How the 64 Types test works: items, scoring, norms and borderline flags

The six letters

A 64 Types code looks like INTJ-CM: four familiar letters, a hyphen, then two new ones. Each position is one trait.

The mapping of the first four positions follows research showing that the MBTI’s E/I, S/N, T/F and J/P scales correlate with Big Five Extraversion, Openness, Agreeableness and Conscientiousness respectively.^[2] Emotional stability is the Big Five trait that no MBTI scale captures,^[2] and Honesty-Humility is the sixth factor of the HEXACO model.^[4] Six two-way letters give 2⁶ = 64 types. See MBTI vs Big Five and Honesty-Humility.

Where the 62 questions come from

All items come from the International Personality Item Pool (IPIP), a public-domain collection that anyone may use for any purpose, including commercially:

• 50 items are the IPIP Big Five factor markers developed by Lewis Goldberg, 10 per trait, for Extraversion, Openness, Agreeableness, Conscientiousness and Emotional Stability.^[1]
• 12 items come from the IPIP-HEXACO Honesty-Humility scales, three for each facet: Sincerity, Fairness, Greed Avoidance and Modesty.^[3] We chose the three items per facet with the highest corrected item–total correlations in a public dataset of 22,782 responses.

You rate each statement, such as “I am always prepared” or “I worry about things”, from 1 (strongly disagree) to 5 (strongly agree). Half the items (31 of 62) are reverse-keyed: agreeing with “I leave my belongings around” lowers your Conscientiousness score. Mixing directions reduces the effect of simply agreeing with everything.

Step by step: from answers to percentiles

1. Reverse-key. For reverse-keyed items, an answer of 5 becomes 1, 4 becomes 2, and so on.
2. Average. Your raw score on each trait is the mean of its items, from 1 to 5.
3. Compare with real people. The raw score is located within the distribution of the norm sample. We store each trait’s percentiles from 0 to 100 and interpolate between them, so your result is the percentage of the norm sample scoring below you.
4. Assign the letter. At or above the 50th percentile you get the higher-pole letter (for example E); below it you get the lower-pole letter (I).
5. Flag borderlines. If your percentile is within 10 points of 50 (between the 40th and 60th), the trait is marked borderline and we show the alternative code with that letter flipped.

Scoring happens entirely in your browser. Your answers are not sent to a server (see our privacy page).

Why percentiles, not raw scores

Raw scores are misleading because people do not use the scale evenly across traits. In our norm data the average Agreeableness score is 3.85 out of 5, while the average Extraversion score is 3.01. As a result, the same raw score means very different things:

Without norms, a 3.8 would look “high” on both. With norms, it is clearly extraverted but slightly below average in Agreeableness.

The norm samples and their limits

Big Five percentiles come from an Open Psychometrics public dataset of 19,718 online volunteers who completed the same 50 IPIP items. Honesty-Humility percentiles come from a public IPIP-HEXACO dataset of 22,782 responses, collected on a 7-point scale and linearly rescaled to our 5-point scale. Both were gathered around 2012.

Reliability (Cronbach’s α) ranges from .79 to .89, which is good for short research scales. The main limitation is the sample: online volunteers who skew young (median age about 22) and English-speaking. Your percentile means “compared with people who take online personality tests”, not “compared with everyone on Earth”. We plan to re-norm on our own users and publish the results.

Borderline flags: the honest part of the result

Scores on personality traits are center-weighted: most people are near the middle, and the distributions are not split into two camps.^[5][2] Anyone near the median can get a different letter on another day simply because of normal measurement error. This is the main reason four-letter types “change” on retest.^[6]

Instead of hiding that, we show it. If your Conscientiousness is at the 55th percentile, you will see your code, a borderline marker on J, and the code you would have with P instead. Read the percentile, not just the letter. By definition 20% of people are in the 40–60 zone on any given trait, so it is normal to have one or two borderline letters.

Where the profile content comes from

The 16 base names (such as The Strategist for INTJ) and the four variants (Steady & Sincere, Steady & Striving, Sensitive & Sincere, Sensitive & Striving) are our own descriptions. Everything we say about careers, relationships, friendships and wellbeing is built from trait-level research and cited on each type page, with uncited lines clearly presented as practical tips. Career suggestions use meta-analytic links between personality and vocational interests.^[7]

We do not use “cognitive functions”, because they lack empirical support.^[2] We also do not produce compatibility percentages, because personality similarity explains very little of relationship satisfaction.^[8] Our compare tool shows differences instead.

The optional Deep Report: 29 facets

Each Big Five trait is made up of narrower facets, and two people with the same code can differ underneath. The optional Deep test uses 116 items from the public-domain IPIP-NEO-120, which measures 30 facets with four items each.^[9] We omit the Liberalism facet because its items ask about voting for liberal or conservative candidates. Percentiles come from Johnson’s public dataset of 410,376 complete responses, where facet reliabilities range from about .66 to .88, lower than for the broad traits because each facet has only four items.

The Deep Report costs $7.99 on the web; your five most distinctive facets are shown free. It is useful if you want to understand why you got a particular letter, for example whether low Emotional Stability comes from anxiety, anger or low mood.

What the test is not

• It is a self-report measure of how you see yourself. People who know you well can add information, especially about evaluative traits such as intellect.^[10][11]
• It is not a clinical or diagnostic tool and not a hiring instrument. Trait–outcome links are real but modest.^[12][13]
• It is not a fixed label. Traits shift over the lifespan and with deliberate effort.^[14][15]

Frequently asked questions