First impression statistics: the real numbers (with sources)

People form a first impression of your face in about 100 milliseconds — one tenth of a second (Willis & Todorov, 2006). They agree on who's attractive far more than the "beauty is subjective" crowd admits: a meta-analysis of over 900 studies found strong cross-cultural consensus (Langlois et al., 2000). And a 30-second glimpse of someone predicts how you'll rate them after far longer exposure (Ambady & Rosenthal, 1992). Those are the headline numbers. Below is the full set, sourced, with what each one actually means for you.

This is a data page. If you came here to grab a stat, the table is right under the key numbers and every figure traces to a named study. No invented percentages, no "studies show" with no study.

Key numbers

• First impressions of a face form in about 100 milliseconds; judgments at 100ms correlate strongly with judgments made with no time limit (Willis & Todorov, 2006).
• A meta-analysis of more than 900 studies found strong agreement on who is attractive, both within and across cultures (Langlois et al., 2000).
• 30-second silent video clips predicted end-of-term evaluations of teacher effectiveness — the "thin slice" finding (Ambady & Rosenthal, 1992).
• Most face-based first impressions reduce to two axes: trustworthiness (valence) and dominance (power) (Todorov).
• The "what is beautiful is good" halo — attractive people are assumed to have better traits — was first demonstrated in 1972 (Dion, Berscheid & Walster, 1972).
• Mate-preference patterns replicate across 37 cultures sampled, showing first-impression criteria aren't just a local fashion (Buss, 1989).

The numbers at a glance

Statistic	Value	What it measures	Source
Time to form a face impression	~100 ms	Trust, competence, attractiveness judged from a tenth-second exposure	Willis & Todorov (2006)
Extra time effect	Confidence ↑, verdict ≈ same	Longer viewing doesn't flip the call	Willis & Todorov (2006)
Studies on attractiveness consensus	900+	Cross-cultural agreement on who's attractive	Langlois et al. (2000)
Thin-slice clip length	~30 s (and shorter)	Predicts much longer evaluations	Ambady & Rosenthal (1992)
Core face-judgment dimensions	2 (trust + dominance)	What the brain extracts first	Todorov
Halo effect demonstrated	1972	Attractive = assumed better traits	Dion, Berscheid & Walster (1972)
Cultures sampled for mate preference	37	Cross-cultural stability of criteria	Buss (1989)

How fast is a first impression, really?

About a tenth of a second. Willis & Todorov (2006) flashed faces for as little as 100 milliseconds and asked people to rate trustworthiness, competence, likability, aggressiveness, and attractiveness. The ratings lined up tightly with ratings given by people who had unlimited time to stare.

Read that twice. More time didn't change the verdict much. It mostly raised people's confidence in the verdict they'd already reached at 100ms.

That's the part most men miss. The judgment isn't slow and deliberate, waiting for you to win it over. It's fast, automatic, and then it spends the rest of the encounter looking for evidence it was right. We unpacked the real-world version of this — the swipe, the walk-past, the door-open moment — in the first-impression window.

Do people agree on who's attractive?

More than you'd hope. Langlois et al. (2000) ran a meta-analysis pooling over 900 studies and found that raters — strangers, across ages, across cultures — agree substantially on who is attractive. The "everyone has their own type" idea is real but it sits on top of a large shared baseline.

Here's the brand stance, and it matters: consensus on attractiveness is not the same as a fixed 0-100 facial-geometry score. The agreement is on a perceived read, and the read is built from cues you partly control — grooming, expression, framing, body composition — not just bone structure. High consensus, controllable inputs. Those two facts coexist.

The same body of work documents the halo effect: "what is beautiful is good" (Dion, Berscheid & Walster, 1972). People assume the attractive person is also kinder, smarter, more competent. It's a bias, it's well-replicated, and it runs in your favor or against you before you've said a word.

What is a "thin slice"?

A thin slice is a tiny sample of behavior that predicts a much larger judgment. Ambady & Rosenthal (1992) showed people 30-second silent video clips of teachers — no audio, no content — and those snap ratings predicted the teachers' end-of-semester evaluations from actual students. Clips as short as a few seconds carried real signal.

The implication for you is brutal and useful at the same time. A stranger watching you cross a room for a few seconds is extracting most of what they'll conclude about you. Posture, gait, where your eyes go, whether your face is doing anything — all of it loads in fast. This is exactly why a frozen frontal selfie undersells you: it kills the motion, the eye contact, and the expression that the thin slice runs on.

What is the face actually judged on?

Two things, mostly. Todorov's research shows that the huge tangle of trait inferences people make from a face collapses onto two main dimensions:

1. Trustworthiness (valence) — should I approach or avoid this person? Built heavily off expression cues: a slight smile, relaxed brow, open face read as trustworthy; a furrowed, tense, downturned face reads as a threat.
2. Dominance (power) — how much can this person impose on me? Built off cues like jaw, brow heaviness, and — critically — posture and how you carry the face.

Almost every snap read — "he seems nice," "he seems intense," "he seems harmless," "he looks like trouble" — is some blend of those two axes. The good news: trustworthiness is largely expression, which is the most controllable variable on your face. You can move your trust score in a single photo by relaxing your face and adding a real, eyes-involved smile.

Are these criteria universal?

Largely. Buss (1989) surveyed mate preferences across 37 cultures and found stable patterns — enough that you can't write off first-impression criteria as a quirk of one place or decade. Singh's work on the waist-to-hip ratio and Little's work on averageness and facial attractiveness point the same direction: there are recurring, cross-cultural cues the eye keeps returning to.

This is why "just be confident, looks don't matter" is half a lie. The cues are real and they replicate. But the cope on the other side — that you're stuck with a fixed number — is also a lie. The replicable cues include a lot you can move: body composition, grooming, framing, expression. The brand's full argument for that is in perceived vs. objective beauty.

The stat that should change your behavior

Speed of judgment is not permanence of judgment. People keep quoting "you never get a second chance at a first impression" as if the 100ms read is carved in stone. It isn't carved — it's a first draft, written from the worst-case input: a still, silent, frontal view.

Every later input rewrites it. Motion. Voice. A face that moves. The way you hold eye contact. Whether your shoulders are back or curled. Dutton & Aron (1974) is the famous proof that context can hijack the read entirely — people on a high, scary bridge misattributed their racing heart to attraction. Arousal got relabeled. The point: the read is malleable to context, not a fixed property of your face.

So the leverage isn't in chasing facial geometry. It's in the controllable inputs that feed the 100ms read and every read after it. Photos first — most men lose on bad lighting and angle, not bad faces. Then grooming and presentation. Then expression and posture, which move the trust and dominance axes directly.

Where these numbers come from (and where they don't)

A note on honesty, because this page exists to be cited. Every figure above traces to a named, real study. We did not include the viral "55-38-7" communication breakdown, the "7 seconds to make a first impression" business-seminar number, or any of the round percentages that float around LinkedIn with no source. They don't survive a citation check, so they're not here.

If a claim on this page isn't followed by a study name, it's our reasoned read of the evidence, labeled as such — not a smuggled-in statistic.

The bottom line

The real first impression statistics are tighter and more useful than the internet's folklore. A face is judged in about 100ms (Willis & Todorov, 2006), people agree on attractiveness across 900+ studies (Langlois et al., 2000), a 30-second slice predicts far more (Ambady & Rosenthal, 1992), and the judgment runs mostly on two axes — trust and dominance (Todorov). None of that means you're stuck. The fastest judgment in the dataset is also the most rewritable one, because it's built from a static, worst-case input. Fix the inputs — photos, expression, presentation — and you change the number. Want to see your current perceived read before the next stranger does? Run the test; it takes a minute and the score is honest.