Org, people & execution
Team Psychological Safety
Edmondson's validated seven-item scale and three-stage leader playbook for building a team where people can flag mistakes, risks and half-formed ideas without fear of punishment.
Also known as Psychological safety, The Fearless Organization model, Edmondson's seven-item scale. First set out by Amy C. Edmondson in 1999; the primary source is cited in full below.
Where this is contested
The widely used 'Fearless Organization Scan' is guided by Edmondson's work but is a separate commercial product, not her original research instrument, and carries the self-report and league-table risks common to any proprietary engagement tool.
- Format
- Checklist / audit
- Level
- Team · Business unit
- Best for
- Assess risk · Plan execution
- Decision stage
- Diagnose · Review
- Difficulty
- Intermediate
- Time to apply
- The pulse survey itself takes about ten minutes to complete. A proper diagnose-to-intervention cycle runs four to eight weeks. Building the three leader behaviours into how a team actually operates is an ongoing habit measured in quarters, not a one-off fix.
Plate · The model
The components
Diagnostic baseline (the seven-item scale)
Edmondson's original seven-item survey, giving the team a numeric baseline for how safe people feel to fail, question and disagree. It's the entry point into the model, not a compliance box to tick, and it only earns its keep if the results get discussed openly.
Signals of strength
Team members answer the reverse-scored items ('held against you', 'difficult to ask for help') honestly rather than defensively · Scores vary meaningfully across sub-groups rather than sitting suspiciously uniform · A low or uneven score triggers a facilitated conversation, not just an updated dashboard
Setting the stage
The leader frames the work as uncertain and interdependent, states plainly why every voice is needed, and admits what they don't know before problems surface. This is the groundwork that makes speaking up feel expected rather than risky.
Signals of strength
Leader names the task's uncertainty and interdependence out loud, not just in a slide · Leader states clearly what's at stake if a problem goes unspoken · Leader admits knowledge gaps ('I don't know how this will land') before being challenged on them
Inviting participation
The leader actively pulls input out of the room rather than waiting for it to arrive, through genuinely open questions, visible listening, and standing structures that make raising a concern routine rather than exceptional.
Signals of strength
Leader asks specific, open questions ('what are we missing here?') rather than 'any questions?' on the way out the door · Leader visibly listens, paraphrasing back and following up, before offering their own view · A standing forum exists (retro, huddle, pre-mortem) where raising a concern is the explicit purpose of the meeting
Responding productively
How the leader reacts to what gets raised. This is the component that decides whether anyone speaks up a second time: thank the messenger, treat well-intentioned failure as information rather than blame, and still sanction genuine violations clearly so safety doesn't slide into a free pass.
Signals of strength
Leader thanks people for raising bad news, not just good ideas · Failure gets discussed as diagnostic information whilst genuine negligence or rule-breaking is still addressed directly, not waved through · People who raised something difficult in the past are seen doing it again, rather than going quiet after the first attempt
When it earns its keep
- A team keeps repeating the same class of mistake or near-miss despite individually competent people, and you suspect the problem is silence rather than skill
- A new team is forming or a new leader is taking over, and the working norms are still up for grabs
- After a visible failure, incident or missed deadline, to check whether the risk was actually spotted early and just never surfaced
- Innovation, process improvement or quality feedback has gone quiet, particularly from junior staff towards senior ones
- Onboarding or restructuring has mixed people from different sub-cultures and you need to know whether norms are shared or just assumed
And when it doesn't
- As a substitute for clear performance standards or accountability; psychological safety is not the same as low standards or letting poor work slide, and treating it that way is the single most common misapplication
- As a one-off engagement survey question bolted onto an annual review with no intention of changing leader behaviour off the back of it
- In an acute crisis that needs rapid, directive decisions rather than open debate; this builds a longer-term climate, it doesn't replace decisive command in the moment
- Where the leader commissioning the diagnostic isn't genuinely willing to hear and act on criticism of their own behaviour, since a survey with no follow-through does more damage than not asking at all
- As a like-for-like comparison tool across very different national or professional cultures without adapting how the questions are asked and interpreted
How to run it
Before starting, gather the inputs the analysis depends on:
- A team that has worked together long enough to have shared, if unspoken, expectations (typically a minimum of several weeks)
- Genuine willingness from the team's formal leader to have their own behaviour examined and to change specific things about it
- A channel to administer the seven-item scale, or a lighter pulse version of it, anonymously and confidentially
- Protected time to discuss results with the team directly rather than just distributing a dashboard
- A realistic time horizon; this is a quarters-long habit, not a workshop
- 1
Administer the baseline
Run Edmondson's seven-item scale, or a shortened pulse version of it, anonymously across the team. A generic org-wide 'trust' question in an engagement survey is not a substitute; the specific items about mistakes being held against you, risk-taking and asking for help are doing the work.
- 2
Read the pattern, not just the average
Break results down by sub-group where sample size allows: tenure, seniority, shift, function. A healthy-looking team average can be hiding one cohort, often the newest or most junior, sitting well below everyone else.
- 3
Put the results in the room
Share findings with the team directly, including the uncomfortable items, and ask what's actually driving the lowest scores before proposing fixes. Discussing the results openly is itself a psychological-safety-building act; hiding them undoes the exercise.
- 4
Work the three leader-behaviour stages on purpose
Build specific, observable habits across setting the stage, inviting participation and responding productively (see components below). Pick two or three concrete behaviours per stage rather than trying to overhaul everything at once.
- 5
Compare the leader's self-rating to the team's rating
Ask the leader to score the same seven items as they think the team would. The size of the gap between their self-rating and the team's actual rating is usually the single most useful number to come out of the exercise.
- 6
Re-run and watch behaviour, not just the score
Re-survey after a meaningful interval, a quarter rather than a fortnight, and track whether people who raised something difficult the first time round do so again. That repeat behaviour matters more than any movement in the headline number.
Reading the result
A numeric baseline (with sub-group breakdown and a leader-versus-team gap) for how safe the team feels taking interpersonal risks, paired with a concrete plan of leader behaviours across the three stages.
- A high average with wide variance across sub-groups is a red flag in itself, not a pass mark
- A large gap between the leader's self-rating and the team's actual rating is usually the most useful single data point in the whole exercise
- Trend and behaviour over a quarter (repeat reporting, disagreement surfacing in meetings) tell you more than any single score
- Treat the score as the start of a conversation with the team, never as a result to publish and move on from
A worked example
Line 4, Meridian Fasteners (near Coventry)
Meridian Fasteners runs three shifts on a mid-sized automotive components line. On Line 4, a forklift came within a foot of an operator threading a machine guard, no injury, no damage. It wasn't reported. Eleven days later a second, more serious near-miss on the same line, this time a dropped pallet, forced the first incident into the open when an operator mentioned it in the follow-up interview almost as an aside. The shift supervisor, Dave, is well liked and technically excellent, and was genuinely shocked nobody had told him. The plant's HR lead brought in a coach to find out why, and used Edmondson's scale across the 14-person shift before touching the safety process itself.
- Diagnostic baseline (the seven-item scale)
- Shift average came out moderate (3.4/5) but two items dragged it down hard: 'if you make a mistake on this team it is often held against you' and 'it is difficult to ask other members of this team for help', both scoring worst among agency and newer starters. Dave's self-rating on the same seven items averaged 4.6, nearly a full point clear of his shift.
- Setting the stage
- Safety talks on Line 4 were compliance briefings, incident stats read out, RIDDOR reminders, sign the sheet. Nobody had ever explicitly said that near-misses were wanted information rather than an admission of fault. Dave assumed this was obvious. It wasn't.
- Inviting participation
- The only formal route for raising a near-miss was a paper form that went into a folder in Dave's office and, as far as the shift could tell, nowhere else. There was no huddle, no five-minute end-of-shift check-in, nothing that made raising something small feel like part of the job rather than going over someone's head.
- Responding productively
- One operator, Marek, had reported a near-miss two years earlier on a different line. He described it as 'noted and filed', no discussion, no visible follow-up, and said he was moved to the less popular night shift within a month, unrelated according to the rota records but read by the whole team as cause and effect at the time. That single episode had quietly set the norm for years.
The read. The plant introduced a two-minute end-of-shift near-miss check-in and Dave started opening every one of them by naming something he'd nearly got wrong himself that week. Reported near-misses on Line 4 quadrupled within six months, which is the result you want, more information surfacing earlier. Reporting of minor equipment faults, a separate and smaller-stakes category, barely moved, a reminder that a scale score and a new forum fix the loudest problem fastest and leave the quieter ones for the next cycle of asking.
Pitfalls
- Treating the seven-item score as a one-off measurement exercise rather than a repeated, discussed conversation
- Confusing psychological safety with comfort or the absence of conflict; a genuinely safe team still argues hard about the work, it just doesn't punish the person for raising it
- Rolling a commercial 'safety score' out as a league table across teams or sites, which reliably produces gaming of the survey rather than honesty
- Talking about wanting people to 'speak up' without ever visibly rewarding someone who does something difficult, so the stated value and the lived experience diverge fast
- Running the diagnostic and then sitting on the results; a survey with no visible follow-through leaves a team less safe than if you'd never asked
- Assuming higher psychological safety automatically means higher performance; the 1999 research links it to learning behaviour, which in turn predicts performance, it is necessary but on its own it is not sufficient
What the critics say
A large meta-analysis found psychological safety robustly correlated with performance, learning and engagement across many studies, but most of the underlying evidence remains cross-sectional and self-reported, so the causal direction, does safety produce better outcomes or do already-strong teams simply report feeling safer, is not fully settled.
Frazier, M.L., Fainshmidt, S., Klinger, R.L., Pezeshkan, A. and Vracheva, V. (2017) 'Psychological Safety: A Meta-Analytic Review and Extension', Personnel Psychology, 70(1), pp. 113-165.
Badly implemented, psychological safety slides into conflict avoidance and tolerance of underperformance dressed up as kindness; Edmondson herself has had to repeatedly correct the reading that it means low standards or no accountability, and it is the most common misapplication practitioners raise.
Widely debated in practitioner and coaching literature; see Edmondson's own clarifications in interviews and The Fearless Organization (2018).
The construct and its scale were developed and validated in Western, individualist, English-language organisational settings. The specific behaviours coded as 'speaking up' don't map cleanly onto higher power-distance or more indirect communication cultures, where silence can signal respect rather than fear, which limits how safely the same seven items can be used unmodified across cultures.
General cross-cultural organisational behaviour literature; see also commentary on cultural variation in psychological safety research.
Proprietary derivatives such as the 'Fearless Organization Scan' are marketed as measuring the construct but compress it into a small number of composite domains, remain self-report instruments with the same bias risks as any survey, and get used as league tables between teams, close to the opposite of what the underlying research recommends.
Critique of commercial psychological safety index products, see independent commentary at psychsafety.com.
Sources and further reading
- Edmondson, A. (1999) 'Psychological Safety and Learning Behavior in Work Teams', Administrative Science Quarterly, 44(2), pp. 350-383. ↗
- Edmondson, A.C. (2018) The Fearless Organization: Creating Psychological Safety in the Workplace for Learning, Innovation, and Growth. Hoboken, NJ: Wiley.
- Frazier, M.L., Fainshmidt, S., Klinger, R.L., Pezeshkan, A. and Vracheva, V. (2017) 'Psychological Safety: A Meta-Analytic Review and Extension', Personnel Psychology, 70(1), pp. 113-165. ↗
- Duhigg, C. (2016) 'What Google Learned From Its Quest to Build the Perfect Team', The New York Times, 25 February.