How reliable is MBTI? What the research says, and how to use type anyway

The 16-type framework is used by millions and criticised by most personality researchers. Both facts are compatible: it is a useful vocabulary and a weak measurement. This page summarises the main findings without hostility or hype, and ends with a way to use type descriptions that survives the criticism.

The four main criticisms

1. Test-retest: your type changes

Reviews in the 1990s (Pittenger, 1993) reported that a large share of people, on the order of half, received a different four-letter type when retested after about five weeks. Most of the change comes from people near the midpoint of a scale: a small shift in answers flips the letter. The Myers-Briggs Company reports higher agreement for the full instrument administered by trained practitioners; free online tests have no published figures. Either way, the letters are less stable than the descriptions imply.

2. Dichotomies that are not dichotomies

Scores on each MBTI scale are distributed roughly like a bell curve, with most people in the middle, not in two clusters. Cutting a continuous trait at the midpoint discards most of the information and makes the "I vs E" distinction meaningless for the many people near the centre (Pittenger, 2005).

3. It overlaps with a better-validated model

McCrae and Costa (1989) showed that four of the MBTI scales correlate with four of the Big Five traits (extraversion, openness, agreeableness, conscientiousness), while the fifth Big Five trait, neuroticism, is missing from MBTI entirely. The Big Five has decades of evidence for predicting outcomes (job performance, health behaviours, relationship satisfaction); the MBTI's predictive evidence is thin. Online "A/T" suffixes exist to patch exactly this gap (see INFJ-A vs INFJ-T).

4. Descriptions are written to feel true

Type profiles are positive, general and flattering, which produces the Barnum effect: people rate vague favourable descriptions as accurate for themselves. The sense of recognition when reading your type is real; it is just not evidence that the type is correct.

What the framework does well

• Vocabulary. "I'm an introvert who needs a plan" is a compact, non-pathologising way to explain a preference to a partner or team.
• Legitimising difference. In workplaces, type language gives people permission to work differently without one way being "wrong".
• Starting reflection. A description that is 70% right is a useful prompt for working out the other 30%.

How to use type without fooling yourself

1. Treat the four letters as a rough summary of continuous traits; if you are near the middle on one, say so ("I lean I").
2. Prefer the trait language of the Big Five when the stakes are real (hiring, clinical decisions): it has the evidence.
3. Never use type to predict what a specific person will do; use it to ask them.
4. Read the descriptions critically: what is observable about you, what is interpretation, what is flattery.

With that in mind, the profiles on this site are written as descriptions of preferences with their costs, not as destinies: INFJ · INTJ · INFJ vs INTJ.