forthcoming in Internalism and Externalism in Semantics and Epistemology, ed. by

Privileged access is the thesis that a thinker S has a priori access to the propositional contents of her occurrent thoughts, where by ‘a priori ’ we stipulate the inclusion of knowledge based on introspective deliverances. Semantic externalism is the thesis that such contents depend in part for their individuation on features of S’s external environment, where the internal/external distinction is typically drawn around the skin and skull. Incompatibilism is then the view that privileged access is incompatible with semantic externalism. Compatibilism is the negation of incompatibilism. Incompatibilists have deployed various strategies, but here we shall only probe into what Brown (2004: 194) called the consequence problem: the joint assumptions of semantic externalism and privileged access pave the way for a priori knowledge of those external environmental features which partly determine the thought contents in question. But it is supposedly in-tolerable that armchair philosophical theorizing should lead to a priori knowledge of the external world once the thinker has a priori knowledge of her present thought contents.1 So, if the reasoning is cogent, either privi-leged access or semantic externalism has got to go. Inspired by McKinsey’s seminal 1991 paper, Davies (1998) and others have taken semantic external-ists to be faced with an incompatibilist charge posed by the (MC) form:2 (1) S has mental property M; (2) S meets non-mental condition C if she has mental property M; (3) S meets non-mental condition C.

The argument known as the 'McKinsey Recipe ' tries to establish the incompatibility of semantic externalism (about natural kind concepts in particular) and a priori selfknowledge about thoughts and concepts by deriving from the conjunction of these theses an absurd conclusion, such as that we could know a priori that water exists. One reply to this argument is to distinguish two different readings of 'natural kind concept': (i) a concept which in fact denotes a natural kind, and (ii) a concept which aims to denote a natural kind. Paul Boghossian has argued, using a Dry Earth scenario, that this response fails, claiming that the externalist cannot make sense of a concept aiming, but failing, to denote a natural kind. In this paper I argue that Boghossian's argument is flawed. Borrowing machinery from two-dimensional semantics, using the notion of 'considering a possible world as actual', I claim that we can give a determinate answer to Boghossian's question: which concept would 'water ' express on Dry Earth? I Semantic externalism, as understood in this paper, is the view that the propositional content of