When Amy Edmondson started her PhD research at Harvard in the mid-1990s, she had a clean hypothesis. Better hospital teams, she assumed, would make fewer medication errors. Strong teamwork meant fewer mistakes. It made intuitive sense.
Then the data came back, and it said the opposite.
The highest-performing nursing teams in her study reported more errors than the worst teams. Not slightly more. Significantly more.
Edmondson’s first instinct was that she’d made a mistake — that her hypothesis was simply wrong, and good teamwork somehow caused more errors. But she didn’t stop there. She went back to the units. She talked to the nurses. And what she found changed how we think about teams, failure, and the courage it takes to say “I messed up” out loud.
The teams that looked safe were the most dangerous
The better teams weren’t making more mistakes. They were reporting more of them.
On the lower-performing units, errors still happened — probably at the same rate or higher. But nurses on those teams buried them. They corrected mistakes quietly, minimized what went wrong, or simply never mentioned the incident to anyone who might judge them for it.
The high-performing teams had something different. Not better protocols. Not more experienced nurses. They had an atmosphere where admitting a mistake didn’t feel like career suicide.
Edmondson called it psychological safety — “a shared belief held by members of a team that the team is safe for interpersonal risk taking.”
The word “shared” is doing heavy lifting in that definition. This wasn’t about individual bravery. It was about the climate of the group. One nurse being willing to speak up doesn’t create safety. Every nurse believing that speaking up won’t be punished — that’s the threshold.
Fifty-one teams and a new framework
Edmondson formalized her findings in a 1999 paper published in Administrative Science Quarterly. She studied 51 work teams at a manufacturing company and tested a specific chain: psychological safety leads to learning behavior, and learning behavior leads to performance.
The chain held. Teams that felt safe asked more questions, experimented more, admitted errors more freely, and sought feedback more often. Those behaviors — not talent, not resources, not experience — predicted which teams actually improved over time.
“Without an ability to be candid, to ask for help, to share mistakes,” Edmondson has said, “we won’t get things done. Not the complex, interdependent things that matter.”
Edgar Schein, the MIT organizational psychologist and one of Edmondson’s intellectual mentors, framed it even more starkly. “In most organizations, the single greatest inhibitor of learning is fear,” he wrote. “Fear of looking stupid. Fear of being punished. Fear of being different.”
Edmondson’s research put numbers on that fear — and showed exactly what it costs.
Google spent two years proving Edmondson right
In 2012, Google launched Project Aristotle — a study of 180 teams designed to answer a deceptively simple question: what makes some teams dramatically better than others?
The researchers expected the answer to be about composition. Put the smartest people together, they assumed, and you get the best results.
That wasn’t it. Not even close.
After two years of analysis, the single most important factor in team performance was psychological safety. It wasn’t just the top predictor — it was, in Google’s words, “far and away the most important” of the five dynamics they identified. The other four — dependability, structure, meaning, and impact — couldn’t even function without it.
Julia Rozovsky, who led the project, explained the finding this way: “There’s no team without trust.” Teams where people couldn’t take risks simply couldn’t learn. And teams that couldn’t learn couldn’t adapt.
The silence that looks like harmony
Here’s the part most managers get wrong. They look at a quiet team and assume things are fine. No conflict. No complaints. No drama.
Edmondson’s research says that silence is the most dangerous signal of all.
When people don’t speak up, it doesn’t mean there’s nothing to say. It means the cost of saying it feels too high. And in that silence, errors compound. Bad ideas survive. Good people leave — not with a bang, but with a resignation letter that says “pursuing other opportunities” and means “no one here listens.”
A 2023 Gallup survey found that only three in 10 employees strongly agree that their opinions seem to count at work. That number has been stagnant for years. And the link to retention is direct — employees who don’t feel heard are 4.6 times more likely to leave.
Edmondson would argue that this isn’t an engagement problem. It’s a safety problem.
What psychologically safe teams actually look like
They’re not soft. They’re not conflict-free. In fact, they argue more — but about ideas, not people.
Edmondson has identified four specific behaviors that signal psychological safety: speaking up with ideas, asking for help, admitting mistakes, and challenging the status quo. These are not personality traits. They’re team norms — and they’re shaped almost entirely by the leader.
If a manager reacts badly to one mistake, one question, one dissenting opinion, the message propagates instantly. Everyone recalculates the risk. And the next error goes unreported.
The fix isn’t a poster on the wall or a town hall about “openness.” It’s what you do in the three seconds after someone tells you something you don’t want to hear.
Do you lean in or pull back? Do you ask a follow-up question or change the subject? Do you thank them or explain why they’re wrong?
Those micro-moments are where safety lives or dies.
Starting with one question
You don’t need a two-year Google study to start building safety. Edmondson recommends starting absurdly small.
In your next team meeting, ask one question: “What did we miss this week?”
Not “what went wrong” — that sounds like blame is coming. “What did we miss” is exploratory. It assumes imperfection. It invites honesty without triggering defensiveness.
Then — and this is the part that matters — respond to whatever comes with curiosity, not judgment. If someone admits they dropped the ball on a deliverable, your first words need to be “thanks for flagging that” before anything else. The quality of your people doesn’t change with one meeting. But the quality of their honesty can.
Edmondson’s hospital teams taught us something that sounds paradoxical but isn’t. The teams that reported the most errors had the best outcomes. The ones that looked clean were the ones hiding the mess.
Your team is making mistakes right now. The only question is whether you’ll hear about them in time to do something useful — or whether you’ll find out the way the low-performing units did.
After the damage is already done.