Psychological Safety

Psychological safety is the shared perception that a team is safe for interpersonal risk taking, in Amy Edmondson’s definition as presented by Google’s re:Work guide. The risks include asking a basic question, admitting a mistake, disagreeing, or offering an uncertain idea. The concept concerns the expected social consequences of those acts. It does not mean that everyone agrees, every proposal is accepted, or performance standards disappear.

Google’s Project Aristotle account places psychological safety first among five dynamics it identified in its own teams, followed by dependability, structure and clarity, meaning, and impact. These findings describe that research population. Variables such as team size and colocation did not show significant associations in that analysis, which the guide cautions does not make them unimportant elsewhere. Use the findings to inform investigation in another organization, not as a universal ranking of causes.

How it is measured, and what it is not

The re:Work guide reproduces Edmondson’s survey statements about mistakes being held against people, raising difficult issues, rejection for being different, interpersonal risk, asking for help, deliberate undermining, and whether skills are valued. Respondents rate their agreement. Some statements run in the opposite direction from others, so an instrument must define its response scale, reverse coding, and aggregation before scores are interpreted. A set of selected statements is not automatically a validated replacement for the original measure. The guide separately gives an example item from Google’s team survey about whether a respondent’s mistakes are held against them; it does not establish that this was the entire psychological-safety scale.

Interpret responses with their context: who was invited, who responded, when the data was collected, and whether team members have different experiences. Explain confidentiality and reporting limits, particularly where small groups or comments could identify individuals. Behavioral evidence adds another view: what happened when someone raised a concern? One response may reveal a problem worth addressing, but it does not measure every member’s perception. Silence can reflect missing information, lack of time, or another reporting channel as well as fear. Neither a survey score nor an isolated observation has automatic priority over the other.

Safety and accountability can support each other. Dependability is another dynamic in Google’s account, and the UK Civil Service standards combine supportive discussion with prompt attention to poor performance and behavior. Fear of punishment can suppress information needed to learn, correct problems, and judge conduct fairly. It is neither the only reason for missing information nor a guarantee that people will remain silent. Responsibility still exists when a problem has not been disclosed; better information improves how it is understood and acted on. Respectful disagreement and an honest conversation about a missed commitment are compatible with psychological safety.

Google’s guide reports Edmondson’s recommendations to frame uncertain work as a learning problem, acknowledge one’s fallibility, and ask curious questions. In a fictional migration, an engineer learned of a business freeze only after six weeks. His earlier silence cannot establish fear when he did not yet know that constraint; ask what he knew and how he experienced the manager’s response. The related expectations and delegation practices appear in Fully Delegated, Then Overruled. The blameless postmortem applies a related principle to incident learning: make candid reporting possible while investigating causes and following through on improvements.

References

• Google re:Work, Understand team effectiveness.
• UK Civil Service, The Line Management Standards: Core stage (April 2025, V1.2).