Abstract
A thumbnail sketch of the philosophical thinking about the a priori would surely include that it has been dominated by two major approaches: the Kantian absolute conception of it and the Millian-Quinean absolute rejection of it (section 2). Yet, one can find in the literature claims about the existence of a ›functional a priori‹, a ›relative a priori‹, a ›relativised a priori‹ and suchlike. They are all meant to carve a space between the two extremes. An important thought behind the search for a middle ground is that the supposed coincidence between the constitutive and the unrevisable is wrong. The entitlement to accept a principle as being constitutive of experience prior to any empirical justification of it is compatible with an entitlement to revise or abandon such a principle on empirical grounds. If a priori principles are meant to be independent of experience, how should this claim of independence be understood so that room is left for the possibility that a principle is both independent of experience and revisable on empirical grounds (section 3)? A straightforward and natural way to approach this issue is to think of constitutive principles along the lines of Poincaréan conventions, which can be seen as delineating a new sense of the a priori – the conventional a priori principles. These are substantive principles that are constitutive of theoretical frameworks – in the sense that they define (or constitute) the object of knowledge – without being either synthetic a priori or empirical generalisations. Still, their negation is conceivable and they are revisable on empirical grounds (section 4).