Error culture: teams only learn when nobody is afraid

The statistics measure courage

In the mid-nineties, the researcher Amy Edmondson studied eight wards in two hospitals to find out how often nurses administered medication incorrectly. She expected the better-managed wards to make fewer mistakes. The result was the opposite: the wards with the best relationships within the team and the most committed ward managers reported the highest error rates.

A large share of the errors recorded came from voluntary reports. Where nurses trusted one another and feared no punishment, they reported their mistakes. Where fear prevailed, the mistakes remained invisible, yet still happened. The error statistics measured less the care taken than the willingness to talk about mistakes. Out of this grew the field of research known today as psychological safety.

What psychological safety is

In 1999, Edmondson defined the term as a team's shared belief that it is safe to take interpersonal risks there. What is meant is: asking a question, admitting a mistake, contradicting the boss, without being shown up for it. In a study of 51 teams at a manufacturing company, she showed that teams with this safety display more learning behaviour, meaning they discuss mistakes, seek feedback and adapt processes. Through this learning behaviour, safety led to better performance.

Google later tested this on itself and examined 180 of its own teams. Who sits on the team turned out to matter less than how the team treats one another. Of five characteristics of effective teams, psychological safety came first, ahead of dependability, clarity, meaning and impact. Google reports that members of such teams leave less often, bring in more revenue and are rated as effective twice as often by managers.

A meta-analysis from 2017 brought together 136 studies with over 22,000 participants. According to it, psychological safety is linked to information sharing, learning behaviour, creativity, engagement and performance. On the causal side stand the behaviour of leaders and a supportive working environment.

Fear in German offices

A 2024 YouGov survey for AXA with around 2,100 respondents shows what the picture looks like here. A quarter do not dare to openly admit a mistake at work. Among the under-25s the figure is 44 per cent. 23 per cent expect negative consequences if they disclose a mistake, and 27 per cent believe they would have to carry the can for work mistakes alone. After a mistake, five per cent turn first to their manager. The partner at home hears about it far more often than the boss.

The trend offers little hope. The DGUV Barometer Arbeitswelt 2025, a forsa survey for the statutory accident insurance scheme, shows that 51 per cent report greater time pressure, 43 per cent a more irritable working atmosphere and 29 per cent a declining culture around mistakes. The three figures appear in the same breath, and that fits together: anyone under pressure who gets snapped at prefers to keep a mistake to themselves and hope nobody notices.

What employment law has long known

The law is further ahead here than some executive floors. Anyone who causes damage at work through slight negligence is as a rule not liable for it. Only in cases of gross negligence, meaning where someone disregards what should have been obvious to anyone, or of intent is an employee fully liable. A single mistake also rarely justifies a dismissal. In most cases a warning must come first, that is, a formal reprimand. Employment law has therefore distinguished for decades between the slip and the breach of duty. That distinction is missing everywhere that every mistake is treated the same way: as a blemish.

Hiding things also has a price that goes beyond the team. If an employee sends customer data to the wrong recipient, that is a data breach. The company must as a rule report it to the supervisory authority within 72 hours of becoming aware of it. That deadline only helps if the mistake reaches the right people internally straight away. A team that keeps quiet out of fear turns a wrong click into a breach carrying the risk of a fine. Fear within a team is therefore also a compliance risk.

A culture around mistakes is not a comfort zone

The most common objection is that a good culture around mistakes means indulgence towards sloppiness. Edmondson herself gives the clearest answer to that. She distinguishes three kinds of failure: preventable failures in routine processes, complex failures in opaque systems and intelligent failures when trying out something new. Anyone who deliberately breaks known rules stands at one end of her scale, and there consequences are appropriate. At the other end stands the experiment, whose failure yields the most valuable information. Psychological safety means a team talks openly about which kind of mistake is at hand. It does not mean that every kind goes without consequences.

The second objection is more serious: managers have little time, and time pressure is increasing. Edmondson's observation from twenty years of research fits here. The managers she accompanied wanted to learn from mistakes and held debriefings for that purpose. Little changed nonetheless, because in every mistake they first looked for the culprit. The effort was there, it simply came to nothing.

Leadership decides the level of fear

On the hospital wards, the ward manager was the strongest influencing factor: where they coached and gave direction, the nurses reported their mistakes. The 2017 meta-analysis confirms that leadership behaviour is among the most important causes of psychological safety. A manager creates fear by dressing down the bearer of bad news in front of everyone. They reduce fear by naming their own mistakes first, taking questions seriously and, after a report, asking first what the team can learn from it.

For me, how a team works together matters more than any metric, because the metrics only arise from it. Trust produces performance, and trust shows itself nowhere more clearly than in the moment when someone says: that was me. Anyone introducing something new, new software, new processes, new AI tools, will see mistakes at the start. Whether the team reports them or hides them depends on what happened last time.