In his first paper, ‘Must We Mean What We Say?’, Stanley Cavell defended the methods of ordinary language philosophy against various charges made by his senior colleague, Benson Mates, under the influence of the empirical semantics of Arne Naess.1 Cavell’s argument hinges on the claim that native speakers are a source of evidence for 'what is said' in language and, accordingly, need not base their claims about ordinary language upon evidence. In what follows, I maintain that this defence against empirical (...) semantics applies equally well to experimental philosophy's attack on doing philosophy from the armchair. In so doing, I attempt to clarify – and adjust – Cavell's claim that statements about ordinary language are rule-descriptions that are neither analytic nor synthetic. (shrink)