Can You Tell Someone's Personality From Their Face? What New Research Shows

The short answer: People form impressions of a stranger's personality from their face within a fraction of a second, and they largely agree with each other about who looks friendly, competent or threatening. But agreement is not accuracy. A new study in Nature Human Behaviour shows that these snap judgements are organised along a handful of broad dimensions, led by a general sense of whether someone seems approachable or threatening, and that they are shaped by gender and race stereotypes. A neutral photo of a face says very little about someone's actual personality.

Key findings

• In a new study, more than 3,400 people sorted photos of unfamiliar faces by how similar they thought the people's personalities were, without being given any trait words.
• Their judgements could be summarised by five dimensions. With mostly White faces, the researchers interpreted them as approachability versus threat, gender-specific warmth, competence, morality and youthfulness.
• The strongest dimension was a broad, holistic sense of approachability, which blended trustworthiness, warmth and competence into one overall impression.
• When the faces included White, Black and Asian people, two of the five dimensions mainly reflected race stereotypes.
• The study measured how people perceive faces, not whether those perceptions are correct. Other research shows that first impressions from faces are often inaccurate, yet still influence elections, hiring and even court sentences.
• The five face dimensions are not the Big Five personality traits.

How fast do we judge faces?

Very fast. In a classic set of experiments, Janine Willis and Alexander Todorov showed people unfamiliar faces for just one tenth of a second. Their judgements of attractiveness, likeability, trustworthiness, competence and aggressiveness were already highly correlated with judgements made with no time limit. Looking longer mostly made people more confident, not substantially different in what they concluded.

These impressions also matter. In a study of US congressional elections, Todorov and colleagues found that judgements of competence based on candidates' faces alone, made within a second, predicted the winner of 68.8 percent of the 2004 Senate races. In a study of convicted murderers in Florida, John Paul Wilson and Nicholas Rule found that people with less trustworthy-looking faces were more likely to have been sentenced to death rather than to life in prison.

The new study: judging faces without words

Most earlier research asked people to rate faces on words chosen by the researchers, such as trustworthy, dominant or competent. That raises a problem: the dimensions you find may depend on which words you ask about, and on the language you ask in.

A team led by Yue Xu and Ralph Adolphs at the California Institute of Technology (Caltech) took a different approach. In two pre-registered online experiments, adults in the United States dragged photos of unfamiliar faces around a circle on the screen. They were asked to place people they thought had similar personalities close together, and different personalities far apart. No trait words were provided at any point.

By combining thousands of these arrangements, the researchers built a map of how similar every face seemed to every other face. They then checked that people were judging personality rather than just visual similarity: the arrangements matched trait ratings from earlier studies far better than they matched simple facial measurements or a face-recognition neural network.

• Experiment 1: 1,751 participants and 100 faces, mostly White.
• Experiment 2: 1,683 participants and 96 faces, one third each White, Black and Asian.

The five dimensions of first impressions

In both experiments, five dimensions described the judgements best. In the first experiment, they explained 76 percent of the variation in how people arranged the faces. The researchers interpreted them as:

Dimension	What it captured
Approachability versus threat	The largest dimension. Faces at one end were linked to words like affectionate, at the other end to words like criminal.
Gender-specific warmth	Warmth judged differently for women's and men's faces, with negative stereotypes such as grumpy or jealous linked mainly to women's faces.
Competence	Resembled the competence and dominance dimensions found in earlier research.
Morality	Morally charged impressions such as humble versus condescending.
Youthfulness	Seeming youthful and unconventional rather than traditional and conservative.

The first dimension is the most interesting. Instead of judging trustworthiness, warmth and competence separately, people seemed to form one holistic overall impression: would I approach this person, or keep my distance? The researchers argue that this approachability or warmth dimension shows up as a universal factor across studies.

How stereotypes shape the picture

The second experiment shows how much first impressions depend on who is being judged. When the faces included White, Black and Asian people, the first dimension mainly separated White and Black faces, and the third combined stereotypes about Asian faces with youthfulness and femininity. Gender also shaped the dimensions in both experiments.

In other words, the dimensions people use to judge faces are not fixed. They reflect the stereotypes that become relevant in a particular group of faces. As the authors put it in the article's introduction, face judgements "may reflect our biases and stereotypes more than true facts about the person."

Can a face actually reveal personality?

This is where it is important to be clear: the new study did not test whether the impressions were accurate. It only mapped how people perceive faces.

Other research gives a sobering answer to the accuracy question:

• In a major review, Todorov and colleagues concluded that the diagnostic value of judgements from faces "has been greatly overstated." People agree strongly with each other, but they tend to agree on the same mistakes.
• Analysing more than a million judgements from a website where people guessed facts about strangers from their photos, Christopher Olivola and Todorov found that people were generally less accurate than if they had ignored the face and simply guessed based on how common each characteristic was.
• Photos that show more than a neutral face do contain some real information. Laura Naumann and colleagues found that observers who saw full-body photos with a natural pose and expression judged several traits better than chance. When pose and expression were standardised, only a few traits, such as extraversion, were judged accurately. Much of the signal comes from expression, posture and clothing, which people partly choose themselves, rather than from the shape of the face.
• Machine learning can find weak patterns too. In a study of more than 12,000 volunteers who submitted their own photos, a neural network predicted self-reported Big Five scores with an average correlation of 0.24. That is a statistical tendency across many people, far from enough to read an individual's personality.

The research suggests that people who actually know you are much better judges. In a meta-analysis of 263 samples, Brian Connelly and Deniz Ones found that ratings by family and close friends agree substantially more with a person's self-ratings than ratings by strangers do, especially for less visible traits such as emotional stability.

Perceived personality is not the Big Five

It is easy to mix the two up, but the five dimensions in the new study are not the Big Five personality traits.

• The five face dimensions describe how observers sort strangers' faces. They are about the person doing the judging and the stereotypes they bring.
• The Big Five traits, openness, conscientiousness, extraversion, agreeableness and neuroticism, describe how a person actually tends to think, feel and behave across many situations. They are measured with tested questionnaires and confirmed by people who know the person well.

Someone can have a face that strangers read as cold or threatening and still score high on agreeableness. Someone can look approachable and be very introverted.

Limitations

• The task was artificial. Arranging photos on a screen is not the same as meeting someone in real life, where voice, movement and context play a role.
• All participants lived in the United States, so the results may not generalise to other cultures.
• The faces were mostly between about 20 and 40-something years old, with neutral expressions.
• The results are based on data aggregated across many participants. Each person only arranged a subset of the faces.
• The labels for the dimensions are the researchers' interpretations.

What this means for you

• Notice your first impressions, but hold them loosely. A face triggers a quick judgement of approachability, warmth and competence, but that judgement often reflects stereotypes.
• Be aware of how others judge you. Expression and posture affect how approachable you seem, often more than the features you were born with.
• Measure personality, do not guess it. If you want to know your personality, a validated questionnaire is far more informative than how others react to your photo.

What does your personality really look like?

First impressions do not tell the whole story. Our free test measures all five traits and 30 facets, based on how you actually tend to think, feel and act.

Take the free Big Five personality test and compare your results with how others see you.

Sources

• Connelly, B. S., & Ones, D. S. (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
• Kachur, A., Osin, E., Davydov, D., Shutilov, K., & Novokshonov, A. (2020). Assessing the Big Five personality traits using real-life static facial images. Scientific Reports, 10, 8487. doi:10.1038/s41598-020-65358-6
• Lin, C., Keles, U., & Adolphs, R. (2021). Four dimensions characterize attributions from faces using a representative set of English trait words. Nature Communications, 12, 5168. doi:10.1038/s41467-021-25500-y
• Naumann, L. P., Vazire, S., Rentfrow, P. J., & Gosling, S. D. (2009). Personality judgments based on physical appearance. Personality and Social Psychology Bulletin, 35(12), 1661–1671. doi:10.1177/0146167209346309
• Olivola, C. Y., & Todorov, A. (2010). Fooled by first impressions? Reexamining the diagnostic value of appearance-based inferences. Journal of Experimental Social Psychology, 46(2), 315–324. doi:10.1016/j.jesp.2009.12.002
• Todorov, A., Mandisodza, A. N., Goren, A., & Hall, C. C. (2005). Inferences of competence from faces predict election outcomes. Science, 308(5728), 1623–1626. doi:10.1126/science.1110589
• Todorov, A., Olivola, C. Y., Dotsch, R., & Mende-Siedlecki, P. (2015). Social attributions from faces: Determinants, consequences, accuracy, and functional significance. Annual Review of Psychology, 66, 519–545. doi:10.1146/annurev-psych-113011-143831
• Willis, J., & Todorov, A. (2006). First impressions: Making up your mind after a 100-ms exposure to a face. Psychological Science, 17(7), 592–598. doi:10.1111/j.1467-9280.2006.01750.x
• Wilson, J. P., & Rule, N. O. (2015). Facial trustworthiness predicts extreme criminal-sentencing outcomes. Psychological Science, 26(8), 1325–1331. doi:10.1177/0956797615590992
• Xu, Y., Keles, U., Lin, C., & Adolphs, R. (2026). Trait judgements of faces using a task without word labels. Nature Human Behaviour. doi:10.1038/s41562-026-02605-2