How fast do we judge a new face? In an experiment at Princeton University, people saw photos of strangers for only one tenth of a second. That is about as fast as a blink.

Even so, their judgments were quite similar to those of people who looked for a long time. When people looked longer, they mostly felt more sure, but their judgments matched no better. And a quick first impression is not always right.

Psychologists Janine Willis and Alexander Todorov wanted to know how quickly people form impressions from faces. In five experiments published in 2006, participants saw photos of unfamiliar faces for 0.1, 0.5 or 1 second.

Then they judged one trait, such as how trustworthy or competent the person looked. These judgments were compared with ratings from people who had no time limit. Even after just 0.1 seconds, the two sets of judgments were already quite similar, especially for trustworthiness. More time hardly improved this match, but it made people more confident.

So longer viewing may simply confirm what we have already decided. But the study only showed that impressions form quickly. It did not show that they are accurate.

Princeton psychologists Janine Willis and Alexander Todorov wanted to know how long it takes to form an impression of a stranger's face. Their answer, published in 2006, was about a tenth of a second. Across five experiments, participants viewed unfamiliar faces for 100, 500 or 1,000 milliseconds and rated a single trait, such as attractiveness, likability, trustworthiness, competence or aggressiveness. A separate group rated the same faces with no time limit.

Even at 100 milliseconds, the two sets of judgments were substantially correlated, and trustworthiness showed the strongest agreement. Longer exposures barely raised the correlation. They did make participants more confident, so extra time seems mainly to confirm a first impression. In a related 2005 study, the candidate whose face was judged more competent won U.S. congressional elections about 70 percent of the time.

But these studies measured how fast and how consistently impressions form, which is a separate question from whether they are right. The benchmark was agreement with other raters. Nobody checked the targets' real personalities. Research suggests that reading character from faces is only weakly valid, and a 2010 study found that relying on facial appearance could sometimes make predictions worse.