Psychological safety survey questions

The worst psychological safety survey question is the obvious one: Do you feel psychologically safe at work? People read the phrase differently, some have never heard it, and those who have know which answer is expected. You learn almost nothing.

Good surveys ask about specific situations instead: making a mistake, asking for help, disagreeing with someone senior. Situations are easier to answer honestly, and they point at something you can change. This guide gives you the standard research questions, a practical set grouped by behaviour, and the part most templates leave out: how to run the survey without making things worse, and how to read what comes back.

Edmondson's seven questions

Psychological safety, as teams use the term, comes from Harvard professor Amy Edmondson. Her 1999 paper in Administrative Science Quarterly, "Psychological Safety and Learning Behavior in Work Teams", defined it as a shared belief that the team is safe for interpersonal risk-taking, and introduced the seven-item scale that is still the most widely used measure. Google later found the same thing in its own teams: its re:Work research, known as Project Aristotle, named psychological safety the most important of five dynamics that set its effective teams apart.

The seven statements, as they are usually reproduced:

How to score it

Each person rates every statement on a seven-point scale, usually from 1 (strongly disagree) to 7 (strongly agree). Reverse the three negative items by subtracting the answer from 8, so a 2 becomes a 6. Average the seven numbers for each person, then average the people to get the team score. Higher is safer.

A worked example. One team member answers 2, 6, 3, 5, 2, 6, 5. Reversing items 1, 3 and 5 turns those into 6, 5 and 6. The seven scores are now 6, 6, 5, 5, 6, 6, 5, which add up to 39. Divided by seven, that person's score is 5.6.

Why three items are reversed, and why you should leave them alone

Reversed items catch people who agree with everything without reading. Many online templates quietly rewrite them so every statement is positive, which feels friendlier but changes the instrument: the numbers are no longer comparable with anyone else's, and the agree-with-everything habit goes undetected. If you use Edmondson's scale, use it as written.

Twelve more questions, grouped by what you can change

Edmondson's scale gives you one reliable number. It does not tell you where to start. For that, group questions by behaviour, so a low score lands on something a team can actually practise. These twelve are written for a working team, three per behaviour, all phrased so that agreeing is the healthy answer:

If you would rather not build and score this yourself, our free psychological safety check asks twelve questions across these same four behaviours, scores them on your device and gives you a written report: each behaviour explained, the one to fix first, and three things to try this week. Have everyone on the team take it separately, including you.

How to run the survey so people answer honestly

A psychological safety survey is itself a test of psychological safety. If people do not trust where the answers go, they will answer politely, and a polite survey is worse than none because it reassures the wrong person. Seven rules:

1. Say why, and what happens next, before you send it. We will look at the results together on the 14th and pick one thing to change. Then do exactly that.
2. Make anonymity real. Use a tool where the manager cannot see individual answers. A common rule is not to report any group smaller than five, so small teams should be combined or skip the breakdown.
3. Drop identifying questions. In a team of eight, asking for tenure and role identifies people. Leave demographics out unless the group is large.
4. Have the manager answer separately. The gap between the manager's view and the team's is the most useful number you will get.
5. Keep it short. Seven to fifteen statements, under five minutes. Longer surveys get worse answers.
6. Book the results conversation before you send the survey. It signals the answers will be used, and it stops the results sitting in a folder.
7. Repeat it, but not too often. Quarterly, or after a major change such as a reorganisation or a new manager. Safety moves slowly, so monthly surveys mostly measure fatigue.

A low response rate, or a suspiciously perfect score, is a result in itself. It usually means the survey was not trusted.

How to read the results

The average is the least interesting number

There is no universal benchmark for psychological safety scores. Survey vendors publish bands, but those come from their own customers and their own wording. Compare a team with itself over time, and compare the behaviours within one team. A move from 4.6 to 5.2 in two quarters means more than any league table.

Look at the spread

A team average of 4.5 can come from eight people who all answered 4 or 5, or from four people at 6 and four at 3. The second team is really two teams: one that feels safe and one that does not, and the average hides both. Look at who is in the low group before you look at anything else.

Look at the lowest item, not the lowest person

The aim is to find the situation that feels unsafe, not the individual who said so. Trying to work out who gave the low answer is the fastest way to guarantee the next survey is polite.

More reported problems can be good news

Edmondson's earliest work on this came from hospital units, where she expected better-led teams to make fewer medication errors and found they reported more. The likely reason was not that they made more mistakes but that people felt able to admit them. The same applies to your team: when safety improves, the number of problems people raise often goes up first.

What to do in the first meeting afterwards

The meeting after the survey matters more than the survey. Done badly, it teaches people that honesty is risky. Done well, it is the first proof that it is not.

1. Share all the results, including the ones that reflect badly on you. Selective sharing is noticed.
2. Thank people for answering honestly, and mean it.
3. Name the lowest-scoring item and say what you think it might mean, so you go first.
4. Ask: what would have to be true for this to score higher next quarter? Not who feels this way?
5. Agree one change, with an owner and a date. One, not five.
6. Say when you will measure again, and do it.

If the conversation stalls, the structure of a good retrospective works well here, and a structured debrief helps people talk about a shared experience rather than about each other. For the longer work of changing the score, see how to build psychological safety in a team.

Questions to leave out

• Do you feel psychologically safe? Too abstract, and everyone knows the right answer.
• Is your manager supportive? Too broad to act on. Ask about a specific behaviour instead.
• Who on the team makes you feel unsafe? A survey is not the place for this, and asking it destroys the anonymity of everything else.
• Double-barrelled statements such as I can raise problems and they get fixed. Someone who can raise problems that never get fixed has no honest answer.

Psychological safety is not the same as everyone being nice, and a good survey keeps that distinction. The challenge items exist because, as Kim Scott's radical candor and the glossary entry on psychological safety both argue, a team where nobody disagrees is not safe. It is quiet.

Tour De Force runs this as live, experiential training for teams worldwide, online and in person. Talk to us, or play The Weekly Challenge from our gamified learning apps to see the method in ten minutes.