A warranted belief may derive inferential warrant from warranted beliefs which support it. It may possess what I call coherential warrant in virtue of being consistent with, or lacking improbability relative to, a large system of warranted beliefs. Finally, it may have foundational warrant , which does not derive from other beliefs at all. I define and distinguish these sources of warrant and explain why all three must be included in the true and complete account of the structure of knowledge, and why the first two sources are significant at all levels of knowledge. Only foundherentism and a weak version of foundationalism can satisfy this criterion. My analysis has significant, and happy, consequences for the epistemological tradition. The project of describing the structure of knowledge is nearly complete. Those who have pronounced the death of epistemology are partially correct, not because epistemology has failed, but because it has been so successful.