Most safety controls are designed once and lived with every day after. A walkway is painted, a speed limit is set, a crossing is placed where the risk assessment said it should be. On the day, all of it fits the site. The interesting question is what happens over the following two years.
Rasmussen, a Danish engineer who spent his career studying how complex work goes right and wrong, offered a picture that has lasted. Every job has freedom in it. Within limits set by cost, workload and safety, people choose how to do the work, and those choices vary from day to day, like "the 'Brownian movements' of the molecules of a gas", in his phrase.
The variation is not random for long. People find what he called "an effort gradient", the easier way, and organisations supply a "cost gradient", the cheaper and quicker way. Both point the same direction. In his words: "The result will very likely be a systematic migration toward the boundary of functionally acceptable performance."
That is drift. Nobody chooses it, and no single step crosses a line.
Drift is not a failure of character. It is what a working system does under steady pressure.
Rasmussen's sharpest point is about why drift stays invisible. Well-run sites protect people with several layers: the layout, the rules, the training, the alertness of the driver. That redundancy is a strength, but it has a side effect. In his words, "a local violation of one of the defences has no immediate, visible effect and then may not be observed in action."
The shortcut works. Nothing happens, because the other layers are still there. So the shortcut becomes the way. He concluded that in such systems "the defences are likely to degenerate systematically through time, when pressure toward cost-effectiveness is dominating."
Diane Vaughan, a sociologist, reached a similar place from a different direction. In The Challenger Launch Decision (1996) she spent years reconstructing how NASA came to launch with a known problem in the booster joints. What she found was not rule-breaking. Erosion of the seals had been seen on earlier flights, assessed, and judged acceptable, because the flights had come back safely. Each judgement made the next one easier. A deviation from the design became, through ordinary engineering process, part of the accepted experience base.
Her name for it, the normalisation of deviance, has travelled far beyond aerospace because the pattern is so familiar. Sidney Dekker later built on both strands in Drift into Failure (2011), arguing that drift comes from normal processes of balancing competing pressures, not from broken parts or careless people.
Take three ordinary controls on a site with vehicles and people on foot. These are illustrations, not real cases.
Every step was reasonable on the day. Each one is exactly what Rasmussen's gradients predict. And because the other defences held, none of them produced an event to investigate. The gap between the written method and the real one widened quietly.
Rasmussen was generous about safety campaigns, and precise about their limits. They create a "counter gradient" that pushes back against the pressure to cut corners, and they work. But, he wrote, "an empirical struggle for a good 'safety culture' will never end because it only works as long as it acts as a continuous pressure compensating the functional pressure of the work environment."
That is not a criticism of the people running the campaigns. It explains why a toolbox talk improves things for a few weeks and why the effect fades: the pressure it counters never stops.
Rasmussen's own answer was the most hopeful part of his paper. The most promising approach, he wrote, is "an explicit identification of the boundaries of safe operation together with efforts to make these boundaries visible to the actors."
On a site, that turns into a few practical habits:
Rasmussen added a point that is easy to miss: making boundaries visible "may also increase system effectiveness", because people can work confidently close to a limit they can see. Safety and efficiency stop pulling against each other. The field now has far better ways to look at ordinary work than it did in 1997, and the thinking he left gives that looking a purpose: to notice the small step before it becomes the method.
A term coined by sociologist Diane Vaughan in The Challenger Launch Decision (1996). It describes how a departure from a standard, once it has happened without harm, gradually comes to be treated as acceptable, until the deviation is the new normal. It happens through ordinary decisions, not through deliberate rule-breaking.
The gradual movement of work practice towards the edge of safe operation under everyday pressures of time, cost and effort. Jens Rasmussen described the mechanism in 1997, and Sidney Dekker's book Drift into Failure (2011) developed it. Each step looks reasonable locally, which is why drift is hard to see.
In his 1997 paper, Rasmussen pictured work moving within limits set by economic failure, unacceptable workload and acceptable performance. Pressure towards least effort and towards efficiency pushes practice towards the safety boundary. Safety campaigns create a counter gradient, but he argued that the most promising approach is to make the boundaries visible to the people doing the work.
No. A violation is a deliberate departure from a rule. Normalisation of deviance describes how a departure stops being seen as a departure at all, because it has become routine and nothing went wrong. The people involved usually believe they are doing the job properly.
Compare current practice with the purpose of each control, look at ordinary shifts rather than audit days, repeat the same check in the same place over time so movement shows up, and ask newer staff how they were taught to do the job.
Published 18 November 2026 ยท Last updated 18 November 2026
This piece starts from the ideas set out in How We Think About Safety.
Wondering what a camera would see at your gate? Request a Safety Mapping โ one camera, two weeks, at no cost.