What the research actually says, what the caveats are, and what it means for how entrepreneurs build teams

The finding that surprised Google

In 2012, Google launched Project Aristotle — an internal research programme designed to identify what made its teams effective. The researchers analysed more than 180 teams and 250 variables. They expected the answer to involve team composition: the right mix of skills, experience, and personality types. What they found instead was that who was on the team mattered far less than how the team worked together. And the single most important factor in how teams worked together was psychological safety — the shared belief that the team is safe for interpersonal risk-taking.

Amy Edmondson at Harvard Business School had introduced and defined this construct in 1999, based on research on hospital nursing teams. Her finding then was equally counterintuitive: better-performing teams reported more errors, not fewer. The explanation was that higher-performing teams had sufficient psychological safety to surface and discuss errors rather than conceal them, which produced both better learning and better outcomes. The construct she named has since been validated across industries in a 2017 meta-analytic review by Frazier and colleagues in Personnel Psychology.

One caveat the article owes: Project Aristotle was an observational study, never peer-reviewed or published in an academic journal. It identified associations between team dynamics and performance, but could not establish causal direction with certainty. The finding should be treated as compelling observational evidence from a large, well-resourced study rather than as definitive experimental proof.

What psychological safety actually means — and what it does not

Psychological safety is frequently misread as a prescription for niceness, conflict avoidance, or the lowering of standards. Edmondson’s research is explicit in the opposite direction. In psychologically safe teams, productive conflict is normal — colleagues challenge ideas directly, dissent is welcomed as part of finding the best solution, and the absence of disagreement is itself treated as a warning sign. If no one is willing to express dissent, that silence is more likely to reflect fear than agreement.

What psychological safety removes is a specific set of fears that Edmondson identifies as the mechanism by which silence accumulates across organisations. People routinely withhold contributions because they are afraid of being seen as ignorant, incompetent, negative, or disruptive. Each of these is a social identity threat — a risk that their status, competence, or belonging within the group will be damaged by speaking. The cost-benefit calculation a team member performs before contributing is weighted by threat circuitry that is neurologically more powerful than the benefit side, as established in the leadership neuroscience article in this series. In a low-safety environment, that calculation consistently produces silence. The questions that go unasked, the problems that go unsurfaced, and the ideas that go unproposed represent the invisible performance cost — invisible to the leader, but cumulative and compounding across the team.

Collective intelligence versus individual intelligence

Woolley and colleagues’ 2010 research in Science found that a group’s ability to perform well across a range of tasks was not strongly predicted by the average or maximum intelligence of its members. Instead, collective intelligence was predicted by three factors: equality of conversational turn-taking, the social sensitivity of group members, and the proportion of women in the group — the latter attributed by the researchers to women’s higher average social sensitivity in the measurement used.

Project Aristotle found the same pattern at organisational scale. The composition of the team — the credentials, experience, and individual performance ratings of its members — was far less predictive of team performance than the quality of its interpersonal dynamics. Individual intelligence is a fixed input that a leader can hire for but cannot substantially change. Psychological safety is a variable that the leader designs through their behaviour.

The learning zone: safety and standards together

The most important applied nuance for entrepreneurs is that psychological safety alone does not produce high performance. Research consistently finds that when psychological safety is high and performance standards are low, the result is a comfort zone rather than a high-performance zone. The learning zone — where both psychological safety and accountability are genuinely high — is where the best performance is observed.

The goal is not to make the team comfortable. It is to make the team safe enough to be productively uncomfortable — to challenge ideas, surface problems, and admit ignorance in service of the work rather than in avoidance of it. Psychological safety and high standards are complementary, not competing. One without the other produces either silence or complacency.

Three leadership behaviours Edmondson’s research identifies

Edmondson’s practical framework organises the building of psychological safety around three clusters. Setting the stage — framing the work as a learning problem rather than a pure execution problem, and openly acknowledging one’s own fallibility and the limits of what is known. Inviting participation — asking genuine questions, creating structures for input, and demonstrating situational humility rather than projecting certainty. Responding productively — expressing appreciation for candour, distinguishing honest mistakes from clear violations, and destigmatising the experience of being wrong.

None of these behaviours is extraordinary. Their consistent absence in most organisations is what makes their presence so significant. The counterintuitive finding — consistent with the advice-seeking research established earlier in this series — is that leaders who model uncertainty and invite correction are experienced as more trustworthy and more effective, not weaker.

The middle manager finding most entrepreneurs overlook

A 2025 Harvard Business Review analysis found that middle managers feel the least psychological safety of any group in most organisations. Caught between leadership demands from above and team needs from below, they inhabit the most constrained interpersonal risk environment in the hierarchy. For entrepreneurs building organisations, this finding complicates the assumption that creating psychological safety at the leadership level is sufficient. If the middle layer is psychologically constrained, the safety being modelled at the top is filtered before it reaches the teams where it would produce performance outcomes. Psychological safety requires deliberate attention at every layer of the hierarchy, not just the founding team.

A book worth reading alongside this

The Fearless Organisation by Amy Edmondson is the definitive applied treatment of the psychological safety research programme and the most important single text an entrepreneur can read before hiring their first team. Edmondson moves from the foundational research — including the hospital nursing studies, the Columbia disaster analysis, and the Project Aristotle findings — into a concrete operational framework for building psychological safety deliberately rather than hoping it emerges. Her learning zone matrix, her three-cluster leadership behaviour model, and her case studies across industries all give the research a practical resolution that academic papers alone cannot provide. For any entrepreneur who wants to understand not just why psychological safety matters but specifically what to do on Monday morning to build it, this is where to start. The book also addresses the question Edmondson takes most seriously: how to hold both high safety and high standards simultaneously, which is the precise tension that most management frameworks fail to resolve.

Have questions about this article?

If any part of this article raised questions you want to explore further, courbot.co is built for exactly that. It is courben.co’s AI assistant, designed around the psychology of entrepreneurship. Ask it anything from this article.

This article is for educational and informational purposes only. Sources: Edmondson, A.C. (1999), Administrative Science Quarterly, 44(2), 350–383. Frazier, M.L. et al. (2017), Personnel Psychology. Woolley, A.W. et al. (2010), Science, 330, 686–688. Google re:Work (2016), Project Aristotle.