Epistemological foundationalism has typically been thought to hold that in order to account for human knowledge we must countenance the direct Justification of some specific kind of beliefs, such as one's beliefs to the effect that one is having a certain sensation. How else, it may be thought, can one analyse Justification without confronting an infinite regress or a vicious circle? I believe that this conception of foundationalism has been so influential that most foundationalists and nearly all their critics have failed to appreciate that foundationalism may be plausibly construed as a thesis mainly about the structure of a body of Justified beliefs. Central to the thesis, so interpreted, is that one's Justified beliefs divide into foundations and superstructure; but no particular content on the part of either set of beliefs need be required. This latitude regarding content is altogether appropriate; for if we use, as a guide to understanding foundationalism, the famous regress argument, from which the thesis derives much of its plausibility, then the only foundations required by the thesis are beliefs whose Justification does not depend on that of other beliefs. Precisely what beliefs these are is a controversial matter on which foundationalists may differ.
What is an Infinite Regress Argument?