Cognitive function stacks

Spend any time in type communities and the four letters stop being the interesting part. What people argue about is the stack: eight cognitive functions, four of them active per type, in a fixed order. INFJ is Ni-Fe-Ti-Se. ENTP is Ne-Ti-Fe-Si. The dominant runs the show, the auxiliary supports it, the tertiary is adolescent, and the inferior erupts under stress.

It is by some distance the most sophisticated-looking thing in the type world, and it is where the four letters go from being a label to feeling like a theory of mind.

It is also the part with the least evidence behind it, and the gap is not small. The four letters have thin and contested support. The eight functions have less. The ordering has essentially none. When type dynamics has been tested directly, the tests have not supported it, and the main published defence of the framework is that it is useful rather than that it is confirmed.

That does not make it worthless, and this entry is not an argument to stop reading about it. It is an argument to be clear about what kind of object it is: a generative vocabulary, not a measurement.

Two authors, sixty years apart, and the second added the part people argue about.

Jung, 1921. Psychological Types proposes two attitudes — introversion and extraversion — and four functions: thinking, feeling, sensation, intuition. Combine them and you get eight orientations. Jung described a dominant function and a less developed opposite, and he wrote about the unconscious compensating for conscious one-sidedness. He was not building an instrument. He was writing observational psychology from clinical practice, with no systematic data and no intention of producing a classification anyone would score.

Myers, 1980. Gifts Differing and the instrument behind it turned Jung's scheme into something administrable, and added the machinery: the J/P letter as a pointer to which function is extraverted, and from that, a rule generating a full four-function order for each of sixteen types. The tertiary and inferior positions, and the whole apparatus now called type dynamics, are largely this layer rather than Jung's.

Everything after that — the eight-function "shadow" stacks, the developmental ages at which each function supposedly comes online, the grip experiences — was added later still, mostly by practitioners and communities, and with even less connection to anything measured.

Less than the elaborateness suggests, but not nothing, and the results point one way.

The dichotomies are not categorical. Arnau and colleagues ran taxometric analyses on Jungian preference measures and found dimensional rather than categorical structure — no latent types, just continua. That undercuts the stack before it starts, because the stack assigns a function order by type, and the types are not discrete. See the cutoff problem for what that does to any type-indexed rule.

Type dynamics does not outperform the letters. Reynierse examined the framework across a series of studies and argued it fails on its own terms: the predictions that distinguish type dynamics from a simple additive account of the four preferences do not hold up, and where the two make different predictions, the simpler one does at least as well. His summary phrase — that type dynamics is a "category mistake" — is blunt, and it comes from inside the type literature rather than from a hostile outsider.

The interaction claims do not appear. The stack's distinctive claim is that functions combine non-additively: Ni in an INFJ is supposed to behave differently from Ni in an INTJ because of what sits beside it. That is a testable interaction, and it is the kind of effect that shows up reliably when it is real. It has not shown up.

The order has no independent measurement. There is no validated instrument that reads your function order off your answers. Every test that reports a stack derives it from your four letters using Myers' rule. So a "function test" result is not a second measurement agreeing with the first; it is the first measurement restated.

Pittenger's review reaches the same place from the psychometric side: the instrument's problems are in the dichotomies, and adding a dynamic superstructure on top of unstable dichotomies does not stabilise anything.

Worth taking seriously, because "it persists despite the evidence" is an incomplete explanation and a slightly lazy one.

It is generative. The stack gives you a lot of vocabulary for a small input. Four letters become four functions, eight if you count the shadow, each with a name, a flavour and a story about when it appears. That is a rich language for talking about yourself, and richness is genuinely useful even when the underlying claims are not confirmed.

It explains within-type variation. Two INFJs are not alike, and the stack offers an account: different development, different function maturity. Any framework that can explain why two members of a category differ will feel more accurate than one that cannot — which is also exactly the property that makes it hard to falsify, and the property the Barnum effect entry describes.

It names real experiences. The inferior-function description — the competent, reflective person who under sustained stress becomes uncharacteristically impulsive, or fixated on the body, or obsessed with detail — is recognisable. People do have a characteristic way of coming apart, and it often is the opposite of their usual mode. That observation is Jung's, it predates any instrument, and it does not require the ordering rule to be true.

Unfalsifiability feels like explanatory power. If the theory covers both your typical behaviour (dominant) and your atypical behaviour (inferior, in the grip), then nothing you do can count against it. That feels like a theory that explains everything. It is the signature of one that predicts nothing.

A defensible position, if you like the framework, looks roughly like this.

Treat it as a vocabulary, not a diagnosis. "I default to internal logic and I'm bad at reading a room in real time" is a useful sentence. "My Ti is dominant and my Fe is tertiary" is the same sentence with an unearned mechanism attached. The first can be checked against your week. The second cannot.

Do not use it to make decisions about other people. A hiring, pairing or team-composition decision made on a function stack is resting on a layer with no validated measurement behind it. If you want something with predictive evidence for work outcomes, the trait scores in the Big Five have it and the stack does not.

Be suspicious of the retrofit. The stack's most common use is explaining something after it happened. Any framework can do that. The test of a theory is a prediction made before, and this one is rarely put in a position to make one.

Know what our tests do and do not give you. The MBTI-style test reports four dimensions and a code. It does not report a function stack, and that is deliberate rather than an omission we plan to fix — publishing a stack would mean presenting a derived label as though it were a second, independent reading. The Shadow Archetype test draws on Jungian language for self-reflection and is not a measurement of function order either; it is a prompt, and it says so.

If what drew you to the stack is the sense that the four letters are too crude to describe you, that instinct is right, and type frameworks is the entry about what to do with it. The answer is usually more resolution, not more superstructure.