Abstract

This dissertation deals with several aspects of the physical interpretation of theories: the use of idealizations in scientific theorizing, the use of theories as conceptual devices and the problem of scientific realism. These issues are considered in the light of the semantic conception of theories and a case history. The case history concerns the development of models of stellar stucture from the earliest models to the "standard model" of Arthur Eddington. ;The semantic conception of theories makes it possible to give formal analyses of several concepts related to idealization: approximation, simplification and abstraction. The concept of idealization stands out as having intrinsic pragmatic and epistemological dimensions which cannot be analyzed in terms of the semantic conception. In light of the case history it is suggested that there is an organization to the concepts which determine the structure of model. Following ideas taken from the artificial intelligence literature, it is argued that the concepts are organized into a kind of associative network which is called a "theoretical network." ;Eddington's "standard model" of the stars is looked at as an example of the use of theories as conceptual devices. The standard model acquired physical significance by degrees over two decades. Its history sheds light on the process by which theories acquire physical significance. An account of the physical interpretation is introduced which relies upon the idea of the theoretical network. On this view theories acquire physical significance by the filling of "slots" within the theoretical network. ;The ideas developed concerning the interpretation of theories are finally used to discuss the problem of scientific realism. A "quasi-realistic" interpretation of theories is developed in contrast to the anti-realistic interpretations of Nancy Cartwright and Ian Hacking. The quasi-realistic position defended asserts that the laws of physics give accurate representations of the world, but these representations are counterfactual. The devices introduced in the dissertation make it possible to understand how such counterfactual representations are evaluated