Articles · Research

Is it accurate? Mostly not, and still useful.

它准吗?大体不准,但仍然有用。

Two things are true at once: the instrument has a real reliability problem, and the vocabulary it gives people is genuinely useful. The trick is not confusing the second for the first.

If you ask whether the MBTI is “accurate”, you get two camps shouting past each other: one citing peer-reviewed reviews, the other citing a decade of personal recognition. Both are describing something real. They are not describing the same thing.

The criticism that lands

Test–retest reliability. Studies that re-administer the instrument a few weeks apart routinely find that a large share of respondents — around half in several widely cited samples — land on a different type the second time. For a system whose central claim is “you are one of sixteen”, that is a serious problem, not a rounding error.

Bimodality. The scoring treats each dimension as two categories. But the underlying scores are roughly continuous and clustered near the middle, which means a great many people are being sorted by a coin-flip’s worth of difference.

The Barnum problem. Type descriptions are warm, general and flattering. Almost everyone recognises themselves in “values harmony, needs downtime, sometimes struggles to say no”. That recognition feels like evidence. It is not.

Predictive validity. For predicting job performance, the Big Five — five continuous dimensions, freely available in the public domain — consistently does better. That is why academic personality psychology moved on decades ago.

The defence that also lands

The four letters are a compression format, not a measurement. If you treat them as a measurement you will be misled; if you treat them as a vocabulary you can get something real out of them.

What they are good for: naming a difference before it becomes an accusation. “You decide on consistency and I decide on impact” is a sentence a couple can use. “You are a T and I am an F” is the same sentence with less blame in it.

They are also good for memory. A typed description is easier to hold than five percentile scores, and easier to discuss with someone who has not read the manual.

Three rules for using it without fooling yourself

  1. Answer about the last month. Not the person you are at your best, not the person you were at nineteen. Recency reduces the aspiration bias that drives most of the retest flips.
  2. Score dimensions, not just the code. “62% toward I” and “51% toward I” are different claims. A near-the-middle score means the letter should carry almost no weight in how you describe yourself.
  3. Never use it to explain away behaviour. “I can’t do deadlines, I’m a P” is not a type fact. It is a preference being promoted into an excuse, and it is the single most common way this framework makes people worse.

What a better instrument looks like

Continuous scores on each dimension, reported with the uncertainty that a self-report deserves. Multiple frameworks rather than one, so that a weak reading in one place can be checked against another. And a plain statement of what the tool is for: reflection, not diagnosis, not hiring, not a verdict on your marriage.

That is the standard we hold ourselves to. It is also why the reports show scores rather than only a code, and why the copy on this site keeps saying “reflective tool” instead of “scientific personality assessment”.

The short answer

No — not in the sense a psychometrician means. It will not survive a retest reliably, it will not predict performance better than the Big Five, and a lot of what feels accurate is a well-written sentence doing its job.

Yes — in the sense that matters for self-reflection. It gives you a compact vocabulary for real differences in how people take in information and make decisions, and that vocabulary is more useful than the vague sense that “we are just different”.

Hold both. Use the words, distrust the certainty.

Keep reading