# When Feeling Safe Isn't Enough: Contextualizing Models of Safety and Learning in Teams

*Sanner · Critique & Boundary, Team Learning · Organizational Psychology Review · 2015 · Paywalled*

A meta-analysis (51 studies, 48 papers, roughly 3,700 teams) arguing that psychological safety's effect on team learning and performance is not unconditional, as almost all prior research had implicitly modelled it (17 of the 20 papers reviewed treated safety as an unmoderated antecedent), but bounded by a 'motivation to learn' that psychological safety itself doesn't supply. The distinction is drawn from classic expectancy theory: psychological safety addresses expectancy (team members believe they CAN take interpersonal risks safely) but says nothing about instrumentality or valence (whether team members see any point in doing so). Because no study in the sample actually measured learning motivation directly, the paper operationalises it entirely through proxy: three task characteristics (creativity requirements, sensemaking requirements, complexity) coded from national occupational databases (O*NET, the Dictionary of Occupational Titles) rather than from the studies themselves, based on context descriptions in the original papers that the authors admit were often brief and vague, and in nearly half the cases, on information solicited directly from study authors after the fact. Bias-corrected correlations across the full sample were fairly strong (safety-learning rho = .58, safety-performance rho = .32, learning-performance rho = .39), but varied considerably across studies, and this variance tracked the three task characteristics closely: complexity alone explained roughly a third of the variance in the safety-learning correlation and nearly half in the safety-performance correlation. The paper's most defensible and least contestable finding is a genuine sampling-bias corrective, separate from the moderation story: the large majority of published psychological safety and learning studies (78-81% depending on the dimension) have sampled from above-median-complexity, above-median-creativity task settings, meaning the field's headline effect sizes are probably inflated relative to the general population of work. The moderation mechanism built on top of that finding is less airtight than the framing suggests, though. The 'motivation to learn' construct doing all the theoretical work is never actually validated against anything; it's an inference chain running from occupational task codes to a psychological state the paper never measures, a limitation the authors themselves acknowledge but don't fully reckon with. An equally plausible alternative reading is that complex, creative work simply has more room for learning behaviour to vary in the first place, so almost any antecedent, not just psychological safety, would correlate more strongly with learning there, for reasons that have nothing to do with motivation specifically. There is also a confound the paper doesn't rule out: high-complexity, high-creativity settings like hospitals and R&D labs are disproportionately where Edmondson's own measurement tradition and research collaborators have concentrated their work, so the observed 'moderation by task type' could partly reflect which research programmes study which populations, rather than a task-level psychological mechanism operating independently of who happened to study it. The regressions testing the core moderation hypotheses also rest on a small number of studies (22 to 27) as the unit of analysis for reported effect sizes that are fairly large by the standards of organisational-behaviour meta-regression, which is worth some caution given how much leverage a handful of occupational codings could have on a regression that size. None of this undermines the paper's more modest central point, that psychological safety alone is an incomplete account of team learning and that the field has been sampling from a systematically unrepresentative slice of workplaces, but the specific mechanism proposed to explain the moderation is more speculative than its reputation as 'the' contextualising paper on psychological safety implies.

- **This page:** https://explore.psychsafety.com/n/sanner-bunderson-2015/
- **View the source paper:** https://doi.org/10.1177/2041386614565145
- **Interactive map:** https://explore.psychsafety.com/?mode=papers&node=sanner-bunderson-2015

## Connected concepts (9)

- [Psychological Safety and Learning Behavior in Work Teams](https://explore.psychsafety.com/n/edmondson-1999.md) (paper)
- [Managing the Risk of Learning: Psychological Safety in Work Teams](https://explore.psychsafety.com/n/edmondson-2002.md) (paper)
- [Implicit Voice Theories: Taken-for-Granted Rules of Self-Censorship at Work](https://explore.psychsafety.com/n/detert-edmondson-2011.md) (paper)
- [Learning Teams](https://explore.psychsafety.com/n/learning-teams.md)
- [Making it Safe: The Effects of Leader Inclusiveness and Professional Status on Psychological Safety and Improvement Efforts in Health Care Teams](https://explore.psychsafety.com/n/nembhard-edmondson-2006.md) (paper)
- [PS Isn't Enough](https://explore.psychsafety.com/n/psychological-safety-isnt-enough.md)
- [PS Index Critique](https://explore.psychsafety.com/n/the-psychological-safety-index-a-critical-look.md)
- [Slacking Off in Comfort: A Dual-Pathway Model for Psychological Safety Climate](https://explore.psychsafety.com/n/deng-leung-lam-huang-2019.md) (paper)
- [Utility of PS](https://explore.psychsafety.com/n/the-utility-of-psychological-safety.md)