Ask a room of executives if they create a psychologically safe environment for their teams, and most say yes without hesitation. Ask the people who report to them, and the number drops. Perceptyx found that 87 percent of executives reported positive perceptions of psychological safety, compared with 69 percent of individual contributors and 66 percent of hourly workers. A 2024 workplace benchmark found the gap even wider: 45 percent of leaders described their workplace as psychologically safe, compared with 23 percent of frontline employees. Safety inside the boardroom itself came in at just 40 percent.
That gap has a specific shape, worth understanding on its own terms, because the shape explains why it resists most of the obvious fixes.
A claim you can't check from where you're standing
Tasha Eurich's research splits self-awareness into two kinds: internal, how clearly someone sees their own values and intentions, and external, how accurately they understand their actual impact on others. "I create psychological safety" sounds like an internal claim. It isn't. It's a claim about someone else's internal experience, whether another person feels safe enough to speak up. No amount of honest reflection verifies that. The answer lives inside the other person, which means the leader is asking a question they are structurally unable to answer on their own.
The evidence arrives pre-filtered
Amy Edmondson's work on psychological safety gets at the mechanism. A leader's sense of how safe the team feels comes from what they can see: who speaks up, who pushes back, who brings bad news early. But that's only the group that has already decided it's safe enough to act. It says nothing about everyone who ran the same calculation and stayed quiet. A leader watching plenty of pushback can reasonably conclude the room is open, while missing the larger, silent group that bet the other way. The evidence was filtered by the exact dynamic being measured, before it ever reached the leader.
Authority edits the data before it arrives
Manfred Kets de Vries adds the second layer: authority itself changes what people are willing to say, independent of the leader's actual intentions. A direct report weighs career risk before offering a hard truth, and that calculation runs whether or not the leader would have welcomed the honesty. So a leader's own sense that people speak freely around them isn't neutral evidence. It's evidence gathered inside the one dynamic built to suppress the exact signal they're trying to read.
Why the survey doesn't change the belief
At this point, a bad employee engagement survey should be enough to correct the picture. Often it isn't. Robert Kegan and Lisa Lahey's work on immunity to change explains why: people carry hidden competing commitments, goals working quietly against the change they claim to want. For many leaders, "I create psychological safety" isn't just a claim about the team. It's protecting something more personal: the self-image of being open, approachable, fair. Disconfirming data directly threatens that image, so it gets absorbed as noise: a bad quarter, a difficult team, an unfair question. Rarely as an accurate signal about behavior.
The reframe
None of this makes psychological safety unmeasurable, nor does it make the effort pointless. It means the measurement can't come from the leader's own vantage point, not occasionally, but structurally. A leader cannot self-assess a condition defined entirely by someone else's experience. What they can build is the outside view, deliberately: real, anonymous, specific feedback on whether people actually feel safe raising a hard truth, gathered often enough that one bad quarter can't explain it away. The gap between belief and reality isn't a character flaw. It's what happens by default when the only evidence in the room has already been filtered by the room itself.