In the decade since Amy Edmondson’s concept of psychological safety entered mainstream leadership development vocabulary, it has become one of the most cited constructs in discussions of team performance, innovation, and inclusion. It has also, in the process of its popularisation, undergone a degree of conceptual stretching that has left many organisations with a version of the idea that is simultaneously easier to endorse and less useful to act on than the original research supports.
The most common distortion is the equation of psychological safety with comfort. The organisation that asks “do our people feel psychologically safe?” and measures this through survey items about whether people feel respected and valued, whether they would describe their workplace as friendly, whether their opinions are solicited in meetings, is measuring something, but it is not primarily measuring psychological safety in the specific and consequential sense that Edmondson’s research identifies. Psychological safety is not about feeling good at work. It is about being willing to take interpersonal risks: to speak up, to challenge, to admit errors, to ask for help, to express genuine uncertainty, in contexts where these behaviours create real social exposure. The distinction between comfort and psychological safety is the distinction between a pleasant work environment and one that produces genuine learning and genuine performance.
This distinction matters practically because many organisations that have invested in creating comfortable, respectful, collegial working environments have been surprised and frustrated to find that those investments have not produced the levels of honest communication, constructive challenge, and genuine learning that psychological safety theory predicts. The reason is that they have produced comfort, not psychological safety. The two are related but not identical, and the developmental work required to produce one is quite different from the developmental work required to produce the other.
What Edmondson’s research actually shows
Edmondson’s original research, conducted in hospital settings and published in 1999, found that teams with higher psychological safety made more errors, in the sense that they reported more errors. This initially counterintuitive finding turned out to be precisely the point: teams with higher psychological safety were surfacing errors that were occurring in all teams but that lower-psychological-safety teams were not reporting. The higher-reporting teams had better learning outcomes and, over time, better performance outcomes, because they were actually engaging with the information needed to improve.
The research identifies psychological safety not as a feature of individuals but as a property of the collective: it describes the shared belief within a team that the environment is safe for interpersonal risk-taking. This collective character is important because it means that psychological safety cannot be built through individual attitude change alone. It requires the specific social conditions and leader behaviours that produce the shared belief at the group level, and it can be undermined by a single high-impact incident that violates the group’s sense of safety regardless of the accumulated investment that preceded it.
The most practically important finding in Edmondson’s subsequent research on psychological safety is the identification of specific leader behaviours that either build or undermine it. Leaders build psychological safety when they model vulnerability, acknowledge their own uncertainty and limitations, invite input before sharing their positions, frame mistakes as learning opportunities rather than performance failures, and actively protect team members who raise concerns from informal retaliation. Leaders undermine psychological safety when they respond to dissent with defensiveness or dismissal, when they signal impatience with the expression of doubt, when they allow the informal social cost of honest communication to go unaddressed, and when their stated openness to input is contradicted by their actual responses when that input is delivered.
The performance paradox
One of the most important and least well-understood features of psychological safety in high-performance contexts is what Edmondson calls the learning zone: the specific combination of high psychological safety and high performance standards that produces the conditions for genuine organisational learning. This is the zone that her research most consistently associates with sustained exceptional performance, and it is the zone that is most difficult to occupy because its two components are, in the short term, in tension with each other.
High performance standards, applied without psychological safety, produce the anxiety zone: people work hard, they are motivated by genuine concern about consequences for failure, but they do not surface problems, do not challenge assumptions, and do not share the information that would allow collective learning. The performance is real but fragile, because it is based on an information environment that is actively filtering out the evidence of its own limitations.
High psychological safety without high performance standards produces the comfort zone: people feel safe, collaboration is genuine, but the absence of performance pressure means that the learning potential of psychological safety is not fully activated. People are willing to take interpersonal risks but there is insufficient urgency to take the risks that produce genuine organisational learning.
The learning zone requires both: the safety to take the interpersonal risks and the performance pressure that makes taking those risks feel consequential and worth the effort. Creating and sustaining both simultaneously is harder than creating either alone, because the natural organisational response to performance pressure is often to reduce the safety that makes honest engagement with performance problems possible. The leader who can hold both, who can maintain high expectations while genuinely protecting the psychological safety that makes those expectations honest rather than merely pressuring, is doing something that is both rare and consequential for the quality of the team’s performance over time.
The specific failure modes in high-performance teams
High-performance teams, the ones where pressure is high, stakes are real, and the expectation of excellence is genuine, produce specific psychological safety failure modes that are different from those of lower-pressure environments and that require specific developmental attention.
The first is the hierarchy of performance as a proxy for hierarchy of ideas. In high-performing teams with strong track records, the ideas of the highest performers tend to receive more deference than their quality necessarily warrants, and the ideas of lower performers tend to receive less deference than their quality sometimes warrants. This is rational in a narrow sense, because past performance is a reasonable predictor of future quality. It becomes a failure mode when it produces a systematic discounting of perspectives that have not yet been validated by performance track record, which includes the newest members of the team and anyone whose performance style is less visible than the dominant norms of the team reward.
The second failure mode is the pressure to seem confident. In high-performance contexts, the expression of doubt or uncertainty is often implicitly coded as a performance deficit rather than as intellectual honesty. The team member who says “I’m not sure this is right” in a context where the social norm rewards certainty is taking a real reputational risk. The result is the systematic suppression of genuine uncertainty, which means the team makes decisions based on the expressed confidence of its members rather than on their actual epistemic state. This is a subtle but consequential degradation of the information quality that the team’s decisions are based on.
The third failure mode is the confusion of interpersonal harmony with intellectual agreement. High-performing teams frequently develop strong relational bonds that are genuinely valuable for the quality of collaboration they enable. Those bonds can also create the specific social dynamic in which challenging a colleague’s position feels like a betrayal of the relationship, because the closeness of the relationship makes intellectual disagreement feel more personal than it would in a more distanced professional relationship. This is the ironic failure of genuine cohesion: the very quality that makes the team function well interpersonally can undermine its capacity for the honest intellectual challenge that genuine learning requires.
Building genuine psychological safety in high-performance contexts
The interventions that build genuine psychological safety in high-performance contexts are not primarily about reducing pressure or softening expectations. They are about ensuring that the interpersonal architecture of the team is designed to make honest challenge genuinely safe in the specific and highly pressured context where it is most needed.
The most important single intervention is the explicit and consistent demonstration by the leader that genuine challenge, including challenges to their own positions and decisions, is not merely tolerated but actively welcomed and specifically rewarded. This demonstration needs to be visible and repeated, because the social evidence that genuine dissent is safe in this team accumulates slowly and can be set back by a single incident in which the cost of speaking up becomes visible. The leader who wants to build genuine psychological safety in a high-performance team needs to treat every moment when they respond to honest challenge as an investment in or withdrawal from the psychological safety of the team, and to make those investment choices deliberately rather than reactively.
The structural intervention that is most effective alongside this relational work is the deliberate creation of processes for surfacing doubt and uncertainty that do not require individual interpersonal courage to activate. Pre-mortem exercises, structured devil’s advocacy, after-action reviews that are facilitated by someone without a stake in the evaluation, all create the specific conditions in which honest intellectual engagement with the team’s actual performance is possible without requiring the kind of social courage that is in short supply in high-pressure, high-stakes team environments.
Psychological safety is not about being comfortable. It is about the willingness to take the specific interpersonal risks that genuine learning and genuine improvement require. Comfort without that willingness is pleasant. It is not the engine of performance.