Psychological Safety Examples You Can Actually Observe
Psychological safety shows up as observable moments, such as a junior person correcting a senior one and nothing bad happening afterward. Its absence shows up as the silence where that moment should have been. Here are twelve of each, paired.
What does psychological safety look like in practice?
Psychological safety looks like a junior person correcting a senior one in a meeting and nothing bad happening afterward. Its absence looks like the silence where that correction should have been, followed by a private message forty minutes later. Every example below is a variation on that one pattern: somebody takes a small interpersonal risk in public, and the room's response makes the next one either cheaper or more expensive.
Amy Edmondson defined the term in 1999 in Administrative Science Quarterly as a shared belief held by members of a team that the team is safe for interpersonal risk-taking. The definition is precise, and it is also invisible from the outside, which is the trouble a manager hits the moment they agree with it. You cannot watch a shared belief. You can watch what people do with it.
So this page is a list of behaviors, paired. Each example gives the moment when safety is present and the same moment when it is missing. Read the second half of each pair first. Leaders recognize the absence long before they recognize the presence, because absence is quiet and presence looks like an ordinary meeting going fine.
Twelve paired examples you can watch for
These cover the six places where the signal is strongest: regular meetings, incident reviews, one-to-ones, hiring debriefs, escalations, and handovers.
- An analyst says the demand number in the deck is wrong, the VP asks which input is off, and the deck changes in the room. Missing: the analyst sends the correction privately afterward, and one person decides whether it travels.
- A director asks what they are missing and three people answer inside ten seconds. Missing: the question gets a pause, then someone restates the plan approvingly.
- In an incident review, the engineer who pushed the change says "that one was mine, here is what I did not check" in the first two minutes. Missing: forty minutes of timeline reconstruction and nobody uses the word "I".
- Someone says the runbook was wrong while the runbook's owner is sitting there, and the owner agrees. Missing: process problems only ever get raised about teams who are not in the room.
- A report tells their manager eleven days out that the date will slip. Missing: the slip appears on the due date with a recovery plan already attached, so it arrives as competence instead of as bad news.
- A report says they did not understand what their manager meant last week and have been guessing since. Missing: people guess.
- In a hiring debrief the most junior interviewer argues against a candidate the hiring manager liked, and the hire does not happen. Missing: scores drift toward whatever the senior person said first, and you can watch it happen in the order people speak.
- A "no" goes into the file with its reason attached and stays there. Missing: the notes record a consensus that did not exist in the room.
- Someone escalates around their manager to unblock a customer, tells the manager they did it, and that is the end of it. Missing: the escalation gets relabeled a communication issue and discussed later without that person present.
- A leader hears about a problem from the person closest to it rather than from a dashboard or a peer. Missing: bad news moves sideways faster than it moves up.
- An engineer going on leave documents the parts of their own system they do not fully understand. Missing: handover notes describe only the parts that work.
- A new hire asks a basic question in a public channel in week three and gets an answer with no joke attached. Missing: new hires ask in direct messages, and the channel history shows it.
The three moments that tell you the most
If you only get to observe three things, observe the first disagreement, the first mistake, and the first piece of bad news after a team forms or a new leader arrives. Those three set the price of every one that follows.
The first disagreement settles whether dissent costs anything. Watch what the most senior person does in the four or five seconds after someone contradicts them. A follow-up question is a different signal from a defense, and everyone present reads it correctly.
The first mistake settles whether blame and analysis are separated. If the review opens with who, the team learns to make the next mistake quieter. If it opens with what happened and stays there until the mechanism is clear, the team learns that an error is a source of information.
The first piece of bad news settles the route. Did it arrive early, from the person holding it, while it was still useful? Or late, from a system, already too shaped to act on? A team tells you its safety level in how it delivers the first schedule slip, and confirms it in the reaction that slip gets.
Why observed examples beat a survey score
A climate survey asks people how they feel about the room. Edmondson's original instrument does that well and it is still the standard way to measure the construct she defined. What it cannot do is tell you which meeting to sit in on Thursday.
Observation and survey scores disagree more often than people expect, and the disagreement is useful in both directions. A team scoring high while the behavior looks thin usually has a polite culture rather than a safe one: nobody gets punished because nobody is taking a risk worth punishing. A team scoring low while the behavior looks strong usually has one relationship going badly, and the average is hiding it.
When the two disagree, trust the behavior and go find out why the number says otherwise. Ask what happened the last time someone raised something uncomfortable. People remember the specific incident, and they will describe it in more detail than any scale will give you.
What it looks like on a remote team
Remote and hybrid teams give you fewer moments and better records. The observable unit shifts from body language to routing: who asks in the channel and who asks in a direct message, who answers a question from someone three levels below them, whether a correction gets posted in the thread or sent quietly.
Two things are worth counting for a month. First, the ratio of questions asked in public channels to questions asked in private, broken out by tenure. New people asking privately while long-tenured people ask publicly is the clearest remote signal I know. Second, what happens in the ninety seconds after a leader posts a proposal. Instant agreement from three people is usually positioning rather than agreement.
What to do with what you observe
Keep a note. Not a scorecard, a note: the date, the moment, who took the risk, what happened next. Ten entries over a month will tell you more than a quarterly engagement report, and unlike a score they are specific enough to act on.
Then separate the two questions, because they need different fixes. The climate question is whether people believe speaking up is safe here. The behavioral question is whether what gets said, once it is said freely, turns out to be reliable: commitments kept, hard things raised early, loops closed. That second half is what the 12 Cs of Trust in the TrustFlow™ methodology score across four quadrants, and what the Team TrustFlow Index rolls up at team level on a 1.00-5.00 scale. The two are separate constructs, and as I argue in Is Psychological Safety the Same as Trust?, a team can hold plenty of one and very little of the other.
Safety is necessary and it is not sufficient. Edmondson's work is among the better-supported bodies of research in organizational behavior and I am not arguing with it. My point is narrower: the belief and the behavior get measured differently, and a leader who only ever collects the belief is guessing at the half that ships the work. To score the behavioral half, start with the trust score. To work on the room itself, that is what psychological safety training is for.
FAQ
What are examples of psychological safety at work?
A junior analyst correcting a senior leader's number in the meeting rather than in a private message afterward. An engineer opening an incident review by naming their own mistake. A report flagging a slipped date eleven days early instead of on the due date. A new hire asking a basic question in a public channel and getting a straight answer. Each one is a small interpersonal risk taken in front of other people, followed by a response that makes the next risk cheaper.
What does low psychological safety look like?
It looks like an ordinary meeting that runs on time. Corrections arrive privately after the fact. Nobody uses the word "I" in an incident review. Hiring scores drift toward whatever the most senior person said first. Bad news travels sideways between peers faster than it travels upward. New hires ask their questions in direct messages. Low safety is quiet, which is why it is usually diagnosed as something else, like a process problem or a communication problem.
How can you tell if your team feels safe?
Watch three moments rather than surveying for a general feeling: the first disagreement, the first mistake, and the first piece of bad news after a team forms or a new leader arrives. In each case the signal is the reaction in the four or five seconds afterward, not the invitation that came before. A follow-up question reads differently from a defense, and everyone in the room reads it correctly the first time.
Is psychological safety the same as being nice?
No, and the two often work against each other. A polite team can look safe because nobody is punished, while the real reason nobody is punished is that nobody is taking a risk worth punishing. Amy Edmondson's 1999 definition is about interpersonal risk-taking, which means disagreement, admitted error and unwelcome news. A team that never produces any of those is comfortable rather than safe.
Can a team look safe and not be safe?
Yes. The common version is a team that scores well on a climate survey while the observable behavior is thin, usually because agreement is fast and dissent never surfaces. The opposite happens too: a team scores low while the behavior looks strong, and one damaged relationship is dragging the average down. When the score and the behavior disagree, go and ask what happened the last time somebody raised something uncomfortable.
How do you measure psychological safety?
The standard method is a climate survey built on Edmondson's original instrument, which asks team members whether they agree the team is safe for interpersonal risk-taking. That measures the belief. It does not measure whether what people say, once they are willing to say it, gets acted on reliably, which is a behavioral question scored separately across the 12 Cs of Trust and rolled up at team level by the Team TrustFlow Index.
You will not find psychological safety by asking for it in a meeting. You will find it in what happened the last time somebody took you up on the invitation. Watch ten of those moments and write them down, and you will know more about your team than a quarterly score will ever tell you.