This paper aims at a systematical analysis of complex methods of measurement. By considering an historical example, namely the determination of the velocity of light, c, by O. Roemer in 1676, we are led to systematical concepts about measurement which seem to be applicable in the philosophy of science in general: "measuring models" and "chains of measurement". We claim that these concepts provide an efficient and adequate frame for the analysis of measurement, especially in more complex cases of "theory-guided" measurement. The claim is supported by the example of the determination of c which nicely fits into the general concepts. We thus apply the strategy so often recommended, namely to use results and examples from the history of science as material, stimulus, and "basis of test" for the philosophy of science.
We do not aim at describing the course of events in a historically complete and satisfactory way. This would require a paper of its own. 2 Rather, we want to make explicit the theoretical, factual and operational assumptions which are necessary in order to draw a coherent, theoretical picture in which the concept and method of determination of c make sense. In this way we obtain a rather complicated picture of the "operational structure" underlying such a "simple" measurement as that of c (compare figure 7 below). Actually, the "real" picture is still a bit more complicated than ours because in a few places (which are indicated) we Use simplifying assumptions which in the 17th century were already known as such. Even with these simplifications the picture is complex enough, and only our general concepts prevent us from loosing the overall survey.
We will use the so-called method of informal set-theoretic predicates in order to state a set of theoretical assumptions (a system of axioms). That is, instead of simply writing down the assumptions, say, A1 .... , A=, we define a set-theoretic predicate P such that P applies precisely to those structures (entities) for which statements A1 ..... A,, are true. 3 Thus what usually is called an axiom, in our treatment will occur as a part of definition (of P). But, of course, defining a predicate P in this way is equivalent to an ordinary axiomatisation (with AI ..... Am