Anywhere from 8 to 72. The free MBTI-style formats we compared ran from 8 to 60 questions; this test asks 72 statements, 12 per letter across six letters, in about 10 minutes. Count is only half the story: 12 five-point statements per letter gave the same letter back 78% of the time, while 15 yes/no questions managed 75%.
Two separate costs
When people ask how long a test is, they count questions. But a test spends your attention in two ways, and they are priced separately.
The first is how many statements each letter gets. One statement about parties cannot tell a mild lean towards E from a strong one; twelve statements about different situations can.
The second is how you answer. A yes/no question keeps one bit: this side or that. A five-point scale, from “Very inaccurate” to “Very accurate”, keeps how far you lean. Throwing that away is like photographing a slope in black and white: you still see which way it goes, but not how steep it is.
The two costs are independent. A 60-question yes/no test pays the first and still pays the second.
The formats side by side
We checked this on 60,000 real answer sheets from Johnson’s public IPIP-NEO data. For each format we drew two non-overlapping sets of research statements, 40 times over, and counted how often the same person landed on the same side of each scale in both sets. Only the four MBTI letters are compared here, so the totals are for four letters.
| Questions (per letter) | Answer format | Same letter | All four the same |
|---|---|---|---|
| 8 (2) | five-point | 62% | 15% |
| 20 (5) | yes/no | 65% | 18% |
| 32 (8) | five-point | 74% | 31% |
| 48 (12) | yes/no | 73% | 28% |
| 60 (15) | yes/no | 75% | 32% |
| 48 (12) | five-point, random research statements | 78% | 38% |
| 48 (12), this test’s four MBTI letters | five-point, hand-picked | about 81% (estimated) | about 44% (estimated) |
Read down the yes/no rows and the count does its job: 5, 12 and 15 questions per letter climb from 65% to 73% to 75%. Then compare the two 48-question rows. Same length, the same kind of research statements, and the five-point version beats yes/no by five points per letter and ten points on the whole type. Format bought what 12 extra yes/no questions could not.
What length is enough
Eight statements per letter on a five-point scale already reach 74%, close to what 60 yes/no questions manage. Twelve per letter adds a few more points. Past that the gains get smaller.
The ceiling is set by where the line is drawn. A letter is the side of the middle you land on, and some people land close to it. For them, no number of extra statements settles the question, because there is not much to settle: they genuinely sit near the line, and on another day a few answers shift and so does the side. Piling on statements past that point makes everyone sit through more, and still mislabels the people near the line.
That is why this test does not keep piling on statements. It asks 12 per letter, then a few more only on the scales that came out close, and if a scale is still close the result says in between rather than forcing a side. Spending questions where the doubt is beats spending them everywhere.
Why 72 and not 48
The table says 48 for the four letters, and the test has 72 statements. Both are right. Twelve statements for each of the four MBTI letters make 48. This test adds two more letters, measured the same way: whether you stay steady or react under pressure (A or T), and whether you lean on others or handle things yourself (C or S). Two more letters at 12 statements each make 72, and six letters give 64 types such as INTJ-AC or ENFP-TS.
The four MBTI scales themselves line up with four traits from the Big Five, the most studied model of personality: E/I with extraversion, S/N with openness, T/F with agreeableness and J/P with conscientiousness (McCrae and Costa, 1989). Most statements here, 65 of the 72, come from the International Personality Item Pool, a public research collection started by Lewis R. Goldberg, which is what made a like-for-like comparison of formats possible. MBTI is a registered trademark of The Myers-Briggs Company, and this site is not affiliated with it.
How we measured
These figures come from a simulation and lean in a known direction.
- Both sets were answered in the same sitting, so every row is an upper bound. A real retake weeks later agrees less.
- Letters were split at the middle of each format’s own results. That flatters yes/no formats.
- The short formats were built from validated research statements. Many quiz sites write their own questions instead, so the short rows are kinder to them than reality.
The last row is a model estimate, not a measured figure. The same model reproduced the random-12 row within half a point (78.0% against 78.3% measured). Note also that per-letter figures multiply: even at about 81% per letter, fewer than half of people get all four letters back. What that means for a result that “changes” is covered in How accurate is an MBTI test?, and the sources are on the method page.
In short
Count questions, but count the answer format too. Two per letter is too few; 12 on a five-point scale is about where the gains get small; a yes/no test needs more questions to reach less. And a test that admits a close call is worth more than one that asks twice as much and forces a side.
Seventy-two statements, about 10 minutes, nothing to sign up for: take the test.