comprised of four components, four so-called 'intellects' - the Logistical Intellect, the Tactical Intellect, the Strategic Intellect, and the Diplomatic Intellect. Each 'temperament', as Walter Geldart brought my attention to in his review of the new Keirsey book in the current issue of this Journal, can now be defined by how individuals of that temperament rank their preference for these 'intellects'.

As Walter points out, these are remarkably similar in structure to the 'functional preference orders' that define MBTI type. For instance, whereas the INFJ is the introverted MBTI type that is defined by an N-F-T-S functional preference order, the Keirsey idealist temperament (which he continues to call the 'NF') is defined by prefering Diplomacy over Strategy, Strategy over Logistics, Logistics over Tactics - which we could abbreviate using the following formula - 'Dip-Strat-Log-Tac'.

Note that Keirsey is inserting his new 'infrastructure' at a level of theory that is once removed from the level at which the 'four functions' reside. The four functions underwrite the 16 types - by rearranging 'functional preference orders' in different permutations, one gets all 16 MBTI types. The Keirsey theory of 'intellects' underwrites the four groups into which he parses the 16 types, so that by rearranging 'intellect preference orders' in different permutations he arrives at the four temperaments - which are now defined in the following way -

Temperament	Intellect preference order
SP	Tac-Log-Strat-Dip
NF	Dip-Strat-Log-Tac
NT	Strat-Dip-Tac-Log
SJ	Log-Tac-Dip-Strat

Look closely at this table. The four variables that underwrite the 16 MBTI types generate EIGHT permutations (which, when combined with the 'I or E' designation, generates 16 Types). Why, then, do permutations of Keirsey's four variable set generate only FOUR 'intellect preference orders'? What happened to 'Tac-Strat-Log-Dip', for example? No explanation is given for why these permutations are left out of the scheme, or why the ones that are included are associated with each of the temperaments in the manner illustrated above.

No explanation is given as to why human beings would not utilize a a Tac-Strat-Log-Dip preference order. We must assume that in Keirsey's theory they do not, however, because if they DID Keirsey has ommitted a viable 'intellect preference order' - and this would mean that there is an additional 'temperament' that he has failed to mention in the list of four that he has specified!!!

There are many aspects of Keirsey's new theory that are left unexplained, not the least of which is how the new parameters (Tac, Strat, Log, Dip) relate to the Jungian parameters (S, T, F, and N). One would hope that he has some mathematical or logically CONSISTENT way of mapping one set onto the other, lest his system, which continues to use the 'S', 'T', 'N' and 'F' labels, fall into internal inconsistency and contradiction. But this is, in fact, dubious - as is revealed by the fact that the expectations that logically follow from the definitions and rules specified in one system, the MBTI (ie, that the NF is 'diametrically opposed' to ST) is explicitly contradicted by conclusions the follow from the definitions and rules of the other system, Keirey's (ie, that the NF is 'diametrically opposed' to the SP).

2) MBTI pairs

The way that Keirsey has grouped the types also forces him to emphasize a different set of pairs than the pairs that Jung (and the MBTI, following him) chose as the significant ones. Look at the MBTI types that fall into each of Keirsey's group of four -

Temperament	MBTI Types in that Group
SP	ISFP, ISTP, ESFP, ESTP
SJ	ISFJ, ISTJ, ESFJ, ESTJ
NF	INFJ, INFP, ENFJ, ENFP
NT	INTJ, INTP, ENTJ, ENTP

Once Keirsey has identified the above four groups as the significant ones, what choice does he really have but to create pairs in which the partners are members of the same group of four? And within those groups, what choice does he really have except to pair those types that differ only in terms of whether they are introverts of extraverts - for example, the ISFP and ESFP? Indeed, this is the only KIND of pair that exists ACROSS ALL FOUR GROUPS! Once we recognize this we see why he has chosen in his new book to de-emphasize the I-E distinction!

This, however, brings him into stark contradiction to Jung, who emphasized the fact that individuals with extraverted iNtuition will be distinctly different from those who have introverted iNtuition, to the extent that it is much more difficult than one might at first assume for one to adopt the perspective of the other. Jung thus considered them 'opposites', indeed, in one sense of the word.

In order to fend off this kind of criticism, Keirsey would have to downplay the importance of the distinction between Ni and Ne - which, in fact, he does. One begins to wonder what will remain of the MBTI system under Keirsey's watch. This is not to say that the maverick direction in which he has chosen to go is wrong, or better or worse than the MBTI. But it does explain why MBTI spokespersons have distanced themselves from the theory by saying, in their new manual, that the two are "separate systems for explaining personality that are independent of each other".

Section Four: The Enneagram

skip to section four

What does any of this have to do with the Enneagram? Does the Keirsey groupings of the 16 MBTI types (NT, NF, SJ, SP) help us to understand the relationship between the Enneagram and the MBTI? No, I don't think so. But if, instead of focusing on the specific groupings - the four 'temperaments' - we look at THE STRATEGY THAT GENERATED THE GROUPINGS, and ask if we can use the same strategy to arrive at a new way of grouping MBTI types that will shed more light on how they distribute across the Enneagram, then I think that the answer is a resounding 'yes'!

When, back in 1994, Pat Dinkelaker and I originally proposed our theory regarding what 'principles' guided the distribution of 16 MBTI types across the 9 points of the Enneagram, most people were trying to associate a 'Jungian' type with each Enneagram Point. By 'Jungian type' we mean, for instance, the 'Introverted iNtuitive', or 'IN', which is comprised of two MBTI types, the INFJ and the INTJ, the MBTI pair which is often referred to as the 'Jungian pair'. This exercise was driven, in part, by the fact that there are 9 Enneagram points and 8 Jungian types, in combination with Occam's Razor (which states that the simpler the theory, the better)- constraints that naturally led to the search for a system in which one Jungian type was associated with each Enneagram Point, and some extra-Jungian explanation was give for the remaining Point.

We found this to be an excellent but difficult exercise. Difficult because the eight Jungian types are mutually exclusive and exhaustive, and this imposes on one a very limited number of options. Could the exercise be done successfully? The theory that Pat and I presented at that time followed those rules. But realizing that in addition to the two MBTI types that we might thereby be 'assigning' to each Enneagram Point, additional MBTI types did IN FACT test as those Enneagram types, we sought to incorporate that fact into our theory. We described the Jungian pairs that we associated with each Enneagram point as 'prototypes' that manifested 'issues' in which people who had tested as other MBTI types might 'cluster' at those Points, in a way similar to the way in which members of a family 'resemble' each other without having to look identical to each other.

Using this approach, Pat and I, in a series of papers published in '95, were able to come up with Jungian prototypes for each Enneagram Point. Evidence that these Jungian types were indeed associated with those points was later provided by a survey on enneagram and MBTI type that showed that in 12 or 13 out of 16 cases the types that we had identified as 'prototypical' were the MBTI types with the highest statistical concentration at those Enneagram Points. Not a bad start, given the very strict self-imposed conditions (described above) under which the exercise was conducted, and the theory generated. But what about the 3 or 4 MBTI types that seemed to distribute in an anomalous fashion?

Seeking to better understand the actual statistical distribution of MBTI type across the Enneagram, we devised methods for measuring this and discovered that Enneagram tests seemed to fail to discriminate well between S and N. We called this 'S-N blindness', and realized that it was intimately associated with the fact that the theoretical (ie, meta-psychological) assumptions on which the Enneagram was based was a 'three function' system deriving from Gurdieff and Ouspensky - which described 'feeling', 'thinking', and a third function which was variously described as 'sensing' or in other ways that made it an amalgamation of S and N.

With this idea of 'S-N blindness' in mind, Keirsey's stratagem occured to me and I began to wonder if I could parse the 16 MBTI types into 8 groups that were different than the Jungian pairs, but (like Keirsey had done) nevertheless mutually exclusive and exhaustive pairs. What I was looking for was not four groups, like Keirsey's, but eight, like the Jungian pairs. The answer to this puzzle was obvious - reduce the sixteen to eight by failing to discriminate between S and N. This generated the following pairs, half of which turned out, unexpectedly to be 'Jungian' pairs!

	S-N Blindness Pairs	equivalent Jungian pair
	I(N/S)FJ	none
	I(N/S)TJ	none
	E(N/S)FJ	EF= ENFJ & ESFJ
	E(N/S)TJ	ET= ENTJ & ESTJ
	I(N/S)FP	IF= INFP & ISFP
	I(N/S)TP	IT= INTP & ISTP
	E(N/S)FP	none
	E(N/S)TP	none

Instead of using the eight Jungian pairs as 'prototypes', could we repeat the exercise using the above pairs? Same exercise - one pair per Enneagram point, each pair is mutually exclusive, and the set of all pairs is exhaustive (no MBTI types are left out). The only difference distinguishing this exercise from the original one would be that the eight pairs were not Jungian ones.

Assigning these 'S-N blindness' pairs to the Enneagram points was not difficult. The four pairs that were Jungian ones could be left at the points at which they were originally assigned, for statistical evidence showed that that was where they belonged. If the remaining pairs were assigned in the way illustrated below, 12 out of the 16 MBTI types remain in the zones to which we originally assigned them as prototypes (the twelve that are most strongly supported by the statistical data). In addition, the statistical anomalies in zones 7 and 3 disappear:

Enneagram Point	S-N Blindness Pairs	equivalent Jungian pair
4	I(N/S)FJ	none
6	I(N/S)TJ	none
2	E(N/S)FJ	EF= ENFJ & ESFJ
8	E(N/S)TJ	ET= ENTJ & ESTJ
9	I(N/S)FP	IF= INFP & ISFP
5	I(N/S)TP	IT= INTP & ISTP
7	E(N/S)FP	none
3	E(N/S)TP	none

After looking further into the matter, we discovered that we could create another list of 8 mutually exclusive and exhaustive pairs (not all of them S-N blind) and arrange them in such a way that maintained the original 'Jungian' flavor of E6 and E4, which was suggested by the narrative profiles traditionally associated with these zones in the Enneagram literature) -

INXJ-4, ISXJ-6, IXFP-9, IXTP-5, EXFP-7, EXTP-3, EXFJ-2, EXTJ-8.

In January of 1997 we suggested the above 'hybrid' assignments, purely on the basis of 'formal' considerations. Sixth months later, the Richards/Flautt study provided evidence that these assignments were 91% accurate in predicting which MBTI types concentrated in which Enneagram groups, as we have shown elsewhere.

Keirsey had unknowingly provided us with a strategy that proved useful in describing how MBTI types distribute across the enneagram. Previous researchers had become either 1) stuck in the box of trying to assign Jungian pairs across the Enneagram, or 2) having broken out of that box, also inadvertently did away with the requirement that the assignments should be mutually exclusive and exhaustive ones, thereby satisfying Occam's razor. Although, unbeknownst to us at the time, Larry Gabbard had devised a theory that assigned non-Jungian pairs to Enneagram points in an irregular manner (one that permitted the same MBTI type to be associated with more than one Enneagram point, etc), no one, to our knowledge had tried to locate a mutually exclusive and exhaustive set of non-Jungian MBTI pairs - one assigned to each Enneagram point - by which distribution across the Enneagram could be described with such a high degree of statistical accuracy.

[Editor's note: It should be noted that further research² in this area, reported at this site, has since focused attention on the significance of 'pure type' in the distribution of MBTI types across the Enneagram. We now have reason to believe that pure types (MBTI types that have the same 'attitude' for the dominant and auxiliary function - eg, Ni and Ti) are not only possible but more frequent than previously suspected, even though they are excluded in MBTI definitions that emphasize opposite attitudes between dominant and second Jungian functions.]

Conclusion

skip to footnotes

My purpose in writing this paper was to explain the theoretical strategy underlying Keirsey's unusual but logically consistent manner of grouping the 16 MBTI types. Using that explanation, I provided a partial critique of his system that was sensitive to the manner in which these abstract strategic moves, made early in his career, continue to shape choices that he is now making, more than twenty years later. I also demonstrated how we imported Keirsey's stratagem into our study of the relationship between the Enneagram and the MBTI, with beneficial results.