What can these inventories actually tell me?
A percentile is a comparison, not a measurement
Your raw score is what you answered. The percentile places it against a particular group of people, and changing the group changes the percentile without changing anything about you.
The comparison group is self-selected
Everyone in it found a personality inventory online and chose to take it. No sample here is random, and a larger sample does not fix that — it measures the same skewed population more precisely.
Low, average and high are bands, not types
The cut is at the 30th and 70th percentiles. Two people either side of a boundary get different words and are barely different. Nothing here sorts you into a kind of person.
Scores move between sittings
Retake an inventory in a month and the numbers will shift a little; a score near a boundary can cross it. Treat a single result as one reading, not a fixed quantity.
These traits are descriptive, not diagnostic
They summarise how people describe themselves. They are not a clinical assessment, they do not diagnose anything, and no score here is a finding about your health.
They predict groups better than individuals
Trait scores show real average differences across large numbers of people. They are weak at saying what any one person will do on any given day, and should not be used to decide who gets hired or admitted.
Who am I being compared against?
A percentile says where you fall among the people in the reference data — not among people in general. That distinction does more work than it looks like it does.
Everyone in that data found a personality inventory on the internet and chose to take it. Nobody was sampled at random. People who spend more time online, who enjoy introspection, or who were taking a psychology course at the time are over-represented, and there is no way to measure by how much.
A bigger sample does not fix this. A million self-selected respondents measures the same skewed population more precisely. It does not turn it into the general population.
- Mini-IPIP — 1,009,226 respondents, collected 2016-2018.
- MFQ-2 — 5,312 respondents, collected 2021-05 (national samples), 2021-09 (online panel).
- IPIP Big-Five Factor Markers — 1,009,184 respondents, collected 2016-2018.
- IPIP-NEO-120 — 619,150 respondents, collected 2001-2011.
- IPIP-HEXACO-240 — 22,786 respondents.
- IPIP-NEO-300 — 307,313 respondents, collected 2001-2011.
So read “75th percentile on Extraversion” as more extraverted than 75% of the people who took this inventory online. That is a real comparison and a useful one. It is not the same claim as “more extraverted than 75% of people”.
This is not a flaw peculiar to us. Most published personality norms work the same way, including commercial ones, which often do not say how their samples were recruited at all. We would rather tell you.
Why does changing sex or age change my result?
Because it changes who you are being ranked against, not what you answered. Your raw score never moves. Personality scores differ systematically between groups — women score higher on Neuroticism and Agreeableness on average, and Conscientiousness rises through adulthood while Neuroticism falls — so the same raw score can be unremarkable in one group and high in another.
Neither ranking is more true. They answer different questions. This is also why the raw score is always shown next to the percentile, and why you can compare against everyone rather than only against your own group.
What do low, average and high mean?
They are bands of the distribution: low is below the 30th percentile, average is the middle 40%, high is above the 70th. They are not kinds of person. Someone at the 69th percentile and someone at the 71st are described differently and are barely different.
Scores near a band edge are the least stable part of any result. If you retook an inventory next month the numbers would move a little, and a score sitting on a boundary can move across it.
Is this the same as Myers-Briggs or 16Personalities?
No, though they measure overlapping things. Those tests sort you into one of 16 types by cutting each trait in half. This one reports where you fall on each trait as a continuous score.
The difference matters because these traits are not naturally divided into two camps — they are spread smoothly, with most people bunched in the middle, which is exactly where a cut does the most damage. That is why retaking a four-letter test so often returns different letters: a score near the middle flips on ordinary measurement noise.
Why “inventory” and not “test”?
Because they measure different kinds of thing, and the distinction is worth keeping.
A test has correct answers and you try to do well on it: an exam, an IQ test, a driving test. Your score says how close you got to right. An inventory has no correct answers and asks what you are usually like. Your score is a description, not a grade. The word comes from the same root as a stocktaking — a list of what is on hand.
This is not a distinction we invented. Almost no serious instrument in this field calls itself a test: the NEO Personality Inventory, the Big Five Inventory, the Minnesota Multiphasic Personality Inventory, the HEXACO Personality Inventory, the Sixteen Personality Factor Questionnaire. Even the Myers-Briggs Type Indicator avoids it, and deliberately — it was named to suggest that it points at a preference rather than measuring an ability.
In ordinary speech everyone says “personality test”, psychologists included, and nobody needs correcting for it. We use the precise word here because this site is about reading these numbers carefully, and half of that is knowing what kind of number you have. The one place you will still see “test” is where we are talking about the four-letter type products, which is what they call themselves.
Where do the questions come from?
The International Personality Item Pool, a public-domain set of items written by Lewis Goldberg so that personality measurement would not depend on commercial instruments. The selection and the reference data come from John A. Johnson's published work.
The items are phrases rather than sentences by design — each one completes an implied “I …”.
The one exception is the MFQ-2, the Moral Foundations Questionnaire-2 by Mohammad Atari, Jonathan Haidt, Jesse Graham and colleagues. Its items are full sentences, are not public domain, and are used here under the Creative Commons Attribution licence the authors published them with. The reference data for it is theirs too, from the paper that introduced it.
The narrative text describing each trait is after John A. Johnson for the IPIP-NEO forms, and written for this service for the others.
Which inventory should I take?
- Mini-IPIP — 20 items, about 3 minutes.
- MFQ-2 — 36 items, about 5 minutes.
- IPIP Big-Five Factor Markers — 50 items, about 7 minutes.
- IPIP-NEO-120 — 120 items, about 15 minutes.
- IPIP-HEXACO-240 — 240 items, about 30 minutes.
- IPIP-NEO-300 — 300 items, about 35 minutes.
Longer forms ask more questions per scale, so their scores move around less; shorter ones cost less of your afternoon. Each is scored against people who took that same form, never against a different one.
They do not all measure the same things. The IPIP-NEO forms and the Big-Five Factor Markers measure five traits; the IPIP-HEXACO measures six, adding Honesty-Humility, which the five-factor model has no scale for. Watch the direction of the fourth trait too: some forms report Neuroticism and others report Emotional Stability, which runs the opposite way. Each result says which it is reporting.
The MFQ-2 is not a personality inventory at all. It measures how much weight you give six moral concerns — Care, Equality, Proportionality, Loyalty, Authority and Purity — and its scores are not comparable with any trait on the other forms.