Psychological Safety and Employee Retention: What Amy Edmondson's Research Reveals

In 2016, Google published the results of a two-year internal research project called Project Aristotle, an effort to identify what distinguished their highest-performing teams from others.

They examined 180 teams, measuring every variable they could think of: composition, expertise, seniority mix, management style, co-location.^[2]

The strongest predictor of team effectiveness, by a significant margin, was psychological safety — not talent, not experience, not working hours, but the degree to which team members felt safe to take interpersonal risks: to speak up, disagree, ask a basic question, or admit a mistake without fear of embarrassment or rejection.

Psychological safety was also, not coincidentally, among the strongest predictors of team-level retention.

→ Take the free CommStyle Assessment

What Psychological Safety Is — and Isn't

Harvard Business School professor Amy Edmondson, whose 1999 paper established the concept as a research field, defined it as a shared belief held by team members that the team is safe for interpersonal risk-taking.^[1]

This is not comfort. Psychologically safe teams aren't teams where everyone agrees or conflict doesn't happen — they're teams where people feel safe to disagree and raise concerns because they believe the environment will handle that honestly rather than punitively.

Psychological safety isn't the same as avoiding difficult conversations — research finds psychologically safe teams actually have more difficult conversations, because people believe those conversations are possible.

The Connection to Retention

Edmondson's research documents multiple pathways through which psychological safety reduces turnover. In psychologically unsafe environments, mistakes are hidden rather than acknowledged, and hidden mistakes compound — the stress of concealment is a sustained drain on wellbeing and commitment.

When people feel unable to raise concerns, they experience "voice behavior" suppression — the systematic stifling of their contribution, producing disengagement.

Teams with high psychological safety also show greater trust in their managers and organizations, because the quality of interpersonal safety reflects the quality of the overall environment.

And psychologically safe environments support more active learning — people try new approaches and course-correct more quickly, producing faster development and more engagement with the work itself.

Behaviors That Build Psychological Safety

Psychological safety isn't a property of the individual — it's a property of the team environment, shaped by the consistent behavior of all members and particularly those with higher status or authority.

Modeling vulnerability — when managers acknowledge uncertainty and admit mistakes — establishes that intellectual honesty is safe, with an outsized impact on team psychological safety.

Inviting input explicitly ("what am I missing here?", "does anyone see a problem with this approach?") changes the social equation, making contribution the expected response instead of requiring individuals to volunteer a potentially risky one.

Responding to concerns with curiosity ("tell me more about that") builds safety, while responses that minimize or dismiss reduce it.

And separating failure from character — focusing on "what do we understand now?" rather than "who is responsible?" — is the behavioral foundation of a team environment where errors get disclosed rather than hidden.

Behaviors That Erode Psychological Safety

The most common safety-eroding behaviors are often unintentional, a consequence of stress or time pressure.

Research identifies: visible or audible negative reactions to ideas in group settings, inconsistent follow-through on promises to address concerns, public criticism that focuses on the person rather than the situation, nonverbal signals of disinterest when someone is speaking, and "safe" responses to unsafe questions that close off further inquiry.

The cumulative effect of these behaviors is a social learning process — people observe what happens when someone takes an interpersonal risk, and calibrate their own future behavior accordingly.

Psychological Safety and the Retention Power Assessment

In the CommStyle Retention Power Assessment, psychological safety is Dimension ②, the second layer of Maslow's hierarchy as applied to workplace retention.^[4] Its position reflects the hierarchy's logic: physical wellbeing must be reasonably addressed before psychological safety becomes primary, but psychological safety needs to be reasonably established before higher-level dimensions can fully function.

Recognition from a manager who's unpredictably reactive means much less than recognition from someone in whose presence you genuinely feel safe.

A low score on Dimension ② is therefore a signal to prioritize this dimension before investing heavily in recognition or growth support.

→ Take the free CommStyle Assessment

Frequently Asked Questions

Is psychological safety the same as being nice?

No. Psychological safety doesn't require uniformly positive dynamics or the avoidance of difficult feedback. Research finds psychologically safe teams have more direct conversations, including critical ones, not fewer.

What it requires is that these conversations happen in an environment where people can engage without fear of being personally attacked or penalized for honesty.

Can one person build psychological safety for a whole team?

Individual behavior matters significantly — research consistently shows the behavior of higher-status members has a disproportionate influence on team psychological safety.

A manager or senior colleague who consistently models vulnerability and responds constructively to concerns can materially shift the team's safety level.

How long does it take to build psychological safety?

Edmondson's research suggests psychological safety is relatively stable within a team environment but sensitive to behavioral changes.^[3] Consistent positive behavior tends to build safety gradually over weeks and months, while a single significant safety-eroding incident can reduce it substantially — the asymmetry means maintaining safe behavior consistently is more effective than recovering after breaches.