How accurately do you know what you look like to other people?

You look in the bathroom mirror before work and think, “I’m better-looking than I give myself credit for.” Later, a phone photo catches you at the wrong angle and you think the opposite.

Which version is right? The measured answer is that your self-rating and strangers' ratings overlap only modestly: Feingold's meta-analysis found a weighted correlation of .24 between them across male, female and combined samples. Feingold 1992 Your opinion carries signal. It is not a substitute for the way unfamiliar people read you. (If the mirror-and-camera half of this is what you actually came for, that question lives in mirror vs photos.)

That result is less insulting than the internet's usual verdict, and more useful. You are not deluded, and you are not secretly objective. You are answering a different question from the panel of strangers.

Key numbers

• .24 — weighted mean self–judge correlation for physical attractiveness. Feingold’s meta-analysis
• 21 studies, 43 independent samples, N = 5,226 — the evidence behind that self-versus-judge estimate. Feingold’s meta-analysis
• .54 vs .83 — strangers’ weighted interrater correlation versus average composite reliability. Feingold’s meta-analysis
• 48% / 52% — common preferences versus individual face preferences in one large twin-based analysis. Germine et al.
• 100 ms — an exposure long enough for attractiveness judgments to correlate .69 with judgments made without a time limit. Willis and Todorov
• .08 — the self-rating/interviewer-rating correlation in a nationally representative sample of 4,882 emerging adults. Ehlinger and Blashill

Table of contents

• How closely does self-rating match what other people say?
• Do people overrate themselves?
• Do strangers even agree with each other?
• Who is the furthest off?
• Are you a good judge of your own photos?
• So what is the number worth?
• What this literature cannot tell you
• What we could not verify
• How to cite this page
• The bottom line
• Sources

How closely does self-rating match what other people say?

Not very closely, but not randomly either. The core estimate is a correlation of .24: men’s and women’s self-ratings tend to move in the same broad direction as judge ratings, while leaving plenty of room for a personal miss.

The estimate rests on 21 studies, 43 independent samples and N = 5,226 people; Feingold described self-rated and judge-rated attractiveness as largely orthogonal. Feingold 1992 Knowing your own rating gives a stranger some information — not enough to reconstruct his judgment. How that kind of correlation should be read is unpacked in our attractiveness research methods glossary.

That is a population relationship, not a verdict on your face. One man is roughly calibrated, the next rates himself far too harshly, and the average cannot tell you which one you are.

Do people overrate themselves?

Often there is a modest flattering tilt, but “everyone overestimates his looks” is not the measured conclusion. Langlois and colleagues found attractive adults saw themselves more favourably overall, and the effect was small: d+ = .26, or 56% versus 44%. Langlois et al. 2000

The direction depends on what is being measured. Epley and Whitchurch morphed each person's photo toward an attractive and an unattractive target in 10% steps; the versions people picked as their real face averaged 6.30%, 12.00% and 13.33% toward the attractive end across experiments. The same tilt appeared for a close friend's face (10.09%) and not for relative strangers (1.3% and 3.3%, non-significant). Epley & Whitchurch 2008 That is recognition shaped by familiarity, not proof that every self-rating is inflated.

The gap points either way at the individual level, and a correlation of .24 is far too modest to justify a universal correction in either direction.

Do strangers even agree with each other?

Yes — more than you agree with them, and still not unanimously. Feingold reported a weighted interrater correlation of .54 and an average composite reliability of .83, against the .24 you get for self-versus-others. Feingold 1992

A twin study split it further: 48% of the variation in face-attractiveness ratings came from common preferences and 52% from individual ones, while the mean ratings given by men and women correlated at r = 0.99. Germine et al. 2015 Faces carry a shared signal; the observer still supplies half the answer.

The headline moves with the method. Large panels make agreement look impressive, and Kramer and colleagues showed why: a trait with an inter-rater agreement of just 0.10 reaches a Cronbach's alpha of 0.90 once 80 raters are used. Kramer et al. 2018 Hönekopp's beholder index makes the same point from the other side — it swings from .84 with homogeneous faces to .35 with heterogeneous ones, so any single “X% of beauty is objective” figure is partly an artefact of which faces were shown. Hönekopp 2006

The honest picture is neither “beauty is completely objective” nor “everyone sees a different face.” Both common perception and personal preference are present.

Who is the furthest off?

There is no single psychological type at the far end. Self-esteem was Feingold's largest personality correlate of self-rated attractiveness, r = .34 — confidence moves the rating without touching the face being rated. Feingold 1992

Appearance distress deserves care here. In 68 men with elevated body-dysmorphic and muscle-dysmorphia symptoms, severity predicted more negative self-face evaluations but did not predict objective self-face recognition sensitivity. Portingale et al. The distortion was evaluative, not perceptual. Feeling unable to judge yourself fairly is not the same as seeing your face less accurately; more on that in body dysmorphia statistics.

The opposite bias exists too: narcissistic traits predicted overestimation of one's own attractiveness in 85 non-patients who were also assessed objectively. Lobbestael et al. 2016 That does not make ordinary confidence pathological. It shows why one self-rating diagnoses neither humility nor vanity.

Are you a good judge of your own photos?

You can be a bad picker of your own photos and still be roughly right about your face. Across two internet studies (n = 610), images chosen by strangers made more favourable impressions than the ones people chose of themselves, F(2, 429) = 77.2. White et al. 2017 The honest caveat, from the paper itself: that penalty was specific to trustworthiness and competence, and was not observed for attractiveness.

Self-selected photographs were also less well calibrated to strangers' impressions (M = 0.165) than other-selected ones (M = 0.227). That is a practical warning about picking a profile picture, not evidence that your whole self-concept is wrong.

You know the backstory of every picture: the haircut, the bad week, the joke just before the shutter. A stranger gets the frame, not the explanation.

Photo choice is a separate skill from self-rating. Someone else may choose a clearer signal even when you and the stranger agree about your general appearance.

So what is the number worth?

Worth something, and not a rank. It correlates with how others judge you, so it is not empty noise — but it also carries mood, history and the accumulated habit of seeing your own face.

Call this the Familiar Face Gap: self-rating and stranger-rating are two measurements of two different things. The gap is not a mistake you can eliminate by staring harder in the mirror. A stranger reads a moment; you read a face you have carried through years of memories.

So an outside reading is useful the way a second opinion is useful. It can interrupt a reflexively harsh or flattering story. It does not own the final word.

What this literature cannot tell you

It cannot tell you where you stand in a live conversation. These studies use photographs, timed exposures, questionnaires and rating scales; none of them measures voice, movement, warmth, style or context. Even the 100 ms finding — attractiveness judgements after a tenth of a second correlate .69 with unlimited-time judgements — was made on static head-shots of 70 amateur actors. Willis & Todorov 2006 See first impression statistics for what that does and does not establish.

The samples also vary. Some classic studies used students or narrow photo sets, while the Add Health analysis used 4,882 American emerging adults. Even there, self-ratings used a 1-to-4 scale and interviewers used a separate 1-to-5 scale, a real mismatch between the measurements.

What the literature gives you is a spread, not a conversion formula. It can tell you self-knowledge carries both signal and error. It cannot tell you which one you are made of.

What we could not verify

We checked the cited meta-analyses and primary studies for a universal male-specific correction, a live-conversation self–other correlation, and a rule that identifies who is overrating versus underrating himself. We could not find the original literature supporting those simple formulas. What the sources provide is a distribution of relationships, not a personal answer disguised as precision.

How to cite this page

Real World Appeal, "How accurately do you know what you look like to other people?," realworldappeal.com, September 2026, https://realworldappeal.com/blog/self-perception-accuracy-statistics. Research links sit in the text; data checked 21 September 2026.

The bottom line

Your self-rating has real signal, but its average match with strangers is only .24, while strangers often agree with one another more than you agree with them. That is a reason for calibration, not panic.

Your reflection is familiar; a stranger’s reaction is immediate. Neither is the whole face.

If you want the outside half of that pair, the free Real World Appeal test gives a structured perception reading — not a clinical instrument, and not a verdict.

Sources

• Feingold, A. (1992). Good-Looking People Are Not What We Think. Psychological Bulletin, 111(2), 304–341. Linked source
• Langlois JH, Kalakanis L, Rubenstein AJ, et al. (2000). Maxims or myths of beauty? A meta-analytic and theoretical review. Psychol Bull 126(3):390-423 Linked source
• Epley N & Whitchurch E (2008). Mirror, mirror on the wall: enhancement in self-recognition. Pers Soc Psychol Bull 34(9):1159-70 Linked source
• Germine L, Russell R, Bronstad PM, et al. (2015). Individual Aesthetic Preferences for Faces Are Shaped Mostly by Environments, Not Genes. Curr Biol 25(20):2684-9 Linked source
• White D, Sutherland CAM & Burton AL (2017). Choosing face: The curse of self in profile image selection. Cogn Res Princ Implic 2(1):23 Linked source
• Ehlinger PP & Blashill AJ (2016). Self-perceived vs. actual physical attractiveness: Associations with depression as a function of sexual orientation. J Affect Disord 189:70-6 Linked source
• Portingale J, Krug I & Butler D (2026). Self-Face Perception in Men With Elevated Body Dysmorphic and Muscle Dysmorphia Symptoms: Subjective Recognition Difficulty and Negative Appearance Evaluation. Eur Eat Disord Rev 34(2):486-499 Linked source
• Willis J & Todorov A (2006). First impressions: making up your mind after a 100-ms exposure to a face. Psychol Sci 17(7):592-8 Linked source
• Hönekopp J (2006). Once more: is beauty in the eye of the beholder? Relative contributions of private and shared taste to judgments of facial attractiveness. J Exp Psychol Hum Percept Perform 32(2):199-209 Linked source
• Kramer RSS, Mileva M & Ritchie KL (2018). Inter-rater agreement in trait judgements from faces. PLoS One 13(8):e0202655 Linked source
• Lobbestael J, de Bruin A, Kok E, et al. (2016). Through rose-coloured glasses: An empirical test of narcissistic overestimation. Personal Ment Health 10(4):305-316 Linked source