The journal Safety Science has had a remarkable history and turns fifty this year. This esteemed journal began in 1976 as the Journal of Occupational Accidents, and its anniversary volume published recently carries Jean-Christophe Le Coze’s stocktake of what half a century of research has delivered (Le Coze, 2026). His assessment is candid and should be somewhat confronting for safety practitioners. The field of safety science has produced an enormous body of models and methods, but its research traditions have drifted apart rather than converged, and he suggests we may still be in the “prehistory of safety science”. For safety practitioners, that verdict likely lands close to home in how we establish and manage our safety systems. Fifty years of research has coincided with fifty years of accumulating lots of safety documentation (while many valuable), and one of the sharpest questions in its recent volumes is whether all that documented activity actually protects the people doing the work. Have we cluttered our spaces so much that we can’t see the wood for the trees.
The research answer is largely uncomfortable and confronting but two strands of work have matured past academic curiosity. That is the gap between work as imagined and work as done, and safety clutter. Neither is an argument working less to improve work safety but both are arguments for doing work safety that actually works and has impact.
Work as imagined vs work as done
Every safety management system contains an implicit theory of how work happens. Procedures, SWMS, permits, and checklists describe work as the organisation and the drafter of these documents imagine it: a stable sequence of steps under predictable conditions. Erik Hollnagel gave this idea its modern vocabulary, distinguishing work as imagined from work as done and arguing that safety management studies failure closely while paying little attention to why work usually succeeds (Hollnagel, 2014).
The strongest objection to this line of thinking has been the absence of hard field evidence that this process works but it is evident that objection is weakening. A study published this year in the journal Safety tracked chemical filling operations at a petrochemical facility over five years, using helmet-mounted cameras to record 1,422 procedural steps across 40 procedures (Ashraf, Peres and Sasangohar, 2026). Coders compared each step against the written procedure. The result: 32.9 per cent of steps showed some form of adaptation, whether skipped, resequenced, or done by a different method. Every observed task was completed successfully and without incident.
One in three steps departing from the written word sounds alarming until the adaptations are unpacked. Interviews with operators and supervisors produced three categories, published as the Routine-Efficiency-Safety framework. Routine adaptations are the normalised practices a work group settles on (“that’s just how we do it”). Efficiency adaptations resequence or combine steps to manage workload without compromising controls. Safety adaptations add verification beyond what the procedure requires, such as inspecting a railcar’s bottom hatches before loading rather than after. Workers were not cutting corners as a rule but were often they were exceeding the written standard.
The implication for auditors and advisors is that a binary compliance lens, where every departure from procedure is somewhat of a violation, misreads much frontline behaviour and may teach workers to hide adaptations rather than explain them. Recurring adaptations are better read as data about where the procedure and the work have separated significantly.
There is a limit condition, and it matters as much as the finding. Repeated successful adaptation is also the observational signature of normalisation of deviance, the pattern Diane Vaughan traced through the Challenger disaster, where each uneventful deviation quietly became the baseline for the next (Vaughan, 1996). Forty procedures without incident says little about low-probability, high-consequence failure, and Rasmussen (1997) argued that adaptation under production pressure migrates work steadily toward the boundary of safe operation unless something pushes back. Hale and Borys (2013), reviewing three decades of research on safety rules, found neither paradigm wins outright: rules defended against every deviation fail, and so does uncritical trust in frontline judgement. Adaptation is resilience while it stays inside risk boundaries the organisation has verified. Once nobody is checking where the boundary sits, the same behaviour is drift. The task is not to celebrate adaptation or suppress it, but to surface it so someone competent can ask which kind it is.
Safety clutter | when work safety work stops serving safety
If a third of procedural steps do not survive contact with real work, why do the procedures keep growing? Griffith University’s Safety Science Innovation Lab put a specific name to the phenomenon: safety clutter, the accumulation of safety procedures, documents, roles, and activities performed in the name of safety that do not contribute to the safety of operations (Rae et al., 2018).
Duplication treats the same risk in overlapping documents, so a single task might sit under a procedure, a SWMS, a permit, and a pre-start checklist. Generalisation rolls a control developed for one context out everywhere, whether or not the hazard exists there. Over-specification prescribes detail beyond what the risk requires, converting professional judgement into box-ticking. Behind all three sits an asymmetry the authors describe precisely: adding a safety activity is easy and carries no personal risk, while removing one exposes whoever signs off. Clutter therefore climbs after every incident, audit, and management change, and almost never comes back down to reality.
Rae and his colleagues document cynicism and surface compliance among workers completing paperwork they can see makes no difference. Rae and Provan (2019) distinguish safety work, the visible activity of forms, audits, and meetings, from the safety of work, the actual level of risk people face. An organisation can excel at the first while the second quietly deteriorates. Every hour a supervisor spends completing a checklist that changes nothing is an hour not spent on supervision of how the job is actually going.
The counterargument also deserves its place in this article. Cooper (2022), writing in Safety Science itself, argued that Safety-II thinking lacked empirical support and offered practitioners few workable methods. Field studies like this year’s petrochemical work are a direct response, but the honest position in 2026 is that the evidence base is building and it has not yet settled. Nothing in the clutter literature licenses removing controls that manage real risk. Obviously, the underlying target is activity that manages no risk at all.
What this means under Australian WHS law
Some WHS managers hear “declutter” and worry about legal exposure. The concern deserves a straight answer, because the legal position cuts both ways for safety practitioners.
We think that businesses should start with what cannot be removed. The WHS Regulation mandates specific records: a safe work method statement before high risk construction work commences (s 299 of the Work Health and Safety Regulation 2011 (Qld)), asbestos registers, and emergency plans among them. Part of what safety practitioners experience as clutter is legally required and can only be consolidated, never deleted. Decluttering therefore begins by separating mandated documents from discretionary ones, and the discretionary pile is usually far larger than assumed.
For the discretionary pile, the primary duty under s 19 of the Work Health and Safety Act 2011 (Qld) requires a PCBU to ensure health and safety so far as is reasonably practicable, with s 18 defining that standard by the likelihood and degree of harm and the availability and suitability of ways to eliminate or minimise the risk. The duty attaches to risk, not to document volume, and a procedure no worker follows does little to discharge it. One caveat: Victoria frames employer duties and officer provisions differently under its Occupational Health and Safety Act 2004, so organisations there need specific advice.
The officer duty under s 27 is often raised as a reason to keep everything. It reads more naturally as a reason to keep what works. Due diligence includes taking reasonable steps to ensure the PCBU has available for use, and uses, appropriate resources and processes to eliminate or minimise risks, and has processes for receiving information about hazards and responding in a timely way. Candour requires two admissions. No Australian court has yet found an officer’s due diligence deficient because the safety system was too voluminous, so it would overstate matters to present clutter as an established legal exposure. The defensible claim is narrower. Where an officer knows, or could readily discover, that a large share of prescribed steps is routinely adapted in the field, a system that keeps producing paperwork instead of examining that gap becomes harder to present as appropriate processes in use.
The second admission is that removal carries somewhat of a hindsight risk. A document deleted eighteen months before an incident it arguably touched will be examined ruthlessly through an investigation, however sound the reasoning was at the time of removing it. That risk partly explains the add-easy, remove-hard asymmetry, and it is not irrational. It is also the strongest argument for removing with full discipline. Consultation records, a documented risk decision, and a named decision-maker turn a removal from an unexplained absence into evidence of the risk-based reasoning the Act requires.
Where to start safe decluttering
The research suggests a sequence rather than a purge and its the same sequence Safetysure applies in its own audit work.
- First, observe work as done before touching any document. Watch a handful of routine high-risk tasks, note where practice departs from procedure, and ask the workers why. Frequent efficiency adaptations usually signal a sequencing problem. Frequent safety adaptations usually signal a gap in the risk assessment that workers have already found and filled.
- Second, audit the safety work itself. For each procedure, form, and recurring activity, the clutter test is blunt: what risk does this control, and what would change if it stopped? Rae et al. (2018) found that many activities survive on the assumption that someone else needs them. Asking the question aloud, with operations and safety in the same room, dissolves a surprising amount of clutter.
- Third, remove with the same discipline used to add. Decluttering is a change to the safety management system and warrants the same consultation, risk assessment, and records as introducing a control, for the hindsight reasons above.
- Fourth, keep the severity firewall intact. A control that manages a real risk poorly needs fixing, not deleting. Clutter means activity with no risk-management function, and conflating the two is how decluttering earns a bad name.
Summing it up
Fifty years in, Le Coze’s verdict is that safety science has converged on less than its practitioners would typically like to hear. On this question, separate research traditions likely point the same way. The volume of work safety activity is a poor proxy for the safety of work, and adaptation in the field is information before it becomes a compliance problem. The organisations best placed to demonstrate due diligence over the next decade will not be the ones with the thickest safety manuals spewing out procedures. They will be the ones able to show that every work safety activity earns its place, and that someone is watching the boundary that it may cross.
Safetysure is an ISO 9001, 45001, and 14001 accredited workplace health, safety, and occupational hygiene consulting firm operating across Queensland, New South Wales, Victoria, Western Australia, the ACT, and Commonwealth jurisdictions. For more information, visit safetysure.com.au.
References
Ashraf, A.M., Peres, S.C. and Sasangohar, F. (2026). Towards a worker-centered framework for categorizing procedural adaptations. Safety, 12(1), 28. https://doi.org/10.3390/safety12010028
Cooper, M.D. (2022). The emperor has no clothes: A critique of Safety-II. Safety Science, 152, 105047. https://doi.org/10.1016/j.ssci.2020.105047
Hale, A. and Borys, D. (2013). Working to rule, or working safely? Part 1: A state of the art review. Safety Science, 55, 207–221. https://doi.org/10.1016/j.ssci.2012.05.011
Hollnagel, E. (2014). Safety-I and Safety-II: The Past and Future of Safety Management. Boca Raton: CRC Press.
Le Coze, J.C. (2026). Half a century of Safety Science (1976–2026). Divergence and convergence in the field. Safety Science, 203, 107343. https://doi.org/10.1016/j.ssci.2026.107343
Rae, A.J. and Provan, D.J. (2019). Safety work versus the safety of work. Safety Science, 111, 119–127. https://doi.org/10.1016/j.ssci.2018.07.001
Rae, A.J., Provan, D.J., Weber, D.E. and Dekker, S.W.A. (2018). Safety clutter: the accumulation and persistence of ‘safety’ work that does not contribute to operational safety. Policy and Practice in Health and Safety, 16(2), 194–211. https://doi.org/10.1080/14773996.2018.1491147
Rasmussen, J. (1997). Risk management in a dynamic society: a modelling problem. Safety Science, 27(2–3), 183–213. https://doi.org/10.1016/S0925-7535(97)00052-0
Vaughan, D. (1996). The Challenger Launch Decision: Risky Technology, Culture, and Deviance at NASA. Chicago: University of Chicago Press.
Work Health and Safety Act 2011 (Qld), ss 18, 19, 27.
Work Health and Safety Regulation 2011 (Qld), s 299.