As finding root causes of human error in safety-critical systems is a cognitively demanding and time-consuming task, it
is particularly necessary to develop a method for improving both the quality and efficiency of the task. Although a number
of methods such as CREAM (Hollnagel, 1998) have suggested causal linking between error causes (or performance
shaping factors) as a method for enhancing the quality and efficiency of human error analysis, no published research to date
has evaluated how useful the causal links are. This paper presents a study for evaluating the effectiveness and efficiency of
the causal links between over 100 error causes by a meta-analysis of 78 railway accident investigation reports from the UK.
Two measures, coverage and selectivity, were used to evaluate the effectiveness and efficiency of the links, respectively.
About 96% of error causes actually included in the accident reports were found by just following the causal links, and
among the total of 121 error causes, the number of error causes to be examined further was reduced to one-tenth (about 13)