Psychological Safety: The Four Leader Behaviours That Actually Build It
Almost every leader will say their team has psychological safety. Far fewer can describe what they did this week to create it.
That gap is the whole problem. Psychological safety is not a state of mind your team chooses to adopt. It is a working pattern that gets reinforced, or quietly eroded, by the small behaviours their leader repeats in meetings, in 1:1s, and in the moments where someone surfaces something uncomfortable.
This article is a practical look at the four leader behaviours the research keeps coming back to. It is grounded in primary sources, not in management self-help, and it is written for leaders who are willing to look honestly at what they do, not just what they intend.
What Psychological Safety Actually Means
The phrase comes from Amy Edmondson, then a doctoral student at Harvard, in a 1999 paper in Administrative Science Quarterly titled "Psychological Safety and Learning Behavior in Work Teams." Edmondson defined it as the shared belief, held by members of a team, that the team is safe for interpersonal risk-taking. In practical terms: it is safe to ask a question that might sound stupid, admit a mistake, push back on a decision, or raise an issue that nobody else wants to raise.
It is not the same as being nice. It is not the absence of conflict. It is not consensus. Edmondson has been clear, including in her later work The Fearless Organization (2018), that psychological safety raises the standard of what gets discussed, not the bar for who gets to speak comfortably. High-safety teams disagree more openly, not less.
The reason this matters is that the cost of low psychological safety is invisible. People do not file a ticket when they decide not to flag a concern. They just stop flagging concerns. By the time the consequences show up in a missed deadline, a quality failure, or a regretted hire, the original silence is months in the past and impossible to trace.
What the Research Actually Says
Three pieces of evidence are worth holding in mind before we get to the behaviours.
The first is Google's Project Aristotle, a two-year internal study published in 2015. Google researchers analysed more than 180 teams across the company, looking at over 250 attributes, trying to find what separated high-performing teams from average ones. The variables they expected to matter (composition, intelligence, tenure mix) did not. Psychological safety came out as the single strongest predictor of team effectiveness, with team dependability, structure and clarity, meaning, and impact rounding out the top five.
The second is Edmondson's broader body of work in healthcare and engineering teams, which has consistently found that higher psychological safety correlates with more error reporting, faster learning, and better patient or product outcomes. The counterintuitive finding from her early hospital studies was that the best teams reported more errors, not fewer, because their members felt safe to say what had actually happened.
The third is more recent. A 2026 paper in the International Journal of Discrimination and the Law by Clarke and colleagues, "Safe to Stay," found that a climate of high psychological safety positively predicted employees' decision to remain in their organisation, even in high-demand workplaces. Retention is downstream of safety, not upstream of it.
The headline does not change. Teams that feel safe say more, learn faster, and stay longer.
The Four Behaviours That Actually Build It
The behaviours below are the ones that show up across Edmondson's work, the Project Aristotle write-ups, and the leadership behaviour studies. They are not abstract values. They are things you can see yourself doing, or not doing, on Tuesday at 10am.
1. Frame the Work as Learning, Not Performance
The first behaviour is the way you describe the situation your team is operating in. If you frame the work as a test of competence, with a clear right answer and a clear wrong one, people will optimise for not being wrong. They will stay quiet when uncertain. They will hedge their language. They will not tell you that they think the strategy has a hole in it.
If you frame the same work as a learning problem, where the answer is not yet known and finding it requires several attempts, the calculation flips. Now staying quiet is the failure mode. Surfacing a half-formed objection is the contribution.
Edmondson's framing in The Fearless Organization is that leaders should explicitly say, in front of the team, things like: "I may miss something. Please speak up if you see an issue." It sounds small. The reason it works is that it does two things at once. It signals that you, the leader, expect to be wrong sometimes, which lowers the perceived cost of being wrong for everyone else. And it gives the team a verbal protocol they can use later: when someone says "I think there's an issue with the plan," they are using your invitation, not breaking ranks.
The opposite behaviour, the one to watch for in yourself, is framing every meeting as a status report against a known target. Status framing is fine for execution-mode work. It is the wrong frame for any work where the team's collective judgment matters more than the leader's individual judgment, which is most strategic work.
2. Separate Contribution from Position
The second behaviour is whether you reliably hear ideas without first weighting them by who said them. This is harder than it sounds. Almost every leader filters input through a fast, mostly unconscious model of who is reliable. Senior people get listened to. Quieter people get nodded at. New hires get heard but not acted on.
The Project Aristotle data found that the highest-performing teams had what the researchers called "equality in conversational turn-taking" — over the course of a meeting, every member ended up speaking roughly as much as every other. Nobody dominated. The teams that violated this pattern, where one or two people did most of the talking, performed worse, even when those one or two people were the most experienced.
The behaviour to practise is direct. In a meeting, after the senior voices have spoken, ask the quietest person in the room what they think, by name, about the specific question on the table. Do not ask "any thoughts?" to the room. The quietest person will not volunteer. Ask them. Then listen without interrupting, and visibly take what they said into the next part of the conversation.
The first three or four times you do this, it will feel mechanical. Within a month, two things will happen. The quieter members of the team will start volunteering input without being asked, because they have learned the room is for them too. And you will start catching errors of judgment that the senior voices were missing, because the senior voices were operating from the same set of assumptions.
3. Respond Productively to Bad News
The third behaviour is the one that disproportionately determines whether the first two stick. It is what you do in the first 30 seconds after someone tells you something you did not want to hear.
If a team member says "we missed the launch date," or "I think the analysis we presented to the board was wrong," or "I disagree with the call you made on Friday," the team is watching what happens next. Not the words. The body language, the tone, the pace of the response.
Three responses kill psychological safety reliably. The first is visible irritation, even if the words that follow are mild. People read frustration faster than they read content, and they will encode "do not raise this kind of thing again." The second is performative calmness that does not actually engage with the content, the kind where the leader thanks them and moves the meeting on. People read that as dismissal. The third, and most damaging, is asking who else knew and for how long, before asking what the actual issue is. That signals that the leader is solving for blame, not for the problem.
The behaviour that builds safety is small and almost embarrassingly simple. Say "thank you for raising this" and mean it. Then ask what we know, what we do not know, and what the next sensible step is. Do not look for who to hold accountable in the same conversation. Hold accountability conversations later, separately, and only when the facts are clear.
The reason this works is that you are not pretending the bad news is good. You are demonstrating that the team can bring you bad news without the act of bringing it being held against them. Over time, that distinction is the entire foundation of the team's willingness to tell you things early, while there is still time to act on them.
4. Make Your Own Errors Visible
The fourth behaviour is whether the team has ever heard you say "I got that wrong." Not in a performative way. Not as a setup for a story about how you fixed it. Just as a flat statement of fact, in the same tone you would use to describe the weather.
Edmondson's hospital research found that nursing units with leaders who openly discussed their own errors had higher error reporting overall, which counterintuitively correlated with fewer adverse patient outcomes. Errors that get reported can be addressed. Errors that get hidden compound.
For most leaders, the resistance to this behaviour is rooted in a worry that admitting error in front of the team will reduce their authority. The data, and the lived experience of leaders who have tried it, suggests the opposite. A leader who can say "I made the wrong call on the December restructuring, and here is what I would do differently" gives the team permission to say the same thing about their own work. A leader who never visibly errs forces the team to perform the same level of inerrancy, which they cannot achieve, which means they have to hide their errors, which means the leader cannot see them coming.
The frequency does not need to be high. One genuine, specific, non-performative admission of error per month, in front of the team, will shift the culture more than any value statement on a wall.
Where Most Leaders Get This Wrong
Two failure patterns are worth naming directly.
The first is treating psychological safety as a soft concept that lives in offsites and team-charter conversations. It does not. It is created and destroyed in the texture of normal weeks. The leader who runs a great offsite about candour and then visibly bristles in a Wednesday meeting when someone raises a concern has just told the team that the offsite was theatre.
The second is conflating safety with comfort. Edmondson is direct on this in The Fearless Organization: psychological safety is not the absence of discomfort. It is the willingness to be uncomfortable in service of being honest. Teams with high psychological safety often have more friction in their meetings, not less, because real disagreements get aired rather than smoothed over. The leader's job is not to remove the discomfort. It is to ensure the discomfort is productive and survivable.
If your team meetings feel uniformly pleasant, that is not necessarily a sign of safety. It is at least as likely a sign that the team has learned what is and is not safe to say.
What This Looks Like in Practice This Month
A simple test you can run on yourself, without telling anyone you are running it.
For the next four weeks, at the end of each week, write three things in a note you do not share. The most important question someone on the team asked you. The most useful piece of pushback you received. The one piece of information you wish you had heard earlier than you did.
If two of those columns are mostly blank by week four, the team is probably not telling you what you need to hear. The work is to figure out which of the four behaviours above you are not yet doing, and to start.
Most of the leverage in leadership is in things you do not feel doing in the moment. Psychological safety is one of them. The team feels it. You usually do not. And the only way to know it is working is that the bad news arrives early, the quiet voices speak up, and the meeting after the mistake is shorter than the meeting before it.
For more on the trust dynamics that underpin psychological safety in distributed teams, see our piece on building trust in remote teams. For the related skill of running 1:1s where this kind of candour can actually surface, see running 1:1s that actually matter.
Sources: Amy C. Edmondson, "Psychological Safety and Learning Behavior in Work Teams," Administrative Science Quarterly (1999); Amy C. Edmondson, The Fearless Organization (Wiley, 2018); Charles Duhigg, "What Google Learned from Its Quest to Build the Perfect Team," The New York Times Magazine (February 25, 2016), reporting on Google's Project Aristotle; Daniel Goleman, "Leadership That Gets Results," Harvard Business Review (2000); Emma Clarke, Sanna Malinen, Katharina Näswall, Annick Masselot, "Safe to Stay: The Role of Leader Behaviors and Psychological Safety in Employee Retention in High-Demand Workplaces," International Journal of Discrimination and the Law (2026).
Join 161,800+ leaders at The Leader's Table for research-backed leadership analysis.