Why your personality test results keep changing
By Depth Profile · Updated November 17, 2026 · 7 min read
A writer for Refinery29 took the 16Personalities test and came out as INFJ-T, the "Turbulent Advocate." Four days later the writer took it again and got INFP-T, a "Mediator." As the piece puts it: "I did the test twice in four days and I got two different results" (Refinery29).
If that has happened to you, you are in a large club. On typing forums you can find people who got ENTP, then ENTJ, then ENFP, then INFP, and now wonder which one is the real them, or whether they have been "mistyped" all along.
The usual explanations are that you answered carelessly or that you are going through changes. Sometimes that's true. More often the cause is simpler and more mechanical, and once you see it, a changed result stops being a mystery and starts being useful information.
The cliff in the middle of every letter
A four-letter type is built from four scores. Behind the I or E there is a number for how introverted or extraverted your answers were. The test then draws a line down the middle of that scale and gives you one letter or the other depending on which side you land.
That works well if people cluster at the two ends, the way the type idea assumes. They don't. Lou Hicks studied the Sensing and Intuition scores of a group of adults in 1984. He wrote that "the distribution of S-N scores was not bimodal," and that MBTI score distributions are typically flat-topped instead of split into two camps (Hicks, 1984). Five years later, McCrae and Costa reported finding no support for the idea that the MBTI measures two-sided preferences or distinct types. The four scales behave like four continuous dimensions (McCrae and Costa, 1989).
So most people sit somewhere in the broad middle of each scale, close to the line. The research group Clearer Thinking put it plainly: "most people fall near the middle," and "for people near the mean, answering even just a question differently can be enough to switch letters" (Clearer Thinking, 2024).
Picture a scale from 0 to 100 with the line at 50. Someone who scores 49 one week and 52 the next has barely moved. Their type report, though, has turned them from an introvert into an extravert, with a new set of paragraphs to match. The letters exaggerate a tiny change in the numbers.
That is why ENTP, ENTJ, ENFP and INFP is a less chaotic sequence than it looks. The N never moved, and the E held for three tests out of four. The last two letters wobbled every time, which suggests that person sits near the middle on those scales. That is a finding about them, and a fairly specific one.
How often types change on a retest
The usual check is to give the same people the same test twice and count how many get the same result. The numbers depend on who you ask, so here is both sides.
The critics' figure comes from David Pittenger, who reported that about half of people land on a different type when retested after only five weeks (Pittenger, 1993, as cited in a 2017 review).
The publisher of the official MBTI says that figure is based on an older version of the test. For the current form, it reports that "over a 4-week retest period, 65 percent of respondents had all four preferences the same," and 93 percent had three or four the same (CPP, 2018).
Those two claims are closer than they sound. Even on the publisher's own numbers, about a third of people come back from a one-month retest with at least one letter changed. The publisher's own researchers put most inconsistent results down to low clarity scores, which means people who sit close to the line (Thompson, The Myers-Briggs Company). A psychologist defending the MBTI in Psychology Today summarized an older study of people whose first scores fell in the intermediate range: on retest, between 25 and 32 percent of them flipped on each scale (Gordon, 2020). Clear preferences tend to stay put. Slight ones move.
Free online type tests are not the official instrument, and we haven't found comparable retest figures published for them, so we can't give you a number. The midpoint problem applies to any test that turns a scale into a letter.
Your answers move too
The cliff explains why small changes flip letters. It doesn't explain why your answers change at all. Two things do most of that work.
Mood leaks into the answers
In one experiment, 98 people filled out a Big Five questionnaire, then went through a procedure designed to make them feel happy or sad, then filled it out again. After the sad induction, their neuroticism scores went up and their extraversion scores went down (Querengasser and Schindler, 2014). Feeling happy had a weaker effect.
Nothing about their personality changed in those few minutes. Their mood changed how they answered. Think of how differently a question along the lines of "I often feel blue" reads on a bad afternoon. If you took one test on a good Sunday and another after a rough week at work, the rough week is in the second set of answers.
You really do vary
The second reason is that people behave differently from one hour to the next, and that's normal. In a set of studies by Fleeson, people reported on their own behavior again and again in daily life over two to three weeks. He found that within-person variability was high: most people showed nearly every level of each trait at some point. Yet each person's average level was almost perfectly stable (Fleeson, 2001).
So when a question asks whether you are talkative, the honest answer is "it depends." Which moments you happen to remember while you answer can change your score. Thinking about last Friday's dinner gives a different answer from thinking about Monday's meeting.
What continuous scores do differently
Big Five tests also contain error and also pick up mood. The difference is in how they report it. They give you a number on each trait instead of a letter, so a small change shows up as a small change.
They are reasonably stable. A review that pooled 848 retest results across many Big Five measures found average retest correlations between .69 and .76 for the five traits (Viswesvaran and Ones, 2000). A correlation of 1 would mean everyone kept exactly the same rank order, and 0 would mean the second test was unrelated to the first. Around .7 means people mostly keep their place, with some movement.
The continuous version is also more useful. Clearer Thinking compared two ways of scoring the same kind of Jungian questions. Keeping the scores as numbers, instead of collapsing them into letters, made them about 36 percent more accurate at predicting people's life outcomes (Clearer Thinking, 2024). Part of what the letters throw away is exactly the information about who sits near the middle.
If you want the longer comparison of the two approaches, our Big Five vs MBTI guide covers it, and how accurate are personality tests explains reliability in more detail.
How to read a result that changed
A changed type is a clue about where you sit, and you can work out a lot from it. Try these in order.
- Find the letters that moved. Write your results side by side. The letters that never change are probably clear preferences. The ones that flip are the scales where you sit near the middle.
- Look for the numbers. Many tests show a percentage or a bar behind each letter. If yours does, compare them. A move from 54 to 47 is a small shift that crossed a line, and not a change of identity.
- Read both descriptions for the flipping letter. If you are near the middle on thinking and feeling, you will probably recognize yourself in parts of both. That is an accurate result. The test is telling you that you use both, depending on the situation.
- Note the conditions. What kind of week was it? Were you at work, at home, tired? If the results differ, ask whether the mood or setting was different too.
- Pick a frame and keep it. Next time, answer every question for how you have usually been over the past couple of years, across most of your life. Avoid answering for this week or for the person you are trying to become.
- Ask someone who knows you. Take the scale that keeps flipping and ask a close friend which way you lean. You know your inner states best, but friends often see your outward behavior more clearly.
- Retake under similar conditions and average. Two or three attempts a few weeks apart, taken in a steady mood, tell you more than any single result.
If your results swing wildly on several scales at once and you also feel unsure who you are from week to week in a way that troubles you, that is worth talking through with a counselor or psychologist. A test can't sort that out, and you don't have to sort it out alone.
Where we come in
Our free Depth Profile assessment gives you a score on each of five core traits instead of a type, so a retake shows you how far you moved, not just which side of a line you landed on. It uses our own short forms of items from the public International Personality Item Pool. Those short forms are not independently validated, and the test is a self-reflection tool, not a clinical one. Take it twice a few weeks apart and compare the numbers. If they move a little, that's the normal amount of wobble. If they move a lot, look at what was different about the two days you took it.
Sources
- Refinery29, a writer retakes 16Personalities four days apart
- Clearer Thinking (2024), How accurate are popular personality test frameworks at predicting life outcomes
- Hicks (1984), Conceptual and empirical analysis of some assumptions of an explicitly typological theory, Journal of Personality and Social Psychology
- McCrae and Costa (1989), Reinterpreting the Myers-Briggs Type Indicator from the perspective of the five-factor model, Journal of Personality
- Pittenger (1993), Measuring the MBTI and coming up short, as cited in Acta de Investigacion Psicologica (2017)
- CPP (2018), Creating clarity around the Myers-Briggs Type Indicator, press release
- Thompson (2020), The Myers-Briggs is no fad, The Myers-Briggs Company
- Gordon (2020), In defense of the Myers-Briggs, Psychology Today
- Querengasser and Schindler (2014), Sad but true? How induced emotional states bias self-rated Big Five traits, BMC Psychology
- Fleeson (2001), Toward a structure- and process-integrated view of personality, Journal of Personality and Social Psychology
- Viswesvaran and Ones (2000), Measurement error in Big Five factors personality assessment, Educational and Psychological Measurement