Why Do First Impressions Last?

You see a face for one-tenth of a second. That is enough.

Research by Janine Willis and Alexander Todorov at Princeton found that exposures as brief as 100 milliseconds are sufficient for people to form confident trait judgments: competent or incompetent, trustworthy or untrustworthy, dominant or submissive. Longer exposure times allow revisions, but the rank order of the judgments remains largely stable. The very fast impression and the considered judgment reach similar conclusions.

This is the thin-slice phenomenon: extremely brief samples of a person’s face, voice, or behavior produce impressions that correlate meaningfully with extended exposure.

The 100-Millisecond Assessment

The speed of first impression formation is not incidental — it reflects how the brain allocates cognitive resources.

Face processing begins in the primary visual cortex, passes through the fusiform face area (which specializes in face identification), and rapidly reaches the amygdala. The amygdala performs threat and social significance assessment extremely fast, before higher-level cortical processing has caught up.

This subcortical route — directly to the amygdala, bypassing extended cortical analysis — is what makes emotional and social responses so fast. By the time you are consciously processing a face, your amygdala has already made a preliminary threat/safety/status assessment.

Nalini Ambady at Harvard, who developed thin-slice research, showed that 30-second silent video clips of teachers were enough for students to rate them in ways that correlated significantly with their actual end-of-semester teacher evaluations. Six seconds was sufficient. Two seconds had predictive value.

Something is being assessed accurately in very brief exposures.

What First Impressions Accurately Capture

Thin-slice judgments are not uniformly accurate or inaccurate. They appear to be reliably good at some things.

Dominance and threat. Humans are particularly accurate at assessing physical dominance (height, muscularity, aggressive cues) from brief exposures — this maps onto evolutionary functions directly. Misjudging a threat is costly.

Emotional state. Basic emotion recognition from face and voice is accurate and fast across cultures. Anger, fear, disgust, joy — these are decoded quickly and reliably.

Competence signals. More surprisingly, rapid face assessments of competence have been shown to predict election outcomes. Todorov and colleagues found that people’s first-impression competence ratings of faces of candidates they had never seen predicted the actual winners of congressional elections at better-than-chance rates. The faces of winners “look more competent” to naive raters who had no other information.

What First Impressions Badly Misread

The same mechanism that is reliably good at detecting dominance and emotional state is unreliable for character judgments.

Trustworthiness assessments from faces are poor predictors of actual trustworthy behavior. Studies by Timothy Levine found that people are roughly at chance at detecting deception from a face — barely better than guessing. Yet confidence in deception detection is high.

The problem: the cues people use to assess trustworthiness (facial structure, resemblance to someone known, baby-faced features, superficial attractiveness) are not actually correlated with honest behavior. The assessment system is real; it’s just tracking the wrong signal.

This is how people with trustworthy-looking faces get more lenient sentences in criminal trials. It’s how a reassuring bedside manner gets mistaken for medical competence. First impressions are not useless — they’re specifically accurate and specifically inaccurate in ways that don’t always match what matters.

The Primacy Effect

Once a first impression is formed, it has disproportionate weight in subsequent processing.

The primacy effect in impression formation, established by Solomon Asch in the 1940s, shows that information presented first about a person shapes how subsequent information is interpreted. Two people described with identical traits but in reverse order are evaluated differently: “intelligent, industrious, impulsive, critical, stubborn, envious” produces a more positive impression than “envious, stubborn, critical, impulsive, industrious, intelligent.”

The early information doesn’t just appear first — it creates a framework within which later information is processed. Positive early information leads to charitable interpretations of subsequent ambiguous information; negative early information leads to hostile interpretations.

This is one reason first impressions are so difficult to update. New information is not evaluated independently — it is evaluated in the context of the existing impression. Information consistent with the impression is accepted readily; inconsistent information is discounted, reinterpreted, or attributed to situational factors (“they were just tired”).

Why Revisions Are Hard

If you meet someone and form a negative first impression, subsequent positive behavior tends to be attributed to effort or circumstance rather than character. If you form a positive impression, negative behavior gets attributed to an off day.

This asymmetry is not simple stubbornness. It reflects the genuinely useful principle that behavior has meaningful signal-to-noise: in any given interaction, a person’s typical character is mixed with situational variation. Weighing earlier impressions heavily and treating later contrary behavior as potentially noisy is not irrational.

The problem is that the initial impression itself may be noisy — based on 100 milliseconds of face exposure, which carries real signal about some things (emotional state, dominance) and carries misleading signal about others (honesty, competence, character).

The first impression is fast, specific, persistent, and partially accurate in ways that don’t fully correspond to what we care about.

You make it in less than a second.

It shapes what you see for months afterward.