Explainer
The traits you can't see in yourself
Every personality test asks you to rate yourself. For some traits that works fine. For others — reliably, predictably — the people who know you are more accurate than you are.
Every personality test you have ever taken shares one assumption: that you are a reliable witness to yourself. You answer the questions, the test scores the answers, and whatever comes out is treated as a measurement.
It mostly works. But it doesn't fail randomly — it fails in a specific, predictable pattern, and the traits where it fails hardest are the ones you would most want to know about.
The pattern
The idea, from work on self–other knowledge asymmetry, is that two things determine whether you or an observer knows you better:
How visible the trait is. Whether it shows up in behaviour other people can see. Talkativeness is visible. Anxiety is often not.
How evaluative it is. Whether having it is flattering or unflattering. Being organised is fairly neutral. Being arrogant is not.
You have the advantage on traits that are invisible and neutral — nobody can see your inner life and you have no particular reason to lie to yourself about it. Observers have the advantage on traits that are visible and evaluative, because they can see the behaviour and have no stake in your self-image.
Self–other agreement follows exactly this shape. It runs around r = .72 for agreeableness, which is highly visible, and drops to about r = .42 for intellect and openness, which are much less so.1
Informant reports are more internally consistent than self-reports on most personality dimensions — and for pathological traits, each source predicts things the other misses.2 3
That last point matters: it isn't that observers are simply better. The two views carry different information, and using only one throws half of it away.
Where it bites hardest
Put the two dimensions together and one region of the map is obviously worst: traits that are highly visible to others and highly unflattering to admit.
That is antagonism — hostility, callousness, assuming bad intent. It is the trait the people around you could describe in a sentence, and the one you are least equipped to see in yourself, because seeing it accurately carries the highest cost.
A meta-analysis of self- and informant reports across roughly 33,000 people found the two sources agree on average levels surprisingly well — people are not uniformly flattering themselves relative to how others rate them.4 The divergence isn't a blanket bias. It's concentrated in particular traits, and in particular people.
This isn't "other people know you better"
The asymmetry runs both ways. On internal experience — anxiety, rumination, how much effort something took — you have information observers have no access to at all. Neither source is authoritative. The useful move is knowing which one to trust for which trait.
A second problem: your current state contaminates the answer
There's a related failure that has nothing to do with honesty.
If you take a personality test during a depressive episode, your scores come out more pathological than they do before it starts or after it lifts. Part of that is genuine overlap between how neuroticism is measured and how depression is measured. Part is mood-congruent recall: feeling bad makes bad examples easier to remember.5
The effect isn't uniform across traits. Neuroticism, extraversion and interpersonal sensitivity all shift toward the normal range as depression improves, while conscientiousness, openness and agreeableness stay roughly put.6 Some elevation persists into remission, so it isn't purely an artefact — but a meaningful chunk of it is state, not trait.
The practical consequence: a test taken during a bad month is partly measuring the month.
What a test can do about it
Three things, none of which require you to become a better witness:
Say which scales are state-sensitive. If a trait is known to move with mood, the result should say so and suggest retaking when things are steadier — rather than silently adjusting the number. Adjusting would remove real variance, because people high in negative emotionality genuinely do have more bad periods.
Ask someone who knows you. The same items, worded in the third person, sent to two or three people. The interesting output isn't their rating — it's the gap between theirs and yours. "You put yourself mid-range on antagonism; two people who know you put you in the top 15%" is close to the most useful sentence a personality test could produce, and almost none of them attempt it.
Prefer evidence over self-concept in the questions themselves. "I am blunt to the point of unkindness" asks you to agree to something unflattering. "People have told me I'm blunt to the point of unkindness" asks about an event that either happened or didn't. The second is easier to answer honestly and harder to argue yourself out of.
How to read your own result
When a personality profile comes back, sort the scales into two piles: things only you can see (worry, rumination, effort) and things everyone around you can see (how you handle conflict, whether you follow through). Trust your scores more in the first pile. For the second, the honest move is to ask someone.
Why this matters for any test that claims biology
Tests that claim to read your brain chemistry from a questionnaire inherit this problem and then compound it. The answers come from the same imperfect self-report — and are then treated as a measurement of something physical, which lends them a precision the input never had.
This test has the same limitation. Its questions are self-report, its antagonism scale is the weakest in the bank for exactly the reason above, and that is stated in the method notes rather than glossed over. An informant module is the fix. It isn't built yet.
Sources
- Self-Other Knowledge Asymmetry extended to pathological traitsSelf-Other Knowledge Asymmetry extended to pathological traits ↩
- Informant vs self report internal consistencyInformant vs self report internal consistency ↩
- Self-other knowledge asymmetries in personality pathologySelf-other knowledge asymmetries in personality pathology ↩
- Self-other agreement meta-analysis (N≈33,000)Self-other agreement meta-analysis (N≈33,000) ↩
- Influence of depressive state features on trait measurementInfluence of depressive state features on trait measurement ↩
- Decoupling personality and acute psychiatric symptomsDecoupling personality and acute psychiatric symptoms ↩