Psychological safety has become one of those terms that gets used to mean almost anything comfortable: a friendly team, a boss who doesn’t yell, a culture where nobody gets criticized. Harvard’s Amy Edmondson, who coined the term in a 1999 study published in Administrative Science Quarterly, would say that version of the idea has it almost backwards.

Edmondson’s original research looked at hospital nursing teams and set out to test something that seemed obvious: better teams should make fewer medication errors. What she actually found was the opposite, and it’s the part of the study that gets left out of the LinkedIn-slide version of psychological safety. The teams rated highest by independent observers on process and outcomes reported more errors, not fewer. Further digging showed why: those teams hadn’t made more mistakes. They were simply willing to talk about the ones that happened, because the team climate made it safe to say “I got this wrong” out loud. The lower-performing teams weren’t making fewer errors — they were burying them.

What psychological safety actually means

Edmondson’s own definition is precise, and it isn’t about comfort: a shared belief that the team is safe for interpersonal risk-taking. It’s the confidence that you won’t be punished, humiliated, or quietly marked down for speaking up — asking a question that might sound basic, flagging a problem before it’s someone else’s job to find it, admitting an error, or challenging an idea a more senior person just proposed.

None of that requires a team to be pleasant. Some of the psychologically safest teams on record argue hard, disagree openly, and push back on each other’s work constantly. What makes it safe isn’t the absence of friction — it’s the absence of a personal penalty for raising something uncomfortable.

Why this is a change-management problem, not just a culture one

Every change program depends on information flowing upward: which process actually isn’t working the way leadership thinks it is, which part of the new system the frontline already knows will break, which objection is real rather than reflexive resistance. A team without psychological safety doesn’t stop having that information. It just stops sharing it — exactly the dynamic Edmondson documented in the nursing units, except the stakes for a stalled change initiative are a quietly failed rollout instead of a quietly unreported error.

This is also why “resistance to change” is so often misdiagnosed. What looks like resistance is frequently silence from people who have real concerns and no safe channel to raise them. The organization reads the silence as buy-in, proceeds, and is surprised months later when the concerns turn out to have been correct all along.

Building it — what leaders actually do differently

Frame the work as a learning problem, not an execution problem. Edmondson’s later research found this framing shift alone changes how much people are willing to say. “We don’t fully know how this will go” invites input; “we have a plan, execute it” discourages it.

Respond to bad news as data, not as an indictment. The single fastest way to destroy psychological safety is to visibly punish, even mildly, the first person who surfaces a problem. Teams learn from what happens to the person who spoke up, not from the policy that says speaking up is welcome.

Ask explicitly, and ask specifically. “Does anyone have concerns?” in a group setting rarely works. Leaders who build safety ask individuals directly, in lower-stakes moments, what they’re seeing that isn’t being said in the room.

The diagnostic question

For any organization mid-change, the useful question isn’t “is our culture nice?” It’s narrower and more uncomfortable: if someone on the front line already knows this change is going to fail in a specific, fixable way, is there a real channel for that to reach the people who could act on it — and would they actually use it?