Psychological safety

In plain terms

Psychological safety is a shared belief, held by the members of a team, that the team is safe for interpersonal risk taking (Edmondson, 1999). The risks are ordinary ones: asking a question that reveals what you do not know, admitting a mistake, reporting a problem or disagreeing with the boss.

The idea matters for change because change runs on information that someone has to be willing to say out loud. A plan that is not working, a step that will not fit the work, a cost nobody priced in: each is more likely to reach the people who can act on it when someone judges it safe to raise.

Where it came from

The idea is older than its current popularity. Edmondson and Lei (2014) write that it was first explored by pioneering organizational scholars in the 1960s, and that it had a renaissance starting in the 1990s that continues today.

The 1960s: safety as a condition for change. Edmondson (1999) traces the idea to early research on organizational change, in which Schein and Bennis (1965) discussed the need to create psychological safety for individuals if they are to feel secure and capable of changing. Their book, Personal and Organizational Change Through Group Methods: The Laboratory Approach, is about laboratory training, the T-group method that grew out of Kurt Lewin's work (see what is organizational development). Schein later made psychological safety the third part of what he called unfreezing. In his account, people who receive information that contradicts what they expect can feel learning anxiety, the fear that admitting a problem will cost them effectiveness, self-esteem or identity. Safety lowers that anxiety enough for them to take the information in; without it, the information is denied (Schein, 1996). See Edgar Schein.

1990: safety as a condition for engagement. William Kahn's study of personal engagement at work named three psychological conditions for bringing yourself to a work role: meaningfulness, safety and availability. A summary of the paper describes safety in his sense as being able to show yourself without fear of negative consequences to self-image, status or career (Kahn, 1990, summary read).

1999: safety as a team belief. Amy Edmondson's paper in Administrative Science Quarterly introduced team psychological safety and tested it in a field study of 51 work teams in a manufacturing company (Edmondson, 1999; a field study not graded in the Evidence Library). Team psychological safety was associated with learning behavior, and team efficacy was not once safety was controlled for. Learning behavior, in turn, carried the link between safety and team performance.

How it works

Three features of the 1999 definition do most of the work.

• It is shared. Psychological safety describes the team, not one person's confidence (Edmondson, 1999).
• It is about interpersonal risk. The question is what happens to you when you say something others may not want to hear. On this definition a team can be friendly and still unsafe, if disagreeing with the leader is costly.
• It works through learning. In Edmondson's study, the link between safety and team performance ran through learning behavior, such as asking for help, seeking feedback and discussing errors (Edmondson, 1999). That is the link to change: a team that does not talk about what is going wrong cannot adjust.

The four stages of psychological safety

A practitioner model breaks psychological safety into stages. In The 4 Stages of Psychological Safety (2020), Timothy R. Clark names four: inclusion safety, learner safety, contributor safety and challenger safety (Clark, 2020). His firm, LeaderFactor, describes them as sequential stages that teams move through as members feel safe to belong, to learn, to contribute and to challenge the status quo, and it sells training and an assessment built on the framework (LeaderFactor, n.d.).

The stages are a way of naming what people are willing to risk, from simply being accepted to questioning how things are done. Read next to Edmondson's definition, they name similar ordinary risks, such as asking a question or disagreeing, but that is a reading of the framework, not a tested link. The Evidence Library holds no test of the four-stage sequence, so treat it as a useful vocabulary rather than a finding.

What the research says

Psychological safety has a large research literature. In the Unfrozen Research Index (24,463 abstracts of research articles and reviews supplied by OpenAlex, covering 28 of the 29 journals (21 core, 7 expansion), 2007 to 2026, 2026 to date), 100 titles or abstracts mention psychological safety or being psychologically safe. Leadership & Organization Development Journal accounts for 20 and the Journal of Applied Psychology for 17. The yearly count rose from 1 in 2007 to 11 in 2025 and 13 so far in 2026. These are keyword counts of titles and abstracts, not full texts.

Safety travels with voice, engagement and learning. Frazier and colleagues (2017) pooled 136 independent samples, more than 22,000 individuals and nearly 5,000 groups, to map what goes with psychological safety and what it goes with (abstract read; not graded in the Evidence Library, and no result from it is used here). A meta-analysis of 47 Korean journal papers on corporate employees found the largest links with innovation behavior, job engagement and voice (Moon & Lee, 2022, rated Medium in the Evidence Library from its abstract). At the group level, it found positive links with team effectiveness and team performance. Among the factors that went with higher safety were self-efficacy and empowering, participative and inclusive leadership; a climate of silence went with lower safety. These are pooled correlations. They show that safety travels with speaking up and engagement, not that raising one causes the other.

It matters more for some work than others. A meta-analytic review, rated Medium from its abstract, found that psychological safety was more strongly associated with team learning and performance in knowledge-intensive settings involving complexity, creativity and sensemaking (Sanner & Bunderson, 2015). The authors suggest safety may be insufficient to stimulate learning where the task does not require it.

In one experiment, speaking up was valued only by clinicians who felt safe. In a vignette experiment with about 100 emergency clinicians in one US hospital, a nurse who spoke up was rated as contributing more to the team than one who stayed silent (Weiss et al., 2023, rated Medium after a full-text read). That held only among clinicians who themselves felt psychologically safe in their department. Among clinicians who felt less safe, speaking up and staying silent were rated alike. Safety was measured in the raters, not assigned, so this part of the result is correlational.

Tested ways of raising it are few. The controlled tests in the Evidence Library are few, and their effects are small.

• In a randomized trial of 56 hybrid work teams (478 employees), two coach-led team sessions raised self-rated psychological safety slightly, and the gain held at 9 weeks (Althammer et al., 2025, rated Medium). In raw terms the change was tiny, about 0.06 on a 5-point scale, and much of the gap came from the comparison teams drifting down.
• In a small randomized comparison of six workplace teams (51 people), a LEGO Serious Play team coaching day raised team cohesion more than conventional facilitation did, but not psychological safety (Passmore et al., 2025, rated Medium). The abstract claims a safety gain; the full text reports that the difference was not significant.
• A related test, not of raising safety directly: being listened to by a supervisor was associated with more creativity, partly through psychological safety, across five studies by the same authors; in their vignette experiment the effect was small and based on imagined supervisors (Castro et al., 2018, rated Medium).

Where the evidence is thin

Four gaps stand out in the studies reviewed here.

1. Cause. Most of the evidence is correlational. Teams that feel safe also speak up and learn more, but the direction can run both ways, and a third factor such as a good manager can drive both.
2. Change outcomes. None of the studies above tests whether raising psychological safety makes an organizational change more likely to succeed. The Loop treats it as the ground every part sits on, and the Loop page says plainly that no field trial in the studies graded so far tests that.
3. Who does the work. Humble leadership is one leader style studied in relation to safety. In a meta-analysis of 53 correlational studies, humble leadership was associated with creativity, engagement and voice, but its link with psychological safety was not significant (Luo et al., 2022, rated Medium after a full-text read). That does not mean leaders do not matter; Moon and Lee (2022) found participative and inclusive leadership among the stronger correlates. It does mean that no single leader style is a proven lever.
4. Stage models. The four-stage framework above has no graded test in the Evidence Library, so whether teams move through the stages in order is not shown here.

How it maps to the Unfrozen Loop

1. Week 0
   Before you start
2. 1
   Surface both ways

   Bad news and ideas travel up more readily when people judge it safe to send them. In one vignette experiment, emergency clinicians rated a nurse who spoke up as contributing more to the team only if they themselves felt safe in their department.

3. 2
   Make sense together, top included
4. 3
   Start it close to the work
5. 4
   Review on a rhythm, then hold the pace
6. 5
   The middle’s own weather
7. Ring
   Safe to speak, and speaking changes something

   The Loop's ground is a culture where it is safe to speak and speaking changes something. Psychological safety is the first half of that line.

Highlighted parts are the ones this idea speaks to. The mapping is my reading, not a tested finding. See the whole Loop and the evidence for each part.

Common misreadings

"Psychological safety means everyone is comfortable." The definition is about taking interpersonal risks, which are uncomfortable by nature (Edmondson, 1999). A team where nobody disagrees may be comfortable and unsafe at once.

"It is an individual trait." It is a belief about the team. Edmondson defined it as shared by team members, and in one vignette study (Weiss et al., 2023) the same act of speaking up was rated differently depending on how safe the rater felt.

"A workshop will fix it." The two randomized tests above found a tiny effect and no added effect. Treat a session as a start, and judge it by whether people raise problems in the weeks after.

"It is enough on its own." Sanner and Bunderson (2015), read here from the abstract, found it more strongly associated with learning and performance in knowledge-intensive settings. Safety is a condition for people to bring problems forward, not a substitute for asking for them.

Using it well

Two things a manager can try this week.

1. Ask for the risky information directly, and on a schedule. At the end of your next team meeting, ask: "What is one thing about this change that is not working yet?" Write down every answer without arguing with it. The Loop's first move for Surface both ways is the same idea: add one question to every team meeting for a month and send the answers up unedited.
2. Watch what happens to the first person who answers. Thank them, say what you will do with the information, and report back at the next meeting. In Schein's account, safety is what lets people take in unwelcome information instead of denying it (Schein, 1996). The same applies to you as the listener.