← Leadership research

Psychological Safety

Amy C. Edmondson · The Fearless Organization (2018)

The most misquoted idea in management. It is not about comfort, and the study everyone cites for it is not the study the finding comes from.

What it actually means

Edmondson's definition is narrow and worth keeping narrow: a shared belief that the team is safe for interpersonal risk-taking.

Interpersonal risk is the specific thing at stake. Not physical safety, not job security, not whether people are nice to each other. The risk is to your standing in the room: of looking foolish for asking, of looking incapable for admitting you are stuck, of looking difficult for disagreeing.

So the test is not whether your team is happy. It is whether someone on it would say the inconvenient thing in front of you, on a bad day, when they are not completely sure they are right.

The finding everyone quotes, and the paper it is not in

You will have heard the result. Better hospital teams reported more drug errors, not fewer. The usual citation attached to it is Edmondson's 1999 paper in Administrative Science Quarterly.

That citation is wrong, and it is wrong almost everywhere.

The error-reporting finding comes from Edmondson 1996, "Learning from mistakes is easier said than done", in the Journal of Applied Behavioral Science. That is the hospital nursing study. The 1999 ASQ paper, "Psychological Safety and Learning Behavior in Work Teams", is a study of 51 teams in a manufacturing company. There are no hospitals in it. What it did was turn psychological safety into something measurable, which is why it became the paper people cite for everything.

The distinction matters because the 1996 study is the better story anyway. Edmondson had expected better-led units to make fewer mistakes. The data said the opposite, which looked like a failed hypothesis. By her own account, what resolved it was sending in a researcher who did not know what she was looking for: the units with higher reported error rates were not making more mistakes, they were willing to say so. The nurses on the worst-reporting units described a climate in which admitting an error was dangerous.

That is the whole concept in one accident. Psychological safety was not designed as a theory of pleasantness. It arrived as the explanation for why good data and bad data come from different rooms.

The four things people are protecting

In The Fearless Organization, Edmondson lays out what silence is actually buying. Nobody wants to look:

• Ignorant, so they do not ask the question.
• Incompetent, so they do not admit the mistake or ask for help.
• Intrusive, so they do not offer the idea.
• Negative, so they do not raise the problem.

Each of those is individually rational. Staying quiet protects you now, and the cost lands on someone else later, usually on the organization, usually when it is expensive. This is why the problem does not solve itself: everyone is behaving sensibly.

Safety is half of it

The most common misreading is that psychological safety means lowering the bar. Edmondson's answer is a two-by-two, and it is the part worth remembering.

Put psychological safety on one axis and standards on the other. Low safety with low standards gives you apathy. High safety with low standards gives you a comfort zone, pleasant and unproductive. Low safety with high standards gives you anxiety, which is where a lot of hard-driving teams actually sit. Only high safety with high standards gives you the learning zone.

Safety is not the opposite of accountability. It is the thing that makes demanding standards survivable, because people can say they are struggling before the deadline rather than after it.

What a leader actually does

Edmondson's recommendations are unglamorous and mostly about framing.

Frame the work as a learning problem rather than an execution problem. If the situation is genuinely uncertain, say so out loud. People calibrate how much deviation is allowed from how certain you sound.

Acknowledge your own fallibility. "I may have missed something, I need to hear from you" costs very little and does a lot of work.

Ask real questions. Not questions you already have the answer to. People can tell the difference, and a rhetorical question teaches them that input is decorative.

And the one that decides it: how you respond the first time someone brings you something you do not want to hear. The team sets the price of speaking up by watching that moment. Not what you said in the all-hands about wanting candor.

Where it comes from, and how far it stretches

Amy Edmondson is a professor at Harvard Business School. She began measuring this in the 1990s; the idea reached a mass audience in 2015 and 2016, when Google's Project Aristotle studied its own teams and reported psychological safety as the strongest single factor in team effectiveness.

Two fair criticisms. The construct is measured by self-report survey, which is a real limitation when what you are measuring is whether people feel able to be honest. And most of the evidence is correlational: safe teams learn more, but teams that are already doing well are also pleasanter places to speak up in.

The bigger practical risk is dilution. "Psychological safety" is now used for almost any workplace kindness, including some that Edmondson explicitly excludes. If the phrase is being used to mean nobody gets criticized, it has stopped referring to her idea.

Reading on

• The Fearless Organization (2018) — the general-audience book.
• Teaming (2012) — earlier, and more practical about groups that form and reform.
• Edmondson 1996, Journal of Applied Behavioral Science 32: 5–32 — the hospital study.
• Edmondson 1999, Administrative Science Quarterly 44: 350–383 — the manufacturing study that made it measurable.

The questions this produces

Knowing the definition is not the same as knowing where you stand. These are the five questions LeaderWise draws from this framework.

Speaking up

Can your team tell you the truth right now, or just the version that's easy to hear?

Interpersonal risk and voice

Does speaking up on your team feel riskier than staying quiet?

Learning from failure

How did you respond the last time someone admitted a mistake to you?

Leader inclusiveness

Are you actively seeking different perspectives, or just tolerating them?

Belonging

As AI changes roles on your team, do people still know they matter?