I like a passage in a book by the (very) mathematical physicist Robert Georch:
"It seems to me that "theories of physics" have, in the main, gotten a terrible press. The view has somehow come to be rampant that such theories are precise, highly logical, ultimately "proved". In my opinion, at least, this is simply not the case - not the case for general relativity and not the case for any other theory in physics. First, theories, in my view, consist of an enormous number of ideas, arguments, hunches, vague feelings, value judgements, and so on, all arranged in a maze. These various ingredients are connected in a complicated way. It is this entire body of material that is "the theory". One's mental picture of the theory is this nebulous mass taken as a whole. In presenting the theory, however, one can hardly attempt to present a "nebulous mass taken as a whole". One is thus forced to rearrange it so that it is linear, consisting of one point after another, each connected in some more or less direct way with its predecessor. What is supposed to happen is that one who learns the theory, presented in this linear way, then proceeds to form his own "nebulous mass taken as a whole". The points are all rearranged, numerous new connections between these points are introduced, hunches and vague feelings come into play, and so on. In one's own approach to the theory, one normally makes no attempt to isolate a few of these points to be called "postulates". One makes no attempt to derive the rest of the theory from postulates. (What, indeed, could it mean to "derive" something about the physical world?) One makes no attempt to "prove" the theory, or any part of it. (I don't even know what a "proof" could mean in this context. I wouldn't recognize a "proof" of a physical theory if I saw one.) "
