Is the MBTI scientifically valid? What the research actually says

What the MBTI claims to measure

The MBTI was developed by Katharine Briggs and Isabel Briggs Myers from Carl Jung’s theory of psychological types. It sorts people on four two-way preferences: Extraversion or Introversion (E/I), Sensing or Intuition (S/N), Thinking or Feeling (T/F), and Judging or Perceiving (J/P). Combining one letter from each pair gives 16 four-letter types, such as INTJ or ESFP.

Two quite different claims are bundled together here, and it helps to judge them separately:

• A measurement claim: the four scales capture stable, meaningful differences between people.
• A typology claim: people come in 16 qualitatively different kinds, each with its own fixed pattern of mental “functions”, and your type is something you are rather than a set of scores you have.

The research verdict on these two claims is very different.

What holds up: the four scales measure real traits

The most influential independent test of the MBTI is a 1989 study by Robert McCrae and Paul Costa, two of the main architects of the Big Five model. They gave 468 adults both the MBTI and the NEO Personality Inventory, using self-reports and ratings by peers. Each MBTI scale correlated with a different Big Five trait: E/I with Extraversion, S/N with Openness to Experience, T/F with Agreeableness and J/P with Conscientiousness. The first two links were strong and the last two moderate. The authors concluded that the MBTI “did measure aspects of four of the five major dimensions of normal personality”.^[1] A British study found the same overall pattern.^[5]

Score consistency is also respectable. A meta-analysis of reliability estimates reported across published MBTI studies found that the scales generally yield scores with strong internal consistency and test–retest reliability, although values varied from study to study.^[2]

So the letters are not random. If you score clearly toward Introversion, you really are, on average, quieter and less socially assertive than most people, and that difference is measurable with other instruments and visible to people who know you.

What does not hold up: 16 distinct types

If people truly came in two kinds on each scale, scores should cluster in two humps, with few people in the middle. That is not what the data show. McCrae and Costa found “no support for the view that the MBTI measures truly dichotomous preferences or qualitatively distinct types”; instead, the instrument measures four relatively independent continuous dimensions.^[1]

Earlier reports of two-humped (bimodal) MBTI score distributions were later traced to a technical artifact. When Bess and Harvey re-scored about 12,000 people with modern item response theory and enough computational precision, the bimodality disappeared and the distributions became “strongly center-weighted”: most people scored near the middle.^[3]

This matters because a cut-off placed in the middle of a center-weighted distribution splits the most crowded region. Two people who differ by a single answer can receive opposite letters, while two people at opposite ends of one side share the same letter.

Here A and B are practically identical, yet one is “an introvert” and the other “an extravert”, while B and C share a letter despite being far apart. Continuous scores avoid this problem.

Why many people get a different type on retest

A common experience is taking the test twice and getting a different four-letter code. Reviews of the MBTI have flagged this classification instability as a major reason for caution in interpreting the type formula.^[4]

Instability follows directly from the center-weighted distribution. Every test score contains a little measurement error. For someone near a cut point, that small error is enough to flip a letter. And because the type needs all four letters to match, the chance that at least one letter flips is much larger than the chance for any single scale. Reliable scales can still produce unstable types.

Real change also plays a part. On average, people become more conscientious, more emotionally stable and more socially dominant between their 20s and 40s,^[6] so a J/P letter that was borderline at 19 may settle at 35. See Can your personality type change? for the evidence.

What the MBTI leaves out

In McCrae and Costa’s data, the fifth Big Five trait, Neuroticism (its opposite pole is emotional stability), was not captured by any of the four MBTI scales.^[1] That is a significant gap. Neuroticism is strongly linked to anxiety and depressive disorders,^[7] and it is the Big Five trait most strongly (negatively) related to job satisfaction.^[8]

Cross-language studies of personality vocabulary have also found a sixth broad factor, Honesty-Humility, covering sincerity, fairness, greed avoidance and modesty.^[9] It predicts ethical and cooperative behaviour beyond the Big Five,^[10][11] and neither the MBTI nor the Big Five measures it directly. Our guide to Honesty-Humility covers this trait.

What about cognitive functions?

Many online type communities interpret the letters through Jungian “cognitive functions” (for example, “introverted intuition”) stacked in a fixed order for each type. McCrae and Costa concluded that their data suggest Jung’s theory “is either incorrect or inadequately operationalized by the MBTI and cannot provide a sound basis for interpreting it”.^[1] Function stacks are not part of mainstream personality research, and we do not use them.

Why it feels so accurate anyway

Feeling that a description fits you is weak evidence that a test is valid. In a classic 1949 demonstration, psychologist Bertram Forer gave students what they believed were personalised profiles. Everyone received the same text, compiled largely from a newsstand astrology book, and the students rated it as a very good description of themselves.^[12] Later reviews showed that acceptance of such feedback rises when it is favourable, vague and presented as tailored to the person.^[13]

Type descriptions are mostly positive, written in general terms, and delivered as “your” type, which is the recipe for this effect. That does not make the underlying scales meaningless, but it means the sense of recognition you feel cannot settle the question. More in Are personality tests accurate?

How to use type language sensibly

How 64 Types handles these problems

We keep the familiar four letters because people find them useful, but we measure each one as a continuous Big Five trait using public-domain research items, report your percentile against roughly 20,000 real responses, and flag any trait within 10 points of average as borderline, showing the code you might get on a retest. We also add the two missing traits, emotional stability (C/R) and Honesty-Humility (M/A), giving 2⁶ = 64 types. The details are on our science page and in How the 64 Types test works.

Frequently asked questions