In this paper, we investigate the constitutive problems and other several aspects of what a research entitled 'formal epistemology' should be. The interest in this subject has to do with the possibility of reaching a privileged point of view or axis of research - i.e., the 'formal' one - that would allow, a better grasp of the richness and variety of the facts and problems tackled by precise (local) epistemology of theories (for example, in physics). This approach is likely to (...) enable one to hold the main structural lines that organize those theories according to a more com prehensive, unifying and synthetic intelligibilit. By the same token, it would eventually allow a better handling of the changes required in the organization of knowledge, putting emphasis on its main directions, drawing up a rational inventory of this knowledge, and perhaps anticipating others. At first, we deal with the `thought of changes' that no approach of the 'form' can afford to leave aside, since the meaning of this concept is inseparable from the contents that come with constructions and modifications. We examine then the notion of 'epistemic operation' as an instrument to create new forms on the theoretical as well as on the meta-theoretical levels. ln the wake of it, we analyze the characteristics of the form and of the formal, as well as their relationship with the contents of knowledge. We also take the notion of object into account, since it depends upon the decision of a subject and upon conventional choices. We finally inquire about the link between `epistemic operations' as specified above and algorithmic functions for knowledge statements, and emphasize the risk of reductionism that might follow from a naturalistic conception of representation. (shrink)