Safety talk: what is your safety system really saying?

We say we want people to speak up. Then, when something goes wrong, we investigate them, audit them and, sometimes, discipline them. Is it any wonder that people think twice before telling us the truth about how work really gets done?

In recent years, much attention has been given to the idea of psychological safety. Harvard professor Amy Edmondson, who has become synonymous with the concept, describes it as a person’s willingness to take an interpersonal risk.¹ In other words, how comfortable they feel sharing information that others might not want to hear. If we suspect there is a negative consequence waiting for us, such as embarrassment, humiliation or discipline, most of us will simply keep quiet.

As health and safety professionals, we produce a vast number of documents, and we interact with people in all business functions. We build management systems to plan, control and monitor safety. 

We track statistics, display posters and teach models that explain how accidents happen. But have you ever stopped to ask: are these systems sending the message we think they are?

Safety is a system of signals
Safety isn’t only managed through procedures and controls. It is constantly communicated: through the words we use, the numbers we track and the symbols we display. Incident statistics, safety slogans and the models we teach all shape how people interpret what safety means in an organisation. Our safety systems can have unintended impacts: they create meaning, and that meaning influences attitudes and perceptions.

Research increasingly backs this up. Common safety metrics don’t only track performance. 
They also act as signals that tell people what the organisation values and how it expects them to behave.² There is even evidence that subtle changes in the wording of a policy can shift how people perceive the safety climate, particularly when it comes to trust and fairness.

Words that assign blame
Human beings rely on mental shortcuts: rules of thumb that help us make quick judgements.³ Without them, every action would feel like learning to do it for the very first time. But these shortcuts can lead us astray. One well-known example is the ‘fundamental attribution error’: our tendency to blame an individual’s actions while overlooking the situation and environment that shaped their behaviour.⁴

Language can make this worse. ‘Agentive’ language emphasises a human actor – for example, compare “he smashed the vase” with “the vase smashed”. In a fascinating study, researchers Fausey and Boroditsky analysed almost 200,000 Old Bailey court cases and found that agentive wording increased the likelihood of blame being assigned, and even influenced legal judgements.⁵ In our world, safety researcher Crista Vesel has shown that this kind of language in investigation guidance nudges practitioners towards finding ‘human error’, shifting the purpose of an investigation from learning to blame.⁶ Terms like ‘unsafe acts’ assume choice, control and avoidability. They presuppose that the problem lies with the person, not the system.

Words also carry associations learned from the places we usually hear them. ‘Investigation’ and ‘officer’ bring to mind the police, courtrooms and establishing culpability. ‘Audit’ suggests inspection, scrutiny and being caught out. ‘Accident’ implies bad luck; ‘incident’ hints at determinable causes, and both reinforce a reactive view of safety.

Psychologists call these effects priming and framing: the way information is presented influences how we think, feel and act, often without us being aware of it.⁷ The point is simple: the language we use in health and safety is not neutral. In my own (unpublished) research, changing the language of a health and safety policy, from traditional safety language to supportive learning-centred language, positively affected people’s perceptions of safety climate and how accurately they reported their errors following a forced error exercise.

The stories we tell about safety
Safety is a social construct. It’s a word we use to describe a temporary moment in time when nothing bad is happening. A state of ‘safety’ doesn’t mean risk has gone away. It may still be present, just not yet materialised. Our judgement of whether something is ‘safe’ is shaped by perception, context and hindsight. It cannot be definitively proven.

Because safety is such a complex idea, we reach for metaphors to explain it – and those metaphors matter.⁸ Describing safety as a ‘journey’ suggests continuous progress, but can imply improvement without ever spelling out what needs to happen, and when. Talk of ‘defences’ and fighting a ‘war’ on hazards frames safety as a battle. Calling it a ‘system’ conveys structure and control. Each of these can be useful, but each can also obscure accountability or oversimplify reality.

Metaphors often reflect how leaders imagine safety, rather than how work is actually experienced on the front line. This is often called the gap between ‘work-as-imagined’ and ‘work-as-done’.⁹ When the story leadership tells doesn’t match workers’ lived experience, trust suffers.¹º

When numbers become meaning
We often treat safety metrics and KPIs as objective measures of performance; notwithstanding the well-documented problems with the validity and reliability of the data behind them.¹¹ But metrics are signals too. Measures such as lost time injury frequency rates are open to manipulation, can strain relationships (think of league tables) and offer a false sense of control.

A low number can signal “we are safe”. But it can just as easily signal “we don’t want bad news”, depending on how it squares with what the workforce sees around them. This is particularly true of mantras like ‘Zero Harm’. Zero symbolises perfection. It leaves no room for natural human fallibility or the everyday variability of work, and it creates pressure not to report when things go wrong.

The unintended consequences are predictable. When performance is judged against injury rates or targets, people spend more energy protecting the metric than improving the work. Events get re-classified to keep the numbers low. Research shows this under-reporting gets worse when safety performance is tied to financial rewards.¹²

Our best-loved models send signals too. The Heinrich/Bird triangle is often used as a simple picture of accident causation. But it is also a powerful symbol. It presents a stable, predictable relationship between minor events and serious outcomes, implying that if we control the base of the triangle, we prevent catastrophe at the top. That reinforces a belief in neat, linear cause and effect, despite established evidence that the factors behind minor injuries are usually not the same as those behind disasters.

James Reason’s famous Swiss cheese model has a similar issue. It is more sophisticated, but it still pictures accidents as failures lining up across layers of defence. This is the same chain-of-events mindset. In reality, complex work systems don’t fail in neat, sequential ways. They are dynamic, non-linear and unpredictable.

Models like these visually encode a belief system that shapes what organisations measure, what they prioritise and how they respond to risk.

How does this affect people?
Language and symbols shape trust because they signal how people are likely to be treated when things go wrong. Blame-laden language, failure-focused metrics and courtroom vocabulary all imply judgement and consequence. These are things we naturally seek to avoid. Supportive, human-centred language, on the other hand, can increase perceptions of trust, fairness and openness: the very foundations of psychological safety.

One influential model of trust suggests it rests on three perceptions: ability, benevolence and integrity.¹³ In other words, people ask themselves: is this organisation (and its representatives) competent? Does it care about me? Will it act in line with the values it claims to hold? The language and symbols we choose in our conversations and safety management systems answer those questions in every interaction, whether we intend them to or not.

People respond not only to what systems and procedures say, but to how they are made to feel. 

If safety processes feel punitive, evaluative or like a tick-box exercise, they undermine perceptions 
of care and integrity. If they feel supportive and genuinely curious, they strengthen trust and encourage people to speak up.

From control and blame to curiosity and learning
Many contemporary safety thinkers now emphasise the importance of learning, particularly from frontline workers, who understand better than anyone the challenges, trade-offs and workarounds of real work. This is nothing new: consultation and participation have been included in legislation and ISO standards for a long time; however, they are often undervalued. But none of that learning is possible without trust and psychological safety. Organisations, leaders and safety practitioners must create environments where people can talk openly about what’s difficult and what went wrong, without fear.

Language and symbols play a critical role here. How we choose to communicate can strengthen or destroy the foundations we are trying to lay. The good news is that small shifts can change how systems are experienced:

• Replace ‘investigation’ with ‘learning review’: signalling a shift in purpose from fault-finding to understanding14
• Change ‘audit’ to ‘reflection’: emphasising collaboration over inspection
• Move from ‘unsafe acts’ to ‘insights’: shifting the focus from the individual to the situation
• Reframe metrics: measuring ‘learning insights’ rather than only injury rates moves the conversation from avoiding failure towards actively understanding work.15

A word of caution
Changing language alone is not enough. In fact, if new words are not matched by genuine changes in purpose and leadership behaviour, superficial rebranding can do real damage. If an organisation adopts the term ‘learning review’ and then responds to the next event with blame and discipline, trust will evaporate – and be even harder to rebuild. People are remarkably perceptive. They can tell whether your presentation is really aligned with your intent.

What is your organisation really saying?
So, here is the question to take back to your workplace. Is your safety management system saying “we want to understand” or “don’t make mistakes”? Every policy, procedure, metric, risk assessment, investigation, town hall, word and symbol sends a signal about what is valued and expected.
Ask yourself: are your signals encouraging people to speak up, or quietly telling them to stay silent?

Grant Tickle is a wellbeing health and safety auditor and consultant at British Safety Council.
Connect with him on LinkedIn:
linkedin.com/in/granttickle/

For more information see:
britsafe.org

References
1.    Edmondson, A. C. (1999). Psychological safety and learning behavior in work teams. Administrative Science Quarterly, 44(2), 350–383.
2.    Pomeroy, J., & Pilbeam, C. (2025). Signs of safety: An investigation of how OHS professionals interpret injury metrics. Journal of Safety Research, 95, 87–100.
3.    Tversky, A., & Kahneman, D. (1974). Judgment under uncertainty: Heuristics and biases. Science, 185(4157), 1124–1131.
4.    Ross, L. (1977). The intuitive psychologist and his shortcomings: Distortions in the attribution process. In Advances in Experimental Social Psychology (Vol. 10, pp. 173–220). Academic Press.
5.    Fausey, C. M., & Boroditsky, L. (2010). Subtle linguistic cues influence perceived blame and financial liability. Psychonomic Bulletin & Review, 17, 644–650.
6.    Vesel, C. (2012). Language bias in accident investigation (MSc thesis, Lund University); Vesel, C. (2020). Agentive language in accident investigation: Why language matters in learning from events. ACS Chemical Health & Safety, 27(1), 34–39.
7.    Tversky, A., & Kahneman, D. (1981). The framing of decisions and the psychology of choice. Science, 211(4481), 453–458; Bargh, J. A., Chen, M., & Burrows, L. (1996). Automaticity of social behavior: Direct effects of trait construct and stereotype activation on action. Journal of Personality and Social Psychology, 71(2), 230–244.
8.    Pomeroy, J., & Pilbeam, C. (2026). Metaphors we’re safe by: Analysing conceptual metaphors within corporate safety reporting. Safety Science, 199, 107190.
9.    Hollnagel, E., Wears, R. L., & Braithwaite, J. (2015). From Safety-I to Safety-II: A White Paper. University of Southern Denmark, University of Florida and Macquarie University.
10.    Mayfield, M., & Mayfield, J. (2021). Sound and Safe: The Role of Leader Motivating Language and Follower Self-Leadership in Feelings of Psychological Safety. Administrative Sciences, 11(2), 51. https://doi.org/10.3390/admsci11020051
11.    Poole, G., & Dekker, S. (2024). Random Noise: Measuring Your Company’s Safety Performance. Routledge.
12.    Probst, T. M., Bettac, E. L., & Austin, C. T. (2019). Accident under-reporting in the workplace. In Increasing Occupational Health and Safety in Workplaces (pp. 30–47). Edward Elgar Publishing.
13.    Mayer, R. C., Davis, J. H., & Schoorman, F. D. (1995). An integrative model of organizational trust. The Academy of Management Review, 20(3), 709–734.
14.    Bown, S. in Lloyd, C. (2020). Next Generation Safety Leadership: From compliance to care. CRC Press, Taylor & Francis Group.
15.    Due Diligence Index Council. (2021). DDI-S Standard: Standard & guidance for use. https://www.duediligenceindex.online/