Everyone knows the safety rule. So why is it still skipped?

The COM-B model describes behaviour in terms of capability, opportunity and motivation. Can someone complete the action? Does the environment allow it? Are the reflective and automatic drivers aligned with doing it? Michie, van Stralen and West developed the Behaviour Change Wheel to link such diagnoses to possible intervention functions [S01]. It is a useful organising framework, not an occupational safety standard or an automatic explanation for every incident.

Consider three workers who omit the same inspection. One does not know the correct sequence and needs competent instruction. Another cannot obtain the necessary equipment during the shift and needs a process change. A third expects ridicule for reporting uncertainty and may need a different social climate as well as clear protection against retaliation. The outward behaviour is identical. The appropriate response is not.

Begin with a proper hazard assessment and competent safety expertise. Investigate task demands, equipment, staffing and training before proposing an association study. Only where meaning or social expectation is a plausible and distinguishable part of the problem should association research be added to observation, interviews and existing safety records.

A missing check is an observation rather than a complete explanation. Keep alternative hypotheses visible until evidence allows you to narrow them down.

These causes can operate simultaneously. Limited time can encourage a shortcut, which later becomes habitual and socially normal. A one-box diagnosis may miss the very system that keeps the behaviour in place.

Being able to describe the right action does not guarantee doing it. Sheeran and Webb synthesised research on the intention-behaviour gap and the conditions that prevent intentions from becoming action [S09]. Their review covers various behaviour domains and cannot be treated as a quantified estimate for every industrial safety task. The defensible lesson is that education and stated intentions alone are incomplete measures of performance.

Risk judgments also have an emotional and experiential dimension. Someone who has completed a shortcut hundreds of times without an obvious incident may perceive the immediate demand of the job as more pressing than a remote hazard. Work on the affect heuristic shows how feelings can inform judgments under uncertainty [S10]. It does not demonstrate that we can predict an individual violation from an emotional response.

Instead of asking only 'Do you understand why the procedure matters?', ask what changes when the line is behind schedule, when the supervisor is absent and when a second worker is unavailable. Ask when people carry out the complete procedure and what made those moments different. That is where behavioural hypotheses begin.

Imagine a team with twenty minutes to finish dispatching an order. The prescribed procedure requires a stop, a separate check and a second colleague. The schedule was written without allowing for those minutes. The worker faces two organisational demands presented as equally essential. Repeating 'safety is our first priority' cannot make that conflict disappear.

A meta-analysis of workplace safety research integrated person-based and situation-based antecedents of performance and safety outcomes [S02]. Longitudinal work has examined relationships between safety climate, motivation, behaviour and accidents across levels and time [S04]. Neither establishes the cause of a particular incident. Together they provide good reasons to investigate the working situation instead of automatically blaming awareness.

Measure the actual time required for the safe process. Examine staffing, access to equipment and the consequences of a justified stop. Does the supervisor protect that decision or ask why the target was missed? A good safety intervention may have to change the plan, equipment or authority structure before it changes any wording.

Managers can give convincing talks about safety. Employees also watch what happens when a person delays a job for a legitimate safety reason. Are they thanked, quietly undermined or openly reprimanded? Which outcome receives genuine recognition at the end of the shift? Safety climate concerns the shared meaning of organisational safety priorities; Zohar's early work is a foundation for this research tradition [S06].

Griffin and Neal distinguish safety compliance from safety participation [S03]. An employee may follow rules without feeling able to speak up about a defective process. Someone who raises concerns may be displaying active safety involvement rather than resistance. Edmondson's research on psychological safety and team learning suggests why people need room to take interpersonal risks, such as admitting uncertainty [S12]. That study is not evidence of a fixed reduction in accidents after a campaign.

A credible safety message therefore needs visible leadership behaviour: people must be able to stop a task safely, report a problem and trust that a justified intervention will not become a career penalty.

These are diagnostic possibilities, not measured Neurofactor findings and not fixed psychological profiles for employees.

A team may endorse 'protection' in a questionnaire while encountering 'delay' in everyday work. Combine reported meanings with observation. One negative phrase is a useful clue but not proof of the causal mechanism.

Entirely fictional illustration. It is not a Neurofactor client result or research involving Defence, BUKO or another identifiable employer. In a distribution centre, staff must inspect a loading platform before a truck departs. The procedure requires a stop, a visual check and confirmation from a second worker. Everyone can explain why this matters. During busy shifts, however, the second confirmation is occasionally omitted.

Management's first reaction is to repeat the instructions. A diagnostic team instead observes several shifts, reviews the schedule and talks to staff confidentially. Three competing explanations emerge. The second worker is sometimes allocated elsewhere. Dispatch targets do not always allow time for the check. And a colleague who stops the line is jokingly called 'the brake'. Those observations have different meanings, and none alone proves causation.

In this fictional design, the response is not a single campaign. The business creates a protected check point in the workflow, ensures that a second person is available and makes legitimate stopping an explicitly supported decision. It then measures whether complete inspections actually become more consistent. Without that follow-up, an appealing story remains just a story.

This comparison shows why 'employees need to care more' is not a useful intervention plan.

More than one row may apply to the same worker or shift. Gather repeated evidence rather than reducing complex work to a single reason chosen after an incident.

Step 1: define one observable safe behaviour, such as completing every required loading check before release. Step 2: compare the documented procedure with how work is actually carried out. Step 3: conduct competent hazard assessment and address urgent risks immediately; do not wait for behavioural research. Step 4: observe multiple shifts and conduct confidential interviews.

Step 5: set out alternative hypotheses about knowledge, skill, opportunity, habits, team norms and negative meanings. Step 6: choose additional measures only when they can distinguish the hypotheses. Step 7: change the process, training or social expectations in a controlled and ethically acceptable manner. Step 8: monitor complete safe behaviour, process quality and unexpected consequences over a meaningful period.

Before collecting data, decide who protects participants, who owns operational safety decisions and who may access identifiable records. No research design should ask a worker to face an unmanaged hazard merely to create a control group.

'Why don't you follow the rule?' sounds like an accusation, particularly when managers have formal authority over respondents. A more productive prompt is: 'Tell me about the last occasion on which completing the procedure was difficult.' Follow with: 'What was happening to the schedule?', 'Who was available?', and 'What would colleagues have said if you had stopped?'.

Explore exceptions as carefully as failures. On which shifts did the complete safe sequence work well? What made the difference: accessible equipment, a realistic workload, a nearby colleague or a supportive supervisor? People may also describe situations in which a written procedure conflicts with real physical conditions. Those reports require professional assessment rather than dismissal.

Make feedback confidential and separate it from routine performance appraisal wherever possible. Psychologically safe conversations support learning, but an interviewer using friendly language does not prove that the organisation is safe to challenge [S12].

An expensive instrument is not necessarily better than a well-designed observation.

EEG cannot read whether a worker intends to break a safety rule. If a question does not require a valid, carefully defined physiological measure, such equipment should not be included for appearance's sake.

Removing a hazard is generally stronger than repeatedly asking employees to remember to avoid it. NIOSH describes the hierarchy of controls: elimination, substitution and engineering controls are preferred before relying on administrative instructions or personal protective equipment [S15]. This is a basic safeguard against turning occupational safety into a communications exercise.

If workers can enter a dangerous machine zone, ask first whether the hazard can be eliminated or access technically controlled. Only after adequate controls are in place should additional communication address remaining safe actions. A positive attitude cannot compensate for a missing interlock, and a catchy slogan does not remove an employer's safety responsibilities.

Behavioural scientists and safety engineers need to work together. One may help explain why people do not use an available measure; the other must ensure the measure really controls the physical hazard. Neither role should pretend to replace the other.

Where knowledge is missing, teach and verify understanding. Where skills are missing, practise the procedure in realistic conditions. Where opportunity is lacking, adjust staffing, equipment and the sequence of work. Where a habit dominates, redesign the cue and make the safe response easier at the decisive moment. Where a social norm is the problem, leaders and respected peers must consistently demonstrate that safe stopping is supported.

Negative associations require more than new wording. If reporting is perceived as 'betraying the team', protect people who report, show how concerns are resolved and remove punitive incentives that encourage silence. If the organisation's behaviour contradicts the campaign, a different slogan is unlikely to earn trust.

For some manageable routines, a concrete if-then plan can help: 'If the handover is incomplete, I do not restart the machine and I notify the responsible colleague.' Research on implementation intentions provides a rationale for specifying responses to familiar cues [S13]. Such plans complement but never replace necessary engineering controls.

The first decision should be the type of control required, not the creative format of the message.

A successful intervention tackles an evidenced barrier. It should not merely look impressive in a workshop or produce an attractive internal poster.

Every team develops informal expectations. Some help people anticipate hazards and coordinate reliably. Others reward shortcuts. A new worker often learns quickly whether colleagues admire careful checks or celebrate never slowing the line. That learning can make the formal procedure socially costly to follow, even when everyone agrees with its stated purpose.

Do not infer the team norm from age, job title or apparent confidence. Observe reactions when work is stopped for safety. Ask about situations when someone challenged a shortcut successfully. Use individual confidential interviews where hierarchy could discourage honest responses. Mullen's study of factors influencing individual safety behaviour supports considering organisational and interpersonal context alongside the worker [S14].

A norm is a hypothesis about shared expectations that can be examined. It is not a licence to categorise a whole shift as unsafe, nor does changing language remove the need for suitable staffing and working equipment.

You may compare two approved ways of explaining a properly safeguarded procedure. You may compare whether an additional reminder is better placed before a safe action or at the end of training. You must not deliberately withhold required protection, introduce a known hazard or ask workers to perform an unsafe routine for experimental purity. If a serious risk is discovered, correct it before continuing an evaluation.

Define acceptable test variants with safety professionals, select measures before starting and consider differences in shift, work type and production volume. Outcomes might include complete checks, timely stops, accurate reporting and unintended consequences. Randomisation can sometimes strengthen causal inference, but it is not always feasible or ethical in field operations.

If you use a before-and-after comparison, disclose limitations. An increase in near-miss reports may reflect better reporting rather than a deterioration in safety. Interpret incident counts together with exposure, reporting quality, direct compliance measures and qualified operational judgment.

A beautiful safety campaign can achieve high approval and no meaningful change in performance. Conversely, a safer process may become more reliable even if no one finds the redesign particularly inspirational. Separate meaning, observable behaviour and safety outcomes rather than folding them into one sentiment score.

An association question can explore how stopping is perceived. Reliable observation or process data can show whether the required action occurs. Reporting records can reveal changes in concerns raised, with careful interpretation of reporting culture. Serious incidents may be too rare for a short study to detect a meaningful change, but they must not be ignored in the wider safety management system.

Follow changes over time. Does the safer routine persist through night shifts, peak periods and temporary staffing? Are there tasks where the procedure still creates friction? These questions keep an attractive communication result from being mistaken for a demonstrated reduction in risk.

Use different evidence for different claims.

The quality of a safety evaluation depends less on the number of metrics than on whether each one answers a clearly defined question.

Operations should verify that the safe procedure fits a real shift, including busy periods. Safety professionals should prioritise hazard assessment, elimination, engineering controls and evidence that remaining risks are adequately managed. HR should protect learning, competence and reporting channels from retaliation. Communications should make safe practices understandable without covering up contradictions in the work process.

The behavioural research contribution is to connect these perspectives. What does 'stopping the line' mean to employees? Does that interpretation recur in independent evidence? Which change to the process or supervisory response makes the safe choice credible at the precise moment of action? These questions are more useful than choosing the most memorable slogan.

Neurofactor's knowledge base explains the behavioural diagnostic framework and the role and limits of a nudge. Neither article should be read as permission to substitute messaging for compulsory safety measures.

One: supplying information where working equipment is missing. Two: treating a physical hazard as a marketing problem. Three: describing every error as a personal failing. Four: praising stopping in a poster while punishing delay in practice. Five: treating an implicit association score as an employee-level injury prediction. Six: celebrating campaign reach without measuring safe performance. Seven: interpreting fewer reports as automatic evidence of lower risk.

These mistakes share an assumption: that the campaign's intended meaning automatically becomes the meaning of safe work on the floor. In reality, people compare promises with what managers, colleagues and systems do when circumstances become difficult. An unfulfilled promise can strengthen the very cynicism a campaign was meant to remove.

A strong investigation is allowed to reach an uncomfortable conclusion: the most useful intervention may be a change to workflow, authority or machinery rather than a new communication concept. That is an achievement, not a failed campaign.

The most important question is not whether workers can recite the hazard in a training session. It is whether the organisation makes the correct action possible when schedules tighten, routines take over and colleagues are watching. Someone may know the rule and still face a shortage of equipment, a strong habit, social pressure or an unwanted meaning attached to stopping. Those explanations lead to different responses and must be examined rather than guessed.

Start by removing and adequately controlling hazards. Then investigate what safe behaviour means in the working environment and how the system can support it. When technical protection, job design and credible team expectations work together, safety communication becomes something employees can verify in practice. That is far more valuable than a campaign people merely remember.