Concept/Organizational Concept/No. 0794
Psychological Safety
Psychological safety is a shared belief that a team is safe for interpersonal risks, such as asking for help or reporting errors. Amy Edmondson defined the team concept in organizational research; Edgar Schein and Warren Bennis earlier described its role in learning and change.
- Evidence
- Well established
- Read
- 6 min
- Links
- 13 connections
01You've seen this when…
- at work
A new analyst finds a broken formula in the budget forecast. Her manager thanks her, corrects the spreadsheet, and asks where else the same mistake might appear.
- in life
Your weekend hiking group agrees to take the steep route. You are unsure you can manage it, but the last person who asked to slow down became the running joke.
- out in the world
During a city emergency drill, a dispatcher notices that the backup radio instructions are wrong. The debrief moves on before she decides it is safe to interrupt.
02The idea
Before someone asks a basic question or challenges a decision, they make a quick social calculation. Will this make them look incompetent? Will the manager become defensive? Will admitting a mistake damage their prospects?
Psychological safety describes the team’s shared answer. People expect that reasonable questions, concerns, and admissions will receive a fair hearing. They can expose gaps in their knowledge while keeping their standing in the group.
That expectation develops through repeated interactions. A leader’s response to an inconvenient warning carries more weight than a statement about openness. Colleagues matter too: eye-rolling, interruptions, and jokes about someone’s mistake teach people what happens here.
The central issue is the expected social cost of speaking up. A psychologically safe team can argue intensely, reject weak proposals, and address misconduct. Its members can tolerate the discomfort of disagreement because disagreement does not automatically threaten their membership or reputation. When belonging depends on agreement, pressure to conform can keep useful information out of the discussion.
03Why it matters
Many teams depend on information that only one person has: an operator hears an unfamiliar vibration, a junior engineer sees an impossible deadline, a nurse notices a confusing instruction. That information becomes useful when the person can bring it into the conversation.
Psychological safety supports several parts of learning:
- Warnings arrive earlier. People can raise a concern while there is still time to investigate or change course. This matters especially in high-reliability organizations, where small signals can precede serious failures.
- Errors become discussable. A team can examine how a mistake happened and whether the same conditions exist elsewhere. Concealment leaves those conditions untouched.
- Assumptions face scrutiny. Challenging the rule behind a decision makes double-loop learning possible. The team can reconsider its approach as well as its execution.
- Help becomes easier to request. People can expose confusion before it turns into rework or a failed handoff.
These benefits require follow-through. A team that welcomes concerns and repeatedly ignores them teaches a different lesson: speaking is permitted, but it changes little. Safety creates an opening for learning; attention, expertise, and action determine what follows.
04A worked example
Amy Edmondson’s research on hospital nursing units, published in 1996, encountered a puzzling pattern. Units judged to work well together appeared in the data with higher detected medication-error rates. The apparent result challenged the expectation that better teamwork would accompany fewer errors.
What it looks like The stronger teams are doing worse on a critical safety measure. A manager reading the figures might conclude that those units need closer supervision.
What’s actually going on Follow-up investigation suggested that the units differed in how readily mistakes became visible. In more open environments, nurses could acknowledge and discuss errors. In others, concerns about blame made disclosure harder. The recorded rate could therefore reflect both the occurrence of errors and the likelihood that someone detected or reported them. The study could not reveal every hidden error, so it did not establish that the more open units made fewer mistakes.
What would have helped A manager interpreting such a dashboard should examine how reports are produced, how supervisors respond, and whether reported problems lead to corrections. Independent record checks, where feasible, can provide another source of evidence. An increase in reports deserves investigation before it becomes grounds for punishment.
The case illustrates a measurement problem with practical consequences: a team that exposes more problems can initially look worse than one that leaves problems concealed.
05Where people trip up
- The first response undermines the invitation. A manager asks for objections, then cross-examines the first person who offers one. Others learn from that exchange. Start by acknowledging the concern, clarifying what the person observed, and agreeing how to assess it. Assess the claim on its merits afterward.
- Comfort becomes the success measure. Learning can involve embarrassment, disagreement, and unwelcome evidence. A pleasant meeting can still suffer from groupthink. Make room for competing explanations and ask people to identify what would change their conclusions.
- Standards fade into the background. Safety works alongside clear expectations about quality, effort, and conduct. Explain what good work requires, distinguish a good-faith error from reckless behavior, and make correction part of the response. People should know both what they can raise and what they remain responsible for.
- Silence gets treated as agreement. Several people may privately doubt a plan while each assumes the others support it. That is pluralistic ignorance. Ask people to write concerns before discussion, invite less senior members to contribute early, and offer a private route for sensitive issues.
- A survey becomes the entire diagnosis. Anonymous questionnaires can reveal patterns, but feared consequences can also affect the answers. Examine concrete episodes: who asked for help, who challenged a decision, how others responded, and what happened afterward. Protect confidentiality when gathering examples.
A useful starting point is one recurring interaction. Improve the response to questions during handoffs, or the way the team reviews mistakes. Repeated, observable changes give people evidence that speaking up is becoming safer.
06Where it doesn’t reach everyone equally
A team average can hide a sharp divide. Senior engineers may speak freely while contractors, newcomers, or people with less status calculate every sentence. Check whose questions receive a patient answer. Separate survey results by role only where confidentiality allows.
Safety is also local. Someone may freely question technical choices yet fear raising a staffing concern, or feel secure with one supervisor and guarded with another. This is a matter of what the measure actually captures: a favorable score does not establish that every member can raise every concern.
07Roots
Edgar Schein and Warren Bennis were examining a difficult part of change in the 1960s: getting adults to reconsider familiar ways of behaving. In the laboratory training groups they wrote about, participants learned by examining their own interactions and receiving feedback from others. That experience could threaten a person’s sense of competence. Their 1965 book described psychological safety as a condition that helped people tolerate the anxiety of learning and change.
Three decades later, Amy Edmondson approached the issue through teams and their everyday work. The nursing-unit puzzle made the consequences of disclosure especially visible: what people felt able to admit affected what an organization could learn about itself.
In 1999, Edmondson defined psychological safety as a shared team belief and studied it in 51 teams at a manufacturing company. Her research connected the climate for interpersonal risk-taking with learning behavior. The idea subsequently spread through management, healthcare, and discussions of the learning organization. It gave managers a way to examine how ordinary interactions influence whether questions and warnings reach the people who need them.
08How solid is this?
Field studies and a substantial meta-analysis support links between psychological safety, learning behavior, and performance. Much of the evidence is observational and survey-based, so the size and direction of causal effects remain less certain than the associations.
09Connections
- Helps counter Authority Bias, Conformity, Evaluation Apprehension, Groupthink, Normalization of Deviance, Normative Social Influence, Pluralistic Ignorance
- Can lead toDouble-Loop Learning
- Part ofHigh-Reliability Organization, Learning Organization
- See also Intellectual Humility, Construct Validity, Red Teaming
+ 3 more in the list
10Origin and sources
Edgar Schein and Warren Bennis described psychological safety in relation to learning and organizational change in 1965. Amy Edmondson developed and tested the shared team-level construct in 1999.
- [1]Schein, E. H., & Bennis, W. G. (1965). Personal and Organizational Change through Group Methods: The Laboratory Approach. Wiley.
- [2]Edmondson, A. C. (1996). Learning from Mistakes is Easier Said Than Done: Group and Organizational Influences on the Detection and Correction of Human Error. The Journal of Applied Behavioral Science, 32(1), 5–28.
- [3]Edmondson, A. (1999). Psychological Safety and Learning Behavior in Work Teams. Administrative Science Quarterly, 44(2), 350–383.
- [4]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.
Suggest an edit· Updated 2026-10-02