Psychological Safety by Edmondson: How to Measure It and Not Kill It With a Conversation

Psychological safety isn't "we're all nice," isn't "we have an open culture," and isn't "we're all about respect." It's a specific thing: can I say "I failed," "I don't understand," "I think we're doing this wrong," and not get buried? Amy Edmondson has been researching this for 30 years. CHROs still confuse it with soft skills.

In conversations about team culture, the phrase "psychological safety" has become trendy. And like any trendy phrase, it loses its meaning. People use it for everything: from "let's hug it out" to "no one should ever criticize." The actual definition Edmondson gave in her research from 1999 onward is different and considerably more concrete.

This article is an attempt to restore the term's working meaning. What it actually is, how to measure it (with a real instrument), what a manager does, and which anti-patterns kill safety instead of building it.

What Psychological Safety Means According to Edmondson

Edmondson defined psychological safety as "a shared belief held by members of a team that the team is safe for interpersonal risk-taking." It's not "a pleasant atmosphere." It's about specific risks.

Her work identifies four types of interpersonal risk a person can take in a team:

• Looking ignorant — asking a question that reveals you don't know something. "Wait, could you please explain?"
• Looking incompetent — admitting you didn't manage. "I failed, I need help."
• Looking negative — criticizing a decision or direction. "I think we're doing the wrong thing."
• Looking like a troublemaker — questioning the status quo. "Why do we even do it this way?"

In a psychologically safe team, people are willing to take these risks. In an unsafe one, they shut down. And it shows immediately: in meetings no one asks clarifying questions, no one admits mistakes, no one criticizes, no one suggests "let's try a different way." Everyone stays silent and nods. A year later such a team goes through a real crisis, because all the accumulated "this doesn't work for us here" breaks through at once.

What matters — psychological safety doesn't contradict high standards. You can be highly demanding and psychologically safe at the same time. Edmondson has a famous 2×2 matrix: high standards + high safety = a "learning team" with the best results. High standards + low safety = a "fear team" with short-term effectiveness and long-term burnout. This is often confused.

The Four Stages of Safety According to Clark

Timothy Clark, building on Edmondson's work, proposed four sequential stages of psychological safety. This model doesn't supersede Edmondson; it adds detail — giving a clearer diagnosis of which stage a team is at right now.

Stage 1 — inclusion. The base layer. "I'm accepted as a person, regardless of experience, opinions, or views." Without this, a team isn't a team, it's an assembly of people. Cultures usually fail here when they have an aggressive practice of "hazing" new hires ("you're new — pay your dues"), group exclusion, or a rejection of difference.

Stage 2 — learning and questions. "I can ask, make mistakes, not understand — I won't be punished." This is about the ability to learn out loud. Teams stuck here are teams where new hires stay silent for three months, then suddenly break production, because they didn't ask a question. In our experience, this is the most common stalling point: stage 1 is passed, stage 2 isn't.

Stage 3 — contribution. "My contribution is valued and used." This is about my ideas genuinely influencing the team's work, not just being politely heard. Teams at stage 3 work well in stable times. At stage 4 they're needed in a crisis.

Stage 4 — challenge. "I can disagree with the status quo." This is about the ability to say "let's do it differently" to a manager or the organization. The hardest stage. If a team has it, it's self-learning, capable of innovation, of critical feedback, of early signals about problems. If not, sooner or later the team misses changes and loses.

Most teams, in our observation, sit between stage 2 and stage 3. At stage 4 — only a few.

How to Measure It — Edmondson's 7-Question Instrument

The main advantage of the Edmondson approach is that she has a working measurement instrument, published in her 1999 paper and validated many times over. It's seven statements an employee rates on a 5-point scale (from "strongly disagree" to "strongly agree"). Some statements are worded in reverse to detect auto-agreement.

You can embed this scale into a team survey as a separate section. A few rules, without which it doesn't work.

Anonymity strictly at the SQL level. If a team has fewer than 5 people, results for that team aren't aggregated. Otherwise an employee who answered "3" to the contribution question could be de-anonymized. And then there won't be honest answers.

Cycle — once every six months. More often and the measurement becomes self-inflicted fatigue. Less often and it lags reality.

Compare with yourself, not the industry. Psychological-safety benchmarks by industry depend heavily on methodology and culture. Far more honest is comparing "our figure today" with "our figure six months ago."

The result isn't a number, it's a conversation. An average score of 3.8 on its own means nothing. What does mean something is the trend and what we do in response. Without action, the measurement is useless.

What a Manager Does — Five Behaviors

Psychological safety isn't built through "let's speak openly." It's built through repeated manager behaviors. Edmondson identifies several key ones in her work; I've boiled them down to five that are easier to remember.

First — frame the work as learning. Not "we need to do X by Friday," but "we need to figure out how to do X by Friday." A small difference in wording, a big difference in perception. In the first case, the task is to do. In the second, to figure it out first — and that legitimizes questions and uncertainty.

Second — acknowledge your own fallibility. A manager who says "I can be wrong too, and on this project I'm not sure about several things myself" removes the team's illusion that they're waiting for the "right" answers. It's hard, because the culture of the manager as "all-knowing" dominates. But without it, there's no safety.

Third — ask questions, especially about the unknown. Not "who has what ready?", but "what should we understand about this project that we don't yet?" That's an invitation to openness, without the threat of evaluation.

Fourth — thank people for criticism, literally. When someone says "I think we're doing this wrong," the manager's reaction should be "thank you for saying that." Not "let's discuss," not "interesting thought" — but an explicit grateful acknowledgment of the risk the person took. This is the strongest signal that criticism is safe.

Fifth — respond to failure productively. When something didn't work out, the reaction is "extract the lesson," not "find the culprit." On a safe team, the retro after a failure looks for the system; on an unsafe one, for a specific person. The manager sets the tone — which kind of retro it will be.

These five behaviors aren't "phrases for an annual offsite." They're daily practice. If they're absent from everyday life, no declaration about "we have a safe team" works.

Anti-Patterns and the Link to Innovation

A few traps that masquerade as psychological safety but actually destroy it.

"We're all nice" as imitation. Social safety ("we love each other, we don't criticize") isn't psychological safety. It's often its opposite. When you can't criticize "because we're nice," criticism moves to the smoking area, and the team loses its ability to learn from mistakes.

Toxic positivity. "No need to raise problems, let's look to the future positively." This destroys stage 2 (learning and questions) and stage 4 (challenging the status quo). Within a quarter the team lives in the illusion that "everything's fine" and misses real signals.

Fake openness. The manager asks for criticism, then punishes the critic — literally or socially (through tone, exclusion from the circle, small sanctions). This is the most destructive pattern. One such case on a team, and psychological safety rolls back to stage 1 within a week.

Performative vulnerability. The manager says "I make mistakes too," but carefully hides their real mistakes or pins them on others. The team reads this and understands: declaration ≠ practice. Trust falls.

And now about the link to innovation. This is perhaps the most important point. The famous Google Project Aristotle study (2012–2015) showed that psychological safety is the top predictor of team effectiveness. Not skill level, not KPIs, not team size. Safety is what determines whether a team can perform at a high level sustainably.

The logic is direct. Without psychological safety:

• No innovation — no one proposes anything new, because new = the risk of criticism.
• No quality — no one says "we have a bug in production" until it blows up.
• No retention — for people who can't speak up, it's easier to leave than to endure.
• No learning — no one asks questions, no one admits gaps, the team stagnates.

This isn't a "soft skill." It's infrastructure for a team's ability to perform. And measuring it is a task no less important than measuring revenue.

The Bottom Line

Psychological safety according to Edmondson isn't atmosphere or warm relationships. It's a specific shared belief that the team is safe for four types of interpersonal risk: looking ignorant, incompetent, negative, or like a troublemaker.

Clark's four stages are a sequential diagnosis: inclusion → learning → contribution → challenge. Most teams are stuck at stage 2 or 3. Stage 4 is rare and valuable.

Edmondson's seven questions are a working measurement tool. With k≥5 anonymity, once every six months, compared with your own past.

Five manager behaviors — frame the work as learning, acknowledge fallibility, ask questions about the unknown, thank people for criticism, respond to failure productively. Daily practice, not phrases.

And the key point — psychological safety is infrastructure for team effectiveness, not a "soft skill for the employer brand." Without it, everything else is an illusion.

Next week we'll look at the data on the four-day workweek — the results of the 2022–2025 pilots and what they say about sustainability.

If you'd like the Edmondson survey (with the scale and scoring instructions), it's at the link in the card. No signup, in a spreadsheet, copy it for yourself.