MBTI and Personality Trait Correlations Explorer

Methodology

The primary data source for these infographics are people who took the Statistical "Which Character" Personality Quiz on this website between September 2019 and February 2026. In that survey people are asked to rate themselves on contrasting pairs of traits, a screenshot of how that looked was below.

This survey had an item bank of 525 different items, although the most any one user could answer in one session was 245 (and most answered far fewer). The items were designed to be extremely diverse and cover a wide range of topics. This is good because it makes the data very interesting to explore.

The items were rated by the user on a 1 - 100 point scale, but for this data explorer those responses have been recoded to a binary classification. Ratings < 45 were categorized as the description on the low pole of the scale, ratings > 55 were categorized as the description on the high pole of the scale and ratings between 46 - 54 were recoded as NA (excluded from the analysis).

At the end of the SWCPQ, users we asked if they would be willing to complete an optional survey to help make the test better. About 35% volunteer. Several different supplemental surveys are distributed here, one of which includes an option where the user could self-report their MBTI type. Shown below.

Using this survey there have been 301,005 people who self-ID'ed a MBTI type. This is likely the largest sample size of Myers-Briggs types ever. However, each person only got a subset of all the possible trait questions so effective sample size is lower ranging from n=2,273 to n=243,632. The sample size per trait varies so much for two reasons: (1) the item bank expanded over time, items that have been in the quiz for longer have more responses and (2) items that were more useful for the SWCPQ matching algorithm were also more likely to be given to the users.

A secondary data source was also added to increase sample size. My personality test for couples uses the items from the same bank in the same format and also included an option to self-report MBTI type. This survey has data for 233,580 people who self-reported an MBTI type, although the max length of this survey was only 56 items per person (significantly less than on the SWCPQ). This survey brings us up to a total sample size of more than 500,000 (!!!).

Considerations on interpretation

This is self-report data

Remember that this is self report data, so think critically about what someone describing themselves in a way could really mean. It's not always what it says on the tin. If you look at the graphs for [High IQ] vs [Low IQ], you see that INTJs are the most "High IQ".... But just because they think that, does not neccesarily mean its true. For example, you could imagine a world where all types have equal IQs and only vary in self-reported IQ because the types have different levels of delusional self-importance.

The survey was not "population representative"

A common way that people ended up discovering the SWCPQ was because they saw it going viral on TikTok, so this survey is somewhat biased towards the TikTok demographics: more liberal, more female, more self diagnosed autism etc.

Type averages contain error

The first and most important thing not to do with these charts is overinterpret noise. This is the easiest thing for humans to do. Because of random variation in who takes the test and random variation in how people were feeling when they did, the percentage of a given type endorsing themselves a trait would necessarily be different in the survey were run a second time. Maybe 84% of 296 ESTPs categorized themselves as smug (vs. sheepish) in this survey. But if you ran it the survey again with a similar number of people maybe only 70% of the ESTPs would identify as smug. The graph below, shows the average difference you should expect between the the average score of a small sample compared to the "population mean".

Data source

Some of the data used to construct these infographics are available for download.