Psychological Safety Isn't Niceness. The Original Study Never Said It Was.

Most leaders today meet psychological safety through Amy Edmondson's book "The Fearless Organization" or the keynotes it inspired, and the attention is deserved. The 1999 study where it all began says something sharper than the popular version: psychological safety is a team-level condition for learning under pressure, and it stands next to accountability, not instead of it.

In the mid-1990s, Amy Edmondson was studying medication errors in hospital nursing teams when she ran into a finding that at first looked like bad news for good management: the better-led teams appeared to make more errors. The data resolved the puzzle in an unexpected way. Those teams did not make more mistakes; they reported and discussed more of them. The weaker teams were not error-free, they were silent. Out of that puzzle grew one of the most productive research programs in management, and a 1999 paper that deserves to be read in the original.

What did Edmondson actually find in 1999?

Amy Edmondson studied 51 teams in a single manufacturing company: management teams, product development teams, staff services, production crews. In "Psychological Safety and Learning Behavior in Work Teams" (Administrative Science Quarterly, 1999), she showed that teams differ in a shared belief she called team psychological safety: the sense that the team is, in her words, "safe for interpersonal risk taking."

Two findings from the paper rarely survive the keynote version.

First, psychological safety is a team-level phenomenon. Within one company, under the same strategy, the same HR policies and the same official culture, the 51 teams varied widely. Some had teams where people asked for help and admitted mistakes as a matter of course; a few doors down, people kept their heads down. That single empirical fact settles a debate that still runs today: psychological safety is not a corporate program, and it is not a personality trait. It is produced locally, above all by what team leaders do and how colleagues respond to each other.

Second, the paper specifies the mechanism. Psychological safety does not improve performance directly. It works through learning behavior: asking questions, admitting errors, seeking feedback, experimenting, raising problems early. Every one of these behaviors carries interpersonal risk; nobody wants to look ignorant or incompetent in front of the people who evaluate them. Safety lowers the price of these behaviors, teams learn more, and better performance follows. The hospital wards had already shown the flip side: silence hides exactly the data that improvement depends on.

What happened between the paper and the keynote stage?

The concept's rise took about two decades. Google's Project Aristotle made psychological safety famous in the mid-2010s, naming it the strongest differentiator among its own teams. Edmondson's book "The Fearless Organization" (2018) is her definitive statement on the subject: two decades of research, condensed for practitioners, and well worth reading in full. Somewhere in the popular reception, however, the concept drifted toward niceness: no criticism, no discomfort, no demands. And with that drift came a perceived tension with performance cultures – as if a team could be psychologically safe or held to high standards, but not both.

Both the paper and the book push back on that either/or. The 1999 paper explicitly distinguishes psychological safety from group cohesiveness, noting that cohesive groups can be the least willing to challenge each other; that observation comes from the groupthink literature, not from us. The book is even more direct: psychological safety does not imply a cozy environment, and it is inaccurate to equate it with removing consequences for poor performance. Safety and standards are two different dials. High standards without safety produce anxiety and silence. Safety without standards produces comfort and drift. Learning happens where both are high.

So here is our position, which is also Edmondson's: psychological safety is a performance technology, not a wellbeing perk. Treat it as a team-level leadership discipline, with the same seriousness you would give a production process, and stop treating it as an HR climate topic that lives in the annual engagement survey. In our work with leadership teams, a speak-up problem rarely announces itself as fear; it shows up as meetings that run smoothly and surface nothing new.

A decision rule we use: if, over the next four to six weeks of leadership team meetings, nobody brings you a mistake, a doubt, or a piece of bad news you didn't already know, do not congratulate yourself on harmony. Assume you have a speak-up problem, and treat it as your leadership task. Expect the climate to shift over months of consistent behavior, not after one  offsite.

How do you build it – in real meetings, not in a workshop?

Because psychological safety is made locally, it is also learned locally. It does not respond to proclamations ("my door is always open") or to one-off training events. It responds to concrete leader behaviors, practiced repeatedly in the team's actual meetings, with reflection in between. Edmondson's own toolkit sorts these behaviors into three moves: set the stage by framing the work as uncertain and interdependent, so that speaking up is obviously necessary rather than optional; invite participation by asking real questions and deliberately drawing in the quieter members; and respond productively when someone does take the risk, by appreciating the act, destigmatizing the error, and then dealing with the substance. That last move is where accountability lives. Appreciating candor is not amnesty: a productive response can still be "this can't happen again, and here is what we change."

In our team development work, the pattern that helps most is unglamorous: pick a small number of behaviors, run them across several ordinary meetings, and check with the team afterwards whether anything felt different. Routines beat intentions.

On the Management Kits platform, the Working Topic "Ensure team members feel safe to speak up" (in the Team Effectiveness kit) turns this into a working plan. Its Psychological Safety Leader Work Tool is structured around the same three moves, and the associated milestone spells out five actions across a series of team meetings, under four hours of work in total, ending with a short feedback check-in where the team tells you whether speaking up has become easier.

Where does this meet AI transformation?

Wherever organizations are introducing AI right now, learning behavior is the binding constraint. Adopting these tools means experimenting in front of colleagues, admitting what you don't understand, and surfacing attempts that failed. Surveys across the last two years show a large majority of companies reporting AI adoption while struggling to turn it into results, and part of that gap is conversational: in many teams, people use AI quietly and share only their successes. A team that cannot talk about mistakes will not talk about failed prompts either. The speak-up climate Edmondson measured in 1999 has quietly become an AI-readiness condition.

What the paper can and cannot carry

Honest limits: the 1999 study covers one company, with cross-sectional data and self-reported behaviors; the broader evidence base has grown since, but any single study deserves modest weight. Psychological safety is necessary rather than sufficient; without direction and standards it buys you pleasant meetings, not performance. And it resists direct measurement inside your own team: loudly asking "do you all feel safe?" mostly produces the answer you asked for.

Which brings us back to the hospital wards where the story started. The teams that looked worse in the error statistics were the ones doing the learning; the quiet teams only looked good. That is the reading of psychological safety worth keeping, and it is the one Edmondson has defended from the 1999 paper through to her latest book. The point was never comfort. The point was to make hard things sayable.

Want to see how your leadership team could work on speak-up routines with the Management Kits platform? Schedule a short demo.