What is an INFP in Big Five terms?

You have had the four letters for years. They have been useful — they gave you language for something, and the description landed the first time you read it. Now you have run into the Big Five somewhere and everyone says it’s the one researchers actually use, and the obvious question is whether you have to start over or whether the thing you already know about yourself translates.

It mostly translates. The map is well established and has been since 1989. What is worth understanding is the one dimension that doesn’t appear on your map at all, and why your letters can change without you changing.

The short answer

Four of the MBTI’s dichotomies line up with four of the Big Five traits, with substantial correlations. From McCrae and Costa’s 1989 study of 468 adults:

MBTI dichotomy	Big Five trait	correlation
E–I	Extraversion	.74
S–N	Openness	.72
T–F	Agreeableness	.44
J–P	Conscientiousness	.49

(Those correlation values are as reported in the secondary literature; the original tables are paywalled.)

So an INFP translates, roughly, to: lower extraversion, high openness, higher agreeableness, lower conscientiousness.

And nothing on neuroticism. No MBTI index tracks it. That is not a small omission — neuroticism is arguably the Big Five trait most predictive of day-to-day wellbeing, and your four letters are silent on it.

The 1989 correlations held up when the Big Five scores came from peer ratings rather than self-report, which is a meaningful check: two people who know you agree about the same underlying dimensions.

Does the translation actually replicate?

Mostly yes — with one large and unexplained exception it would be dishonest to leave out.

The four-way pattern has been reproduced several times, including by Furnham and colleagues in 2003 on 900 people, and in earlier work in 1994 and 1996. That is a reasonably solid base.

But the largest study ever run on the question — Furnham (2022), 9,487 employed adults — largely failed to replicate it. E–I correlated with extraversion at only r = .20. S–N did not correlate with openness at all. Of 140 correlations between the MBTI and the Big Five domains and facets, only two exceeded r = .25.

Take that seriously, and also take it with some suspicion. An r of .20 between two instruments measuring introversion, sharing method variance, is implausibly small against the .74 found elsewhere — which points at something specific to that dataset, likely a range-restricted sample of managers in assessment centres, rather than a discovery that the relationship was never there. Furnham does not diagnose it. The honest position: the largest single study is also the most anomalous, and nobody has explained why.

If the dimensions are fine, what’s the problem?

The letters, not the questions. The MBTI’s scales are decent; the arithmetic that turns them into a type is where it goes wrong.

Start with the good news, because it is real. A meta-analysis of 70 reliability coefficients found the MBTI’s scales averaging .815, with E–I at .838, S–N at .843, J–P at .822, and T–F the only one below .80 at .764. Those are respectable numbers. The items work.

The problem is what happens next. Each scale produces a continuous score, and then the instrument cuts it in half to give you a letter. Cutting a continuous variable in two is a known, quantifiable loss: dichotomising one scale removes about 36% of the shared variance with anything you want to predict, and dichotomising both removes about 60%. Pittenger puts the MBTI’s own loss at 26–32%.

And the cut falls in the worst possible place, because most people are near the middle. Which brings us to the thing that bothers people most.

Why did my type change when I retook it?

Because your score moved slightly across a line, not because you changed.

The test publisher’s own figures say it plainly: “about 50 percent of people get the same four-letter or whole type on retest,” and 90% get the same result on three or four of the four scales. They frame this as about eight times better than chance, which is fair on its own terms — but it means a coin-flip’s chance of the label you were given being the label you get next time.

The peer-reviewed picture agrees. Pittenger (2005) — a critical review reporting other researchers’ primary data rather than new results of its own — collects the retest studies: across a five-week interval, 50% of participants received a different classification on one or more scales; the MBTI manual itself reported 35% with a different four-letter type after four weeks.

Crucially, the instability concentrates exactly where you’d expect. For people whose initial score was near the middle of a scale, the proportion whose label flipped was 32% on E–I, 25% on S–N, 29% on T–F and 30% on J–P. If you sit near a boundary — and most people sit near a boundary on at least one dimension — your letter is close to arbitrary.

This is the practical case for the dimensional version. A Big Five score of 43rd percentile on extraversion tells you that you’re slightly below average and how confident to be. “I” tells you you’re on one side of a line, and next month it may tell you you’re on the other.

But aren’t there really sixteen types?

On the evidence, no — and the strongest version of this argument comes from an unexpected place.

The theory requires that scores cluster at the two ends of each dimension. As Stein and Swan (2019) put it: “the most straightforward realization of the 16-type idea would be that the dichotomy scores should tend to have bimodal distributions. Infamously, they do not.”

The MBTI’s newer scoring did appear to produce bimodal distributions for a while. Bess and Harvey (2002) checked, on roughly 12,000 respondents, and found the apparent bimodality was a computational artifact — the scoring software’s default setting. Recompute it properly and the distributions are “strongly center-weighted”: one hump, most people in the middle.

A formal taxometric analysis of Jungian preferences reached the same conclusion: no genuine categories underlie them, only continuous dimensions. That is consistent with the broader literature — across 177 taxometric studies of 533,377 people, once confounds were controlled the estimated prevalence of genuine categories was about 14%, and normal personality came out dimensional.

Stein and Swan note something sharper still: the current scoring uses item response theory in a way that makes midpoint scores harder to obtain, on the assumption that type theory is correct. They call it “a worryingly circular double-down.”

So what do I lose, and what do I keep?

You keep more than the usual takedowns admit, and you gain one whole dimension.

Keep: the four dimensions are real and your scores on them are informative. Being drawn to ideas over concrete detail, preferring depth over breadth of company, weighting harmony over cold analysis — those are measurable things about you and the MBTI measured them tolerably well.

Lose: the letters, the sixteen boxes, and the sense that you belong to a kind. Also lose the idea that INFPs are a community with shared properties — on the evidence, they are a set of people who happened to land on the same side of four lines, many of them by less than a measurement error.

Gain: neuroticism. It is the single biggest practical difference. Two INFPs with identical letters can differ enormously in how much they worry, how hard setbacks land and how quickly they recover — and that difference will shape their lives more than the four letters they share. Your MBTI profile has never said anything about it.

If you want the closest thing to a direct translation, take a Big Five assessment and read your extraversion, openness, agreeableness and conscientiousness scores as the dimensional versions of your four letters. Then read the fifth one, which is new information.