Original research
Only about one in five people are average on all five Big Five traits. Most of us sit near the top or bottom of at least one, and half of us on two or more. Here is how we know, and where the numbers are weak.
The short answer
We looked at 68,988 results from our free Big Five test. On each trait, we marked the outer part of the range, roughly the lowest and highest sixth of scores, as an edge. Then we counted how many results avoided every edge.
inside the ordinary middle range on every trait
in the outer third of scores on one trait or more
about half of all results
| Trait | Ordinary range | At an edge |
|---|---|---|
| Openness | 37 to 51 | 30.8% |
| Conscientiousness | 30 to 48 | 33.4% |
| Extraversion | 28 to 46 | 33.1% |
| Agreeableness | 34 to 49 | 31.8% |
| Sensitivity | 28 to 48 | 33.9% |
Even a middle score has a shape
Each Big Five trait is built from three smaller parts, called facets. Extraversion, for example, is made of sociability, assertiveness and energy level. Two people with the same Extraversion score can get there in very different ways.
In 48.9% of results, at least one trait had one facet in the top 30 percent and another facet in the bottom 30 percent of everyone who took the test. So about half of people have a trait whose parts point in opposite directions. That is how someone can be easy with friends and still dread speaking up in a group.
A middling trait score, in other words, often hides two strong things. Your report shows the facets under each trait, so you can see which parts carry the score.
How we drew the lines
For each trait we found the score at the 15th percentile and the score at the 85th percentile of these results. A result at or below the first, or at or above the second, counts as an edge.
Scores are whole numbers, so many people tie exactly on a cut-off score. We count those ties as edges. That is why between 30.8% and 33.9% of results sit at an edge on each trait, rather than exactly 30 percent. The table above gives every cut-off.
If the five traits were unrelated to each other, about 86 percent of results would be at an edge somewhere. We see 79 percent. The gap comes from traits going together: someone high on one trait is a little more likely to be high or low on another in a matching way, so edges cluster in the same people.
The headline does not depend on the exact line. Draw it more tightly, so that ties at a cut-off count as ordinary, and 28 percent are average on all five instead of 21. Either way, being ordinary on every trait at once is the exception.
One more finding
In about one test in ten, we time every answer. People answer “Worry a lot” in about 2.1 seconds, and “Tend to find fault with others” in about 4.5.
Both are unflattering things to say about yourself, but only one is about other people. “Tend to find fault with others” also draws one of the highest rates of neutral answers on the test, more than one answer in four (27.9%). People seem to hesitate before admitting they judge.
The slowest statement, “Am inventive and find clever ways to do things”, is a different kind of hard: people weighing a compliment rather than avoiding a criticism.
Where these numbers are weak
The study covers 68,988 distinct results from English result pages, 23 June to 8 September 2026. A result is one completed test, not one person: someone who took the test twice can count twice, and two people with identical scores count once. We first reported this sample as 68,909; a recount gave 68,988, with the same shares to one decimal place.
Each result is stored in its own results link as fifteen facet scores, so a profile can be recovered exactly without needing names or emails. None of these results carries an age, gender or name, which is why there is no breakdown by age here. The test has asked for age, optionally, since September 2026.
Everyone here chose to take our online personality test. Compared with the published norms for the questionnaire, these results run lower on Openness by about 0.44 of a standard deviation and lower on Agreeableness by about 0.33. So treat the shares as a description of people who take this test, not of all adults.
When you take the test, each trait is turned into a percentile against 58,401 results from people who took it here between 23 June to 1 September 2026. That is a slightly smaller, earlier slice of the same data, with straight-line answer patterns removed.
The 60 statements are adapted from the BFI-2 (Soto and John, 2017) and scored the way the BFI-2 is: five traits, each built from three facets of four statements. You can download the summary data as a CSV: per-trait score percentiles, averages, edge cut-offs and edge shares, and the timing medians. It holds aggregates only, with no personal data.
Your turn
See where you land.
Sixty statements, about five minutes, and your full report is free. Your scores are compared with 58,401 results from people who took this test here.