INFP in March. INFJ in June. ENFP on a site your friend sent you, taken on a train. The natural reading is that one of them is right and the others were taken badly, or worse, that you do not know yourself well enough to answer twenty questions about your own life consistently.
Neither is what happened. The instability is a property of the model, not of you, and once you can see the mechanism the whole thing stops being unsettling and becomes quite specific information about yourself.
What is usually going on
You are near the middle, and the model has to round. This is the whole answer and everything else is detail. Each of the four dimensions is scored as a continuous quantity and then cut at a line, and the letter you get is which side of that line you landed on. A 52 against 48 split produces a capital letter with exactly the same confidence as an 89 against 11 split. Two answers change on a Tuesday, the score moves four points, the letter flips — and because the letter is part of a four-letter name, the entire profile you get handed flips with it. You went from a detailed description of one kind of person to a detailed description of a different kind of person because of a handful of points on one dimension out of four.
Scores cluster in the middle, so this happens to a lot of people. If the types were genuine categories, the scores would be two-humped and almost nobody would sit near the cut. They are not. On every dimension the distribution is a single mound with the bulk of people somewhere in the centre, which means near-the-line is not an edge case; it is the ordinary case. Roughly speaking, the more average you are on a dimension, the less your letter means and the more it moves.
One specific dimension is doing the flipping, and it is worth knowing which. Line up your results and you will usually find three letters that never move and one that does. That unstable one is where you are near the middle, and it is a real description of you: being genuinely balanced between thinking and feeling, or between planning and improvising, is a way to be. It is just a way the model has no notation for, so it alternates between two wrong answers instead.
You answered as a different person each time. Most of these instruments do not tell you which context to answer for, so you choose without noticing — and the you at work and the you on a Sunday do not give the same answers about planning, socialising or how decisions get made. People who take a test at their desk and again on holiday frequently get different types, and both sittings were honest.
State moved the needle. A stressful fortnight, three bad nights of sleep, the week after a breakup — these shift self-report on questions about energy, sociability and structure by more than four points, easily. The instrument is measuring how you answered today, and today is a variable.
The items were not the same items. Every free version of this on the internet uses different questions to measure the same four dimensions, with different wording and different numbers of items. Someone far from the middle gets the same letter from all of them. Someone near the middle is genuinely at the mercy of which fifteen sentences a particular site chose.
And some of the movement was you being more honest. This one is uncomfortable and it is real. People answer the first sitting somewhat aspirationally — as the person they are working on being — and later sittings, especially after a hard stretch, somewhat more plainly. If your type changed in the direction of a less flattering profile, it may not be noise.
What to do about it
- Look for the number, not the letter. If a test gives you four letters and no indication of how strong each preference was, it is withholding the only part that would let you interpret it. That is a reason to distrust the test rather than yourself.
- Pick a context and say so. Answer for me outside work and keep it. A result that averages your office self and your Saturday self describes a person who does not exist in either place.
- Treat the flipping dimension as a finding. Being mid-range on thinking and feeling is not a failure to have a personality; it means you weigh both and that neither of the two profiles will fit you well. That is more useful than a letter would have been.
- Do not retake it after a bad week. You will get a valid measurement of a bad week.
- If you want stability, use a continuous instrument. The Big Five reports five numbers and rounds nothing, which is why its scores hold up across months where a type label does not. What a percentile actually means covers what those numbers actually mean once you have them.
- Expect a continuous instrument to move a little as well, and know how much. Trait scores drift for reasons that have nothing to do with typology, and why a personality test result keeps changing has the longitudinal figures for how much of that is ordinary.
What it depends on
How much your type moves is decided by one thing: how far from the middle you sit on each of the four dimensions. That is a measurable quantity and you can simply have it. Someone at the far end of a dimension has a letter that has been the same since school. Someone in the middle has a letter that is, in the most literal sense, being generated by rounding.
Our version of the test reports that position for each dimension alongside the four letters, for exactly this reason, and says where the four-letter model is weak rather than hiding it behind a label. If you want the longer version of why the categories behave like this, how accurate the MBTI actually is goes through the retest and distribution research, and the type pages show what a described type looks like when the dimensional scores are put back in. It is the same reason no two people with the same four letters are alike: the letters record which side of each line you landed on and say nothing at all about how far.