Theory Reduction in the Social Sciences. The example of balance theories. A central topic both in philosophy of science as well as in the empirical sciences is the reconstruction of the relations between theories. In the past comparisons of theories by means of traditional linguistic methods have proved to be extremely difficult and complicated. In this article the reconstruction of intertheoretical relations based on model-theoretical terms is propagated, as formulated within the structuralist view of theories. The efficiency of a model (...) theoretical based comparison of theories is demonstrated by two theory elements from the social science research program of balance theories: The basic element by Heider and the transitivity theory by Holland and Leinhardt. First of all both theory elements are introduced informally and reconstructed in the structuralist format. On the basis of these reconstructions can be shown, that the Heider theory can be formally reduced to the Holland-Leinhardt theory and that the theory younger in history means an improvement. However, an integration of all balance theoretical elements into a theory net is not possible. (shrink)

Several case studies and theoretical reports indicate that the structuralist concept of a constraint has a central role in the reconstruction of physical theories. It is surprising that there is, in the literature, only little theoretical discussion on the relevance of constraints for the reconstruction of social scientific theories. Almost all structuralist reconstructions of social theorizing are vacuously constrained. Consequently, constraints are methodologically irrelevant.In this paper I try to show that there really exist constraint-type assumptions in mathematical modelling in the (...) social sciences. Methodologically constraints have exactly the same role in the context of social mathematical modelling as they have in physical theories. In typical cases of mathematical modelling in the social sciences, the related constraints work as empirical hypotheses and should be tested by statistical means. (shrink)