Has Psychological Safety lived up to the hype?

Or did we mistake an outcome for a 'lever'?

Equal talk time is the simplest secret behind every effective team. That’s roughly how the idea reaches us: it fits on a slide, in a playbook (guilty), or a ten-practices-for-fearless-teams carousel on LinkedIn. I championed the narrative for years, and I’ve changed my mind. Not about whether psychological safety is good or whether it’s worth having, it is. About whether you can do what the slide implies: reliably produce this “mythical beast”. Run the workshop, adopt the practices and mindset, et voilà, you’ll have a high-performing team out the other end.

No. Psychological safety isn’t a ‘lever’ we can pull to create an outcome. It’s what happens when a lot of things fall into place, more often by luck than design, and not something you can switch on with a bit of willing or training. The gap, between a thing that emerges and a thing you can produce on demand, is where the consultant and trainers’ psychological safety narrative falls down.

And that gap points to what we can call a ‘category error’. Some things can be categorised as linear. Cause leads to effect - the same input gives the same output - and the whole is the sum of its parts. A rocket is the snazziest example of the linear category. It’s about as complicated as can be, but orderly. You pull a lever or three (cause) and you get a result (effect). Nice and linear.

Organisations and teams, of course, are not like this. And treating them as if they were is our fatal flaw. They’re complex. The same input lands differently depending on history, relationships, timing, and a whole myriad of things that don’t stay still. As such, a complex system can’t be controlled because no lever reliably produces what you want, despite our best efforts and assumptions. This is why the thing that worked a treat in one context falls flat in another.

Psychological safety is a good place to watch this category error happen. We treat the team like a linear engineering problem, identify an input, pull the lever, expect the output, and then wonder why the results don’t reliably arrive. I’ve been there. Too many times.

Found Here, Applicable Everywhere?

The medical link runs deep in how this idea was popularised. Amy Edmondson’s early work (Edmondson, 1996) looked at only eight hospital teams in two hospitals. These are settings where the stakes are higher, more immediate and more visible than your average office. Miscommunicate and someone might die. Her foundational study of psychological safety came soon after, in 51 teams in a single manufacturing company, and found that teams with higher psychological safety showed more learning behaviour, which in turn predicted performance (Edmondson, 1999). The finding makes intuitive sense. People who feel safe to speak up surface problems sooner. But as fascinating as these findings are, I don’t imagine Amy Edmondson intended them to sound as simple and transferable as the change industry makes them out to be.

What happened next is that the finding travelled from the hospital and the factory floor to the open-plan office and remote teams. Google - not your average company - ran Project Aristotle and reported psychological safety as the most important factor in what made its teams effective (Google, re:Work, 2015). Research out of MIT pointed in a similar direction, identifying equal turn-taking in conversation and ‘social sensitivity’ (≈ empathy) as markers of collectively intelligent groups (Woolley et al., Science, 2010). You’ll see this convergence held up as proof. Several serious bodies of research, same answer. What more proof do you want?

But these findings don’t confirm a portable mechanism. Each found that psychologically safe-feeling, well-functioning teams tend to perform better, each in its own setting, often measured through the same crude proxy. That's a correlation observed in particular contexts, not a recipe that transfers to yours. And the ground is less firm than the convergence suggests. When researchers tried to reproduce the turn-taking result with new groups, they couldn't (Bates & Gupta, Intelligence, 2017).

You may have heard of the Replication Crisis, which plagues the field of psychology. The very studies held up as proof of new ways of working are exactly the kind that struggle to be replicated. True to linear thinking, they didn't adequately consider context. An effect that looks robust in one setting turns out smaller or absent in another. Somewhere between the academic research and pop psychological safety practice, a good thing that teams were observed to have has become something we can all 'install' in our teams. This is one of many examples showing the new ways of working movement is, unfortunately, propped up on some very shaky research foundations.

Equal Voice, Unequal Safety

The measure that came to stand in for psychological safety, roughly equal talk time with everyone contributing, is not the silver bullet it’s sold as. It’s a proxy, and a crude one. A team can have equal talk time and still be seething. A team can have one dominant voice and still do excellent work because that voice has earned trust and the others freely defer. Counting who speaks and treating the result as safety is yet more linear thinking applied to a complex context. Find the measurable input, adjust it, and expect the output to follow. It ignores what’s actually producing the pattern, the team’s history, how the leader behaves when challenged, whether people have watched honesty get punished before. None of that shows up in a participation metric.

Notice where the proxy came from. The settings that led to these findings, the ward and the factory floor, had noticeable features the open-plan office and remote team often lack. The stakes of silence were immediate and physical, a nurse who says nothing can affect another person’s health, a line worker who says nothing might make a faulty product. And everyone was in the room, faces visible, turns countable. Turn-taking and social sensitivity were measured because they could be measured. In most modern ‘office’ environments the cost of staying quiet is diffuse and deniable, a slightly worse decision surfaces and it’s unlikely the silent party gets the blame. The reality that gave psychological safety its teeth is weaker in the context where the finding gets sold hardest.

And then, of course, most knowledge work stopped happening in a room at all. Picture the actual conditions now. A standup on video, six or so faces in boxes, two or three cameras off. A decision that gets made in a Slack thread where the quieter people on the team either react with a thumbs-up or say nothing, and nobody can tell which one means agreement. There is no talking time to equalise and no room to read. The whole apparatus, equal voice, social sensitivity, the facilitator watching who speaks, assumes a co-located team most people stopped being part of years ago. The evidence was built almost entirely on people in rooms together. The hype carried on into hybrid and remote work it was never tested in, which is the lived reality for most of the people being sold to.

Wanting It, Bearing It

There’s a particular version of this you can watch play out. A leader notices the trend. They’ve read the LinkedIn posts, maybe Edmondson’s book The Fearless Organization (2018). They start talking about psychological safety, encouraging people to speak up, saying all the right things in the all-hands. And then they dominate the next meeting. They talk first and longest. They interrupt. And when candour finally arrives, when someone actually says the difficult thing the leader claimed to want to hear, they bristle. They get defensive. They explain why the person is wrong. They may even remember it at the next promotion round.

Most of them aren’t lying. They want this secret sauce. They just can’t tolerate it when it shows up. Wanting psychological safety and being able to withstand it are different capacities, and the second is much rarer than the first. It’s the same pattern that runs through every failed attempt at distributed decision-making: the leader likes the principle in theory, then can’t get out of the way when the reality produces something they wouldn’t have chosen.

This is why it can’t be installed in the way LinkedIn would have us believe. The leader’s own behaviour under pressure is one of the conditions, which means the person trying to create the safety is also one of the variables determining whether it appears. You can’t announce your way past that. If the first honest challenge gets punished, however subtly, people read the punishment, not the announcement. They learn what actually happens when you speak up, and they calibrate. One defensive reaction from the boss teaches more than ten psychological safety workshops.

What Margaret Heffernan Said

Asked about how to help people find the courage to speak up, Margaret Heffernan’s response turned my thinking on its head.

First of all, I think one has to be quite careful with the phrase psychological safety. I know this has been much publicised by Amy Edmondson. And I think it’s important. But I think it’s also a little naive, in so far as, I can do my very best to create a work environment that feels safe. But if there’s high unemployment, if there’s high inflation, if people are carrying a very large amount of personal debt, usually in the form of a mortgage, I promise you people won’t feel safe. So there’s a significant amount to do with psychological safety that is entirely beyond the control of the organisation. (Heffernan, Leadermorphosis, 2021)

This outlines precisely where the psychological safety lever assumption falls arse over tit.

The promise of psychological safety as a thing you can produce assumes the organisation is controllable. Run the right workshops, model the right behaviours, measure the right things, and people will feel safe. But someone struggling with a big mortgage and bills and a family to feed doesn’t leave that reality at reception. Nor does the colleague who is simply playing the game (be that coasting or clambering the slippery pole). The conditions that determine whether they feel safe extend well past the team, past the building, and into an economy nobody in the room governs. The team cannot override that by running an inspiring workshop. It’s a hard limit on what any intervention can reach, and it’s almost never mentioned by the people selling the intervention (once again, guilty as charged in the past).

The Category Error

Step back, and the pattern is always the same. A real thing gets noticed in a complex system, where it emerged from conditions nobody fully controls. It gets turned into a measure, the measure becomes a snazzy practice or elusive mindset, and it gets sold as a lever, as though the team were a linear system and safety were the output you get by simply adjusting the input.

That’s the cardinal sin that plagues the new ways of working, transformation and change industry: treating a complex system as if it were a linear one. Treat a team like a rocket. Pull the lever, get the result. Except there is no lever, because psychological safety (or whatever else) was never an input you supply. It’s an outcome. It comes out of a team’s history, the relationships in the room, how the boss reacts the first time someone says something hard, and a good deal that has nothing to do with the workplace at all, the mortgage, the job market, whatever’s waiting at home. All sold (and by well-meaning, good people) as something we can neatly install on Monday. And when it disappoints, nobody blames the category error. They blame the facilitator, the leader, the culture, or whatever else lets them stay blind to the realities of complexity and the unpredictable nature of the system we’re working with (people, teams and organisations).

This is the habit Margaret Heffernan was pointing at when she warned us against hunting for “the silver bullet, that one thing that changes everything.” She doesn’t think it exists, and it doesn’t. A “mythical beast”, in her words, because the organisations we work in are far too complex for one idea to change everything. Psychological safety became one more candidate for the beast, a single lever offered to a system that simply doesn’t bow to levers.

So has psychological safety lived up to the hype? No. As an observation about why some teams do better, sure. As a lever you can pull to produce it on demand, in any context, in a room or on a screen, whatever’s pressing on people from outside, no. That was the category error all along: a complex system yet again mistaken for a linear one. An outcome masquerading as an input. The messy reality doesn’t chime with the slide.

🇨🇭 And here, for no reason, is a pic from Switzerland taken last week as our trip nears its end. We’re now near Annecy, and three weeks from home.

---