Where the idea came from
The term was defined in its current form by Amy Edmondson, a professor at Harvard Business School, in a 1999 paper in Administrative Science Quarterly. She described it as a shared belief that the team is safe for interpersonal risk-taking.
The finding that made the work well known came out of a study she ran on hospital teams. She expected better teams to make fewer medication errors. The data said the opposite: the better-regarded teams reported more errors. On closer examination they were not making more mistakes, they were willing to report them. The weaker teams were making the same errors and hiding them.
That distinction is the whole concept. Psychological safety does not change how often people get things wrong. It changes whether anyone finds out in time to act.
What it is not
The term gets stretched until it means "a pleasant workplace," which drains it of use and makes it easy for skeptical executives to dismiss. Four clarifications do most of the work.
- It is not lowered standards. Edmondson's own framing places psychological safety on one axis and performance standards on the other. High safety with low standards produces a comfortable team that achieves little. The target is high on both.
- It is not the absence of conflict. Psychologically safe teams argue more, not less. What changes is that disagreement attaches to ideas rather than to people's standing.
- It is not everyone being nice. Excessive niceness is frequently a symptom of low safety, because people are managing each other's reactions instead of saying what they think.
- It is not job security. A team can feel free to challenge each other's thinking during a restructure. The two are unrelated.
What it costs when it is missing
Low psychological safety is expensive in ways that rarely appear as a line item, because the cost is always something that did not happen.
- Problems surface late. The engineer who suspected the timeline was wrong in March says so in August, when it is a crisis instead of an adjustment.
- The quietest useful person stops contributing. Usually someone junior, frequently the person closest to the actual work.
- Decisions get worse while looking more decisive. Unanimous agreement in a room of twelve is almost never real. It means the disagreement moved to private channels.
- Attrition concentrates among the people you least want to lose. Strong performers have options and use them; the people who stay in a low-safety team are disproportionately those who cannot easily leave.
Google's Project Aristotle, a multi-year internal study of what distinguished its most effective teams, ranked psychological safety as the strongest of the five factors it identified, ahead of dependability, structure, meaning, and impact. The composition of a team mattered less than how the team treated each other.
How leaders destroy it without noticing
Almost nobody sets out to build a fearful team. Safety erodes through small, repeated signals, most of them invisible to the person sending them.
- Answering first. When the most senior person states a position before anyone else speaks, the discussion becomes about that position. This is the most common single cause and the easiest to fix.
- Reacting visibly to bad news. Not shouting, just tightening. People calibrate to micro-expressions faster than to policy, and they only need to see it once.
- Rewarding certainty over accuracy. If confident wrong answers get more airtime than hedged right ones, the team learns to perform confidence.
- Asking for challenge and then defending. Inviting pushback and responding with justification teaches people that the invitation was decorative. This does more damage than never asking.
- Treating one person's mistake as a character issue in public. Everyone in the room updates their own risk calculation.
What actually builds it
Edmondson's practical framing has three parts: frame the work as a learning problem, acknowledge your own fallibility, and model curiosity by asking a lot of questions. In day-to-day terms that looks like a small number of specific habits.
- Speak last in any discussion where you hold the most authority. Single highest-leverage change available to most managers.
- Say what you are unsure about, by name. Not a general admission of imperfection but a specific one: "I do not know whether this pricing holds at enterprise scale." It gives explicit permission to disagree with something concrete.
- Respond to the first piece of bad news well and visibly. Early reports are a test, and the team is watching what happens to the person who spoke.
- Ask questions you do not know the answer to. Questions with a known answer are examinations, and everyone can tell the difference.
- Separate blame from cause. When something fails, ask what made the failure possible before asking who was involved. Most repeat failures are structural.
- Thank people for inconvenient information, specifically. Naming what someone raised and what it changed does more than a general encouragement to speak up.
Measuring it
Edmondson's original instrument uses seven statements that team members rate. Adapted, they still work as a diagnostic:
- If you make a mistake on this team, it is held against you.
- Members of this team are able to bring up problems and tough issues.
- People on this team sometimes reject others for being different.
- It is safe to take a risk on this team.
- It is difficult to ask other members of this team for help.
- No one on this team would deliberately act in a way that undermines my efforts.
- Working with members of this team, my unique skills and talents are valued and utilized.
Items 1, 3, and 5 are reverse-scored. Run it anonymously or the results measure something else entirely. Behavioral signals are worth as much as survey data: how often someone contradicts you in a group setting, how quickly bad news reaches you, and whether the same two people do most of the talking.
Frequently asked questions
Who created the concept of psychological safety?
Amy Edmondson of Harvard Business School defined it in its current form in a 1999 paper, building on earlier organizational work by Edgar Schein and Warren Bennis in the 1960s. Her hospital-team research established the link between safety and error reporting.
Why did Google rank psychological safety so highly?
Project Aristotle set out to find what made some Google teams more effective than others and expected the answer to involve team composition. It did not. Psychological safety emerged as the strongest of five factors, meaning how team members treated one another predicted performance better than who was on the team.
Does psychological safety mean avoiding difficult conversations?
The reverse. Psychologically safe teams have more direct disagreement, not less, because the cost of speaking is lower. Teams that avoid difficult conversations are usually displaying low safety, not high.
How long does it take to build psychological safety in a team?
Noticeable change within a quarter is realistic if the leader changes their own behavior consistently. It erodes considerably faster than it builds, and a single badly handled incident can undo months of work, particularly early on.
Can you have psychological safety and high performance standards together?
Yes, and that combination is the point. Edmondson maps safety and standards as independent axes: high safety with low standards gives you a comfortable team that underachieves, and high standards with low safety gives you anxiety and hidden problems. High on both is where teams learn fast.
How do you measure psychological safety on a team?
Edmondson's seven-item survey, run anonymously, is the standard instrument. Pair it with behavioral indicators: how long it takes bad news to reach you, whether anyone contradicts you in group settings, and how evenly airtime is distributed in meetings.