Abstract

I argue for an alternative to invariantist, contextualist, and relativist semantics for ‘know’. This is that our use of ‘know’ is inconsistent; it is governed by several mutually inconsistent inference principles. Yet this inconsistency does not prevent us from assigning an effective content to most individual knowledge ascriptions, and it leads to trouble only in exceptional circumstances. Accordingly, we have no reason to abandon our inconsistent knowledge-talk.