ABSTRACT
This chapter describes how complementary triangulation has often been preferred to validatory triangulation. It shows that the preference relies on an epistemological distinction between quantitative and qualitative data. The company are testing the new hypothesis, derived now from numbers, words and more numbers, by examining train despatch procedures at stations on certain routes, and other aspects of use of the sets. 'Triangulating' quantitative and qualitative data traditionally involves viewing these as distinct, and 'adding them together' to produce a 'complete' picture. This is called complementary triangulation and this is the method most often adopted by safety managers. One way to avoid the effort required to deal with discourse, of course, is to use rating scales or questionnaires to turn discourse into numbers. If the incident is the area between actual and acceptable operations and the latter is contextually variable, then incident frequency data are no more or less objective than any other categorical data.
