The Teams That Report the Most Mistakes Are Usually the Best Teams
The nursing units reporting the most medication errors were the best-performing units in the hospital.
That's what Amy Edmondson found in 1999 when she set out to study whether good leadership led to fewer mistakes. The hypothesis was straightforward: better teams, fewer errors. What the data showed reversed it. Teams rated highest by administrators and supervisors were logging more errors, not fewer. The teams rated lowest were logging the fewest.
Edmondson didn't conclude that good teams made more mistakes. She concluded something more useful: good teams reported more of the mistakes they made. The difference is everything.
What Edmondson Actually Found
The 1999 study — published in Administrative Science Quarterly — examined medication error rates across 16 nursing units at a large US hospital. Each unit had a different culture around speaking up, admitting uncertainty, and flagging problems. Edmondson named the variable she was measuring "psychological safety": the shared belief among team members that the environment is safe for interpersonal risk-taking — admitting errors, asking questions, raising concerns.
Teams with high safety surfaced issues. Teams with low safety stayed quiet. Same underlying rate of errors; dramatically different reporting rates. The teams with fewer logged errors weren't making fewer mistakes — they were operating in environments where mistakes didn't get acknowledged.
This is the finding most summaries of Edmondson's work skip past. Psychological safety doesn't reduce errors directly. It makes errors visible. That visibility is what enables correction. And correction, over time, is what produces genuinely better performance.
Why "Psychological Safety" Became a Misused Term
Google's 2016 Project Aristotle — an internal study on what made Google's teams effective — identified psychological safety as the top predictor of team performance across 180 teams. The finding was accurate. The way it was communicated was not.
"Psychological safety" entered management vocabulary as "teams where people feel comfortable." Corporate workshops started selling it as warmth, harmony, niceness. Leadership trainings positioned it as making people feel cared about. The underlying mechanism — information flow — got stripped out.
Amy Edmondson's subsequent work, particularly The Fearless Organization (2018), pushed back on this flattening explicitly. Safety is not comfort. Safety is the structural property of an environment that makes interpersonal risk-taking feel survivable. Raising a concern, admitting you don't know something, challenging a decision in progress — these are risky acts in most organizations. High-safety environments lower that risk. They don't eliminate it, and they don't make everything pleasant. They make truth-telling viable.
The distinction matters because the corporate version of psychological safety is useless and frequently counterproductive. A "safe space" framing focuses on preventing discomfort. The real goal is enabling productive discomfort — the kind that surfaces contradictions before they become expensive.
The Communication Mechanism
A 2024 study published in PLOS ONE examined 580 high-tech employees across multiple organizations, using structural equation modeling to trace the path between psychological safety and innovation. The researchers tested whether safety improved innovation directly, or whether something else mediated the relationship.
The answer was clear: communication behavior was the full mediator. Psychological safety didn't improve innovation outcomes on its own. It worked by changing what people communicated about — specifically, enabling them to surface partial ideas, admit what they didn't know, and voice disagreement early in a process rather than after investment had accumulated.
Teams with high safety didn't talk more. They talked about different things.
Three specific communication behaviors showed up consistently across the literature:
Admitting uncertainty in real time. In low-safety environments, "I don't know" is read as incompetence. In high-safety environments, it's the beginning of a useful conversation. The question "I'm not sure this approach will work — can we test a smaller version first?" sounds obvious. It requires safety to ask, because it implies the original plan might be wrong, and someone owns that plan.
Voicing partial ideas. High-safety teams surface half-formed thinking earlier. This looks inefficient. It isn't. The cost of an undiscussed option discovered after deployment is orders of magnitude higher than the awkwardness of raising it during planning. Safety lowers the bar for what's "ready" to share.
Disagreeing early. This is the one every organization claims to want and consistently punishes in practice. Early disagreement — challenging a direction before the team is committed — feels disruptive. It is also far cheaper than late disagreement. High-safety environments normalize early challenge by making the cost of speaking up lower than the cost of staying quiet.
Where High-Safety Teams Look Different
The organizational tension is real. The cultures that most need psychological safety — high-stakes, high-accountability environments — are often the ones most likely to have built reward systems that punish exactly the behaviors safety requires.
If promotions go to people who project certainty, teams learn to project certainty. If leaders respond to bad news by finding someone to blame, teams learn to filter what reaches leadership. If the person who raises a concern is seen as a problem-maker rather than a problem-finder, concerns stop being raised. The safety erodes. The information stops flowing. Errors accumulate quietly until they're too large to ignore.
Edmondson's documentation of this pattern spans industries. The 2010 BP Deepwater Horizon disaster included documented instances of engineers flagging safety concerns that were filtered or minimized before reaching decision-makers. The 2003 Columbia shuttle disaster showed the same pattern: problems were visible to people with direct knowledge; the environment made surfacing them costly; the cost of silence exceeded the cost of speaking up only in retrospect.
The Volkswagen emissions scandal. The Boeing 737 MAX failures. Wells Fargo's fraudulent account creation. Each involved people with direct knowledge of problems who couldn't or didn't raise them in time to prevent the outcome. The technical and ethical failures are real. So is the structural failure: environments that made accurate information too expensive to surface.
High-performing teams look messier from the outside than low-performing ones. They have more disagreement in meetings. They report more problems. They raise more concerns. This gets mistaken for disorder. It's actually information moving.
The Safest Teams Aren't the Most Comfortable
Psychological safety isn't a personality trait. It isn't a leadership style. It's a structural property of the environment — something built and destroyed by specific behaviors.
Leaders who respond to bad news with curiosity rather than blame build it. Leaders who ask questions rather than issue directives build it. Leaders who admit their own uncertainty give others permission to admit theirs. None of these behaviors require a particular personality — they require understanding that the cost of invisible problems exceeds the discomfort of visible ones.
The teams reporting the most errors in Edmondson's hospital study weren't struggling. They were seeing clearly. That capacity — to look at what's actually happening and say it out loud — is what separates teams that improve from teams that don't.
Productive work requires truth. Safety doesn't make the truth comfortable. It makes the truth possible.
The dynamic of high-performing individuals who can't operate in low-safety environments connects to something worth examining separately: what happens to identity when high-achievers stop being able to tell the truth at work.
Photo by Thirdman via Pexels.