Psychological Safety and Learning Behavior in Work Teams

Edmondson · Origins, Voice & Silence, Team Learning, Measurement & Method · Administrative Science Quarterly · 1999 · Open access

Open this neighbourhood on the live map →

The paper that turned psychological safety from a concept in the organisational change literature into a team-level construct with a measure, and the source of the definition almost every later study cites: a shared belief held by members of a team that the team is safe for interpersonal risk taking. The belief is tacit, taken for granted rather than discussed, and Edmondson is precise about what it is not. It is not group cohesiveness, which can suppress disagreement to the point of groupthink. It is neither permissiveness nor relentless positive affect, but a confidence that the team will not embarrass, reject or punish someone for speaking up. And it involves but exceeds interpersonal trust: trust describes a relationship between people, psychological safety the climate of a team. She traces the construct to Schein and Bennis (1965) and to her own 1996 hospital study, where nurses in one team found it natural to report a drug error and nurses in another, in the same hospital, described being put on trial for one.

The empirical work is a three-phase multimethod study at a pseudonymous office furniture manufacturer of about 5,000 people. Eight team meetings and seventeen interviews came first, and several of the seven psychological safety items descend from quotations gathered there. Then surveys to 53 teams of four types (34 functional, 9 self-managed, 5 product development, 3 project), returned by 427 members from 51 teams, with two or three managers outside each team rating its learning and performance. Then follow-up observation and interviews with seven of the twelve highest and lowest learning teams. The seven-item scale had an alpha of .82 and an intraclass correlation of .39, the highest of any group-level variable in the study, which is what licenses treating it as a property of the team.

Six of eight hypotheses held. Team psychological safety accounted for 63 per cent of variance in self-reported learning behaviour and 35 per cent in observer-rated learning, and when context support, leader coaching and team efficacy were entered alongside it, only psychological safety stayed significant. Learning behaviour fully mediated its effect on observer-rated performance. Psychological safety in turn mediated the effects of context support and leader coaching on learning, although coaching only marginally predicted safety on its own. That is the paper's structural claim: psychological safety is the mechanism through which team design reaches behaviour, not a rival to design. Both failed hypotheses concerned team efficacy, the shared belief that the team can do its work, which added little once psychological safety was controlled and mediated nothing. Edmondson reads the failure as sharpening the argument: what frees people to ask, admit and experiment is confidence about how others will respond, not confidence in the team's ability.

The qualitative cases carry the texture. A self-managed staining team, repeatedly stripped of members for other jobs, thanked those who pointed out drips on a finished part where it had once hunted for a fault to hold against them in return, and sought second opinions by showing parts to outsiders without saying what the concern was. A publications team with the same design and resources as two others scored 3.9 on psychological safety against their 5.0 and 5.9, and a member said its leader did not want to know when things went badly. From seven cases Edmondson draws an asymmetry: high-learning teams overcame poor design and thin support, while low-learning teams sat in a self-sealing pattern of private concern and withheld action that they could not leave without intervention.

The limitations are the author's own and should travel with the findings. The data supported the construct but could not conclusively separate it from trust; the design is cross-sectional and single-company, so nothing here is causal; 51 teams is small for multivariate work; the antecedent findings rest on one survey and are called tentative; leader coaching is three items about availability and says nothing about what leaders actually do; and the model omits the dynamic by which unpunished risk reinforces the safety that permitted it, which she names as the thing to study next. For a field that has built a great deal on this paper, the most useful thing in it may be how carefully its author declined to overclaim.

Edmondson, A.C. (1999) 'Psychological safety and learning behavior in work teams', Administrative Science Quarterly, 44(2), pp. 350–383.

Neighbours (113)

View the source paper →