Most people assume a good team is mostly a casting problem. Find the smartest engineer, the steady operator, the natural leader, mix them in roughly the right ratio, and the work takes care of itself. Talent in, performance out.

Google decided to check. Over roughly two years they studied 180 of their own teams, ran more than 35 statistical models on hundreds of variables, and waited to see which combination of people won. What came back was perhaps suprising; the thing that mattered most was not who was on the team at all. It was something about how the people on it treated each other, and the clearest way to see it was to watch what happened when someone admitted they had got something wrong.

A quick note before going further. I am a curious generalist here, not a psychologist or an organizational researcher, and this piece is me reading the study rather than handing out advice. Project Aristotle is one company’s internal, non-peer-reviewed research on its own staff, so treat what follows as an interesting finding from a particular place, not a settled law about every team everywhere.

Google released the study, named Project Aristotle in 2016. The sample was 180 teams, split into 115 engineering project teams and 65 sales pods, ranging from three to fifty people. They pulled on more than 250 items from internal surveys, ran hundreds of double-blind interviews, and went looking for the magic mix of individuals. The mix turned out not to be the point. Google’s researchers found that “what really mattered was less about who is on the team, and more about how the team worked together”. Composition variables you would expect to matter, like seniority, tenure, team size, and individual performance, did not reliably separate the strong teams from the weak ones.

Five dynamics came out of it, in order of importance: psychological safety, dependability, structure and clarity, meaning, and impact.

The one at the top was psychological safety.

The term itself is older than the study. The organizational scientist Amy Edmondson introduced team psychological safety in a 1999 paper, describing it as a “belief that one will not be punished or humiliated for speaking up with ideas, questions, concerns, or mistakes”.

What makes the Google version concrete is the survey line they used to measure it. One of the statements employees responded to was “If I make a mistake on our team, it is not held against me.” 

It is tempting to file psychological safety under “be nice,” the kind of thing a break-room poster covers. I think that misreads it. The reason admitting a mistake is such a good test is that it is the moment when honesty is most expensive. Anyone will share good news. The question is what the team does when the news is bad and it is your fault.

Google’s own description of a high-safety team puts the bar there. People on such teams, the write-up says, “feel confident that no one on the team will embarrass or punish anyone else for admitting a mistake, asking a question, or offering a new idea.”

When that confidence is missing, people hide errors, sit on questions, and stay quiet about the risky idea that might have been the good one. The team looks calm and is quietly going wrong.

Of course, the finding has its limits. The psychological-safety-as-number-one result was internal to one company and broadcast widely, so it is wise to treat it as a hypothesis worth testing than a proven rule. It does not make the idea wrong. It makes it something to check against your own experience.

So run the test. The next time someone on your team owns a mistake out loud, watch the few seconds after. Does the room move toward the problem, or toward the person? Whatever you see in those seconds is the team you actually work on, not the one on the org chart or in the values deck. If the answer makes you uncomfortable, that discomfort is the finding.