Psychological Safety Scale: Measure Shared Interpersonal Risk, Not Comfort

Communication · Executive Insight Brief · 209

The average can hide the risk

Assess a defined team climate while preserving aggregation evidence, divergent perceptions, and the boundary between safety and actual voice

Premise

Amy Edmondson’s seven-item Psychological Safety Scale measures a shared belief that a defined team is safe for interpersonal risk taking. It concerns expected consequences when members ask for help, admit mistakes, raise difficult issues, differ from others, or take socially exposed action. It is not a general score of comfort, happiness, or managerial virtue.

For executives, the value is diagnostic precision. A team climate may make speaking, questioning, and learning easier or riskier, but a high score does not prove that people voice concerns, that disagreement is gentle, or that decisions improve. Safety is an enabling condition, not the behavior or outcome itself.

The essential model

Edmondson developed the scale in a 1999 field study of team learning. The survey reached 496 members of 53 teams in one manufacturing company; 427 people from 51 teams responded, an 86 percent response rate. Seven positively and negatively worded statements cover mistakes, difficult issues, difference, interpersonal risk, help seeking, undermining, and recognition of unique skills. Items are scored so higher values represent greater psychological safety.

Although individuals answer, the intended unit is the team. Edmondson tested within-team convergence and between-team variation rather than merely averaging responses. An intraclass correlation of .39 supported meaningful team-level variance. Later meta-analysis covered 136 independent samples, more than 22,000 individuals, and nearly 5,000 groups. Broad evidence does not remove the obligation to test whether perceptions are sufficiently shared in the team being assessed.

Why this matters

Imagine a product team with strong delivery metrics but late discovery of technical risks. Its mean safety score is low, yet the distribution is more revealing: senior engineers report moderate safety while new engineers report very low safety. Interviews show that junior staff expect public questioning after mistakes and raise concerns privately after meetings.

That pattern directs attention to status, leader response, meeting practice, and escalation channels. It does not prove that the leader caused late defect discovery or that every junior employee remains silent. Pair the scale with speaking-up incidents, help seeking, error reporting, participation, learning routines, and qualitative accounts before making causal or intervention claims.

Risks and limits

A team average can hide the most important diagnosis. Response coverage, within-team dispersion, subgroup patterns, and the stated level of analysis belong beside the mean. Changing the referent from team to organization, department, class, or another setting is an adaptation; changing anchors, shortening the measure, or blending instruments also alters the measurement claim.

The scale provides no universal cutoff or prescribed intervention. It does not measure mental health, trust in general, absence of conflict, objective leader safety, or guaranteed voice. Define the unit, protect confidentiality, preserve scoring, avoid league tables, and name adaptations. Verify rights for commercial reproduction, translations, proprietary surveys, or publication of the full item set.

Executive takeaway

Use the Psychological Safety Scale to assess shared expectations about interpersonal risk in a defined team. Report distributions and convergence before promoting individual responses into a climate claim. Then connect the result to what people actually say, report, question, and learn. The score can reveal a condition for voice; it cannot certify a safe culture or explain its causes.

Key questions

• What precisely is the team or work unit whose interpersonal-risk climate is being assessed?
• Do response coverage, dispersion, and within-team convergence justify a shared team score?
• Which status or subgroup differences would disappear inside the mean?
• What behavioral and qualitative evidence shows whether people actually speak, seek help, and discuss error?

Selected sources

• Edmondson, A. (1999). Psychological safety and learning behavior in work teams. Administrative Science Quarterly, 44(2), 350–383.
• Edmondson, A. C., & Lei, Z. (2014). Psychological safety: The history, renaissance, and future of an interpersonal construct. Annual Review of Organizational Psychology and Organizational Behavior, 1, 23–43.
• Frazier, M. L., Fainshmidt, S., Klinger, R. L., Pezeshkan, A., & Vracheva, V. (2017). Psychological safety: A meta-analytic review and extension. Personnel Psychology, 70(1), 113–165.