There are many ways of rejecting realism and thus many sorts of anti-realist. In this chapter we shall confine our attention to antirealist positions spawned by the supposed success of the challenge to realism which we looked at in Chapter 4.

If the challenge to realism is successful we cannot give an account of meaning that makes use of a semantic notion that is verification transcendent. In particular, in the presence of undecidable sentences, we cannot appeal to a bivalent notion of truth. Now it might seem (and Dummett often gives this impression) that a consequence of this is a rejection of truth-conditional accounts of meaning. That is, we would give an account of meaning in terms other than truthconditions and then allow an account of truth to issue perhaps in Tarskian fashion. But this isn't obvious. It may be that we can give an account of an anti-realistically acceptable notion of truth and then use this notion in a truth-conditional account of meaning.

What notions might an anti-realist appeal to in providing such accounts? An anti-realist will only make use of a notion which, when applied to a sentence, yields something that can be ascribed as implicit knowledge to speakers and which thus can be manifested by an appropriate exercise of recognitional capacities. So the notions that might be put to work by an anti-realist will be epistemic: for instance, notions of justification, warranted assertion, verification, falsification and so on.