Is There a Trustworthy Face? Todorov's Two Axes of Face Evaluation

Is there really such a thing as a 'trustworthy face'?

The moment you first see someone, a verdict pops into your head almost automatically. "I somehow trust this person," or "something makes me wary" — the feeling arrives in under a second. What is striking is that people tend to agree fairly well on which faces look trustworthy. That consensus is exactly why a 'trustworthy face' can feel like a real, fixed thing.

The real question this article asks starts right there. Does a face that merely looks trustworthy actually belong to a more trustworthy person? The psychologist Alexander Todorov and his colleagues spent years digging into that question with data. We will walk through the two-axis model they uncovered, how the mouth corners and eyebrows build a trust impression, and the important reversal that comes at the end.

Let me be clear about one thing first. This article is not about judging people by their faces; if anything, it shows how easily that impression fools us. An impression is only a surface feeling — it is not evidence of anyone's personality or trustworthiness.

Todorov's finding — face impressions boil down to two axes

The study Oosterhof and Todorov published in PNAS in 2008 took an unusual approach. The researchers showed people emotionally neutral faces, let them describe their impressions freely with no constraints, and then gathered ratings on many traits such as trustworthiness, attractiveness, aggressiveness, and dominance. They then compressed all of that data with a statistical method called principal component analysis (PCA).

The result was strikingly simple. The countless impressions people read from faces mostly boiled down to just two axes. The first was trustworthiness and the second was dominance. Todorov explained that these two axes map onto a very old survival question: does this person mean me harm, and do they have the strength to carry it out?

The crucial point is that these two axes are only a map for organizing impressions. The trustworthiness axis does not measure whether a person is actually honest, and the dominance axis does not measure real strength or personality. Think of it purely as a framework for how impressions get arranged in our heads when we look at a face.

What builds the 'trust impression' — mouth corners and eyebrows

The most compelling insight in Todorov's work is what the trustworthiness axis really is. The reason even neutral faces split on trust is that a face can faintly resemble a particular emotional expression. A neutral face that resembles a slight smile reads as friendly, and one that resembles slight anger reads as threatening. Our brain automatically reads the faint emotional cues that linger even on an expressionless face.

Why the brain guesses trust from a face — overgeneralization

So why does our brain rush to guess trust from a face alone? Todorov and the psychologist Leslie Zebrowitz explain it as overgeneralization. A circuit built to read genuinely useful survival signals ends up firing the same way at unrelated faces that merely echo those signals. Two overgeneralizations come up most often.

This reaction is far more automatic than it feels. Brain-imaging studies found the amygdala responds more strongly to faces that look less trustworthy, even before the face is consciously noticed. That shows how fast and unconscious the reaction is — it does not mean the judgment is right.

The most important reversal — an impression is a bias, not a fact

Here is the most important part of the whole article. The point Todorov drives home in his book Face Value (2017) is this: people's trust impressions agree with one another to a remarkable degree, yet that consensus does not predict who will actually behave honestly. Many people feeling the same way does not make the feeling true.

And yet this bias really does move the real world. Trustworthy-looking faces have been shown to attract larger investments in economic trust games, and competent-looking faces correlated with election outcomes in one study (Todorov et al., 2005). The unsettling part is that this influence rests on bias, not on any legitimate basis. It means an impression can move reality — not that the impression is true.

Historically, too, attempts to read the inner person from the face — physiognomy and phrenology — have a painful past as tools of prejudice and discrimination. And the same risk repeats today in services that promise to analyze faces with AI. So enjoy the 'trustworthy face' purely as a fascinating psychological phenomenon, and never use it as a ruler to sort or judge people.

So how should we handle this impression?

You cannot really stop a first impression from popping up on its own. But you can practice treating it as a tentative hypothesis rather than a conclusion. When a face 'looks trustworthy,' check once more rather than less; when someone gives you a bad feeling, set it aside to ask whether it is really just the lighting, the angle, or that day's expression. Even this alone can meaningfully shrink first-impression bias.

The same goes for your own impression. Relaxing your expression and easing the tension around the mouth and eyes really does bring out friendlier cues. But that only adjusts the visible mood as a styling reference; it does not create trustworthiness or change who you are. Once you remember that an impression is only an impression, you judge others less harshly and dress yourself up with a lighter heart.