The science behind 21 Questions

A meta-analysis of 263 samples covering 44,178 people found that ratings from friends, family and others agree with each other, line up with self-ratings, and predict real behaviour, sometimes better than self-ratings do. How close the rater is to the person matters most.

So in 21 Questions: your Mirror only counts people connected to you, and whose view counts more is up to you: you pick who knows you better, without seeing what anyone answered.

• Connelly & Ones (2010). An other perspective on personality: Meta-analytic integration of observers' accuracy and predictive validity. Psychological Bulletin, 136(6), 1092–1122. doi:10.1037/a0021212
• Ashton & Lee (2025). Self- and observer reports of personality. Annual Review of Psychology, 76, 771–795. doi:10.1146/annurev-psych-020124-115044

The self is the best judge of inner traits like anxiety, while friends are better judges of others, like how sharp you come across. Neither view is complete on its own.

So in 21 Questions: you answer about yourself too, and the Mirror shows where the two views meet and where they don't.

• Vazire (2010). Who knows what about a person? The self–other knowledge asymmetry (SOKA) model. Journal of Personality and Social Psychology, 98(2), 281–300. doi:10.1037/a0017908

Researchers found traits that several people agree describe someone, which that person doesn't recognise in themselves. The finding was replicated in a follow-up study.

So in 21 Questions: the Mirror's blind spots are traits where the people who know you agree with each other, and disagree with you.

• Gallrein, Carlson, Holstein & Leising (2013). You spy with your little eye: People are "blind" to some of the ways in which they are consensually seen by others. Journal of Research in Personality, 47(5), 464–471. doi:10.1016/j.jrp.2013.04.001
• Gallrein, Weßels, Carlson & Leising (2016). I still cannot see it: A replication of blind spots in self-perception. Journal of Research in Personality, 60, 1–7. doi:10.1016/j.jrp.2015.10.002

Two acquaintances' combined judgments predicted people's actual behaviour better than their own self-judgments.

So in 21 Questions: trait results only show once at least 3 people have answered, and only combined. (A few fun "most likely to" picks show on their own, days later, never with a name.)

• Kolar, Funder & Colvin (1996). Comparing the accuracy of personality judgments by the self and knowledgeable others. Journal of Personality, 64(2), 311–337. doi:10.1111/j.1467-6494.1996.tb00513.x

Opening up is tied to liking: people like those who open up to them, and like people more after opening up to them. Taking turns at it builds liking between new acquaintances, and in one study a set of gradually deeper questions created more closeness than small talk.

So in 21 Questions: the quiz is a back-and-forth: friends guess you, then you guess them.

• Collins & Miller (1994). Self-disclosure and liking: A meta-analytic review. Psychological Bulletin, 116(3), 457–475. doi:10.1037/0033-2909.116.3.457
• Aron, Melinat, Aron, Vallone & Bator (1997). The experimental generation of interpersonal closeness. Personality and Social Psychology Bulletin, 23(4), 363–377. doi:10.1177/0146167297234003
• Sprecher, Treger, Wondra, Hilaire & Wallpe (2013). Taking turns: Reciprocal self-disclosure promotes liking in initial interactions. Journal of Experimental Social Psychology, 49(5), 860–866. doi:10.1016/j.jesp.2013.03.017

21 Questions is a game informed by this research. It is not a clinical or diagnostic tool, and it isn't a validated psychological test: our traits, questions and scoring are our own, and they haven't been independently validated yet. The papers above support the ideas we build on, not 21 Questions itself, and their authors aren't affiliated with us.