Abstract

In his recent Philosophers’ Imprint paper “The (mostly harmless) inconsistency of knowledge attributions” [Weiner, 2009], Matt Weiner argues that the semantics of the expression “knows that”, as it is used in attributions of knowledge like “Hannah knows that the bank will be open,” are inconsistent, but that this inconsistency is “mostly harmless.” He presents his view as an alternative to the invariantist, contextualist and relativist approaches currently prevalent in the literature, (e.g. [Stanley, 2005], [DeRose, 1995], [Hawthorne, 2006], [MacFarlane, 2005]) and argues that it avoids important disadvantages of each. Yet in calling the supposed inconsistency of knowledge attributions “mostly harmless”, Weiner implies that his view does not have new disadvantages of its own. My purpose in the present paper is to argue that the inconsistency and harmlessness theses cannot be jointly maintained: if we accept that the semantics of ‘know’—or indeed any word—are inconsistent, then we face a dilemma: one horn is dialetheism, the view that there there are true contradictions, the other is the view that that semantic competence in English requires belief in, or similar commitment to, falsehoods. I will argue that neither of these options is well described as “mostly harmless.” The paper is structured as follows: in the first part I present Weiner’s view and his arguments for it. Then in section 2 I compare the question of whether the semantics of ‘knows that’ are inconsistent to the much older controversy over whether the semantics of the expression ‘is true’ are inconsistent. In section 3 I will present Hans Herzberger’s arguments from the 1960s for thinking that no expression in a natural language can have inconsistent semantics. Finally, in section 4 I argue that although Herzberger’s argument seems anachronistic today, both contemporary ways of avoiding his conclusion have significant disadvantages.