In an essay of the early sixties, Feyerabend presented his point of view as a resumption of certain aspects of Kant's philosophy. He rejected the idea of absolutely valid transcendental conditions of knowledge, but retained that of all-pervasive theoretical assumptions as ways of looking at the world: "what is regarded as "nature" at a particular time is our own product in the sense that all features ascribed to it have first been invented by us and then used for bringing order into our surroundings".¹ This idea is partly incorporated in the network model which I just introduced, since in the network model, while our cognitive claims are empirically testable, a considerable role is set aside for linguistic, theoretical and methodological-axiological epistemic conditions.