Introverted Thinking is a model you maintain in private whose only obligation is not to contradict itself. That is the whole function. Almost everything people find difficult about heavy Ti users falls out of that one property: the fights about what a word means, the refusal to accept an answer that works but doesn’t make sense, the long pause before agreeing to anything, and the habit of arguing a side they don’t actually hold. None of it is contrarianism. It is a referee doing the job it was built for.
The mechanism, stated mechanically
Ti checks a claim against an internal model. Te checks it against an external result or an external authority. Both are “thinking” in the sense that both decide what gets admitted as true. The difference is who referees. Ti’s referee is a private structure of definitions and relations that is not permitted to contain a contradiction; a claim is admitted only once it has been re-derived from parts already inside. Extraverted Thinking’s referee sits outside: did it produce the outcome it predicted, does the record support it, does the person accountable for the result accept it?
Neither is the more rigorous one. They are rigorous about different things and they fail in opposite directions. Ti fails by being flawlessly coherent about a world that isn’t there. Te fails by running something that works without ever asking why, which is fine until the conditions change and nobody on the team can say which part was load-bearing. The function stack rules put these two on the same axis for exactly this reason: whichever one you run by default, the other is the check you don’t apply to yourself.
One claim, two engines
Watch the same sentence hit both. Somebody drops a proposal into a team channel, and the two replies arrive within a minute of each other.
“We should kill the weekly one-on-ones. Every team I have been on that dropped them shipped faster.”
“How many teams, and faster by what measure? Give me cycle time either side of the change for each one. If it holds up I will drop them on my team for a sprint and we look at the numbers in three weeks — and if nothing moves, they go back.”
“Faster than what baseline? And one-on-ones is covering two different things here. There is the status sync, which duplicates standup, and there is the only channel where someone tells you the thing they would never say in a group. If what your teams dropped was the first one, your claim is an argument against status syncs and it says nothing at all about the second. Which one were they dropping?”
“Te ran the sprint and had an answer in three weeks. Ti split the term in two and never did find out whether the original claim was true.”
Look at what each reply actually asked for:
- Te went outside the argument. It wanted the result, a comparison, and a way to find out. Notice it never questioned the concept — it was ready to run a real experiment on a category containing two different things, and would have trusted the muddled number afterwards because the number came from reality.
- Ti went inside the sentence. It found one term doing two jobs and refused to evaluate anything until that was settled. Notice it never asked for data at all. Not once. It wasn’t indifferent to evidence — it just cannot use evidence about a category it can’t define.
- Neither one is stalling. Te is not being shallow and Ti is not being pedantic for sport. Each is doing the only check it trusts, and each finds the other’s check strangely optional.
The four-question Ti/Te rule
Four questions, and unlike most typing questions they are about artefacts and habits rather than self-image. Answer for how you are on a Sunday, not for what your job demands — a compliance role makes everybody sound Te and a research role makes everybody sound Ti.
- What makes you trust a conclusion?Ti: you can rebuild it yourself from pieces you already accept. Trust arrives the moment the derivation closes, and it arrives from nobody in particular. Te: it produced what it predicted, more than once, ideally for someone whose track record you can check.
- What makes you reject one?Ti: a contradiction. A word used two ways, a step that doesn’t follow — and it still counts when the conclusion is fine and everyone is happy with the outcome. Te: it didn’t work, or nobody can show that it did. A conclusion with an unbroken record survives even when the explanation for it is a mess.
- What do your notes look like?Ti: definitions, distinctions, conditions, at least one paragraph starting with “unless”, and a sentence rewritten four times to get the phrasing exact. Frequently no owner and no date, and you don’t notice that until someone points it out. Te: owners, dates, statuses, numbers, a decision log. The reasoning compressed to a clause, because the reasoning was never the deliverable.
- What do you do when an expert contradicts you?Ti: ask for the reasoning and hold your position until you get it. Credentials move the prior, not the verdict. “They’re probably right, I just can’t see why yet” is a pure Ti sentence and it is why Ti users get called arrogant by people who meant to be helpful. Te: update on the spot when the source has the record, then watch what happens. Deference is cheap and the outcome reports back soon enough.
Everyone has both, so mixed answers are the norm. If you split, weight question three: the notes already exist, you wrote them before you had a theory about yourself, and you cannot retroactively perform them. Go open the last plan or brief you produced voluntarily and read what its longest section is about. If questions one and two say Ti but question four says Te, that is often circumstance rather than type — holding out against a senior expert is expensive in some rooms, and plenty of Ti users learn to nod and go re-derive it quietly at home.
Ti in all four slots
Position in the stack is authority, not skill. First-slot Ti runs whether or not anything asked it to and gets the first word on everything. Fourth-slot Ti sits dormant until the top of the stack has failed, then arrives without a volume control. Same function, four completely different-looking people.
1st — INTP, ISTP
The model runs whether or not anyone asked. Terms get pinned before questions get answered, and the pinning is not optional.
2nd — ENTP, ESTP
A filter applied to material the dominant went and got. Precision on demand, then dropped when the exploration moves on.
3rd — INFJ, ISFJ
A large private framework built for the owner’s own satisfaction, never argued aloud, and therefore never tested.
4th — ENFJ, ESFJ
Dormant until stress, then sudden disproportionate pedantry about one small fact that can actually be settled.
Dominant: INTP and ISTP
Ti-Ne-Si-Fe and Ti-Se-Ni-Fe. Identical first and fourth slots, which is why these two share so much surface: the same insistence on pinning a term before answering, the same silence that gets read as disagreement, the same out-of-nowhere emotional event from inferior Fe. The model is always running. It is not switched on by a question; the question just makes it audible.
The whole difference is what feeds it. Ne feeds the INTP hypotheticals, so their model can cover systems they have never touched — sometimes brilliantly, sometimes with a confident opinion about hardware they have never held. Se feeds the ISTP the object in front of them, so their model is assembled from things they have physically taken apart. A quick test: ask each to explain how something works. The INTP gives you the general principle and hedges about the specific unit. The ISTP tells you exactly what this unit does, in what order, and shrugs at the general principle as unnecessary.
Auxiliary: ENTP and ESTP
Ne-Ti-Fe-Si and Se-Ti-Fe-Ni. Second-slot Ti does not own the agenda; it is a pruning tool applied to material the dominant went out and collected. This is why ENTPs look inconsistent to a Ti-dominant. They will dismantle your argument with genuine precision and then, ten minutes later, appear not to care about the conclusion — because the Ti pass was serving Ne’s exploration, not maintaining a permanent structure that has to still be true next Tuesday.
In an ESTP the same slot shows up as troubleshooting under time pressure: a fast, disposable causal model of the machine, the room or the opponent, built in seconds and discarded the moment the situation resolves. Auxiliary Ti is deployed. Dominant Ti is running. If you want to know which one you have, ask whether the model persists when nobody is asking you anything.
Tertiary: INFJ and ISFJ
Ni-Fe-Ti-Se and Si-Fe-Ti-Ne. Third-slot Ti is real, often substantial, and almost entirely invisible: elaborate private frameworks built for the builder’s own satisfaction, systems of categorisation nobody is ever shown, hours of reading before a single opinion is voiced. Its characteristic weakness is that it is unaudited. Because it is never argued out loud, it never meets a counterexample, so it can grow very large and very tidy without ever being tested.
Under Fe strain it becomes a hiding place. The INFJ stops managing the room and starts producing an airtight-sounding analysis of why the person who hurt them was structurally bound to do so. It feels like insight. Functionally it is Ti supplying a coherent story that removes the need to have the conversation Fe is dreading. The ISFJ version is quieter: a precise, accumulated set of criteria about how a thing is properly done, unstated until violated, then delivered as a correction that lands far harder than intended because nobody knew the rule existed.
Inferior: ENFJ and ESFJ, where it shows up as sudden pedantry
Fe-Ni-Se-Ti and Fe-Si-Ne-Ti. Here is the scene, because it is always the same scene. An ESFJ has run a family dinner — who sits where, whose diet, which subject stays off the table. Fe has been spending all evening. Then someone says something mildly dismissive, and what comes out is not an emotional appeal. It is fifteen minutes of establishing that the trip was in 2019, not 2020, with the receipt located on a phone and read aloud. Nobody at the table cares about the year.
That is inferior Ti. The atmosphere Fe was managing stopped responding, and the fourth function offered the only thing it knows how to do: find one small item that can be settled definitively and settle it. Being right about the year becomes a substitute for the sentence they cannot say, which is “I did all of this and you were rude to me.” The ENFJ version runs bigger — a late-night research spiral producing a rigorously argued case delivered as a wall of text at 2am. The argument is often structurally fine. It is simply not what the conflict is about.
Ti is not intelligence and Te is not organisation
This framing needs killing, because it corrupts every self-typing attempt that touches it. Ti is a check, not an IQ. Te is a check, not a filing habit. A Ti user with bad inputs builds an immaculate model of a world that does not exist, and the immaculateness is exactly what stops them noticing. A Te user with good instruments beats a cleverer Ti user routinely, because their referee is reality and it gets consulted daily rather than annually.
A worked example. A support ticket says users lose their draft when they navigate away from the settings page. The Ti-dominant engineer resolves it in two minutes without opening the app: drafts are persisted by a single autosave hook, that hook is mounted in the editor route, the settings route does not mount it, therefore there is no draft on settings to lose, therefore the reporter has confused settings with the editor. Every step of that follows. It is also wrong, because a shared component quietly pulled the hook into settings six weeks ago and the model’s premise expired without sending a notification.
The engineer holds out through three more tickets. Not from ego — from inside the model, three confused users are outnumbered by a proof. What eventually breaks it is not a fourth ticket. It is opening the settings page, typing a sentence, clicking away and watching it vanish. Ninety seconds, and it is a Te move. So the decision rule for anyone who runs on Ti: every model ships with a named falsifier and the cost of checking it — one observation that would kill this, and how long that observation takes to obtain. A model that forbids its own test is not rigorous, it is closed, and that is a contradiction Ti genuinely cannot sit with once it has been pointed at.
Ti in argument: defending a position you don’t hold
This is the behaviour that costs Ti users the most relationships, and it is almost always misread. A Ti user is not advocating for an outcome. They are load-testing a structure. A weak argument for a conclusion they agree with is more offensive than a strong argument for one they reject, because the weak argument will be reused next time on a question where nobody already knows the answer, and then it will produce garbage.
The second thing they are doing is locating the load-bearing premise. The fastest way to find out which assumption everything is resting on is to push from the outside and see which support, when you remove it, drops the whole thing. Taking the opposite side is a lab technique. It is not a belief.
“Obviously remote is better. Nobody wants to commute.”
“Then the claim is about commuting, not about remote. Let me take the other side for a minute. If the only thing holding a team together is a channel where you cannot see anyone hesitate before they answer, what replaces the hallway? I do not think that beats your position. I want to see whether your position survives it.”
“So you want everyone back in the office. Great.”
“No — I would vote exactly the way you would. I just do not want to be holding your reason when someone sharper than both of us leans on it.”
What they were checking: whether the conclusion is supported by the stated reason or merely accompanied by it. The cost is that everyone present now thinks a position was taken, and an Fe-led listener has scored it as disloyalty while an Fi-led listener has scored it as not caring about something that matters. Two rules fix most of it:
If you are the Ti user: label the move first
One sentence before you start. I agree with your conclusion, I am going to argue the other side for two minutes to find where it breaks. It costs nothing and it prevents the entire misread, because the misread is about position, not about tone.
If you are on the other end: ask one question
Are you defending that or testing it? A Ti user answers honestly and instantly, because that distinction is the exact thing they have been tracking the whole time. It ends the ambiguity faster than any amount of arguing back.
There is one situation where the whole move should be switched off: when someone is describing something that happened to them rather than making a claim about the world. Load-testing an experience reads as auditing it, and no amount of “I’m only testing the reasoning” repairs that afterwards.
The mistype worth getting right: Ti vs Fi
Ti gets confused with Ni, with intelligence, and with being difficult, but the expensive confusion is with Introverted Feeling. Both are introverted judging: private, unshared, quietly immovable, and both decline to justify themselves on demand. In our typing study — 1,179 conversations after de-duplication — 42.3% came back INTP and 23.3% INFP. That is two thirds of the sample sitting in the two types whose difference is precisely this call: Ti-Ne-Si-Fe against Fi-Ne-Si-Te. Same perceiving axis, opposite judging axis. (The 42.3% is not a base rate, incidentally — it is a fact about who volunteers years of chat logs for analysis.)
What the objection is made of
Ti: that does not follow — the objection has parts, and fixing the reasoning genuinely answers it. Fi: I am not doing that. Press for justification and you get a restatement with more feeling attached, not a derivation, because there was never a derivation.
What changes your mind
Ti updates on a better argument, from a stranger, in eight seconds. Fi does not update on arguments at all; it shifts slowly when the thing stops matching what you actually value, and it cannot be talked into that on anyone else’s schedule.
Where the precision goes
Both are precise, about different objects. Ti gets exacting about definitions, categories and whether two cases are the same case. Fi gets exacting about the correct word for a feeling and the exact degree of wrongness in something.
What silence means
Ti goes quiet because the model has not closed yet and speaking early would commit it to something unfinished. Fi goes quiet because the thing it would have to say is not up for discussion and putting it into words in front of an audience would cheapen it.
If two of those four still feel ambiguous, use the fourth slot, because inferior functions are louder and far less flattering than dominant ones. Ti-Ne-Si-Fe ends in Fe, so the bad day sounds like does anyone here actually like me, was I cold to them, did I say the wrong thing three weeks ago. Fi-Ne-Si-Te ends in Te, so the bad day sounds like I have produced nothing measurable, everyone else has systems, I should be able to just execute. Nobody performs their inferior function on purpose, which is what makes it the honest half of the function stack.
What Ti leaves behind in writing
Functions are legible in text because they are habits of construction rather than opinions. Ti leaves definitional repair work — “depends what you mean by”, a sentence that revises its own earlier clause mid-paragraph, an answer that arrives as a distinction rather than a yes. Conditional scaffolding: if you mean A then no, if you mean B then obviously. And one thing that shows up as an absence: Ti rarely cites. Te writing carries a name, a source, a number. Ti writing carries authority with no footnotes, because it re-derived the thing instead of looking it up, and it forgets that this is not visible to anyone else.
That is the whole argument for scoring how someone actually writes rather than asking them. Self-report on a question like “do you value logic?” collects nothing — everybody says yes — while a pass over your real messages can count how often you answered a question by splitting it. If you would rather start from a free MBTI test, take it first and then check the result against your own outbox.