The Paddington and Southall rail crashes have focussed public concern – and therefore government attention – on rail safety. This may lead to structural and regulatory changes, signals siting reviews, the introduction of new technology (TPWS, ATP) etc. However the obvious role of human error in both of these accidents has not been directly addressed. The training methodology to avoid and mitigate human error accidents has been developed over 20 years in aviation and adapted in a variety of contexts including US Marine pilots, airport ground handling staff, maintenance engineers, European ATM controllers, power plant operators etc. It is even used as part of hospital Operating Theatre staff training. However rail TOCs have not made any effort to implement CRM-type training for any operational or onboard staff (drivers, guards, ticket examiners).

The only significant training analysis and development effort for rail human factors was undertaken by Scotrail on the eve of privatisation. This has had one post-privatisation result – the CIRAS confidential incident reporting project, which is now well-established in Scotland.

What does Human Factors (HF) Offer? Each accident is the result of a causal chain which includes the active failure, failed defences, deficient procedures, risky workplace practices and an unsafe corporate culture (which includes unsafe policies and inadequate safety resources). Only when all these factors are present is an accident likely.

Each accident is preceded by precursor incidents and detectable failures – warnings that an error chain is building. Typically there will be an accident for each 100 incidents. This is termed the “disaster incubation period”. If the incidents are allowed to persist an accident becomes inevitable – despite the fact that each accident appears unique, a “one in a million” chance.

Every accident can be prevented by breaking any one or two of the links in the error chain.

Human Factors models seek to avoid error and mitigate its effects by examining all of the conditions present in the workplace, as well as technology defences and operational training. Simply, it makes it possible to learn from incidents and accidents and take effective steps to prevent them by breaking error chains before they produce disasters.

How effective is it? By July 1996, Delta Airlines experienced $8million in ground handling damages per 10,000 departures. (That is, direct damage from incidents such as backing a baggage wagon into a wingtip.) That rate was increasing both per 1000 flights, and absolutely with the growth in traffic volume. By introducing a comprehensive ground handling HF policy and training system, damage rates dropped in 2 years to less than $3million per 10,000 departures. Even with a nearly 40% increase in total flights, the absolute number and cost of ground incidents declined significantly.

The Downside The problem is clearly understood, the solution is available, but it has not been implemented in any concerted way by the rail industry. Whether this situation is changing it is too early to tell. The Inquiries from Southall and Paddington will inevitably point to training and operational deficiencies. Individual TOCs have expressed interest in HF programme aspects, but the financial equation requires both stick and carrot – unless the industry can see a way to recover additional training outlay there is no guarantee that HF training will be widely undertaken without explicit regulation of the sort adopted by the CAA.

In this context, it would obviously be helpful if a significant body of knowledge and experience could be brought to bear to show that:

  • the cost of implementing HF programmes is trivial compared to the cost of accidents
  • HF solutions are do-able, do not require enormous new resources, and improve operating efficiency and profitability.
  • At 600+ SPADs/year further disasters are inevitable
  • Technology defences alone are inadequate

If aviation experience and the lessons from Southall and Paddington are reliable, then simply adding to the number and complexity of technology defences will not solve the problem. In some situations it will make it worse (more systems to operate, maintain and fault, less transparency, poorer risk management by drivers). Correctly addressing human error requires consistent effort to maintain effective quality management of the safety process over an indefinite time period.