Truly Empiricist Semantics

In van Fraassen's The Scientific Imagewe are told that the scientific anti-realist need not appeal to some special (e.g. reductionist) semantics for scientific language. He can allegedly hold – just like his direct opponents typically do – that truth-conditional semantics is appropriate both for claims about the observable and claims about the unobservable. However, I shall point out that this kind of semantics goes badly with the anti-realist's epistemic attitude vis-à-vis the unobservable. In this paper an alternative semantics shall be outlined that is perfectly compatible with the rest of the anti-realist's position. Although this alternative semantics will be seen to fly in the face of the established doctrine in the field, I shall argue that there is currently no good reason not to adopt it. It will further be pointed out that this semantics has something to be recommended from a realist perspective as well.