This work is an analysis of F. P. Ramsey's philosophy of science. Twentieth-century philosophy of science was marked by attempts to consider the relation between scientific theories and our knowledge of the empirical world through considerations of abstract mathematical structure. Such considerations led Bertrand Russell to an account of the relation between our theoretical picture of the world and its real nature as a relation of structural similarity. Subsequently, Max Newman gave what has become a well-known logico-mathematical objection to this (...) account. William Demopoulos recently showed that Newman's problem applied not only to Russell's realist account, but also to a variety of otherwise disparate accounts of theoretical knowledge. The common element underlying these accounts is a conception of theories as abstract formal structures. Many such accounts have incorporated key elements of Ramsey's views, most notably the Ramsey-sentence. Moreover, Demopoulos has interpreted Ramsey's own view of theories as sharing the essential features of those abstract views, and therefore their common problem. My analysis aims to show that this abstract conception of theories does not adequately characterize Ramsey's view. Namely, his account of theories was not an attempt to do the epistemology of science in the fashion of Russell or Eddington, or of subsequent structuralist views that have adopted the Ramsey-sentence. I show this by a broader exposition of Ramsey's work on the nature of theories, comparing his seminal paper with his many other remarks on the nature and purpose of theories. I begin by discussing the historical context of Newman's objection, and a generalization of it that shows its broad applicability to abstract characterizations of theoretical knowledge. I then reconstruct Ramsey's view of theories, to show how far it extends beyond the Ramsey-sentence picture. Finally, I discuss the relevance of this view to contemporary debates concerning realism and instrumentalism. I characterize Ramsey's view as focused not on grounding our theoretical knowledge in abstract structure, but instead on demystifying the role of theoretical language and concepts in a theory's application to the world. (shrink)