In perhaps the most fundamental sense of "realism", a realist about certain statements holds their truth or falsity to be independent of our ability to verify or to falsify them. This does not imply that we are not in fact in a position to verify or to falsify them, but that it could happen that they were true or false even though we were not in that position. Thus the idealist thesis that reality is entirely mental, non-material, is not in itself incompatible with realism. Consider Berkeley's theory that the truth about ordinary objects is a matter of perceptions in the mind of God; if it is also held that what occurs in God's mind is not dependent on our ability to verify it, then the theory is realist in the relevant sense. Is this not an eccentric use of "realism"? Not at all, for it directly reflects certain central concerns in epistemology and in the theory of concept formation. For Berkeley (in another frame of mind), Kant, the verificationists, and recently Michael Dummett's anti-realist, two questions about realism are thought to be unanswerable: if the reality about which apparently we think and speak were constituted independently of its epistemic accessibility to us, then (1) how could we know about it? and (2) how could we have a bone fide conception even of its possibility? (It is the latter question with which Dummett is primarily concerned.)