Why your result changes every time you take a personality test

You answered honestly both times, and the machine gave you two answers. It isn't you, and it isn't quite the test: it's a method error.

The second test

A year ago, a test told you that you were INFJ. You recognized yourself. You kept it, you talked about it, you may even have slipped it into a bio somewhere.

Last week, you take another one. The same, or nearly, the same family of test, the same four letters at the end.

Result: INFP.

One letter changed. And there you are, sitting in front of your screen with three questions cascading through your head: which one is the real one? Have I changed in a year? Or do I know myself that badly?

Here's the detail that stings: you answered honestly both times. You didn't cheat, you weren't trying to please yourself, you didn't tick boxes to look like someone. You told the truth, twice, and the machine handed you back two different identities, as if your personality had flipped while you were looking away.

If you're wondering why your result changes every time you take a personality test, here's the promise: you haven't changed. The test isn't (only) bad. What's wrong is more interesting than that, and once you understand it you'll never look at a four-letter result the same way again.

The two false explanations

"I changed between the two tests."

That's reassuring, and it's false. A personality profile is stable over years. It moves, yes, but slowly, pushed by life, never in three weeks, I've given that point a whole article. If your result flipped in a month, it isn't you that moved. It's the measurement.

And that should alarm you: something that gives a different result on every reading isn't measuring your personality. It's measuring something else, the mood of the day, the wording of a handful of questions, chance.

"These tests are worthless."

That's tempting, and it's too quick. Above all, it would rob you of the real lesson.

Some tests really are sloppy, that's true. But even a serious test, well built, will produce this instability if it forces you to pick a category. The problem isn't only the quality of the tool. It's a design decision made upstream: cutting people into boxes. And even a good box-based test does that.

This isn't the place to take the MBTI apart in full, I do that elsewhere, and I run it past the three criteria of a reliable test here. Today's question is narrower and more precise: why did the result change?

So: not you, and not simply a mediocre test.

The cause is structural.

What's actually happening: the type versus the degree

The problem is the border.

A type-based test, MBTI, 16 personalities, all the ones that hand you letters, works by cutting. On each axis it files you on one side or the other. Introvert or extravert. Thinker or feeler. You have to pick a camp; there's no third option.

But real people aren't sorted into two camps. On any trait, most people are in the middle, in a zone where neither of the two words really applies. One big central mass, not two blocks separated by a chasm.

And that's where it all breaks.

  • If you're at 51% toward extraversion, the test shows E.
  • If, on another day, four or five answers put you at 49%, it shows I.

Two points apart. The opposite letter. A different identity.

Take a simple image: an exam, with the pass mark at 70 out of 100. You hand in your paper, it's marked 69.5. Failed. The same paper goes to a second examiner, slightly more generous on one small question: 70.5. Passed.

Your paper didn't change by a single word. What changed is which side of the line it landed on. A mark, a degree, a continuous quantity, was converted into a status, passed or failed, two worlds, two ways of being treated. The verdict looks categorical. It rests on one point.

And notice what the status makes disappear: between the person who got 70.5 and the person who got 95 there's a chasm, yet they're in the same box. Between the person who got 70.5 and the person who got 69.5 there's nothing, and they're filed as opposites.

Your personality is the mark: a continuous quantity, precise, stable. The type is the status stuck on top of it. And you, on the axis that changed, were right on the line.

The consequence is decisive, and counterintuitive: the closer you are to the middle on an axis, the more unstable your type is. And since most people are in the middle on several axes at once, most people have an unstable type. It isn't the people who know themselves badly who get different results. It's the balanced ones, the ones in the center, that the system categorizes worst, precisely because they're where the border cuts.

Your result didn't change because you changed. It changed because a slider was forced to become a box, and you were sitting right on the dividing line.

Why a real profile doesn't flip

So the solution isn't to "find the good type-based test." There isn't one. The flaw isn't in the execution, it's in the principle.

The solution is to stop measuring in types.

A model in degrees, the Big Five, doesn't file you in a box. It gives you your position on a scale. Not "introvert" or "extravert," but "extraversion: 47 out of 100." And that difference, which sounds technical, changes everything.

Stability. If you're at 47 today and 49 in a month, the model shows… 47, then 49. A small variation stays a small variation. There's no pass mark to clear, so nothing can flip all at once. The result moves as much as you do, which is to say barely. Where the type turned two points into a change of identity, the degree shows them as two points. Which is what they are.

The information kept. The type throws away the most important information of all: how much. "Introvert" doesn't tell you whether you're slightly or massively introverted, and that's night and day, two different lives behind the same word. The degree keeps that nuance. And it's exactly that nuance that describes your real life, not the label that flattens it.

The middle finally represented. Where the type forces balanced people to pick a false camp, and to switch camps with the wind, the degree gives them their true place: the center. The people the MBTI tosses endlessly from one type to another are exactly the ones a score in degrees describes best. Your instability in one becomes your clear position in the other.

The right test doesn't give you a camp. It gives you a position, and a position doesn't flip over two points.

What this changes for you

Stop looking for "your real type."

It doesn't exist, because the question itself is badly posed. You aren't a type surrounded by a bit of noise. You're a set of positions, some of them close to the middle. Looking for THE right label is looking for a box for something that, by nature, doesn't have one.

Read the instability as information.

If your type moved a lot between two tests, that isn't pure chance. It tells you something true: that you're close to the middle on the axes that flipped. That's real data about you, your balance on those dimensions, which the type reveals, paradoxically, at the very moment it fails to file you.

Change tools, not tests.

The problem isn't that you happened to take bad type-based tests. It's the principle of the type itself. Move to a model that measures in degrees, and the question "which one is the real one?" disappears on its own, there's only one profile, and it doesn't flip because you answered two questions differently on a tired day.

From INFJ to INFP

Come back to that letter that changed, to that INFJ turned INFP.

That flip wasn't a sign that you know yourself badly. It wasn't even necessarily a sign that the test was junk. It was the symptom of a method error: cutting into boxes what is a question of degree. You weren't indecisive, or unstable, or lost. You were in the middle, and nobody had ever given you a tool capable of seeing it.

A model that sorts you into types will always give you a result that wobbles near the borders. A model that measures you in degrees, the Big Five, gives you a stable position on each of the five traits, the middle included and finally represented. No more "which one is the real one?": one profile, which doesn't flip because you ticked two answers differently.

The test is here. It takes ten minutes.