Psychological Safety: What It Is and Isn't

How Edmondson Found It

Edmondson's discovery was almost accidental. As a graduate student in the early 1990s she was studying error rates on hospital nursing wards, expecting to find that better-functioning teams made fewer mistakes. The data showed the opposite. The wards she had identified as the highest-functioning, on the basis of nurse-manager interviews and observed behavior, had higher reported error rates than the lower-functioning wards. The finding was bizarre.

The resolution, which took her several years to work out, was that the high-functioning wards were not making more mistakes; they were reporting more of the mistakes they made. The low-functioning wards had cultures in which reporting an error invited retribution, so nurses simply did not report. The effect was that the low-functioning wards looked safer in the data while in fact being more dangerous in practice. Edmondson named the variable that distinguished the two kinds of ward 'psychological safety' — the shared belief that the team is safe for interpersonal risk-taking, including the risk of admitting that one has made a mistake.

The Construct, Stated Carefully

The crucial nuance, often lost in popular treatments, is that psychological safety is about interpersonal risk, not interpersonal comfort. A team is psychologically safe if a team member can say, 'I don't understand what you just said,' or 'I think we're heading toward a wall,' or 'I made a mistake on that ticket and the customer is upset,' without expecting retribution. Psychological safety is not about people being nice to each other. It is about people being able to say hard things to each other without political consequences.

The distinction matters because the popular version of the concept has drifted in the direction of niceness — the assumption that a psychologically safe team is one where everyone agrees, no one is criticized, and difficult feedback is avoided. This is precisely the wrong reading. Edmondson's later work makes the point explicit: psychological safety enables candor, not its absence. A team where everyone is polite and no one challenges anyone else is not psychologically safe; it is conflict-avoidant, which is a different condition with worse consequences. Real psychological safety produces more conflict on substance, not less, because the team has the underlying confidence to engage with hard disagreements rather than route around them.

Project Aristotle

Google's Project Aristotle, conducted in 2012-2014 by the company's people-analytics team, was an internal research effort to identify what made certain Google teams more effective than others. The team analyzed 180 Google teams across multiple dimensions — composition, skills, tenure, demographics, leadership — looking for the variables that predicted high performance. The result was unexpected. Most of the variables they had hypothesized to matter — team composition, individual brilliance, even the manager's experience — turned out not to predict performance reliably.

The variables that did predict performance were largely about how the team behaved. Five characteristics emerged as consistent predictors, with psychological safety as the strongest by a substantial margin. The other four were dependability (members reliably finish what they start), structure and clarity (members understand goals, roles, and execution plans), meaning (the work matters personally to members), and impact (the work matters in some larger sense). Project Aristotle's findings were widely publicized, and they brought psychological safety into the corporate vocabulary in a way that Edmondson's academic publication, by itself, never had.

What the popular coverage often missed is that psychological safety was the strongest of five variables, not the only one. A team that is safe but not dependable, or safe but lacking structure, will not perform well — the safety is necessary but not sufficient. The popularization tended to treat psychological safety as the answer; the research treats it as the necessary precondition for the other four to operate.

Where It Gets Weaponized

The misuse pattern is consistent and worth naming. In some organizations, 'psychological safety' has become the language used by underperformers to deflect critical feedback. 'I don't feel psychologically safe in this conversation' becomes a way of refusing to engage with substantive criticism of the work. Managers, fearful of being accused of creating an unsafe environment, soften feedback to the point of uselessness. Performance management decays. The team no longer has hard conversations. The result is precisely the opposite of what Edmondson's research calls for — a team that is conflict-avoidant masquerading as a team that is psychologically safe.

Edmondson herself has written about this misuse extensively, particularly in her 2019 book The Fearless Organization. Her position is that psychological safety must coexist with high standards. A team can be both safe and demanding; in fact, the combination is what produces high performance. A team that is safe but lacks accountability — what she calls the 'comfort zone' — produces nice meetings and mediocre work. A team that is demanding but lacks safety — the 'anxiety zone' — produces good short-term output and burnout. Only the combination of both — the 'learning and high-performance zone' — produces durable excellence. The misuse of safety language to undermine accountability is the most common modern failure mode of the framework.

How to Build It

If you want to build genuine psychological safety on a team, the operational practices Edmondson recommends are unglamorous and effective. Frame work as learning rather than execution: acknowledge that the team is operating under uncertainty, that mistakes will happen, and that the team's job is to learn faster than the competition. Acknowledge your own fallibility as the leader: name a mistake you made recently and what you learned from it. Model curiosity rather than judgment: ask 'what did we learn?' before you ask 'who messed up?' Respond to bad news with appreciation for the messenger before any conversation about the problem.

The single most powerful practice is the leader's own behavior in the moments when bad news arrives. The team is constantly reading the leader for the real rules. If the leader receives bad news with anger, blame, or dismissal, the team learns that bad news will not be reported. If the leader receives bad news with curiosity and a focus on what to do next, the team learns that bad news is welcome and that the team is on the same side. This is, again, leading by example as the actual mechanism by which culture is transmitted (Chapter 24). Edmondson's research is, on close reading, a sustained empirical case for the same claim Schein made operationally: culture is what leaders pay attention to and how they respond.