Two managers describe the same meeting. One says her team does not trust each other. The other says his team has no psychological safety. Both are looking at a room where people stop short of saying what they think, and both are about to spend money on a fix. Only one of them has named the condition accurately, and neither knows which.
The two terms come out of separate research traditions, describe different units, and break in different ways. Treating them as synonyms produces expensive misfires: an offsite built around bonding exercises when the problem is a leader who cuts off dissent in front of everyone, or a norms workshop when the problem is one colleague who took credit for someone else's work last spring and never acknowledged it.
What Edmondson Defined
Amy Edmondson introduced team psychological safety in a 1999 paper in Administrative Science Quarterly. The abstract defines it as "a shared belief held by members of a team that the team is safe for interpersonal risk taking" (Edmondson, 1999, p. 350).
The fuller definition, on page 354, carries more of the load:
Team psychological safety is defined as a shared belief that the team is safe for interpersonal risk taking... The term is meant to suggest neither a careless sense of permissiveness, nor an unrelentingly positive affect but, rather, a sense of confidence that the team will not embarrass, reject, or punish someone for speaking up. This confidence stems from mutual respect and trust among team members.
Three things in that sentence do specific work. The belief is shared, which makes it a property of a group. The object of the belief is consequences, specifically what happens to you after you speak. And Edmondson rules out the warm reading up front: permissiveness and relentless positivity are explicitly excluded.
She also separates the construct from group cohesiveness, citing research showing that cohesion can suppress willingness to disagree and challenge others' views, as in groupthink (Janis, 1982), which implies a lack of interpersonal risk taking. A team that likes each other enormously can be one of the least safe places to raise an objection.
What Trust Means in the Research
The dominant definition of organizational trust predates Edmondson's paper by four years. Mayer, Davis, and Schoorman published it in Academy of Management Review in 1995, and Edmondson cites them directly when she needs the term. Their definition, from page 712:
The definition of trust proposed in this research is the willingness of a party to be vulnerable to the actions of another party based on the expectation that the other will perform a particular action important to the trustor, irrespective of the ability to monitor or control that other party.
A party. Another party. This is a statement about two people and an expectation running between them. The paper spends several pages fencing trust off from three neighbors it gets confused with:
- Cooperation. "Cooperation does not necessarily put a party at risk." Someone can cooperate happily with a colleague they distrust when a punitive manager or an absence of alternatives makes cooperation the only rational move.
- Confidence. Citing Luhmann (1988), the authors locate the difference in whether risk is recognized and accepted. "If you do not consider alternatives... you are in a situation of confidence. If you choose one action in preference to others in spite of the possibility of being disappointed by the action of others, you define the situation as one of trust."
- Predictability. Knowing what somebody will do is a separate thing from trusting them. A colleague who reliably misses every deadline is perfectly predictable.
Ability, benevolence, and integrity
Mayer, Davis, and Schoorman then break the perception of a trustee into three antecedents (pp. 717 to 719):
- Ability is "that group of skills, competencies, and characteristics that enable a party to have influence within some specific domain." They stress the domain specificity: a colleague can be highly competent in one technical area and have little aptitude, training, or experience in another.
- Benevolence is "the extent to which a trustee is believed to want to do good to the trustor, aside from an egocentric profit motive." Their example is a mentor helping a protege with no extrinsic reward for doing so.
- Integrity "involves the trustor's perception that the trustee adheres to a set of principles that the trustor finds acceptable," assessed through consistency of past actions, credible word from others, and congruence between what the trustee says and does.
Those three come apart in practice, which is where most diagnostic confusion starts. A team can rate a colleague at the ceiling on ability and near the floor on benevolence. That combination reads as total confidence in the person's work and none in their motives, and it will suppress speaking up in ways that look exactly like a safety problem from the outside.
How the Two Constructs Fit Together
Edmondson answers this herself, in the same paragraph where she cites Mayer, Davis, and Schoorman:
Team psychological safety involves but goes beyond interpersonal trust; it describes a team climate characterized by interpersonal trust and mutual respect in which people are comfortable being themselves.
"Involves but goes beyond" is the load-bearing phrase in the entire debate. Trust is an ingredient inside psychological safety, and it is insufficient on its own. Psychological safety takes that ingredient and adds a team-level layer about one specific behavior: whether you will pay a social price for speaking.
Different Units, Different Failure Modes
Trust lives in a pair. You can trust one teammate's code review and not another's, trust the CFO's numbers and not their stated priorities. The value changes every time you swap one of the two people.
Psychological safety lives in a group, and Edmondson tested that claim rather than asserting it. Across 51 work teams at a manufacturing company, team psychological safety produced a group-level intraclass correlation of .39 (F(50,427)=6.98, p<.001). Individual responses clustered by team far more than chance allows, which is the statistical signature of a shared belief rather than an average of private opinions.
That difference in unit determines the shape of both the problem and the repair. Trust is repaired through track record: repeated reliable behavior, observed over time, in a specific relationship. Psychological safety is repaired in public and in the moment, because it is read off how a leader handles the next mistake and the next objection while everyone is watching.
| Trust | Psychological safety | |
|---|---|---|
| Defined by | Mayer, Davis, and Schoorman, 1995 | Edmondson, 1999 |
| Unit | A pair of people | The team as a group |
| What the belief is about | Whether a specific person will do a particular thing that matters to you | What happens to you after you speak up in front of everyone |
| Built from | Ability, benevolence, integrity | How the group visibly treats interpersonal risk |
| What it varies with | Which two people are involved | Which team you are on (intraclass correlation of .39 across 51 teams) |
| How it fails | One relationship goes guarded while the rest stay open | The whole room goes quiet regardless of who is talking to whom |
| How it is repaired | Track record, built through repeated reliable behavior over time | A leader's public response to the next mistake or objection |
Signals That Point at Trust
- People double-check specific colleagues' work with no formal reason to, which reads as low confidence in ability.
- Information and resources get withheld from particular individuals rather than from the group, which reads as low confidence in benevolence.
- Somebody remembers a promise broken eight months ago and brings it up unprompted. Integrity assessments are built from consistency, and a single loud inconsistency can dominate the record.
- The pattern is asymmetric. The same person speaks freely with four teammates and guards everything around the fifth.
- It follows people between contexts. Move the two of them onto a different project and the caution moves with them.
Our seven signs a remote team has trust issues breaks these down at the level of specific observable behavior, and the leader's playbook for building trust on a remote team covers the repair sequence.
Signals That Point at Psychological Safety
- People say the hard thing in a one-on-one and go silent about the same topic in the group. The silence attaches to the setting.
- Errors get hidden or reframed even by people who are well liked individually, because the fear is about the room's reaction.
- Dissent dies without a confrontation. Nobody shoots the idea down; the idea simply never gets said, so there is no incident to point at.
- It is consistent across the whole team regardless of who is speaking to whom. That consistency is precisely what the .39 intraclass correlation captures.
- It shifts fast when leader behavior shifts, because one public response to one mistake updates the shared belief for everyone at once.
Edmondson's earlier hospital research produced two quotes from nurses at the same institution, working on different patient-care teams. The first:
Mistakes are serious, because of the toxicity of the drugs [we use], so you're never afraid to tell the Nurse Manager.
The second:
You get put on trial! People get blamed for mistakes... you don't want to have made one.
Same hospital, same profession, same drugs, same stakes. What changed was the team climate, and it showed up as a shared expectation about what happens after you report an error.
The Question That Separates Them
Ask yourself two versions of the same question about the thing nobody is saying.
"Would this person say it to me privately?" If yes, and they still will not say it in the meeting, you have a psychological safety problem. The content is available; the setting is suppressing it.
"Would this person say it to that specific colleague, in any setting?" If no, while they speak freely to everybody else, you have a trust problem in one relationship. Fixing the meeting format will do nothing for it.
High dyadic trust in every pairing plus low psychological safety is common and produces the classic failure where everyone privately agrees the plan is bad and nobody says so before it ships. The reverse also occurs: a newly formed cross-functional team, with no accumulated history between any two members, can speak up freely from the first session when the leader makes the consequences of speaking visibly safe. Psychological safety is read off group signals, so it does not require the time that dyadic trust requires to accumulate.
The two states also degrade each other. Low trust makes psychological safety fragile, because a stated norm of safety is not believed by people who doubt specific colleagues' motives. Low psychological safety makes trust inert, because trust that exists between individuals never gets converted into speaking up when the group setting suppresses it. Our piece on psychological safety in hybrid teams traces how distance accelerates both.
What the Evidence Says About the Stakes
Google's Project Aristotle studied 180 teams (115 engineering project teams and 65 sales pods, sized 3 to 50, median 9 members), reviewed more than 250 survey items, and tested over 35 statistical models. Psychological safety came out as the most important of five team-effectiveness dynamics, ahead of dependability, structure and clarity, meaning, and impact. Individuals on teams with a stronger team culture are less likely to leave Google, bring in more revenue, and are rated as effective twice as often by executives (Google re:Work).
Gallup reports that only 3 in 10 U.S. workers strongly agree their opinions count at work, and models that moving to 6 in 10 would produce a 27% reduction in turnover, a 40% reduction in safety incidents, and a 12% increase in productivity (Gallup). In a case study in the same piece, an HR team that deliberately built psychological safety moved 0.70 points on Gallup's five-point engagement index, against a benchmark where 0.20 is a meaningful improvement.
Which form of safety is hardest to get is also measurable. In LeaderFactor's PSindex study of 3,366 employees across more than 20 organizations, 65% named inclusion, feeling safe to be your authentic self, as the first need satisfied, while 86% named challenging the status quo as the last (LeaderFactor). Permission to dissent arrives long after permission to belong.
Both Constructs Are Measurable
Both have validated instruments behind them. The Psychological Safety Index, the Fearless Organization Scan's instrument, based on Edmondson's original research, reports a Cronbach's Alpha of .82 to .86 (Fearless Organization), which is strong internal consistency for a team-level instrument.
Edmondson's 1999 results show why the measurement is useful. Team psychological safety predicted team learning behavior on both self-reported (β=.76, p<.01) and observer-rated (β=.46, p<.01) measures, and that learning behavior mediated the relationship between psychological safety and observer-rated team performance. Across seven alternative models, psychological safety was the strongest and most consistent predictor, outperforming context support and team leader coaching, and it was the only variable that stayed significant when everything was entered together.
That chain is specific about causation. Psychological safety works on performance indirectly, by licensing the behaviors that move it: asking for help, discussing errors openly, seeking feedback, and experimenting. Trust operates a level below that, making the promise of safety credible enough for anyone to act on it.
Diagnosing It From Behavior Instead of Self-Report
Both constructs are usually measured with surveys, and surveys have a structural weakness on exactly this topic: a team too unsafe to raise an objection in a meeting is not a team that answers an anonymous form candidly. The instrument depends on the thing it is trying to measure.
Behavior under mild, consequence-free pressure sidesteps that. This is the premise QuestWorks is built on. Teams run a fixed weekly 25-minute session on QuestWorks' own cinematic, voice-controlled platform, with Slack or Microsoft Teams handling install, invites, and admin. The whole team plays at the same time, seated dynamically into groups of 3 to 6, and the quests put people in situations where somebody has to raise a concern, override a plan, or admit a mistake while the clock runs.
What comes back is a Team Intelligence Score, 0 to 100, delivered weekly with an eight-week trend line and named sub-scores including Role Clarity and Decision Velocity, plus a "Do this week" list of concrete actions. Surfacing who speaks into a group and who only speaks in pairs, week over week, separates the team-level pattern from the relationship-level one without asking anybody to rate their colleagues. Participation is voluntary and opt-in, and the data is never tied to performance reviews.
Pricing is $199 per month per team with 10 seats included, or $1,990 per year. There is a 30-day free trial covering four game days, no credit card required.
The Practical Upshot
When your team goes quiet, resist the urge to name the condition before you have looked at the pattern. Check whether the silence follows the room or follows a person. Silence that follows the room is a shared belief about consequences, and it changes when a leader responds differently to the next mistake in front of everyone. Silence that follows a person is an expectation about that individual's ability, benevolence, or integrity, and it changes only through a track record built over time.
Two conditions, two diagnoses, two repairs. Edmondson's construct sits on top of Mayer, Davis, and Schoorman's, and a team can be missing either layer while the other holds.
