“Edmondson psychological safety scale” is the name most people use for the seven-statement survey Amy Edmondson published in 1999: everyone on a team rates each statement from 1 to 7, three are reverse-scored, and the answers are averaged into one team score. Try it on your own team below.
Attuned runs inside 300+ companies, including





Edmondson wrote these for a 1999 study of 51 work teams in a manufacturing company. The scale is also cited as the team psychological safety scale (Edmondson 1999) or Amy Edmondson’s psychological safety survey. In her study, the scale had an internal reliability (Cronbach’s alpha) of .82.
Three steps, and a spreadsheet can do all of them. Start with the flip.
Items 1, 3 and 5 are worded so that agreeing means less safety. Flip them before you do anything else: score = 8 − answer. A 2 becomes a 6, a 7 becomes a 1, and a 4 stays a 4. Forget the flip and three of the seven answers pull the score the wrong way, which is, fittingly, the kind of mistake that gets held against you later.
Teal is safer, coral is less safe. The flip only applies to items 1, 3 and 5.
Add the seven scores (after the flip) and divide by seven. Each person ends up with a number between 1 and 7, where higher means they feel safer on this team. Some people report the total instead, from 7 to 49; the average keeps it on the same scale as the answers, which makes it easier to explain.
| Item | 1 R | 2 | 3 R | 4 | 5 R | 6 | 7 | Mean |
|---|---|---|---|---|---|---|---|---|
| Answered | 2 | 6 | 3 | 5 | 2 | 4 | 6 | |
| Counts as | 6 | 6 | 5 | 5 | 6 | 4 | 6 | 5.43 |
Illustrative answers: 38 ÷ 7 = 5.43.
The team score is the mean of its members’ scores. Edmondson defined psychological safety as a belief the team shares, so she measured it person by person and analysed it team by team. Read the team’s mean alongside its spread.
Write down the spread at the same time: the lowest and highest person, or the standard deviation if your tool gives it. You will want it in a minute.
A team mean is the average ripeness of a fruit bowl. Two bowls can both come out at “nearly ready”: one because every banana in it is a sensible yellow, the other because half are still rock-green and the rest are speckled brown and quietly recruiting fruit flies. Same average, and nobody would plan breakfast around them the same way. Read the spread before you read the mean.
Eight people having roughly the same experience. The conversation is with the whole team, about the item that scored lowest for most of them.
Five people feel fine and three do not. The mean hides the three. Look for what they share before you talk to anyone: a shift, a site, a project, or how far they sit from the manager.
Illustrative teams, 8 people each
Edmondson’s paper does not define a score below which a team is “unsafe”, and we are not aware of a validated threshold anyone else has published. A 4.8 means nothing on its own. Compare the team with itself over time, or with other teams in your organization that answered the same seven statements in the same way.
The scale was built to produce one score, and that is how it should be reported. When you sit down with the team, though, the item means are where the conversation starts: a team that scores well overall but low on item 5 has a specific problem with asking for help, and that is easier to talk about than “our safety score is 4.9”.
Seven questions take a couple of minutes. Getting honest answers to them takes a bit more planning.
Use a tool that does not show the manager who answered what, and say so before people start. Asking a team that feels unsafe to put their names on how unsafe it feels gets you a very safe-looking score.
Decide in advance that no result is shown below a floor, such as five responses. On a team of four, an anonymous answer is a guessing game with four possible winners.
One round gives you a starting point. Running it every quarter or every six months, with the same wording and the same tool, is what makes a change readable. Run it again after a reorganization or a new manager, because either can move the score on its own.
Tell the team when they will see the result and that you will discuss it with them. If nothing happened after the first round, say why before you run the second.
Edmondson’s scale was built for research. If you are running a study or comparing teams, use it as written. For a manager, it has one limit: it hands back one number, and that number does not say which behavior to change on Monday.
Our page on the psychological safety assessment compares the main ways to measure it, and our overview of psychological safety software sets Edmondson’s scale beside the tools built on it. Attuned’s own assessment scores each team on six behaviors, including Mistake Reporting and Expressing Disagreement, and ranks a next step for the manager.
The conversation itself still has to happen. Attuned’s AI coach for managers, TalkCoach (in early access), uses shared intrinsic motivator profiles and the context you give it to help you prepare: an opening, another perspective, and what to ask next. For the thinking behind the scale, read Amy Edmondson on psychological safety and our Fearless Organization summary.
Illustrative example
Source: Edmondson, A. (1999), “Psychological Safety and Learning Behavior in Work Teams,” Administrative Science Quarterly 44(2), 350–383. Item wording, the seven-point response scale, reverse-scored items, the 51-team sample and the reliability figure are from this paper.
Bring us one team. We’ll show you how Attuned scores a team on six behaviors and ranks a next step for its manager.
Book A Demo