Rather than attempting to combine the two meta-methodological programmes for justifying the epistemological study of science, which is the case of Laudan's normative naturalism, this paper aims at presenting a third alternative to the controversy between the traditional normativism and the reductionistic naturalism. The paper is a preliminary move in developing a theory of the autopoietic cognitive organization of science. The underlying assumption of this project calls that science is a self-constructing, self-specifying and homeostatic system. The scrutinizing of these three (...) predicates leads to the view that the epistemological propositions about science cognitive organization are neither normative, nor descriptive, but transcendental ones. The final discussion shows the connection between this project and the theory of group rationality. (shrink)